Characterizing the Microbiota of Bovine Digital Dermatitis

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

Graduate Studies The Vault: Electronic Theses and Dissertations

2021-03-24 Characterizing the microbiota of bovine digital dermatitis

Caddey, Benjamin

Caddey, B. J. (2021). Characterizing the microbiota of bovine digital dermatitis (Unpublished master's thesis). University of Calgary, Calgary, AB. http://hdl.handle.net/1880/113188 master 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

Characterizing the microbiota of bovine digital dermatitis

Benjamin Jordan Caddey

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF MASTER OF SCIENCE

GRADUATE PROGRAM IN VETERINARY MEDICAL SCIENCES

CALGARY, ALBERTA

MARCH, 2021

Abstract

Bovine digital dermatitis (DD) is a skin disease affecting cattle worldwide. As a significant contributor to infectious lameness, DD is increasingly an economic and animal welfare concern for dairy and beef producers. DD is a polymicrobial disease, with many different types of anaerobic bacteria strongly associated with lesions. No causative agents or etiopathogenesis mechanisms of DD are yet accepted due to the multiple different species present and microbiota variation among individual lesions. The involvement of Treponema in lesion development is generally accepted, whereas what combination of other anaerobic bacteria are involved is currently debated. Insufficient and incomplete identification and characterization of these additional species are a limiting factor in DD pathogenesis research. This thesis aimed to fill the gaps in knowledge of these additional anaerobes throughout DD lesions, while providing the first comprehensive description of DD microbiota in feedlot beef cattle. Through high throughput sequencing and culturing of DD tissue biopsies, we identified Treponema, Mycoplasma,

Fusobacterium spp., Porphyromas levii, and Bacteroides pyogenes as potential DD pathogens, and used a multiplex qPCR for absolute quantification and reproducible characterization of species population dynamics. We identified T. medium, P. levii, and T. phagedenis as a group strongly associated with early lesion stages, thus potentially being involved in lesion formation.

Through a meta-analysis of all publicly available DD metagenomic studies, we identify

Treponema, Mycoplasma, Fusobacterium, and Porphyromonas as the primary DD-associated microbiota. We recommend focusing future DD research efforts on culturing and characterizing species of these groups in an effort to establish etiopathogenesis mechanisms of DD.

ii Preface

This thesis contains two manuscripts prepared for submission. The following manuscripts included in this thesis are:

Chapter 2

Caddey, B, Orsel, K, Naushad S, Derakhshani H, De Buck J. 2020. Identification and quantification of bovine digital dermatitis-associated microbiota across lesion stages in feedlot beef cattle.

Chapter 3

Caddey B, De Buck J. 2020. Meta-analysis of bovine digital dermatitis microbiota reveals distinct microbial community structures associated with lesions.

iii Statement of Contribution

Ben Caddey (BC), Jeroen De Buck (JDB), and Karin Orsel (KO) designed the study for chapter

2. Biopsy sampling protocols for the collection of skin biopsies were designed by BC, JDB, and

KO. Sampling was conducted by BC, KO, Anice Thomas, Julian Cortes, Lacey Wynnyk, and

Tina Petersen. Biopsy handling and processing, including anaerobic culture of Treponema and other anaerobes, DNA extraction, and sample preservation were performed by BC. Purified biopsy DNA was sequenced against the V3V4 hypervariable region of the 16S rRNA gene by the McMaster Genomics Facility. BC performed all processing and analysis of metagenomic sequencing data. Anaerobic bacteria from DD biopsies were isolated and identified by BC, and

DNA extractions were performed by BC before Hooman Derakhshani sequenced and assembled draft genomes of isolates. Species-specific genes for qPCR targeting were identified by Sohail

Naushad and BC. The novel multiplex qPCR was designed and validated by BC, and all qPCR reactions and analysis were performed by BC. The study for Chapter 3 was designed by BC and

JDB. All studies eligible for inclusion were identified by BC, and data collection and analysis was performed by BC. Data interpretation for both chapters was conducted by BC and JDB. All manuscripts and chapters in this thesis were written by BC under the guidance and feedback of

JDB.

iv Acknowledgements

First and foremost, I would like to thank my supervisor, Dr. Jeroen De Buck, for his role in this transformative journey over the past couple years. The moulding of your supervision has led to immense personal growth, and developed confidence in my aptitude for research. Thank you for trusting in my capabilities and allowing me the freedom to explore my research interests without hesitation. I will fondly remember our meetings that were always filled with creative and ambitious dialogue, and occasionally overflowing with out-of-this-world ideas, usually requiring one of us to bring the other back to reality.

Thank you to my supervisory council. Your knowledge, experience, and insights were invaluable to this work and my personal development. Thank you to Dr. Doug Morck for helping kickstart my progress on this journey by providing anaerobic cultures so that I could get well-acquainted with fastidious organisms. Dr. Karin Orsel, many-many thanks for your constant support and shared interest in my dairy-crazy passions. You were pivotal in improving my critical thinking, scientific reasoning, and conceptualization of research on a grander scale, always with a producer-first mentality. Also, thank you to all the veterinarians, producers, and feedlot workers that supported our research and for your continued enthusiasm.

To my lab mates, thank you for the years of light-heartedness and support to create a welcoming lab environment. To Rakel Arrazuria, you inspired a rigorous criticism for every step of the scientific process. Your mentorship went above and beyond, and I can’t thank you enough for your friendship and support. To my fellow graduate students and the UCVM, who facilitated a community of support and teamwork. Thank you, Lisa Gamsjaeger, for the mutual support

v during the rollercoaster ride that is graduate school. I am incredibly grateful for your friendship and all the adventures we had that made Calgary feel like home. To all my equally crazy mountaineering friends, thank you for all the amazing and insane trips outside of the city. The mountains will never stop calling.

To my parents and grandparents, you have all supported my goals and ambitions unconditionally, even when it means losing a pair of hands for baling small-square hay in 40 ºC weather for a couple summers! You all motivate me to be my best every day.

Thank you, all.

vi Dedication

To Ron and Pat Jackson,

who have always been there for me,

every step of the way.

vii Table of Contents

Abstract ...... ii Preface ...... iii Statement of Contribution ...... iv Acknowledgements ...... v Dedication ...... vii Table of Contents ...... viii List of Tables ...... x List of Figures ...... xi List of Abbreviations ...... xii Chapter 1: Introduction ...... 14 Bovine digital dermatitis ...... 14 DD in beef cattle ...... 15 Related lesions ...... 15 Lesion scoring ...... 16 Risk factors ...... 17 Mitigation and treatment strategies ...... 19 Causation ...... 20 Infection models ...... 21 Treponema ...... 22 Porphyromonas ...... 23 Mycoplasma ...... 24 Fusobacterium ...... 25 Other associated bacteria ...... 25 Gaps in knowledge of DD microbiology ...... 26 Objectives ...... 27 References ...... 28

Chapter 2: Identification and quantification of bovine digital dermatitis-associated microbiota across lesion stages in feedlot beef cattle ...... 41 Abstract ...... 41 Importance ...... 42 Introduction ...... 42 Methods ...... 45 Sampling strategy and biopsy collection ...... 45 Biopsy processing ...... 46 Deep amplicon sequencing and analysis ...... 47 Whole genome sequencing and analysis ...... 49 Multiplex qPCR development and validation ...... 50 Quantitative real-time PCR and analysis ...... 51 Data availability ...... 52 Results ...... 52 Deep amplicon sequencing of the V3-V4 hypervariable region ...... 53 Identification of bacterial isolates from skin biopsies ...... 55

viii Quantitative real-time PCR (qPCR) design ...... 55 Species absolute quantification directly from purified biopsy DNA ...... 56 Absolute abundance of species in healthy skin across farms ...... 57 Comparison of qPCR and deep amplicon sequencing quantification methods ...... 57 Discussion ...... 58 Acknowledgments ...... 64 References ...... 64 Tables ...... 71 Figures ...... 72 Supplemental Material ...... 78

Chapter 3: Meta-analysis of bovine digital dermatitis microbiota reveals distinct microbial community structures associated with lesions ...... 84 Abstract ...... 84 Importance ...... 85 Introduction ...... 85 Methods ...... 88 Data collection and selection criteria ...... 88 Sequence processing ...... 88 Data analysis ...... 89 Results ...... 91 Study selection and characteristics ...... 91 Characterization of microbiota in DD positive and DD negative skin ...... 92 Study variables influencing the microbiota of DD positive skin ...... 94 Discussion ...... 95 Conclusion ...... 101 Acknowledgements ...... 101 References ...... 102 Tables ...... 109 Figures ...... 113 Supplemental Material ...... 118

Chapter 4: Summarizing Discussion ...... 121 DD lesion microbiota: feedlot beef cattle and beyond ...... 121 Key DD-associated microbiota ...... 122 Insights into DD lesion establishment and progression ...... 123 Insights into DD transmission and infection dynamics ...... 125 Insights into future treatment and therapeutics ...... 126 Tools we have developed ...... 127 Next steps ...... 128 Significance for producers and researchers ...... 129 References ...... 130

Appendices ...... 135

ix Appendix A ...... 135

List of Tables

Table 2.1: Bacterial genomic copies (mean ± SD) per mg of biopsy tissue of each bacterial species at each M stage ...... 71 Table 2.2: Percent of samples with detectable amounts of each bacterial species ...... 71 Table S2.1: Primers and probes used to sequence and quantify DD-associated bacteria ...... 78 Table S2.2: Number of species isolates collected from each DD lesion stage ...... 80 Table S2.3: DNA spike-in to evaluate performance of multiplex qPCR detection directly from biopsy DNA ...... 80 Table S2.4: Bacterial genomic copies (mean ± SD) per mg of biopsy tissue of each bacterial species at each farm for healthy (M0) samples ...... 81 Table S2.5: Percent of healthy skin biopsies with detectable amounts of target bacterial species at each farm ...... 81 Table 3.1: Characteristics of studies and sequencing experiments included in meta-analysis ... 109 Table 3.2: Alpha diversity estimates of DD negative and DD positive skin microbiota ...... 111 Table 3.3: Permutation analysis of variance and analysis of multivariate homogeneity of group dispersions on Bray-Curtis distances of DD negative and DD positive skin microbiota grouped at different taxonomic levels ...... 111 Table 3.4: Evaluation of random forest model performance on classifying DD status from microbial composition at three taxonomic levels ...... 111 Table 3.5: Study variable PERMANOVA analysis on Bray-Curtis distances of DD positive skin microbiota grouped at different taxonomic levels ...... 112 Table 3.6: Study variable BETADISP analysis of Bray-Curtis distances between DD positive skin microbiota grouped at different taxonomic levels ...... 112 Table 3.7: Alpha diversity estimates of DD positive skin microbiota across different 16S rRNA hypervariable regions ...... 112 Table S3.1: Proportion of DD negative and DD positive skin samples containing at least one read of each genus ...... 118

x List of Figures

Figure 2.1: Principal coordinates analysis on A) unweighted and B) weighted UniFrac distances ...... 72 Figure 2.2: Percent relative abundance of bacteria within each M stage ...... 73 Figure 2.3: Differential abundance analysis of genera associated with DD lesions ...... 74 Figure 2.4: Percent relative abundance of bacterial species within each M stage ...... 75 Figure 2.5: Absolute abundance of each species across all M stages ...... 76 Figure 2.6: Correlation network analysis of species absolute abundances in DD lesions ...... 76 Figure 2.7: Percent relative abundance of species in each M stage compared by quantification method ...... 77 Figure S2.1: Analysis of Illumina MiSeq read counts in sequencing controls ...... 82 Figure S2.2: Pairwise associations between log copy numbers of species in DD lesions ...... 83 Figure 3.1: Principal coordinates analysis of Bray-Curtis distances of the microbiota in DD negative and DD positive skin ...... 113 Figure 3.2: Mean percent relative abundances of microbiota in DD negative and DD positive skin ...... 114 Figure 3.3: Relative importance ranking of random forest classifier at genus level taxonomy . 115 Figure 3.4: Principal coordinates analysis on Bray-Curtis distances of DD positive skin microbiota ...... 116 Figure 3.5: Mean percent relative abundance of DD negative skin and DD lesions from 16S rRNA hypervariable regions V1V2, V1V3, and V3V4 ...... 117 Figure S3.1: Relative abundance heatmap of bacterial families on each sample included in the meta-analysis ...... 119 Figure S3.2: Heatmap of average relative abundances of DD positive microbiota within each BioProject ...... 120

xi List of Abbreviations

ANOVA Analysis of Variance

ASV Amplicon Sequence Variant

ATM Anaerobic Transport Media

BETADISP Multivariate Homogeneity Of Groups Dispersions (Variances) bp Base pair

Bpyo Bacteroides pyogenes

CODD Contagious Ovine Digital Dermatitis

DD Bovine Digital Dermatitis

DNA Deoxyribonucleic Acid

FAA Fastidious Anaerobe Agar

Fuso Fusobacterium sp.

Fn Fusobacterium necrophorum

ID Interdigital Dermatitis

NCBI National Center for Biotechnology Information nt Nucleotide

OTU Operational Taxonomic Unit

PCoA Principal Coordinates Analysis

PCR Polymerase Chain Reaction

PERMANOVA Permutational Analysis of Variance

Plev Porphyrmonas levii qPCR Quantitative Polymerase Chain Reaction rRNA Ribosomal RNA

xii SRA Sequence Read Archive

Tde Treponema denticola

Tme Treponema medium

Tpe Treponema pedis

Tphg Treponema phagedenis

WGS Whole Genome Sequences

xiii 1 Chapter 1: Introduction

2 Bovine digital dermatitis

3 Digital dermatitis (DD) is a significant cause of infectious lameness in cattle (1). First

4 characterized in 1974 Italy (2), DD has since been identified across the world in most countries

5 with major dairy production systems (3–9). As a multifaceted disease with multiple

6 morphological presentations, DD lesions are described in multiple ways throughout its temporal

7 progression. Commonly separated into two categories, lesions can appear as: 1) moist

8 circumscribed and ulcerative lesions, with a red granular appearance that can be painful and

9 bleed upon external pressure; and 2) chronic lesions that are identifiable from the hyperkeratotic

10 growth and papilliform and/or hair-like raised projections (10). Herd-level and cow-level

11 prevalence estimates vary in different studies, but estimates are as high as 93% of farms and up

12 to 21% of cattle (4). While not always the outcome for every DD case, the primary clinical

13 symptom of interest for DD is lameness (1). Lameness is a significant economic and animal

14 welfare concern, and can vary across a range of severities (11, 12). Economic losses associated

15 with DD and lameness for dairy cattle are multidimensional, including decreased milk

16 production, decreased fertility rate, and increased culling rate (13, 14). Infections are primarily

17 associated with multiple Treponema spp., but DD is a polymicrobial disease with multiple types

18 of bacteria present in lesions (15, 16). After over 40 years since DD was first described, we still

19 do not understand the etiological agents or pathogenesis mechanisms, and consequently we are

20 struggling to identify effective treatment options.

14 22 DD in beef cattle

23 Hoof issues resulting in lameness of feedlot beef cattle are predominantly foot rot (interdigital

24 phlegmon) (17); however, DD is recently emerging in feedlot beef cattle populations (18).

25 Relative to dairy breeds, the scientific literature surrounding beef cattle DD lesions is limited.

26 There are a limited number of DD prevalence estimates on feedlots, but cow-level estimates have

27 ranged from 1–11% (6, 19); whereas, it is estimated that between 25–50% of feedlot animals in a

28 Midwestern USA farm have a DD lesion at least once during their lifetime (20). Economic losses

29 concern reduced average daily gain, which is shown to be approximately 0.08–0.028 kg/day for

30 animals depending on how many active lesions were observed in an animal throughout the

31 feeding cycle (20). The mean time for diagnosis of DD is after 100 days on feed, which means

32 DD mostly affect heavier cattle on feedlots (21). Lameness in heavier feedlot cattle on their

33 finishing diets are problematic for producers due to shipping and processing plant regulations

34 discriminating the transport and sale of lame animals.

36 Related lesions

37 There are a number of other diseases that are related to DD, in terms of bacterial groups involved

38 and/or localization to the hoof area. There are other skin lesions present in dairy cattle, such as

39 ischaemic teat necrosis (22), udder lesions (23, 24), and pressure sores (25) that all contain DD-

40 associated Treponema. The microbiology of contagious ovine digital dermatitis (CODD) appears

41 to be similar to bovine DD; however, the morphology of CODD lesions are often more severe,

42 resulting in massive hoof erosion and frequent severe lameness (26). Treponema-associated hoof

43 disease is recently emerging in Elk populations, with the same main Treponema spp. as in bovine

44 DD, but similar morphological presentations and severity as CODD (27).

15 45

46 Other infectious diseases affecting the hoof area of cattle include interdigital phlegmon (foot rot),

47 interdigital dermatitis (ID), and toe tip necrosis. Interdigital dermatitis in cattle presents in the

48 interdigital space of the foot, and may have a similar microbiota as in DD, but causation is

49 mostly associated with Dichelobacter nodosus presence (28, 29). Outside of animal research,

50 human periodontal disease has a similar microbiology as DD. Periodontal disease is a

51 polymicrobial infection of an anaerobic environment, with strong association of Treponema

52 involvement (30). The better understood infection mechanisms of periodontal pathogens offer

53 useful models for research in DD microbiology.

55 Lesion scoring

56 Bovine DD lesions progress through non-linear temporal changes in morphological

57 presentations. There are two common scoring systems for DD lesions: the M stage system (10,

58 31), and the Iowa DD scoring system (32). Scoring DD into M stages is the most common; it

59 categorizes DD lesions on a visual basis with a well-defined set of characteristics and a

60 satisfactory interobserver agreement for most stages (33, 34). There are 5 different M stages: M1

61 lesions are small ulcerative lesions that are less than 2 cm in diameter; M2 lesions are larger

62 lesions that are also circumscribed and ulcerated; M3 lesions present as circumscribed (healing)

63 lesions with a thick scab; M4 lesions are chronic stages showing hair-like papilliform projections

64 and raised rubbery appearance; and M4.1 lesions are chronic lesions with a small focal ulcerative

65 lesion. Lesions can be separated into clinically relevant statuses as active lesions (M1, M2,

66 M4.1) and inactive lesions (M3, M4), as lesions do not progress linearly throughout the M

67 stages. The Iowa scoring presents a similar grading schematic, but early lesions are split into two

16 68 pathways based on localization of initial lesions before progressing to ulcerated and chronic

69 lesions (32). The different lesion stages may be the result of complex microbial community

70 structure dynamics throughout lesion progression (32, 35); and discerning the microbial species

71 involved and their potential interactions may help explain the complex morphology and

72 pathogenesis of DD.

74 Risk factors

75 Housing style and management practices influence risk of DD prevalence. Pasture access is

76 associated with reduced risk of DD compared to indoor housing (8). Grooved concrete flooring

77 is associated with a higher risk of DD compared to slatted flooring, and dirt or pasture has the

78 lowest risk of DD (8). Farms using a rotary milking system also have higher risks for increased

79 DD prevalence compared to farms with a herringbone parlour system, although these differences

80 may be due to increased ease of DD detection in rotary parlours (36). Co-raising young stock

81 from multiple farms together, as well as buying in cattle results in a higher DD prevalence (36,

82 37). Season is also a factor in DD prevalence, with a higher prevalence observed in spring

83 compared to the summer months (36, 38). Overall leg hygiene and cleanliness of standing areas

84 have significant impacts on DD prevalence, with less well-maintained hygienic standards

85 generally equating to a higher risk of DD (8, 39). Hoof trimming can improve overall claw

86 health; however, bringing in an external hoof trimming on farm and especially if there is a lack

87 of equipment disinfection, hoof trimming can increase DD prevalence on farms (36).

89 Risk factors on individual animals include a wide variety of factors such as age, genetics, breed,

90 and even diet. Parity and lactation stage are well-known risk factors for DD, with cows after their

17 91 first calving and in the mid-late lactation stage being at the highest risk for DD (4). There are

92 potential confounding effects with the parity risk factor, particularly cows with multiple

93 lactations are more likely to be culled for DD and other issues (5). The presence of other hoof

94 diseases such as foot rot and interdigital dermatitis increase likelihood of DD infection (5). The

95 Holstein breed is shown to have a higher risk and prevalence of DD compared to multiple other

96 breeds, including beef cattle (5, 6, 39). Diet also influences DD risk as seen in feedlot beef cattle,

97 where cohorts on diets supplemented with organic and inorganic trace minerals had decreased

98 prevalence compared to diets with solely inorganic trace minerals (20). Individual foot and leg

99 conformation influence susceptibility to DD; larger heel depths and increased claw angles both

100 were associated with animals without DD lesions (40). It is evident that different animals have

101 different levels of susceptibility for DD, and there is a level of heritability for DD, but this

102 estimate varies across studies (41, 42). SNPs have been identified in genes involved in skin

103 proliferation and inflammatory responses in animals with DD, which may help explain variation

104 in animal susceptibilities (43).

105

106 Knowledge of the transmission dynamics of DD from infected to healthy animals is somewhat

107 understood through the above risk factors. Because we do not understand the full etiological

108 factors of DD, it is impossible to identify all risk factors, environmental reservoirs, and routes of

109 transmission. It is estimated that approximately 70% of the average duration of a DD lesion

110 presents as an M4 stage, thus this lesion stage is estimated to be the most influential stage in DD

111 transmission to healthy feet (44, 45). This increases the importance of identifying and treating

112 not only active lesions, but inactive lesions; however, these models are built without knowledge

113 of the true infection and transmission routes, and have consequential limitations. Transmission is

18 114 hypothesized to be through a skin-skin route, but there is also speculation of an environmental

115 reservoir or intermediate in the infection cycle (46, 47). Hoof knives are shown to become

116 contaminated with Treponema after trimming animals with DD lesions (48, 49), amplifying the

117 importance of the role of hoof trimmers in the prevention of DD transmission. Treponema are

118 also found in slurry and feces, suggesting the role of the environment in transmission of DD (50,

119 51). There are also Treponema detectable in oral, rumen, gut, and rectum tissues in cattle on

120 farms with DD lesions (35, 47, 52). In addition, Fusobacterium and Porphyromonas are present,

121 in low abundance similar to Treponema in slurry (50). This evidence is not definitive of slurry as

122 a potential route of transmission, and we need more thorough characterization of DD-associated

123 species in the environment to decide what are the natural infection reservoirs.

124

125 Mitigation and treatment strategies

126 Treatment strategies have primarily relied on topical use of oxytetracycline or copper sulfate in

127 favour of systemic antibiotics that have complications with efficacy, cost, and withdrawal times

128 (53). Definition of treatment success and application protocol causes variation in treatment

129 reports. One aim of treatment is to transition active painful lesions to inactive lesions, and

130 common treatments generally have high success ratios in these terms. Bandaging of lesions is

131 more effective than topical antimicrobial spray without bandaging, and bandaging with

132 tetracycline can result in up to 86% of active lesions transitioning to inactive lesions after 4

133 weeks (54). Spraying lesions in the parlour with tetracycline can heal 69% of lesions after one

134 week, although this treatment method did not reduce rate of recurring lesions (55). Another study

135 reports up to 26% of lesions recurring after tetracycline treatment (56). Moreover, if we change

19 136 our definition of ‘cure’ to a return to normal skin in terms of morphological presentation,

137 treatment success of tetracycline can be reduced down to just 9% (57).

138

139 There are collective treatment strategies that target all animals in affected herds to reduce overall

140 cow-level prevalence. A systematic review on collective treatments does not support the benefit

141 of collective treatments in terms of prevention or treatment of DD on dairy farms (58). Another

142 meta-analysis on footbaths (a form of collective treatment), also largely does not support a

143 scientific efficacy for footbaths, with the only supported footbath protocol being 5% copper

144 sulfate applied 4-plus times a week (59). Low sample sizes and low quality of evidence are

145 limiting factors in establishing efficacy of footbaths on reducing DD prevalence (58, 59). More

146 thorough understanding of DD microbiology, and identification of the etiological agents that

147 need to be targeted may increase efficacy of treatment and mitigation strategies.

148

149 Causation

150 Early efforts to determine causative agents of DD identified spirochetes within lesions, and

151 further characterization lead to the proposal that Treponema were the primary pathogen within

152 lesions (10, 60). Over time, multiple Treponema phylotypes and species were identified in DD

153 lesions through 16S rRNA sequencing and PCR (61, 62), and the hypothesis of a polytreponemal

154 origin became evident. Polytreponemal involvement within lesions likely still remains true;

155 however, there is recent agreement that there are non-Treponema bacteria consistently present in

156 lesions and may also aid in lesion development. In general, a polymicrobial etiology is accepted,

157 but there is no collective agreement on the core microbial constituents that causes the disease.

158 There is a lack of consistency of a single bacterium or group within all lesions, and these bacteria

20 159 are mostly fastidious anaerobes which are difficult to isolate and characterize type strains.

160 Further confusion about a causative agent exists in the complex morphological manifestations of

161 DD, as the different stages of DD lesions have significantly different microbial composition (32,

162 35). It is possible that interactions between the environment, hosts, and pathogens may be

163 driving these shifts in microbial community structure across lesion stages; however, these

164 mechanisms are not currently elucidated. Moreover, there is limited understanding on the

165 interactions and dynamics of bacterial species throughout lesion progression, with only T.

166 phagedenis, T. pedis, T. denticola and T. medium being reliably quantified throughout lesion

167 progression (63). Across different studies, Dichelobacter, Fusobacterium, Campylobacter,

168 Mycoplasma, Bacteroides, Porphyromonas, and other genera have all been associated with DD

169 lesions (15, 32, 64–66). Without knowledge of the individual species within these genera, we

170 cannot fully understand the potential contributions of the different bacteria to DD pathogenesis.

171

172 Infection models

173 Strategies to model the pathogenesis of DD-associated microbiota and reproducing the infection

174 have been performed in murine and cattle models. Murine models have assessed the ability of

175 different Treponema species and strains to produce abscesses following subcutaneous injections.

176 Different strains of T. phagedenis cause different sized abscesses in mice, suggesting that

177 virulence may differ among strains (67). Interestingly, when inoculating mice with T. medium, T.

178 phagedenis, and T. pedis, it is evident that T. phagedenis forms the smallest abscesses and T.

179 medium forms the largest, suggesting that T. medium is more important in lesion formation (68).

180 Mixing of different species also reduce abscess size compared to just single species infections

21 181 (68), suggesting potential species interactions; however, these results may change dramatically

182 with different strains of Treponema (67).

183

184 Attempts to recreate the disease in cattle with Treponema cultures have largely been

185 unsuccessful in terms of accurately and consistently reproducing DD (69, 70). To best recreate

186 DD lesion morphology (circumscribed ulcerative lesions), applying macerated biopsies from

187 naturally occurring DD lesions is the most consistent method of inducing ulcerated lesions (69,

188 70). This evidence suggests that Treponema alone is likely not enough to induce lesion

189 formation, and that additional bacteria and conditions are required to cause DD.

190

191 Treponema

192 Treponema are the most prevalent bacteria in DD lesions (15, 32, 35). There are multiple

193 Treponema species and phylotypes present in lesions, and which species are present is not

194 consistent across individual lesions (63, 71). This inconsistency and variation in species

195 establishes hypotheses of DD as a polytreponemal disease, but another important hypothesis

196 suggests this distribution of Treponema spp. corresponds to opportunistic infections (18). As

197 fastidious anaerobic organisms, Treponema are notoriously difficult to culture and isolate (72),

198 resulting in a lack of type strains available and impairment of our ability to fully understand

199 pathogenesis mechanisms of these Treponema in DD lesions.

200

201 Most commonly identified in DD lesions are T. phagedenis, T. medium, T. pedis, T. vincentii, T.

202 refringens, and T. denticola (16, 35, 63, 73, 74). No T. pallidum, the causative agent of syphilis,

203 has been verified in lesions (75). Treponema have two main morphological changes in their

22 204 growth curves, a spiral form followed by a cystic form, but the differences in morphology has

205 not yet been linked to DD pathogenesis (76). Treponema have been identified in many other

206 tissues in cattle, such as the mouth, rumen, gut, rectum, udder, and in the feces (23, 50, 51, 77,

207 78). Treponema are implicated in diseases of other animals, including CODD, skin lesions in

208 goats, pigs, equine hoof cankers, and Treponema-associated hoof disease in elk (25, 27, 79–81).

209 In another polymicrobial infection, periodontal disease shows involvement of T. denticola in

210 pathogenesis of the disease (30). Virulence factors of T. denticola include major surface proteins

211 and high expression of proteases establishing adhesion to epithelium and facilitating tissue

212 destruction, and these mechanisms may be similar in DD pathogenesis (30). However, little is

213 understood of the pathogenesis of DD-associated Treponema, thus further understanding of the

214 Treponema spp. involved in DD and their individual roles is necessary.

215

216 Porphyromonas

217 Porphyromonas is one of the most commonly identified bacteria in DD lesions, outside of

218 Treponema (32, 74). Porphyromonas levii is a potential etiological agent of DD, but it is also

219 isolated from foot rot and uterine infections in cattle (82, 83). Quantification of bacteria by meta-

220 transcriptomics of DD lesions identify Porphyromonas as one of the most active bacterial groups

221 in DD, behind Treponema (84). Antibodies in dairy cattle serum against P. levii are significantly

222 higher in animals on farms with DD than farms without DD, linking P. levii to DD infected

223 animals (85).

224

225 P. levii is relatively resistant to phagocytosis from neutrophils and macrophages, and has

226 immunoglobulin proteases (82, 86), which can suggest a potential mechanism for it in DD

23 227 lesions in immune regulation. Potential roles for Porphyromonas in DD can be identified from

228 periodontal research. Although P. gingivalis is not identified in DD, it is one of the main species

229 involved in periodontal disease pathogenesis. Keystone taxa are bacteria that have large

230 influences on community structure, and P. gingivalis is recognized as a keystone member due to

231 its ability to initiate tissue destruction and facilitate dysbiosis in oral microbiota (87, 88). There

232 is also strong evidence of metabolic synergy between Porphyromonas and Treponema, causing

233 shifts in amino acid biosynthesis and upregulation of T. denticola virulence factors (89).

234 Synergistic biofilms are also formed by P. gingivalis and T. denticola (90). This evidence of

235 synergistic facilitation by Porphyromonas in periodontal disease may be extrapolated to DD

236 etiopathogenesis, and more research is required to confirm the involvement of Porphyromonas in

237 DD and its synergistic interactions.

238

239 Mycoplasma

240 Mycoplasma is only recently identified with a potential role in DD, even though it was identified

241 from lesions early on in DD research (60, 91). Mycoplasma relative abundance in DD lesions is

242 between 5% and 28% (15, 32, 35). Most often M. fermentans is identified in lesions, and through

243 PCR, is detectable in all DD lesions of one study (15). There is a considerable lack of

244 Mycoplasma culture isolates and genome sequences available from DD lesions or related

245 diseases, as well as a lack of understanding of the species involved in DD lesions. This lack of

246 knowledge surrounding Mycoplasma in DD needs to be addressed to understand if there is any

247 involvement of Mycoplasma in DD etiopathogenesis.

248

24 249 Fusobacterium

250 Fusobacterium, specifically F. necrophorum, is more commonly recognized for its causative role

251 in foot rot than in DD; however, Fusobacterium is present up to 8% relative abundance of the

252 microbiota within DD lesions (15) and in 33.3% of DD samples as measured by PCR (75). In

253 addition, F. necrophorum is present in bacterial clusters mainly in the superficial layers of the

254 skin of DD lesions (15, 74). Similarly to P. levii, antibodies against F. necrophorum are also

255 significantly increased in farms with DD compared to farms without DD (85). It is also possible

256 that F. necrophorum and P. levii generate mixed-species biofilms, which in vitro, can impair the

257 innate immune response (92). There are two well-defined virulence factors of F. necrophorum,

258 endotoxin and leukotoxin, which aid in overcoming immune mechanisms, but the activity of

259 these factors may change with the various subspecies that are isolated from hoof lesions (93, 94).

260 In periodontal disease, Fusobacterium is a secondary pathogen, colonizing oral microbiota after

261 dysbiosis begins, and clusters with other bacteria to generate increased anaerobic conditions (30).

262 Increased abundance of F. necrophorum is linked with increased severity of foot rot in sheep,

263 although it is not crucial for disease initiation (95). The ability of Fusobacterium to generate

264 biofilms, reduce innate immune function, and to generate increased anaerobic conditions may

265 explain a role for it in DD development; however, more work needs to be done to identify the

266 species involved and if Fusobacterium is a primary or secondary invader in DD.

267

268 Other associated bacteria

269 Dichelobacter nodosus, one of the causative agents of ovine foot rot (96), is frequently identified

270 in DD lesions (32, 74). PCR-based identification of D. nodosus shows it is present in up to 55%

271 of DD lesions (75), but does not significantly increase in abundance in lesions compared to

25 272 healthy skin (32, 35). The role of D. nodosus in bovine DD is unclear, but many suggest it is

273 important to lesion formation and epithelial destruction to aid in secondary colonization (32, 64).

274 Guggenheimella bovis was one of the first non-Treponema bacteria with etiological implications

275 after it was isolated from DD lesions (91, 97), but since has not been identified as a causative

276 factor in many large scale DD microbiome studies. Candidatus Aemobophilus asiaticus has been

277 thought as a potential etiological agent of DD lesions (35, 73), but has only been identified in

278 metagenomic studies by one research group. Campylobacter are significantly increased in

279 relative abundance in early stages of DD (32), but are also suggested to be mere a secondary

280 invader due to its superficial localization in the epidermis (10).

281

282 Gaps in knowledge of DD microbiology

283 This thesis addresses significant gaps in our understanding of DD microbiology. There is no

284 comprehensive characterization of DD lesion microbiota in feedlot beef cattle, and no in depth

285 comparison between DD microbiota in dairy and beef cattle breeds. There is a shortage of type

286 strain availability from DD bacterial isolates, and a general lack of validated species-level

287 identification and quantification in DD. Inconsistency in non-Treponema microbial constituents

288 thought to be involved in DD pathogenesis also impedes DD microbiology research. Addressing

289 these gaps will identify a consistent DD-associated microbiota in both beef and dairy cattle, and

290 provide a myriad of new tools and knowledge for future elucidation of DD etiopathogenesis

291 mechanisms.

292

26 293 Objectives

294 In this thesis, we aim to:

295 1. Characterize the microbial composition of DD lesion stages in feedlot beef cattle of

296 Southern Alberta

297 2. Develop and validate a novel multiplex species-specific qPCR for key DD-associated

298 non-Treponema spp.

299 3. Use the novel multiplex qPCR and existing species-specific qPCRs to study species

300 distributions throughout lesion stages

301 4. Identify consistent DD-associated microbiota across studies to generate a narrow subset

302 of bacteria to target in future studies of DD pathogenesis

303

27 304 References

305 1. Read DH, Walker RL. 1998. Papillomatous digital dermatitis (footwarts) in California

306 dairy cattle: Clinical and gross pathologic findings. J Vet Diagnostic Investig 10:67–76.

307 2. Cheli R, Mortellaro C. 1974. La dermatite digitale del bovino, p. 208–213. In Proceedings

308 of the 8th International Conference on Diseases of Cattle.

309 3. Yang DA, Johnson WO, Müller KR, Gates MC, Laven RA. 2019. Estimating the herd and

310 cow level prevalence of bovine digital dermatitis on New Zealand dairy farms: A

311 Bayesian superpopulation approach. Prev Vet Med 165:76–84.

312 4. Solano L, Barkema HW, Mason S, Pajor EA, LeBlanc SJ, Orsel K. 2016. Prevalence and

313 distribution of foot lesions in dairy cattle in Alberta, Canada. J Dairy Sci 99:6828–6841.

314 5. Holzhauer M, Hardenberg C, Bartels CJM, Frankena K. 2006. Herd- and cow-level

315 prevalence of digital dermatitis in the Netherlands and associated risk factors. J Dairy Sci

316 89:580–588.

317 6. Hesseling J, Legione AR, Stevenson MA, McCowan CI, Pyman MF, Finochio C, Nguyen

318 D, Roic CL, Thiris OL, Zhang AJ, van Schaik G, Coombe JE. 2019. Bovine digital

319 dermatitis in Victoria, Australia. Aust Vet J 97:404–413.

320 7. Demirkan I, Erdoğan M, Demirkan AÇ, Bozkurt F, Altındiş M, Navruz FZ, Köse Z. 2018.

321 Isolation and identification of Treponema pedis and Treponema phagedenis-like

322 organisms from bovine digital dermatitis lesions found in dairy cattle in Turkey. J Dairy

323 Sci 101:10317–10326.

324 8. Wells SJ, Garber LP, Wagner BA. 1999. Papillomatous digital dermatitis and associated

325 risk factors in US dairy herds. Prev Vet Med 38:11–24.

326 9. Mamuad LL, Seo BJ, Faruk MS Al, Espiritu HM, Jin SJ, Kim W Il, Lee SS, Cho Y Il.

28 327 2020. Treponema spp., the dominant pathogen in the lesion of bovine digital dermatitis

328 and its characterization in dairy cattle. Vet Microbiol 245:108696.

329 10. Döpfer D, Koopmans A, Meijer FA, Szakáll I, Schukken YH, Klee W, Bosma RB,

330 Cornelisse JL, Van Asten AJAM, Ter Huurne AAHM. 1997. Histological and

331 bacteriological evaluation of digital dermatitis in cattle, with special reference to

332 Spirochaetes and Campylobacter faecalis. Vet Rec 140:620–623.

333 11. Davis-Unger J, Pajor E, Schwartzkopf-Genswein K, Marti S, Dorin C, Spackman E, Orsel

334 K. 2017. Economic impacts of lameness in feedlot cattle. Transl Anim Sci 1:467–479.

335 12. Coetzee JF, Shearer JK, Stock ML, Kleinhenz MD, van Amstel SR. 2017. An Update on

336 the Assessment and Management of Pain Associated with Lameness in Cattle. Vet Clin

337 North Am - Food Anim Pract 33:389–411.

338 13. Gomez A, Cook NB, Socha MT, Döpfer D. 2015. First-lactation performance in cows

339 affected by digital dermatitis during the rearing period. J Dairy Sci 98:4487–4498.

340 14. Dolecheck KA, Overton MW, Mark TB, Bewley JM. 2019. Use of a stochastic simulation

341 model to estimate the cost per case of digital dermatitis, sole ulcer, and white line disease

342 by parity group and incidence timing. J Dairy Sci 102:715–730.

343 15. Nielsen MW, Strube ML, Isbrand A, Al-Medrasi WDHM, Boye M, Jensen TK, Klitgaard

344 K. 2016. Potential bacterial core species associated with digital dermatitis in cattle herds

345 identified by molecular profiling of interdigital skin samples. Vet Microbiol 186:139–149.

346 16. Klitgaard K, Bretó AF, Boye M, Jensen TK. 2013. Targeting the treponemal microbiome

347 of digital dermatitis infections by high-resolution phylogenetic analyses and comparison

348 with fluorescent in situ hybridization. J Clin Microbiol 51:2212–2219.

349 17. Davis-Unger J, Schwartzkopf-Genswein KSG, Pajor EA, Hendrick S, Marti S, Dorin C,

29 350 Orsel K. 2019. Prevalence and lameness-associated risk factors in Alberta feedlot cattle.

351 Transl Anim Sci 3:595–606.

352 18. Orsel K, Plummer P, Shearer J, De Buck J, Carter SD, Guatteo R, Barkema HW. 2018.

353 Missing pieces of the puzzle to effectively control digital dermatitis. Transbound Emerg

354 Dis 65:186–198.

355 19. Brown CC, Kilgo PD, Jacobsen KL. 2000. Prevalence of papillomatous digital dermatitis

356 among culled adult cattle in the southeastern United States. Am J Vet Res 61:928–930.

357 20. Kulow M, Merkatoris P, Anklam KS, Rieman J, Larson C, Branine M, Döpfer D. 2017.

358 Evaluation of the prevalence of digital dermatitis and the effects on performance in beef

359 feedlot cattle under organic trace mineral supplementation. J Anim Sci 95:3435–3444.

360 21. Terrell SP, Reinhardt CD, Larson CK, Vahl CI, Thomson DU. 2017. Incidence of

361 lameness and association of cause and severity of lameness on the outcome for cattle on

362 six commercial beef feedlots. J Am Vet Med Assoc 250:437–445.

363 22. Clegg SR, Carter SD, Stewart JP, Amin DM, Evans NJ, Blowey RW. 2016. Bovine

364 ischaemic teat necrosis: A further potential role for digital dermatitis treponemes. Vet Rec

365 178:71.

366 23. Sobhy NM, Mahmmod YS, Refaai W, Awad A. 2020. Molecular detection of Treponema

367 species organisms in foremilk and udder cleft skin of dairy cows with digital dermatitis.

368 Trop Anim Health Prod 52:815–821.

369 24. Ekman L, Nyman AK, Landin H, Magnusson U, Waller KP. 2018. Mild and severe udder

370 cleft dermatitis—Prevalence and risk factors in Swedish dairy herds. J Dairy Sci 101:556–

371 571.

372 25. Clegg SR, Crosby-Durrani HE, Bell J, Blundell R, Blowey RW, Carter SD, Evans NJ.

30 373 2016. Detection and Isolation of Digital Dermatitis Treponemes from Bovine Pressure

374 Sores. J Comp Pathol 154:273–282.

375 26. Sullivan LE, Clegg SR, Angell JW, Newbrook K, Blowey RW, Carter SD, Bell J, Duncan

376 JS, Grove-White DH, Murray RD, Evans NJ. 2015. High-level association of bovine

377 digital dermatitis Treponema spp. with contagious ovine digital dermatitis lesions and

378 presence of Fusobacterium necrophorum and Dichelobacter nodosus. J Clin Microbiol

379 53:1628–1638.

380 27. Han S, Mansfield KG, Bradway DS, Besser TE, Read DH, Haldorson GJ, Alt DP, Wilson-

381 Welder JH. 2019. Treponeme-Associated Hoof Disease of Free-Ranging Elk (Cervus

382 elaphus) in Southwestern Washington State, USA. Vet Pathol 56:118–132.

383 28. Walker RL, Read DH, Loretz KJ, Nordhausen RW. 1995. Spirochetes isolated from dairy

384 cattle with papillomatous digital dermatitis and interdigital dermatitis. Vet Microbiol

385 47:343–355.

386 29. Knappe-Poindecker M, Gilhuus M, Jensen TK, Klitgaard K, Larssen RB, Fjeldaas T.

387 2013. Interdigital dermatitis, heel horn erosion, and digital dermatitis in 14 Norwegian

388 dairy herds. J Dairy Sci 96:7617–7629.

389 30. Edwards AM, Dymock D, Jenkinson HF. 2003. From tooth to hoof: Treponemes in tissue-

390 destructive diseases. J Appl Microbiol 94:767–780.

391 31. Berry SL, Read DH, Famula TR, Mongini A, Döpfer D. 2012. Long-term observations on

392 the dynamics of bovine digital dermatitis lesions on a California dairy after topical

393 treatment with lincomycin HCl. Vet J 193:654–658.

394 32. Krull AC, Shearer JK, Gorden PJ, Cooper VL, Phillips GJ, Plummera PJ. 2014. Deep

395 sequencing analysis reveals temporal microbiota changes associated with development of

31 396 bovine digital dermatitis. Infect Immun 82:3359–3373.

397 33. Solano L, Barkema HW, Jacobs C, Orsel K. 2017. Validation of the M-stage scoring

398 system for digital dermatitis on dairy cows in the milking parlor. J Dairy Sci 100:1592–

399 1603.

400 34. Vanhoudt A, Yang DA, Armstrong T, Huxley JN, Laven RA, Manning AD, Newsome

401 RF, Nielen M, van Werven T, Bell NJ. 2019. Interobserver agreement of digital dermatitis

402 M-scores for photographs of the hind feet of standing dairy cattle. J Dairy Sci 102:5466–

403 5474.

404 35. Zinicola M, Lima F, Lima S, Machado V, Gomez M, Döpfer D, Guard C, Bicalho R.

405 2015. Altered microbiomes in bovine digital dermatitis lesions, and the gut as a pathogen

406 reservoir. PLoS One 10:e0120504.

407 36. Yang DA, Laven RA, Heuer C, Vink WD, Chesterton RN. 2018. Farm level risk factors

408 for bovine digital dermatitis in Taranaki, New Zealand: An analysis using a Bayesian

409 hurdle model. Vet J 234:91–95.

410 37. Yang DA, Gates MC, Müller KR, Laven RA. 2019. Bayesian analysis of herd-level risk

411 factors for bovine digital dermatitis in New Zealand dairy herds. BMC Vet Res 15:125.

412 38. Yang DA, Laven RA, Chesterton RN. 2019. Effects of climate and farm management

413 practices on bovine digital dermatitis in spring-calving pasture-based dairy farms in

414 Taranaki, New Zealand. Vet J 247:75–80.

415 39. Relun A, Lehebel A, Bruggink M, Bareille N, Guatteo R. 2013. Estimation of the relative

416 impact of treatment and herd management practices on prevention of digital dermatitis in

417 French dairy herds. Prev Vet Med 110:558–562.

418 40. Gomez A, Cook NB, Rieman J, Dunbar KA, Cooley KE, Socha MT, Döpfer D. 2015. The

32 419 effect of digital dermatitis on hoof conformation. J Dairy Sci 98:927–936.

420 41. Palmer MA, O’Connell NE. 2015. Digital dermatitis in dairy cows: A review of risk

421 factors and potential sources of between-animal variation in susceptibility. Animals

422 5:512–535.

423 42. Biemans F, de Jong MCM, Bijma P. 2019. A genome-wide association study for

424 susceptibility and infectivity of Holstein Friesian dairy cattle to digital dermatitis. J Dairy

425 Sci 102:6248–6262.

426 43. Sánchez-Molano E, Bay V, Smith RF, Oikonomou G, Banos G. 2019. Quantitative Trait

427 Loci Mapping for Lameness Associated Phenotypes in Holstein–Friesian Dairy Cattle.

428 Front Genet 10:926.

429 44. Yang DA, Laven RA, Müller KR, Gates MC. 2020. Modelling the transmission dynamics

430 of bovine digital dermatitis in New Zealand pastoral dairy production systems. Vet Res

431 51:16.

432 45. Biemans F, Bijma P, Boots NM, de Jong MCM. 2018. Digital Dermatitis in dairy cattle:

433 The contribution of different disease classes to transmission. Epidemics 23:76–84.

434 46. Evans NJ, Murray RD, Carter SD. 2016. Bovine digital dermatitis: Current concepts from

435 laboratory to farm. Vet J 211:3–13.

436 47. Evans NJ, Timofte D, Isherwood DR, Brown JM, Williams JM, Sherlock K, Lehane MJ,

437 Murray RD, Birtles RJ, Anthony Hart C, Carter SD. 2012. Host and environmental

438 reservoirs of infection for bovine digital dermatitis treponemes. Vet Microbiol 156:102–

439 109.

440 48. Gillespie A V., Carter SD, Blowey RW, Staton GJ, Evans NJ. 2020. Removal of bovine

441 digital dermatitis-associated treponemes from hoof knives after foot-trimming: A

33 442 disinfection field study. BMC Vet Res 16:1–9.

443 49. Sullivan LE, Blowey RW, Carter SD, Duncan JS, Grove-White DH, Page P, Iveson T,

444 Angell JW, Evans NJ. 2014. Presence of digital dermatitis treponemes on cattle and sheep

445 hoof trimming equipment. Vet Rec 175:201.

446 50. Klitgaard K, Strube ML, Isbrand A, Nielsen MW, Jensen TK. 2017. Microbiota analysis

447 of an environmental slurry and its potential role as a reservoir of bovine digital dermatitis

448 pathogens. Appl Environ Microbiol 83:e00244-17.

449 51. Klitgaard K, Nielsen MW, Ingerslev HC, Boye M, Jensen TK. 2014. Discovery of bovine

450 digital dermatitis-associated Treponema spp. in the dairy herd environment by a targeted

451 deep-sequencing approach. Appl Environ Microbiol 80:4427–4432.

452 52. Evans NJ, Brown JM, Murray RD, Getty B, Birtles RJ, Hart CA, Carter SD. 2011.

453 Characterization of novel bovine gastrointestinal tract Treponema isolates and comparison

454 with bovine digital dermatitis treponemes. Appl Environ Microbiol 77:138–147.

455 53. Laven RA, Logue DN. 2006. Treatment strategies for digital dermatitis for the UK. Vet J

456 171:79–88.

457 54. Klawitter M, Döpfer D, Braden TB, Amene E, Mueller KE. 2019. Randomised clinical

458 trial showing the curative effect of bandaging on M2-stage lesions of digital dermatitis in

459 dairy cows. Vet Rec Open 6:e000264.

460 55. Jacobs C, Orsel K, Mason S, Barkema HW. 2018. Comparison of effects of routine topical

461 treatments in the milking parlor on digital dermatitis lesions. J Dairy Sci 101:5255–5266.

462 56. Cutler JHH, Cramer G, Walter JJ, Millman ST, Kelton DF. 2013. Randomized clinical

463 trial of tetracycline hydrochloride bandage and paste treatments for resolution of lesions

464 and pain associated with digital dermatitis in dairy cattle. J Dairy Sci 96:7550–7557.

34 465 57. Krull AC, Shearer JK, Gorden PJ, Scott HM, Plummer PJ. 2016. Digital dermatitis:

466 Natural lesion progression and regression in Holstein dairy cattle over 3 years. J Dairy Sci

467 99:3718–3731.

468 58. Ariza JM, Relun A, Bareille N, Oberle K, Guatteo R. 2017. Effectiveness of collective

469 treatments in the prevention and treatment of bovine digital dermatitis lesions: A

470 systematic review. J Dairy Sci 100:7401–7418.

471 59. Jacobs C, Beninger C, Hazlewood GS, Orsel K, Barkema HW. 2019. Effect of footbath

472 protocols for prevention and treatment of digital dermatitis in dairy cattle: A systematic

473 review and network meta-analysis. Prev Vet Med 164:56–71.

474 60. Collighan RJ, Woodward MJ. 1997. Spirochaetes and other bacterial species associated

475 with bovine digital dermatitis. FEMS Microbiol Lett 156:37–41.

476 61. Evans NJ, Brown JM, Demirkan I, Murray RD, Vink WD, Blowey RW, Hart CA, Carter

477 SD. 2008. Three unique groups of spirochetes isolated from digital dermatitis lesions in

478 UK cattle. Vet Microbiol 130:141–150.

479 62. Klitgaard K, Boye M, Capion N, Jensen TK. 2008. Evidence of multiple Treponema

480 phylotypes involved in bovine digital dermatitis as shown by 16S rRNA gene analysis and

481 fluorescence in situ hybridization. J Clin Microbiol 46:3012–3020.

482 63. Beninger C, Naqvi SA, Naushad S, Orsel K, Luby C, Derakhshani H, Khafipour E, De

483 Buck J. 2018. Associations between digital dermatitis lesion grades and the quantities of

484 four Treponema species. Vet Res 49:111.

485 64. Rasmussen M, Capion N, Klitgaard K, Rogdo T, Fjeldaas T, Boye M, Jensen TK. 2012.

486 Bovine digital dermatitis: Possible pathogenic consortium consisting of Dichelobacter

487 nodosus and multiple Treponema species. Vet Microbiol 160:151–161.

35 488 65. Santos TMA, Pereira R V., Caixeta LS, Guard CL, Bicalho RC. 2012. Microbial diversity

489 in bovine Papillomatous digital dermatitis in Holstein dairy cows from upstate New York.

490 FEMS Microbiol Ecol 79:518–529.

491 66. Yano T, Moe KK, Yamazaki K, Ooka T, Hayashi T, Misawa N. 2010. Identification of

492 candidate pathogens of papillomatous digital dermatitis in dairy cattle from quantitative

493 16S rRNA clonal analysis. Vet Microbiol 143:352–362.

494 67. Elliott MK, Alt DP, Zuerner RL. 2007. Lesion formation and antibody response induced

495 by papillomatous digital dermatitis-associated spirochetes in a murine abscess model.

496 Infect Immun 75:4400–4408.

497 68. Arrazuria R, Knight CG, Lahiri P, Cobo ER, Barkema HW, De Buck J. 2020. Treponema

498 spp. isolated from bovine digital dermatitis display different pathogenicity in a murine

499 abscess model. Microorganisms 8:1507.

500 69. Gomez A, Cook NB, Bernardoni ND, Rieman J, Dusick AF, Hartshorn R, Socha MT,

501 Read DH, Döpfer D. 2012. An experimental infection model to induce digital dermatitis

502 infection in cattle. J Dairy Sci 95:1821–1830.

503 70. Krull AC, Cooper VL, Coatney JW, Shearer JK, Gorden PJ, Plummer PJ. 2016. A highly

504 effective protocol for the rapid and consistent induction of digital dermatitis in holstein

505 calves. PLoS One 11:e0154481.

506 71. Nordhoff M, Moter A, Schrank K, Wieler LH. 2008. High prevalence of treponemes in

507 bovine digital dermatitis-A molecular epidemiology. Vet Microbiol 131:293–300.

508 72. Wardle HM. 1997. The challenge of growing oral Spirochaetes. J Med Microbiol 46:104–

509 116.

510 73. Zinicola M, Higgins H, Lima S, Machado V, Guard C, Bicalho R. 2015. Shotgun

36 511 metagenomic sequencing reveals functional genes and microbiome associated with bovine

512 digital dermatitis. PLoS One 10:e0133674.

513 74. Moreira TF, Facury Filho EJ, Carvalho AU, Strube ML, Nielsen MW, Klitgaard K, Jensen

514 TK. 2018. Pathology and bacteria related to digital dermatitis in dairy cattle in all year

515 round grazing system in Brazil. PLoS One 13:e0193870.

516 75. Staton GJ, Sullivan LE, Blowey RW, Carter SD, Evans NJ. 2020. Surveying bovine

517 digital dermatitis and non-healing bovine foot lesions for the presence of Fusobacterium

518 necrophorum, Porphyromonas endodontalis and Treponema pallidum. Vet Rec 186:450.

519 76. Döpfer D, Anklam K, Mikheil D, Ladell P. 2012. Growth curves and morphology of three

520 Treponema subtypes isolated from digital dermatitis in cattle. Vet J 193:685–693.

521 77. Sullivan LE, Carter SD, Duncan JS, Grove-White DH, Angell JW, Evans NJ. 2015. The

522 gastrointestinal tract as a potential infection reservoir of digital dermatitis-associated

523 treponemes in beef cattle and sheep. Appl Environ Microbiol 81:7460–7469.

524 78. Nascimento LV, Mauerwerk MT, Dos Santos CL, De Barros Filho IR, Birgel EH,

525 Sotomaior CS, Madeira HMF, Ollhoff RD. 2015. Treponemes detected in digital

526 dermatitis lesions in Brazilian dairy cattle and possible host reservoirs of infection. J Clin

527 Microbiol 53:1935–1937.

528 79. Crosby-Durrani HE, Clegg SR, Singer E, Angell JW, Evans NJ, Carter SD, Blundell RJ,

529 Duncan JS. 2016. Severe foot lesions in dairy goats associated with digital dermatitis

530 treponemes. J Comp Pathol 154:283–296.

531 80. Clegg SR, Sullivan LE, Bell J, Blowey RW, Carter SD, Evans NJ. 2016. Detection and

532 isolation of digital dermatitis treponemes from skin and tail lesions in pigs. Res Vet Sci

533 104:64–70.

37 534 81. Sykora S, Brandt S. 2015. Occurrence of Treponema DNA in equine hoof canker and

535 normal hoof tissue. Equine Vet J 47:627–630.

536 82. Walter MRV, Morck DW. 2002. Chemotaxis, phagocytosis, and oxidative metabolism in

537 bovine macrophages exposed to a novel interdigital phlegmon (foot rot) lesion isolate,

538 Porphyromonas levii. Am J Vet Res 63:757–762.

539 83. Cunha F, Jeon SJ, Daetz R, Vieira-Neto A, Laporta J, Jeong KC, Barbet AF, Risco CA,

540 Galvão KN. 2018. Quantifying known and emerging uterine pathogens, and evaluating

541 their association with metritis and fever in dairy cows. Theriogenology 114:25–33.

542 84. Marcatili P, Nielsen MW, Sicheritz-Pontén T, Jensen TK, Schafer-Nielsen C, Boye M,

543 Nielsen M, Klitgaard K. 2016. A novel approach to probe host-pathogen interactions of

544 bovine digital dermatitis, a model of a complex polymicrobial infection. BMC Genomics

545 17:987.

546 85. Moe KK, Yano T, Misumi K, Kubota C, Nibe K, Yamazaki W, Muguruma M, Misawa N.

547 2010. Detection of antibodies against Fusobacterium necrophorum and Porphyromonas

548 levii-like species in dairy cattle with papillomatous digital dermatitis. Microbiol Immunol

549 54:338–346.

550 86. Lobb DA, Loeman HJ, Sparrow DG, Morck DW. 1999. Bovine polymorphonuclear

551 neutrophil-mediated phagocytosis and an immunoglobulin G2 protease produced by

552 Porphyromonas levii. Can J Vet Res 63:113–118.

553 87. Banerjee S, Schlaeppi K, van der Heijden MGA. 2018. Keystone taxa as drivers of

554 microbiome structure and functioning. Nat Rev Microbiol 16:567–576.

555 88. Roberts FA, Darveau RP. 2015. Microbial protection and virulence in periodontal tissue

556 as a function of polymicrobial communities: Symbiosis and dysbiosis. Periodontol 2000

38 557 69:18–27.

558 89. Tan KH, Seers CA, Dashper SG, Mitchell HL, Pyke JS, Meuric V, Slakeski N, Cleal SM,

559 Chambers JL, McConville MJ, Reynolds EC. 2014. Porphyromonas gingivalis and

560 Treponema denticola exhibit metabolic symbioses. PLoS Pathog 10:e1003955.

561 90. Zhu Y, Dashper SG, Chen YY, Crawford S, Slakeski N, Reynolds EC. 2013.

562 Porphyromonas gingivalis and Treponema denticola synergistic polymicrobial biofilm

563 Development. PLoS One 8:e71727.

564 91. Wyss C, Dewhirst FE, Paster BJ, Thurnheer T, Luginbühl A. 2005. Guggenheimella bovis

565 gen. nov., sp. nov., isolated from lesions of bovine dermatitis digitalis. Int J Syst Evol

566 Microbiol 55:667–671.

567 92. Lockhart JS, Buret AG, Ceri H, Storey DG, Anderson SJ, Morck DW. 2017. Mixed

568 species biofilms of Fusobacterium necrophorum and Porphyromonas levii impair the

569 oxidative response of bovine neutrophils in vitro. Anaerobe 47:157–164.

570 93. Tan ZL, Nagaraja TG, Chengappa MM. 1996. Fusobacterium necrophorum infections:

571 virulence factors, pathogenic mechanism and control measures. Vet Res Commun 20:113–

572 140.

573 94. Narayanan S, Stewart GC, Chengappa MM, Willard L, Shuman W, Wilkerson M,

574 Nagaraja TG. 2002. Fusobacterium necrophorum leukotoxin induces activation and

575 apoptosis of bovine leukocytes. Infect Immun 70:4609–4620.

576 95. Witcomb LA, Green LE, Kaler J, Ul-Hassan A, Calvo-Bado LA, Medley GF, Grogono-

577 Thomas R, Wellington EMH. 2014. A longitudinal study of the role of Dichelobacter

578 nodosus and Fusobacterium necrophorum load in initiation and severity of footrot in

579 sheep. Prev Vet Med 115:48–55.

39 580 96. Bennett G, Hickford J, Sedcole R, Zhou H. 2009. Dichelobacter nodosus, Fusobacterium

581 necrophorum and the epidemiology of footrot. Anaerobe 15:173–176.

582 97. Schlafer S, Nordhoff M, Wyss C, Strub S, Hübner J, Gescher DM, Petrich A, Göbel UB,

583 Moter A. 2008. Involvement of Guggenheimella bovis in digital dermatitis lesions of dairy

584 cows. Vet Microbiol 128:118–125.

585

586

40 587 Chapter 2: Identification and quantification of bovine digital dermatitis-

588 associated microbiota across lesion stages in feedlot beef cattle

589 Abstract

590 Bovine digital dermatitis (DD) is a skin disorder that is a significant cause of infectious lameness

591 in cattle around the world. However, very little is known about the etiopathogenesis of the

592 disease, and the microbiota associated with DD in beef cattle. In this study, we provide a

593 comprehensive characterization of DD and healthy skin microbiota of feedlot beef cattle. We

594 also developed and validated a novel multiplex qPCR to quantify the distribution of DD-

595 associated bacterial species across DD lesion stages. We determined the DD-associated

596 microbiota with deep amplicon sequencing of the V3-V4 hypervariable region of the 16S rRNA

597 gene, followed by application of novel and existing qPCR assays to quantify species distributions

598 of Treponema, Porphyromonas, Fusobacterium, and Bacteroides across lesion stages. Deep

599 amplicon sequencing revealed that Treponema, Mycoplasma, Porphyromonas, and

600 Fusobacterium were associated with DD lesions. Culturing of DD biopsies identified

601 Porphyromonas levii, Bacteroides pyogenes, and two Fusobacterium spp. within DD lesions.

602 Through species-specific qPCR on DD lesion DNA, we identified P. levii in 100% of active

603 lesion stages. Early stage lesions were particularly associated with T. medium, T. phagedenis,

604 and P. levii. This study suggests a core DD microbial group consisting of species from

605 Treponema, Fusobacterium, Porphyromonas, and Bacteroides, which may be closely tied with

606 the etiopathogenesis of DD. Further characterization of these species and Mycoplasma spp. are

607 necessary to understand the microbial factors involved in DD pathogenesis, which will help

608 elucidate DD etiology and facilitate more targeted and effective mitigation and treatment

609 strategies.

41 610

611 Importance

612 Previous work, primarily in dairy cattle, has identified various taxa associated with digital

613 dermatitis (DD) lesions. However, there is a significant gap in our knowledge of DD

614 microbiology in beef cattle. In addition, characterization of bacteria at species level in DD

615 lesions is limited. In this study, we provide a framework for accurate and reproducible

616 quantification of major DD-associated bacterial species from DNA samples. Our findings

617 support DD as a polymicrobial infection, and we identified a variety of bacterial species

618 spanning multiple genera that are consistently associated with DD lesions. The DD-associated

619 microbiota identified in this study may be capable of inducing the formation and progression of

620 DD lesions, thus should be primary targets in future DD pathogenesis studies.

621

622 Introduction

623 Bovine digital dermatitis (DD) is an infectious skin lesion affecting cattle around the world, and

624 was first described in the 1970’s (1, 2). DD lesions, primarily localized to the skin between the

625 heel bulbs on the hind legs, are a significant contributor to infectious lameness resulting in major

626 production and economic losses in dairy cattle (3–5). Since its original description, DD is

627 primarily studied as a dairy industry issue, with cow-level prevalence estimates ranging from 3%

628 to 23% (6, 7); however, DD is recently emerging in feedlot beef cattle populations (2), reported

629 by Kulow et al. when that approximately 50% of feedlot cattle on one farm experienced a DD

630 lesion during their study period (8). Management practices mainly focus on broad spectrum

631 antimicrobials, most commonly topical applications of oxytetracycline or copper sulfate;

42 632 however, treatment of DD has unsatisfactory cure rates as low as 9% (9), beckoning calls for

633 greater understanding of DD etiopathogenesis and more targeted treatments.

634

635 The etiological agents of DD are not yet fully identified, as lesions are polymicrobial in nature

636 and contain various fastidious anaerobic bacteria, explaining a lack of dedicated culture methods

637 and the insufficient characterization of species and type strains necessary to understand bacterial

638 pathogenesis mechanisms. Spirochetes are the most consistent bacterial group associated with

639 DD lesions (10–12). Spirochetes from various Treponema species and phylotypes are identified

640 in DD lesions, and individual lesions normally contain different combinations of Treponema spp.

641 at different proportions (13, 14). In addition to Treponema spp., multiple studies of dairy DD

642 lesions identify Porphyromonas (15), Dichelobacter (16), Guggenheimella (17), Bacteroides

643 (18), Fusobacterium (19), Mycoplasma (20), and many other genera as being associated with DD

644 lesions, further supporting a polymicrobial causation of DD.

645

646 Limited knowledge on the presence and the population dynamics of bacterial species throughout

647 DD stages. Most recent DD microbiology studies employ high throughput sequencing strategies

648 and have expanded our understanding of DD, but they fail to reliably classify or quantify species

649 level taxa (21). Recently Beninger et al. quantified four major Treponema spp. in DD lesions of

650 dairy cattle by a species-specific multiplex qPCR, and identified T. phagedenis, T. pedis, and T.

651 medium as highly correlated with DD disease development and progression (22). There still

652 remains a significant number of additional DD-associated Treponema spp. that require further

653 characterization to better understand their potential involvement in disease pathogenesis. In

43 654 addition, no validation of the population dynamics of non-Treponema species exists in DD

655 literature, and thus speculation of their role in DD etiopathogenesis is limited.

656

657 In contrast to dairy breeds, the microbiology of DD lesions in beef breeds is largely unexplored.

658 DD lesions in beef cattle are associated with the presence of different Treponema phylotypes, as

659 well as Fusobacterium necrophorum and Dichelobacter nodosus (23, 24). However, there is no

660 knowledge currently on the quantities and distribution of any DD-associated bacteria in beef

661 cattle DD lesions. Provided the differences in housing and management practices between dairy

662 and feedlot systems, there is insignificant evidence at this time to suggest that the microbial

663 community structure in dairy cattle DD accurately represents beef cattle DD microbiota. Without

664 a full comprehensive identification of DD-associated microbiota in beef cattle, we cannot

665 reliably extrapolate our existing knowledge of DD microbiology in dairy cattle lesions.

666

667 This study aimed to target the gaps in knowledge of DD microbiology in feedlot beef cattle. As

668 DD is an emerging issue in feedlot beef cattle, it was crucial to perform a comprehensive

669 microbiological assessment of DD lesions in this cattle population. Microbiota associated with

670 DD lesions and healthy skin of beef cattle were identified. To better understand which bacterial

671 species are associated with DD lesion formation and progression, bacterial population dynamics

672 were quantified throughout DD lesions stages using existing and newly developed species-

673 specific real-time quantitative PCR assays.

674

44 675 Methods

676 Sampling strategy and biopsy collection

677 Beef cattle from three feedlots in Southern Alberta, Canada were sampled for DD. Forty animals

678 each from a total of 11 separate pens were followed throughout the feedlot cycle starting from

679 fall arrival starting November 2018 to September 2019. Each pen was examined during three

680 separate events throughout the year, in which six study animals per pen were restrained in stand-

681 up chutes and hind-feet were lifted and inspected for DD lesions. All lesions were classified

682 according to the M stage scoring developed by Dopfer et al. (10) and modified by Berry et al.

683 (25). Briefly, there are 5 different M stages: M1 stage is a small (less than 2 cm) circumscribed

684 and ulcerative lesion; M2 lesions are larger ulcerative lesions compared to M1; M3 lesions are

685 described as a healing stage with a rubbery scab covering the lesion; M4 lesions are chronic and

686 have definitive hyperkeratotic growth with raised papilliform projections; and M4.1 lesions are

687 chronic with a small ulcerative focus. All skin free from visible DD lesions were classified as

688 healthy (M0). Upon inspection of feet and classification of lesion stage, the lesion area was

689 washed with water to remove all manure/debris before sample biopsy collection. Because the

690 majority of feet were classified as healthy (M0), these were systematically sampled at every 2nd

691 and 5th animal that passed through the chute. After 3 mL of lidocaine (Lidocaine HCl 2%, Zoetis

692 Canada Inc., Kirkland, Quebec, Canada) was subcutaneously injected, samples were collected

693 with a 4mm biopsy punch (Integra Miltex, Integra Life Sciences Corporation, York,

694 Pennsylvania, USA). All biopsy cores were immediately placed upright in anaerobic transport

695 media (ATM; Anaerobe Systems, Morgan Hill, California, USA), and transported to be

696 processed at the lab within 8 hours of sampling. All animal use was approved by the University

45 697 of Calgary Veterinary Sciences Animal Care Committee (VSACC) under animal care protocol

698 #AC17-0224.

699

700 Biopsy processing

701 All biopsies were removed from ATM tubes and processed in an anaerobic chamber

702 (Bactron3000, Sheldon Manufacturing, Inc., Cornelius, Oregon, USA). To limit environmental

703 contamination, the outer layer of epidermal skin was removed from the biopsies prior to any

704 additional processing. Using sterile scalpels, biopsies were then sectioned longitudinally into 3

705 approximately equal sections; one biopsy section was for anaerobic culture, one for DNA

706 extraction, and one for long term storage at -80 ºC.

707

708 DD-associated anaerobic bacteria were cultured and isolated from biopsies to determine which

709 bacterial species were present in DD lesions and for collecting isolates to aid in qPCR

710 development. Biopsies were cultured for up to 7 days at 37 ºC on Fastidious Anaerobe Agar

711 (FAA; Neogen Corporation, Lansing, Michigan, USA) within an anaerobic chamber (5% CO2,

712 5% H2, 90% N2) to isolate and identify DD-associated species of non-Treponema anaerobic

713 bacteria. FAA media was supplemented with 1 µg/mL vitamin K1 (Sigma-Aldrich, St. Louis,

714 Missouri, USA), 5 µg/mL hemin (Sigma-Aldrich, St. Louis, Missouri, USA), and 5%

715 defibrinated sheep blood (Cedarlane Laboratories, Burlington, Ontario, Canada). Biopsies were

716 first smeared to cover a quarter of the area on each FAA plate, which were then streaked across

717 the rest of the plate to obtain separate colonies. After anaerobic incubation, different colony

718 morphologies were subcultured to isolate single colonies, which were subsequently identified by

719 Sanger sequencing of the full length 16S rRNA gene using 27F (5’-

46 720 AGAGTTTGATCMTGGCTC-3’) and 1392R (5’-CGGAACATGTGMGGCGGG-3’) primers.

721 Colony PCR assays before sequencing had a total volume of 25 µL and used TopTaq Master

722 Mix (Qiagen, Hilden, Germany) and primers at 400 nM. PCR cycle conditions had an initial

723 denaturation at 95 ºC for 5 min, followed by 35 cycles of 95 ºC for 30 seconds, 58 ºC for 30

724 seconds, 72 ºC for 90 seconds, and a final step at 72 ºC for 10 minutes. Sanger sequences were

725 aligned against the BLAST nt database, and bacterial species were identified when full length

726 16S rRNA sequence identity was greater than 98.65%.

727

728 Biopsies were weighed before DNA extraction, and up to 25 mg of biopsy tissue was used for

729 extraction. Biopsies were incubated at 56 ºC overnight in 40 µL of proteinase K, 180 µL of ATL

730 buffer, and 20 mg/mL of lysozyme (Sigma-Aldrich, St. Louis, Missouri, USA) until tissue lysis

731 was complete. Lysis mixture was then transferred to a tube containing 200 mg of 0.1 mm

732 zirconia/silica beads for 3 min of uninterrupted bead beating, then DNA extraction followed the

733 Qiagen DNeasy Blood and Tissue kit (Qiagen, Hilden, Germany) under manufacturer’s

734 recommendations for animal tissue. DNA extraction controls were performed without biopsy

735 tissue to identify contaminants during the extraction process. DNA was stored at -80 ºC in

736 DNase/RNase-free water until use for sequencing and qPCR.

737

738 Deep amplicon sequencing and analysis

739 The V3-V4 hypervariable region of the 16S rRNA gene was sequenced from purified biopsy

740 DNA, along with DNA extraction and blank (DNase/RNase-free water) controls. Primers used

741 for sequencing are described in Table S2.1. In a nested reaction to generate amplicons for

742 sequencing, 15 cycles of amplification were performed using 8F

47 743 (AGAGTTTGATCCTGGCTCAG) and 926R (CCGTCAATTCCTTTRAGTTT) primers,

744 followed by 25 cycles with 341F and 806R primers (26) to amplify the V3-V4 hypervariable

745 region. PCR products were visualized on agarose gels and positive amplicons were sequenced on

746 the Illumina Miseq platform (v3; 600 cycles; 2 x 300 nt). Amplification and sequencing of

747 biopsy DNA was performed by the McMaster Genome Facility (Hamilton, ON, Canada).

748

749 Demultiplexed reads were analyzed and processed using the DADA2 R package v.1.14.1 (27).

750 Forward and reverse reads were truncated approximately after average Phred score dropped

751 below 30. Shorter reads, and reads with ambiguous nucleotides were removed. Next, amplicon

752 sequence variants (ASVs) were inferred, and then paired-end reads were merged with a

753 minimum overlap of 30 nt. After the removal of chimeric sequences, ASVs were taxonomically

754 classified using the SILVA database v.132 (28). Species classification for Treponema,

755 Bacteroides, Mycoplasma, Porphyromonas, and Fusobacterium was performed locally with

756 BLAST+ v.2.10.0 (29) against the NCBI 16S ribosomal RNA RefSeq database (30) using the top

757 hit and a 97% identity cut off against full length sequences for species classification.

758

759 Weighted and unweighted UniFrac distances were calculated and analyzed using Principal

760 Coordinates (PCoA) to identify differences in microbial composition between samples. Variation

761 in microbial composition relative to M stage was measured by permutational ANOVA with 999

762 permutations, and was considered significant with a P value less than 0.05. All diversity analysis

763 was performed using vegan v.2.5.6 (31). Relative abundances were calculated for all samples,

764 and then displayed as a mean percent relative abundance for each M stage. DESeq2 v.1.26.0 (32)

765 was used to normalize sequencing depth and to identify DD-associated taxa across M stages, and

48 766 associations were considered significant at a P value less than 0.01. All analyses of deep

767 amplicon sequencing reads were conducted in R v.3.5.3.

768

769 Whole genome sequencing and analysis

770 Whole genome sequencing (WGS) was performed for isolates of Fusobacterium sp.,

771 Porphyromonas levii, and Bacteroides pyogenes (Table S2.2) to identify species-specific genes

772 for qPCR targets. DNA was extracted from these isolates using the Qiagen DNeasy Blood and

773 Tissue kit (Qiagen, Hilden, Germany), according to manufacturer’s instructions, and normalized

774 to 5 ng/µL using the Qubit dsDNA HS kit (Life Technologies, Carlsbad, California, USA).

775 Sequencing and draft genome assembly was conducted as described in Derakhshani et al. (33)

776 for barcoded Illumina HiSeq reads. Briefly, sequencing libraries were prepared using the

777 NEBNext Ultra II FS DNA Library Prep kit for Illumina (NEB, Ipswich, Massachusetts, USA).

778 Libraries were sequenced at the McMaster Genome Facility (Hamilton, ON, Canada) on the

779 Illumina HiSeq 2500 System (2 x 250 nt). De novo assembly of draft genomes was performed

780 using Unicycler v.0.4.8 (34). Assembled genomes were then annotated using prokka v.1.14.5

781 (35).

782

783 Species-specific genes, defined as genes that are present in all known strains of a particular

784 species but absent in other bacteria, were identified for Porphyromonas levii, Fusobacterium sp.,

785 and Bacteroides pyogenes, according to methods described by Naushad et al. (36). Briefly,

786 potential unique genes were identified by performing a BLASTn search of all open reading

787 frames (ORFs) of a species of interest against an in-house database containing newly sequenced

788 genomes and representative genomes from known bacterial species. The ORFs that were

49 789 detected in a single species but not identified in any other species were considered unique genes.

790 The specificity and copy number of candidate species-specific genes was confirmed by the

791 BLASTn search against the draft genomes and publicly available genomes of each species.

792

793 Multiplex qPCR development and validation

794 A multiplex qPCR targeting species-specific genes of Porphyromonas levii, Fusobacterium sp.

795 (undefined species, BioSample accession number: SAMN16729900), and Bacteroides pyogenes

796 was developed and validated for absolute quantification directly from DD biopsy DNA. Species-

797 specific genes for each target species were IX289_000224 (Fusobacterium sp.), IX335_000626

798 (P. levii), and IX319_002165 (B. pyogenes). Only one copy of each gene is assumed per genome

799 based on BLAST searches of each species-specific gene against all genomes available for each

800 species. Primers and fluorescent probes were designed using Primer3 v.4.1.0 and are shown in

801 Table S2.1. Primer and probe sequences were designed so that primers had a Tm of

802 approximately 60 ºC and probes had a Tm from 67–70 ºC, amplifying targets between 75–175

803 bp. GC content was between 40–60% for all oligonucleotides. Primer specificity in singleplex

804 reactions were validated with melt curve analysis, gel electrophoresis, and Sanger sequencing of

805 the qPCR product. Primer annealing temperature was optimized in a temperature gradient and

806 melting curve analysis ensured specificity. Multiplex reactions with all primer/probe

807 combinations were optimized when singleplex and multiplex results run in parallel were within 1

808 Ct value, standard deviation among triplicate Ct values were less than 0.5, and reaction

809 efficiencies were all between 90–110%. DNA purified from each species (Biosample accessions:

810 P. levii, SAMN16729910; Fusobacterium sp., SAMN16729900; B. pyogenes, SAMN16729906)

811 were used as qPCR standards. The multiplex qPCR was further validated for absolute

50 812 quantification directly from biopsy DNA through a spike-in experiment, in order to determine if

813 accurate absolute quantification was possible using this multiplex qPCR. Standard DNA (5 x 104

814 copies) for each species was spiked into five biopsy DNA samples that didn’t have any prior

815 detectable species DNA, and qPCR output (triplicate copy numbers of target gene) was

816 compared to the expected 5 x 104 gene copies. The final multiplex qPCR contained all primers at

817 500 µM, probes at 250 µM, 10 µL TaqMan™ Fast Advanced Master Mix (Applied Biosystems,

818 Foster City, CA, USA), 2 µL of template biopsy DNA (20 ng), and H2O to a total reaction

819 volume of 20 µL. Final multiplex qPCR cycling conditions was 50 ºC for 2 min, 95 ºC for 20 s,

820 40 cycles of 95 ºC for 10 s and 59.6 ºC for 30 s.

821

822 Quantitative real-time PCR and analysis

823 Absolute quantification by qPCR (CFX96TM Real-Time System, Bio-Rad Laboratories, Inc.;

824 Hercules, California, USA) was performed for microbiota strongly associated with DD lesions.

825 In total, 3 different qPCR assays were used in this study: one multiplex qPCR developed by

826 Beninger et al. (22) targeting 4 different Treponema spp. highly prevalent in DD, a qPCR

827 developed by Witcomb et al. (37) targeting Fusobacterium necrophorum, and the multiplex

828 qPCR developed in this study targeting P. levii, Fusobacterium sp., and B. pyogenes. Standards

829 for the Treponema qPCR were prepared as plasmid copy numbers according to Beninger et al.

830 (22), whereas genomic DNA purified from DD isolates were used for the other qPCR assays

831 (BioSample accessions: P. levii, SAMN16729910; Fusobacterium sp., SAMN16729900; B.

832 pyogenes, SAMN16729906; F. necrophorum, SAMN16729904). All standards were measured

833 with Qubit dsDNA HS kit (Life Technologies, Carlsbad, California, USA) before each reaction.

834 Purified biopsy DNA was normalized to 10 ng/µL prior to qPCR. All qPCR reactions required

51 835 efficiency between 90–110%, and all no-template controls negative. All DNA extraction controls

836 were tested with each qPCR.

837

838 All absolute abundances were normalized by tissue biopsy weight used in DNA extraction so

839 that bacterial quantities were compared as copy numbers per mg of biopsy tissue. Non-

840 parametric Kruskal-Wallis tests and post-hoc Mann-Whitney U tests were used to identify

841 significant differences between mean species abundances. Correlation networks were generated

842 to determine the direction and strength of pairwise associations between individual bacterial

843 species in DD lesions. Correlation matrices were generated by calculating pairwise Spearman

844 correlations between natural log transformed copy numbers of each bacterial species. Only

845 significant correlations were included in the network analysis. Networks were visualized using

846 the R package igraph v 1.2.5 (38). All P values were corrected for multiple hypotheses with the

847 Benjamini-Hochberg method, and P values less than 0.05 were considered statistically

848 significant. All analysis was conducted in R v.3.5.3.

849

850 Data availability

851 Raw fastq reads generated from deep amplicon sequencing are accessible using the BioProject

852 ID PRJNA664530 in the NCBI SRA database (www.ncbi.nlm.nih.gov/sra). WGS for all strains

853 sequenced in this study are accessible in the NCBI Genbank database

854 (www.ncbi.nlm.nih.gov/genbank) under the BioProject ID PRJNA676053.

855

856 Results:

52 857 Altogether, we collected and analyzed the microbial composition of 120 skin biopsies. We

858 collected 40 biopsies from healthy skin (M0), 8 biopsies from M1 lesions, 38 biopsies from

859 ulcerated M2 lesions, 4 biopsies from healing M3 lesions, 20 biopsies from M4 lesions, and 10

860 biopsies from active M4.1 lesions.

861

862 Deep amplicon sequencing of the V3-V4 hypervariable region

863 Out of 120 skin biopsies obtained in this study, 98 resulted in successful amplification, based on

864 gel electrophoresis of the nested V3-V4 hypervariable region PCR assay, and were then

865 sequenced on the Illumina MiSeq platform (M0, n = 20; M1, n = 8; M2, n = 37; M3, n = 4; M4,

866 n = 19; M4.1, n = 10). A total of 4,860,255 sequences passed initial quality filtering, and after

867 inferring ASVs and chimera removal, 2,994,368 sequences were used for taxonomic

868 classification. No sequences from DNA extraction controls remained after quality filtering and

869 DADA2 processing (Fig. S2.1A). Negative controls (DNase/RNase-free water) contained

870 relatively fewer reads, on average, compared to healthy skin and DD samples, and represented

871 only two ASVs from Proteobacteria (Fig. S2.1A and B). After removing sequences with low

872 reads (less than 500), 90 samples remained for microbiota analysis (M0, n = 16; M1, n = 6; M2,

873 n = 37; M3, n = 4; M4, n = 18; M4.1, n = 9). Permutational analysis of variance

874 (PERMANOVA) on weighted and unweighted UniFrac distances each identified a significant

875 difference (P = 0.001) in microbial composition among M stages (Fig. 2.1A and B).

876

877 Healthy skin primarily contained Actinobacteria, Bacteroidetes, Proteobacteria, and

878 predominantly Firmicutes (Fig. 2.2A). Compared to healthy skin, DD lesions across all M stages

879 had a higher relative abundance of Spirochaetes, accompanied by a higher relative abundance of

53 880 Tenericutes (Fig. 2.2A). Fusobacteria were also associated with DD lesions, and were relatively

881 absent in healthy skin, and peaked at approximately 3.4% relative abundance in M2 lesions (Fig.

882 2.2A). At family level taxonomic grouping (Fig. 2.2B), healthy skin had relatively more diverse

883 bacterial populations than DD lesions. Of the taxonomically classifiable sequences,

884 Spirochaetaceae were the most abundant in DD lesions (Fig. 2.2B). There were also higher

885 relative abundances of Porphyromonadaceae, Mycoplasmataceae, Family XI, and

886 Fusobacteriaceae in DD lesions compared to healthy skin (Fig. 2.2B). No obvious visual

887 differences were apparent in relative abundance between M stages for these DD-associated

888 bacteria. Mycoplasmataceae, however, had its highest relative abundance in M3, M4 and M4.1

889 lesions (Fig. 2.2B). Fusobacteriaceae strongly associated with M2 lesions compared to other

890 disease stages, only having a relative abundance greater than 3% in M2 lesions.

891

892 Using DESeq2, differential abundance analysis in DD lesions compared to healthy skin

893 identified a group of genera strongly associated with DD lesions (Fig. 2.3). Treponema,

894 Porphyromonas, Mycoplasma, and Fusobacterium counts were significantly higher (P < 0.01) in

895 most of the DD M stages compared to healthy skin (Fig. 2.3). In addition, many genera that are

896 members of the Firmicutes phyla had significantly higher (P < 0.01) counts in most DD lesions

897 compared to healthy skin (Fig. 2.3).

898

899 Species level classification was pursued for members of Bacteroides, Fusobacterium,

900 Mycoplasma, Porphyromonas, and Treponema. Interestingly, Fusobacterium were the only

901 genus to not have unclassified reads, while all other genera had large proportions of unclassified

902 ASVs. For Bacteroides, only B. pyogenes was identified in M2 and M4 lesions at a relatively

54 903 low relative abundance (Fig. 2.4). F. mortiferum was not detectable in M1 lesions but was

904 relatively more abundant than F. necrophorum in all other DD stages (Fig. 2.4). Of all

905 Mycoplasma spp. identified, M. fermentans was the primary Mycoplasma sp. within M2 lesions;

906 whereas, in other M stages (M4 and M4.1), multiple species of Mycoplasma became abundant

907 (Fig. 2.4). The distribution of Treponema spp. appeared highly dynamic across M stages,

908 beginning with T. medium as the predominant species present in healthy skin, and T. pedis, T.

909 phagedenis, and T. refringens were relatively abundant in all DD lesion stages (Fig. 2.4).

910

911 Identification of bacterial isolates from skin biopsies

912 Bacterial culture was attempted for all skin biopsies obtained. A total of 198 isolates from 69

913 biopsies were successfully cultured and identified. Bacteria that were cultured from DD lesions

914 and were not isolated from healthy skin biopsies are shown in Table S2.2. A number of species

915 from the genera Fusobacterium and Porphyromonas were isolated from all M stages of DD

916 lesions, but were not cultured from M0 biopsies. In addition, B. pyogenes was identified in all M

917 stages of DD lesions, but was not isolated from M0 samples.

918

919 Quantitative real-time PCR (qPCR) design

920 Isolates of B. pyogenes, Fusobacterium sp., and P. levii were collected and identified by full

921 length 16S rRNA sequence alignment (Table S2.2). Fusobacterium sp. 16S rRNA gene sequence

922 best match was against F. mortiferum (96.97% identity). These three species isolated from

923 biopsy cultures matched the DD-associated microbiota identified through deep amplicon

924 sequencing (Fig. 2.4), and thus were targeted for qPCR development. To demonstrate the

925 accuracy of the qPCR, we spiked in 5 x 104 copies of target DNA directly into purified biopsy

55 926 DNA, in which the maximum difference observed between actual DNA copies present and mean

927 qPCR output was 7 x 103, or 14% (Table S2.3).

928

929 Species absolute quantification directly from purified biopsy DNA

930 A total of 120 purified biopsy DNA samples (M0, n = 40; M1, n = 8; M2, n = 38; M3, n = 4; M4,

931 n = 20; M4.1, n = 10) were used for species absolute quantification. Absolute abundance was

932 quantified for a total of 8 DD-associated species. Treponema denticola was only detected in 5

933 samples due to the qPCR reaction requiring at least 103 gene copies for detection. Thus T.

934 denticola was not included in further analysis. All DNA extraction controls had non-detectable

935 levels of each species. Most of the species quantified were significantly more abundant (P <

936 0.05) in DD lesions compared to healthy skin, except for T. medium, which had no significant

937 differences in abundance across any M stage (Table 2.1). The majority of species had their

938 highest abundance in M2 and M3 lesions (Table 2.1; Fig. 2.5).

939

940 All species tested by qPCR were detectable in the majority of DD lesions, except for T. medium,

941 which was only detectable in approximately 40% of DD lesions, but was present in 62% of M1

942 lesions (Table 2.2). In active DD lesions, P. levii was detectable by qPCR in all samples, and

943 detectable in 38% of healthy skin samples (Table 2.2). Similarly, T. phagedenis was detectable in

944 the majority of samples from all M stages, including healthy skin (Table 2.2). Of all species that

945 were detectable in the majority of DD lesion samples, Fusobacterium sp. was detected in the

946 lowest amount of healthy skin samples at 2.5% (Table 2.2).

947

56 948 Spearman rank correlations performed on absolute abundances in DD lesions identified B.

949 pyogenes and P. levii as having the strongest association among all pairwise species

950 combinations (Fig. S2.2, Fig. 2.6). Negative correlations were only identified in T. medium

951 pairwise comparisons with both Fusobacterium species (Fig. S2.2). In addition, T. medium only

952 had significant pairwise correlations (P < 0.05) with T. phagedenis (Fig. 2.6). Treponema

953 phagedenis was the only Treponema spp. to have significant correlations (P < 0.05) with non-

954 Treponema spp. (Fig. 2.6). Fusobacterium sp. and F. necrophorum abundances were not

955 significantly correlated (Fig. 2.6).

956

957 Absolute abundance of species in healthy skin across farms

958 Healthy skin samples from three separate feedlots were acquired; two feedlots had active cases

959 of DD at the time of sampling, and one feedlot had no active cases of DD during the sampling

960 period. In the feedlot with no active DD cases, T. phagedenis, T. medium, F. necrophorum, B.

961 pyogenes, and P. levii had significantly lower abundances (P < 0.05) compared to healthy skin

962 samples from both active DD feedlots (Table S2.4). In addition, T. medium, F. necrophorum, and

963 B. pyogenes were detectable in zero samples from the feedlot without active DD cases (Table

964 S2.5).

965

966 Comparison of qPCR and deep amplicon sequencing quantification methods

967 On average, Treponema made up the majority of detectable DD lesion microbiota, regardless of

968 quantification method (Fig. 2.2, Fig. 2.5). Data from deep amplicon sequencing appeared to

969 consistently overrepresent T. medium across all M stages when compared to qPCR abundances

970 (Fig. 2.7). Abundance dynamics for T. pedis and T. phagedenis were relatively comparable

57 971 across M stages for both quantification methods, except for a relative overrepresentation of T.

972 phagedenis in the M4.1 stage as measured by qPCR (Fig. 2.7). Deep amplicon sequencing also

973 showed a higher relative abundance of Fusobacterium spp. compared to qPCR quantification,

974 which instead favoured P. levii abundance over all other non-Treponema species tested (Fig.

975 2.7).

976

977 Discussion:

978 In this study, we presented a microbiota strongly associated with DD lesions of feedlot beef

979 cattle. The majority of DD lesion microbiota comprised of Treponema spp., namely T.

980 phagedenis and T. pedis; meanwhile, P. levii, B. pyogenes, and the presence of two different

981 Fusobacterium spp. also strongly differentiated DD lesions from healthy skin. Most of these

982 potential DD pathogens were significantly higher in abundance in all DD M stages when

983 compared to normal healthy skin. These data provide evidence of DD lesion formation and

984 development as a potential outcome of the prevalence and abundance dynamics of these key

985 species across DD M stages.

986

987 Overall beef DD lesion microbiota did not have drastic differences compared to the dairy DD

988 lesion microbiota identified in previous studies (15, 20). In beef lesions Treponema were the

989 predominant member in terms of relative abundance, and Porphyromonas, Mycoplasma, and

990 Fusobacterium were more prevalent in DD lesions compared to healthy skin. Similar DD-

991 associated microbiota have been identified in dairy DD studies (20, 39, 40). We also identified

992 significantly higher abundances of some Firmicutes taxa in DD lesions compared to healthy skin

993 in beef breeds. In contrast, there is little evidence for involvement of Firmicutes in DD

58 994 progression of dairy cattle, which have low relative abundance and overall decreasing counts as

995 lesions progress (15, 20).

996

997 Treponema spp. are consistently argued as one of the main causative agents of DD, and we

998 provide further evidence of their presence and potential involvement, particularly for T.

999 phagedenis and T. pedis, which were prevalent in beef cattle DD lesions. We were able to

1000 identify seven different Treponema spp. that were present in DD lesions from classifiable V3-V4

1001 hypervariable region sequences, supporting previous hypotheses that multiple strains and species

1002 of Treponema can play a role in DD development. We were able to accurately quantify the

1003 absolute abundances of three Treponema spp. by qPCR; however, there are additional species

1004 outside of our targets, which have been consistently identified in DD lesions, such as T.

1005 refringens (20, 39), and require further study to validate their associations with DD lesion stages.

1006

1007 In early stage DD lesions, we identified that M1 lesions are strongly associated with the presence

1008 of T. phagedenis and T. medium relative to healthy skin, suggesting the potential importance of

1009 these species in early lesion development. Of note, T. medium prevalence and abundance on

1010 average is lower in most other M stages compared to M1 lesions, suggesting T. medium might

1011 contribute to initiating but not sustaining lesions. Treponema phagedenis was the most prevalent

1012 Treponema spp. tested across all DD lesions, and was the only Treponema spp. to have

1013 significant correlations with non-Treponema spp. abundance. In this analysis, T. phagedenis

1014 appeared to be the potential interactive link between Treponema spp. and other genera, although

1015 this could be a product of relatively high T. phagedenis abundance and prevalence compared to

59 1016 most other species examined, but this still warrants further study on the potential metabolic

1017 interactions between DD-associated species.

1018

1019 Recently, a polybacterial etiology for DD is generally agreed upon, with many non-Treponema

1020 bacteria also being consistently identified in DD lesions (20, 40). The novel multiplex qPCR

1021 developed as part of this study in combination with our deep sequencing results provide strong

1022 evidence supporting a polybacterial etiopathogenesis of DD based on the multiple bacterial

1023 species associated with DD M stages. In particular, P. levii has been identified in DD lesions in

1024 prior studies via high throughput sequencing and fluorescent in situ hybridization (FISH) (15,

1025 39), but until now has not had a sensitive and reproducible method of quantification across lesion

1026 stages. Our novel multiplex qPCR enabled us to identify P. levii in all active stages of DD

1027 lesions sampled in feedlot beef cattle. This finding, combined with P. levii being one of the most

1028 abundant bacteria we quantified outside of the genus Treponema, represents strong evidence of a

1029 possible etiological role in DD. Porphyromonas levii has been previously dismissed as a mere

1030 secondary opportunistic invader based on its predominant superficial location within the dermis

1031 of DD lesions (15, 39), potentially reducing its overall effect in DD lesion formation. However, a

1032 study quantifying bacterial gene expression patterns in DD lesions supports a potential

1033 involvement for P. levii in DD development (41) Thus, it is possible P. levii plays a role in

1034 influencing overall metabolic processes of DD microbiota and certainly warrants further

1035 investigation.

1036

1037 The identification of F. necrophorum, which is considered the primary causative agent of bovine

1038 foot rot (42), in DD lesions is not novel; however, the finding of an additional Fusobacterium sp.

60 1039 by culturing DD biopsies hasn’t been shown before in previous DD literature to the best of our

1040 knowledge. In addition, F. equinum was the only other Fusobacterium species identifiable from

1041 V3-V4 hypervariable region sequences, but it was not included in our qPCR quantification

1042 because it was less prevalent than the other two Fusobacterium spp., and was not cultured from

1043 DD lesions. Both Fusobacterium spp. quantified by species-specific qPCR assays had similar

1044 absolute abundances throughout each M stage, and correlation analysis identified no significant

1045 correlation between the two species. The random pairwise distributions between the two

1046 Fusobacterium spp. abundances suggests that each species is randomly dispersed across different

1047 animals. Therefore, if there is a causative role for Fusobacterium in DD, it is possible that there

1048 may be limited functional differences between Fusobacterium spp. in DD development.

1049 Investigations into homologues of main virulence factors between the two Fusobacterium spp.

1050 can help identify whether or not the species have unique roles in lesions. Given that F.

1051 necrophorum has been shown to generate mixed species biofilms with P. levii to impair

1052 neutrophil response (43), it is possible that the Fusobacterium sp. we isolated in this study may

1053 interact in a similar way.

1054

1055 Healthy skin microbiota varied significantly across different feedlots with and without active DD

1056 cases at the time of sampling. Of particular interest, T. medium, F. necrophorum, and B.

1057 pyogenes were all undetectable in healthy skin only of animals on a feedlot with no active cases

1058 of DD. This finding opens a multitude of avenues for investigating the transmission dynamics of

1059 DD, as well as shedding additional light on potential pathogenesis that establishes early stage

1060 lesions in natural DD progression. These results also illuminate the possible connection of farm-

1061 and animal-based risk factors with the microbial community dynamics from healthy to DD skin.

61 1062 However, with a low number of feedlots included in this study, and the numerous different

1063 management practices between farms, there currently does not exist sufficient validity to make

1064 significant conclusions on these findings.

1065

1066 The methodology and findings presented in this study provide a crucial advancement in our

1067 understanding of bovine DD microbiology. The combined use of deep amplicon sequencing and

1068 culture methods to identify key DD-associated bacterial species, and then developing a novel

1069 multiplex qPCR to validate their population dynamics across lesion stages provides a deeper and

1070 more thorough understanding of DD than any of these techniques alone. Along with the rise in

1071 high throughput sequencing strategies to identify DD-associated microbiota come significant

1072 limitations and biases affecting the validity of findings, including varying copy numbers of target

1073 amplicon (i.e. 16S rRNA gene) across taxa, uneven sampling depth, primer bias, low taxonomic

1074 resolution, and variation in bioinformatic processing (44). Knowledge of the microbial

1075 community is required beforehand to develop an appropriate mock community to quantify some

1076 of these biases, but this is not always possible when studying a novel ecological system. Real-

1077 time PCR can provide a more reliable quantification on a narrow subset of species, and serves as

1078 an appropriate validation of deep amplicon sequencing results (45). Our method of selecting

1079 species-specific genes for qPCR targets leads to a specific reaction targeting individual species.

1080 However, this method is validated on publicly available genomes and locally-derived strains,

1081 thus it cannot be guaranteed to efficiently amplify all global strains of each species, thereby this

1082 qPCR should be consistently tested as more genomes become available for each species.

1083

62 1084 The combination of qPCRs used in this study target a group of bacteria associated with DD

1085 pathogenesis, but not all species of interest could be targeted. It is critical to involve additional

1086 Treponema spp., but perhaps more importantly, we need more understanding on the Mycoplasma

1087 spp. dynamics in DD lesions. Mycoplasma is more recently implicated in DD pathogenesis (15,

1088 20); however, very little validated information of the species involved exists outside of M.

1089 fermentans that was identified in the majority of DD lesions in one study by PCR (15). There is a

1090 lack of targeted culture methods to successfully isolate Mycoplasma from DD tissue, and thus

1091 there are no publicly available genomes from Mycoplasma spp. isolated from DD tissue or even

1092 other related diseases of the hoof area. Genomes derived from DD-relevant isolates are essential

1093 to developing robust methods to study their population dynamics throughout lesion development.

1094 Through deep amplicon sequencing, we identified that the relative abundance of Mycoplasma is

1095 highest in M4 and M4.1 lesions; therefore, further characterization of Mycoplasma spp. may

1096 uncover significant links in DD etiopathogenesis of chronic lesions.

1097

1098 Conclusion

1099 We demonstrated a combination of bacterial species strongly associated with polymicrobial DD

1100 lesions. Abundance of Treponema, Fusobacterium, Bacteroides, and Porphyromonas strongly

1101 differentiated DD lesions from healthy skin. The combination of methodologies and the

1102 multiplex qPCR performed in this study targets a critical need in DD research for identification

1103 of species involved in DD lesions. Through this approach, we provide an accurate and sensitive

1104 method of quantification for these potential DD pathogens from DNA samples. Further

1105 investigations into additional species from other genera and especially further characterization of

1106 additional Treponema and Mycoplasma species can facilitate significant leaps in identifying

63 1107 etiopathological agents. This study along with future characterizations will be necessary to fully

1108 understand the microbiological factors involved in DD progression, and lead to the development

1109 of more effective mitigation and treatment strategies.

1110

1111 Acknowledgments

1112 We thank the Alberta beef producers that were involved in the study, as well as Drs. Hendrick

1113 and Klassen of Coaldale Veterinary Clinic for their assistance in identifying potential study

1114 participants. Thank you to Lacey Wynnyk, Tina Petersen, Anice Thomas, and Julián Cortés for

1115 their assistance in DD biopsy sample collection. Thank you to Dr. Janet Hill and Dr. Douglas

1116 Morck for their constructive feedback on this manuscript.

1117

1118 Research funding was provided by Alberta Agriculture and Forestry, and the Simpson Ranch

1119 Grant at the University of Calgary Faculty of Veterinary Medicine. Additional scholarship

1120 funding was provided by the Canadian Dairy Commission.

1121

1122 References

1123 1. Cheli R, Mortellaro C. 1974. La dermatite digitale del bovino, p. 208–213. In Proceedings

1124 of the 8th International Conference on Diseases of Cattle.

1125 2. Orsel K, Plummer P, Shearer J, De Buck J, Carter SD, Guatteo R, Barkema HW. 2018.

1126 Missing pieces of the puzzle to effectively control digital dermatitis. Transbound Emerg

1127 Dis 65:186–198.

1128 3. Read DH, Walker RL. 1998. Papillomatous digital dermatitis (footwarts) in California

1129 dairy cattle: Clinical and gross pathologic findings. J Vet Diagnostic Investig 10:67–76.

64 1130 4. Gomez A, Cook NB, Socha MT, Döpfer D. 2015. First-lactation performance in cows

1131 affected by digital dermatitis during the rearing period. J Dairy Sci 98:4487–4498.

1132 5. Dolecheck KA, Overton MW, Mark TB, Bewley JM. 2019. Use of a stochastic simulation

1133 model to estimate the cost per case of digital dermatitis, sole ulcer, and white line disease

1134 by parity group and incidence timing. J Dairy Sci 102:715–730.

1135 6. Solano L, Barkema HW, Mason S, Pajor EA, LeBlanc SJ, Orsel K. 2016. Prevalence and

1136 distribution of foot lesions in dairy cattle in Alberta, Canada. J Dairy Sci 99:6828–6841.

1137 7. Yang DA, Johnson WO, Müller KR, Gates MC, Laven RA. 2019. Estimating the herd and

1138 cow level prevalence of bovine digital dermatitis on New Zealand dairy farms: A

1139 Bayesian superpopulation approach. Prev Vet Med 165:76–84.

1140 8. Kulow M, Merkatoris P, Anklam KS, Rieman J, Larson C, Branine M, Döpfer D. 2017.

1141 Evaluation of the prevalence of digital dermatitis and the effects on performance in beef

1142 feedlot cattle under organic trace mineral supplementation. J Anim Sci 95:3435–3444.

1143 9. Krull AC, Shearer JK, Gorden PJ, Scott HM, Plummer PJ. 2016. Digital dermatitis:

1144 Natural lesion progression and regression in Holstein dairy cattle over 3 years. J Dairy Sci

1145 99:3718–3731.

1146 10. Döpfer D, Koopmans A, Meijer FA, Szakáll I, Schukken YH, Klee W, Bosma RB,

1147 Cornelisse JL, Van Asten AJAM, Ter Huurne AAHM. 1997. Histological and

1148 bacteriological evaluation of digital dermatitis in cattle, with special reference to

1149 spirochaetes and Campylobacter faecalis. Vet Rec 140:620–623.

1150 11. Brandt S, Apprich V, Hackl V, Tober R, Danzer M, Kainzbauer C, Gabriel C, Stanek C,

1151 Kofler J. 2011. Prevalence of bovine papillomavirus and Treponema DNA in bovine

1152 digital dermatitis lesions. Vet Microbiol 148:161–167.

65 1153 12. Evans NJ, Brown JM, Demirkan I, Murray RD, Birtles RJ, Hart CA, Carter SD. 2009.

1154 Treponema pedis sp. nov., a spirochaete isolated from bovine digital dermatitis lesions. Int

1155 J Syst Evol Microbiol 59:987–991.

1156 13. Klitgaard K, Bretó AF, Boye M, Jensen TK. 2013. Targeting the treponemal microbiome

1157 of digital dermatitis infections by high-resolution phylogenetic analyses and comparison

1158 with fluorescent in situ hybridization. J Clin Microbiol 51:2212–2219.

1159 14. Klitgaard K, Boye M, Capion N, Jensen TK. 2008. Evidence of multiple Treponema

1160 phylotypes involved in bovine digital dermatitis as shown by 16S rRNA gene analysis and

1161 fluorescence in situ hybridization. J Clin Microbiol 46:3012–3020.

1162 15. Nielsen MW, Strube ML, Isbrand A, Al-Medrasi WDHM, Boye M, Jensen TK, Klitgaard

1163 K. 2016. Potential bacterial core species associated with digital dermatitis in cattle herds

1164 identified by molecular profiling of interdigital skin samples. Vet Microbiol 186:139–149.

1165 16. Capion N, Boye M, Ekstrøm CT, Jensen TK. 2012. Infection dynamics of digital

1166 dermatitis in first-lactation Holstein cows in an infected herd. J Dairy Sci 95:6457–6464.

1167 17. Schlafer S, Nordhoff M, Wyss C, Strub S, Hübner J, Gescher DM, Petrich A, Göbel UB,

1168 Moter A. 2008. Involvement of Guggenheimella bovis in digital dermatitis lesions of dairy

1169 cows. Vet Microbiol 128:118–125.

1170 18. Yano T, Moe KK, Yamazaki K, Ooka T, Hayashi T, Misawa N. 2010. Identification of

1171 candidate pathogens of papillomatous digital dermatitis in dairy cattle from quantitative

1172 16S rRNA clonal analysis. Vet Microbiol 143:352–362.

1173 19. Staton GJ, Sullivan LE, Blowey RW, Carter SD, Evans NJ. 2020. Surveying bovine

1174 digital dermatitis and non-healing bovine foot lesions for the presence of Fusobacterium

1175 necrophorum, Porphyromonas endodontalis and Treponema pallidum. Vet Rec 186:450.

66 1176 20. Krull AC, Shearer JK, Gorden PJ, Cooper VL, Phillips GJ, Plummera PJ. 2014. Deep

1177 sequencing analysis reveals temporal microbiota changes associated with development of

1178 bovine digital dermatitis. Infect Immun 82:3359–3373.

1179 21. Johnson JS, Spakowicz DJ, Hong BY, Petersen LM, Demkowicz P, Chen L, Leopold SR,

1180 Hanson BM, Agresta HO, Gerstein M, Sodergren E, Weinstock GM. 2019. Evaluation of

1181 16S rRNA gene sequencing for species and strain-level microbiome analysis. Nat

1182 Commun 10:5029.

1183 22. Beninger C, Naqvi SA, Naushad S, Orsel K, Luby C, Derakhshani H, Khafipour E, De

1184 Buck J. 2018. Associations between digital dermatitis lesion grades and the quantities of

1185 four Treponema species. Vet Res 49:111.

1186 23. Sullivan LE, Carter SD, Blowey R, Duncan JS, Grove-White D, Evans NJ. 2013. Digital

1187 dermatitis in beef cattle. Vet Rec 173:582.

1188 24. Sullivan LE, Evans NJ, Blowey RW, Grove-White DH, Clegg SR, Duncan JS, Carter SD.

1189 2015. A molecular epidemiology of treponemes in beef cattle digital dermatitis lesions

1190 and comparative analyses with sheep contagious ovine digital dermatitis and dairy cattle

1191 digital dermatitis lesions. Vet Microbiol 178:77–87.

1192 25. Berry SL, Read DH, Famula TR, Mongini A, Döpfer D. 2012. Long-term observations on

1193 the dynamics of bovine digital dermatitis lesions on a California dairy after topical

1194 treatment with lincomycin HCl. Vet J 193:654–658.

1195 26. Klindworth A, Pruesse E, Schweer T, Peplies J, Quast C, Horn M, Glöckner FO. 2013.

1196 Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-

1197 generation sequencing-based diversity studies. Nucleic Acids Res 41:e1.

1198 27. Callahan BJ, Mcmurdie PJ, Rosen MJ, Han AW, Johnson AJA, Holmes SP. 2016.

67 1199 DADA2: High-resolution sample inference from Illumina amplicon data. Nat Methods

1200 13:581–583.

1201 28. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, Peplies J, Glöckner FO.

1202 2013. The SILVA ribosomal RNA gene database project: Improved data processing and

1203 web-based tools. Nucleic Acids Res 41:590–596.

1204 29. Camacho C, Coulouris G, Avagyan V, Ma N, Papadopoulos J, Bealer K, Madden TL.

1205 2009. BLAST+: Architecture and applications. BMC Bioinformatics 10:421.

1206 30. O’Leary NA, Wright MW, Brister JR, Ciufo S, Haddad D, McVeigh R, Rajput B,

1207 Robbertse B, Smith-White B, Ako-Adjei D, Astashyn A, Badretdin A, Bao Y, Blinkova

1208 O, Brover V, Chetvernin V, Choi J, Cox E, Ermolaeva O, Farrell CM, Goldfarb T, Gupta

1209 T, Haft D, Hatcher E, Hlavina W, Joardar VS, Kodali VK, Li W, Maglott D, Masterson P,

1210 McGarvey KM, Murphy MR, O’Neill K, Pujar S, Rangwala SH, Rausch D, Riddick LD,

1211 Schoch C, Shkeda A, Storz SS, Sun H, Thibaud-Nissen F, Tolstoy I, Tully RE, Vatsan

1212 AR, Wallin C, Webb D, Wu W, Landrum MJ, Kimchi A, Tatusova T, DiCuccio M, Kitts

1213 P, Murphy TD, Pruitt KD. 2016. Reference sequence (RefSeq) database at NCBI: Current

1214 status, taxonomic expansion, and functional annotation. Nucleic Acids Res 44:D733–

1215 D745.

1216 31. Dixon P. 2003. VEGAN, a package of R functions for community ecology. J Veg Sci

1217 14:927–930.

1218 32. Love MI, Huber W, Anders S. 2014. Moderated estimation of fold change and dispersion

1219 for RNA-seq data with DESeq2. Genome Biol 15:550.

1220 33. Derakhshani H, Bernier SP, Marko VA, Surette MG. 2020. Completion of draft bacterial

1221 genomes by long-read sequencing of synthetic genomic pools. BMC Genomics 21:519.

68 1222 34. Wick RR, Judd LM, Gorrie CL, Holt KE. 2017. Unicycler: Resolving bacterial genome

1223 assemblies from short and long sequencing reads. PLOS Comput Biol 13:e1005595.

1224 35. Seemann T. 2014. Prokka: Rapid prokaryotic genome annotation. Bioinformatics

1225 30:2068–2069.

1226 36. Naushad HS, Lee B, Gupta RS. 2014. Conserved signature indels and signature proteins

1227 as novel tools for understanding microbial phylogeny and systematics: Identification of

1228 molecular signatures that are specific for the phytopathogenic genera Dickeya,

1229 Pectobacterium and Brenneria. Int J Syst Evol Microbiol 64:366–383.

1230 37. Witcomb LA, Green LE, Kaler J, Ul-Hassan A, Calvo-Bado LA, Medley GF, Grogono-

1231 Thomas R, Wellington EMH. 2014. A longitudinal study of the role of Dichelobacter

1232 nodosus and Fusobacterium necrophorum load in initiation and severity of footrot in

1233 sheep. Prev Vet Med 115:48–55.

1234 38. Csardi G, Nepusz T. 2006. The igraph software package for complex network research.

1235 InterJournal Complex Sy:1695.

1236 39. Moreira TF, Facury Filho EJ, Carvalho AU, Strube ML, Nielsen MW, Klitgaard K, Jensen

1237 TK. 2018. Pathology and bacteria related to digital dermatitis in dairy cattle in all year

1238 round grazing system in Brazil. PLoS One 13:e0193870.

1239 40. Zinicola M, Lima F, Lima S, Machado V, Gomez M, Döpfer D, Guard C, Bicalho R.

1240 2015. Altered microbiomes in bovine digital dermatitis lesions, and the gut as a pathogen

1241 reservoir. PLoS One 10:e0120504.

1242 41. Marcatili P, Nielsen MW, Sicheritz-Pontén T, Jensen TK, Schafer-Nielsen C, Boye M,

1243 Nielsen M, Klitgaard K. 2016. A novel approach to probe host-pathogen interactions of

1244 bovine digital dermatitis, a model of a complex polymicrobial infection. BMC Genomics

69 1245 17:987.

1246 42. Van Metre DC. 2017. Pathogenesis and Treatment of Bovine Foot Rot. Vet Clin North

1247 Am - Food Anim Pract 33:183–194.

1248 43. Lockhart JS, Buret AG, Ceri H, Storey DG, Anderson SJ, Morck DW. 2017. Mixed

1249 species biofilms of Fusobacterium necrophorum and Porphyromonas levii impair the

1250 oxidative response of bovine neutrophils in vitro. Anaerobe 47:157–164.

1251 44. Brooks JP, Edwards DJ, Harwich MD, Rivera MC, Fettweis JM, Serrano MG, Reris RA,

1252 Sheth NU, Huang B, Girerd P, Strauss JF, Jefferson KK, Buck GA. 2015. The truth about

1253 metagenomics: Quantifying and counteracting bias in 16S rRNA studies. BMC Microbiol

1254 15:66.

1255 45. Jian C, Luukkonen P, Yki-Järvinen H, Salonen A, Korpela K. 2020. Quantitative PCR

1256 provides a simple and accessible method for quantitative microbiota profiling. PLoS One

1257 15:e0227285.

1258

1259

70 1260 Tables

1261 Table 2.1: Bacterial genomic copies (mean ± SD) per mg of biopsy tissue of each bacterial 1262 species at each M stage Lesion Tphg Tped Tmed Fn Fs Pl Bp stage M0 2.28 ± 0.333 ± 2.87 ± 0.023 ± 0.004 ± 0.181 ± 0.005 ± 12.5a 0.829a 17.2a 0.110a 0.029a 0.579a 0.026a M1 223 ± 468b 417 ± 69.1 ± 1.24 ± 0.426 ± 9.32 ± 0.700 ± 965ab 160a 2.79b 0.708b 11.4bc 1.14b M2 4.56x103 ± 3.19x103 14.8 ± 36.0 ± 34.6 ± 186 ± 8.88 ± 8.42x103c ± 55.8a 118b 89.2bc 520b 24.8b 8.08x103b M3 5.05x103 ± 3.70x104 1.36x103 37.3 ± 69.5 ± 708 ± 56.6 ± 9.78x103abc ± ± 73.6b 123c 1.34x103bc 99.6b 7.39x104b 2.72x103a M4 1.22x103 ± 1.10x103 1.55 ± 17.2 ± 35.6 ± 112 ± 6.11 ± 3.05x103b ± 5.84a 69.0b 94.8bc 348c 20.3b 4.80x103b M4.1 1.15x103 ± 15.2 ± 12.4 ± 4.70 ± 7.59 ± 147 ± 7.60 ± 2.87x103bc 29.9ab 31.8a 14.1b 21.2bc 384bc 22.5b 1263 Tphg, T. phagedenis; Tped, T. pedis; Tmed, T. medium; Fn, F. necrophorum; Fs, Fusobacterium 1264 sp.; Pl, P. levii; Bp, B. pyogenes. Different letters within a column indicate significant difference 1265 (P < 0.05). 1266 Table 2.2: Percent of samples with detectablea amounts of each bacterial species Percent (%) of samples with detectable species present Lesion Tphg Tped Tmed Fn Fs Pl Bp stage M0 68 32 25 10 2 38 5 (n=40) M1 100 38 62 50 50 100 62 (n=8) M2 97 76 29 50 76 100 60 (n=38) M3 75 100 75 50 100 75 50 (n=4) M4 75 65 40 55 60 85 35 (n=20) M4.1 80 40 40 50 70 100 70 (n=10) Total DD 89 66 39 51 70 95 55 (n=80) 1267 Tphg, T. phagedenis; Tped, T. pedis; Tmed, T. medium; Fn, F. necrophorum; Fs, Fusobacterium 1268 sp.; Pl, P. levii; Bp, B. pyogenes. a Detectable by qPCR

71 1269 Figures

1270

1271 Figure 2.1: Principal coordinates analysis on A) unweighted and B) weighted UniFrac 1272 distances. Samples were coloured based on M stage of DD lesions. 1273

72 1274

1275

1276 Figure 2.2: Percent relative abundance of bacteria within each M stage. A) Bacteria were 1277 grouped based on taxonomy at phylum level, and phyla with less than 1% relative abundance 1278 were grouped together. B) Bacteria were grouped based on taxonomy at family level, and 1279 families with less than 3% relative abundance were grouped together. 1280

1281

73 1282

1283 Figure 2.3: Differential abundance analysis of genera associated with DD lesions. DESeq2 1284 normalized sample counts were used to compare genera fold change in abundance for DD M 1285 stages (M1, M2, M3, M4, M4.1) against M0 (healthy) samples. Only genera that were 1286 statistically significant (P < 0.01) in at least three DD M stages compared to healthy skin are 1287 shown. 1288

1289

74 1290

1291 Figure 2.4: Percent relative abundance of bacterial species within each M stage. ASVs that 1292 could not be classified at the species level were grouped as unclassified within each genus. 1293 Bacterial species outside the five genera of interest (Bacteroides, Fusobacterium, Mycoplasma, 1294 Porphyromonas, and Treponema) are not shown. 1295

1296

75 1297

1298 Figure 2.5: Absolute abundance of each species across all M stages. Species copy numbers 1299 were standardized by weight of biopsy tissue. Each dot and colour represents a different sample 1300 and a bacterial species, respectively. Lines are present for easy visual tracking of mean species 1301 copy number, but do not represent a linear or continuous relationship between M stages. 1302

1303

1304 Figure 2.6: Correlation network analysis of species absolute abundances in DD lesions. 1305 Significant pairwise Spearman correlations (P < 0.05) were included in the network analysis. The 1306 thickness of each line is proportional to the level of correlation. All correlations included were 1307 positive. 1308

76 1309

1310 Figure 2.7: Percent relative abundance of species in each M stage compared by 1311 quantification method. Species were separated according to A) Treponema spp. and B) non- 1312 Treponema species. Relative abundance was compared between deep amplicon sequencing of 1313 the V3-V4 hypervariable region of the 16S rRNA gene, and species-specific qPCR 1314 quantification. Fusobacterium sp. targeted by qPCR and F. mortiferum identified by deep 1315 amplicon sequencing were both labeled as Fusobacterium sp.. 1316

1317

77 1318 Supplemental Material

1319 Table S2.1: Primers and probes used to sequence and quantify DD-associated bacteria Primer Sequence (5’–3’) Target Reference 341F CCTACGGGNGGCWGCAG Bacteria Klindworth et al. 2013 (26) 806R GGACTACHVGGGTWTCTAAT Bacteria Klindworth et al. 2013 (26) FusoF TCTTTCAATGCTGGGATGCTCT Fusobacterium This study sp. FusoR TGATGGTCCACAATTCTCTCTACA Fusobacterium This study sp. PlevF GGGTGTAGTGCCTACAATAG Porphyromonas This study levii PlevR CCTGAGAAGAGCAGATAGTG Porphyromonas This study levii BpyoF ATTGGCGCTTGTCTCCTACC Bacteroides This study pyogenes BpyoR TATTCATCCATCGTGCGGCC Bacteroides This study pyogenes FnecF AACCTCCGGCAGAAGAAAAATT Fusobacterium Witcomb necrophorum et al. 2014 (37) FnecR CGTGAGGCATACGTAGAGAACTGT Fusobacterium Witcomb necrophorum et al. 2014 (37) TmedF AAAGCGCTACGAATCCTAAG Treponema Beninger medium et al. 2018 (22) TmedR ATCATTACCCGTCCACAAAG Treponema Beninger medium et al. 2018 (22) TphgF CCCGCAGGAAGGTATAATC Treponema Beninger phagedenis et al. 2018 (22) TphgR CACAGCTGTTGTGGTATTAAG Treponema Beninger phagedenis et al. 2018 (22) TpedF ACACCGATTGTACTGAATGA Treponema Beninger pedis et al. 2018 (22)

78 TpedR CCACGAGCTTTCTACAGATT Treponema Beninger pedis et al. 2018 (22) TdentF GGAAACTTAGGAATTCGATATGTAG Treponema Beninger denticola et al. 2018 (22) TdentR CCTTCTTTAGTTTCTTTGTGAGG Treponema Beninger denticola et al. 2018 (22) Probe Sequence (5’–3’) Target Reference

FusoP HEX™ / Fusobacterium This study CTCACTTTTGCACTTATTTCCTGCACTGA / 3' sp. IB®FQ PlevP TxR®-X NHS / Porphyromonas This study CTTGTCACCATCAAAGGCGGCG / 3' IB®FQ levii BpyoP 6-FAM™ / Bacteroides This study CTGACAGACGAAACCCTCAGCAGAATACT / pyogenes IB®FQ FnecP 6-FAM™ / Fusobacterium Witcomb TCGAACATCTCTCGCTTTTTCCCCGA / BHQ- necrophorum et al. 2014 1 (37) TmedP CAL Fluor® Red 610 / Treponema Beninger TGCACCCTTGTTTACTACTGCACAGCC / medium et al. 2018 BHQ-2 (22) TphgP HEX™ / AATCCGCCTACGACTGCGATACCA Treponema Beninger / IB®FQ phagedenis et al. 2018 (22) TpedP 6-FAM™ / Treponema Beninger ACTACACGTGGAGTACCGAATGCT / IB®FQ pedis et al. 2018 (22) TdentP Quasar® 670 / Treponema Beninger AGCATACAGCGATTATAACAAAGCCCTCGA denticola et al. 2018 / BHQ-2 (22) 1320

1321

79 1322 Table S2.2: Number of species isolates collected from each DD lesion stagea Lesion B. Fusobacterium F. P. P. Porphyromonas Prevotella stage pyogenes sp. necrophorum levii somerae sp. sp. M0 0 0 0 0 0 0 0 M1 7 3 1 1 1 4 0 M2 4 13 1 13 1 10 1 M3 1 1 1 1 0 0 0 M4 2 2 1 4 0 2 0

M4.1 3 4 0 3 3 5 0

1323

1324 Table S2.3: DNA spike-in to evaluate performance of multiplex qPCR detection directly 1325 from biopsy DNA B. pyogenes Fusobacterium sp. P. levii Target DNA 5.0 x 104 5.0 x 104 5.0 x 104 copies spiked-in Target DNA 0 ± 0 0 ± 0 0 ± 0 copies in biopsy Mean ± SD 5.1 x 104 ± 6.9 x 5.7 x 104 ± 1.1 x 104 4.7 x 104 ± 5.8 x target DNA 103 103 copies in biopsy after spike-in Efficiency 95.7% 91.3% 92.5%

R-squared 0.998 0.996 0.996

1326 Efficiency and R-squared values are based on qPCR standard curve for each species reaction

1327

1328

80 1329 Table S2.4: Bacterial genomic copies (mean ± SD) per mg of biopsy tissue of each bacterial 1330 species at each farm for healthy (M0) samples Farm ID Tphg Tped Tmed Fn Fs Pl Bp (M0) H 0.432 ± 0.570 ± 0.177 ± 0.014 ± 0 ± 0a 0.112 ± 0.002 ± (n=14) 0.690a 1.06a 0.347a 0.041a 0.267a 0.009a K 5.96 ± 0.327 ± 8.021 ± 0.053 ± 0 ± 0a 0.386 ± 0.012 ± (n=14) 21.2b 0.895b 29.0a 0.182a 0.926a 0.044a M* 0.133 ± 0.066 ± 0 ± 0b 0 ± 0b 0.015 ± 0.022 ± 0 ± 0b (n=12) 0.277c 0.112b 0.052a 0.052b 1331 Tphg, T. phagedenis; Tped, T. pedis; Tmed, T. medium; Fn, F. necrophorum; Fs, Fusobacterium 1332 sp.; Pl, P. levii; Bp, B. pyogenes. Different letters within a column indicate significant difference 1333 (p < 0.05). * Farm had no reported active cases of DD during sampling. Farms H and K had 1334 active cases of DD, with an unknown prevalence, during sampling period. 1335

1336 Table S2.5: Percent of healthy skin biopsies with detectablea amounts of target bacterial 1337 species at each farm Percent (%) of samples with detectable species present Farm ID Tphg Tped Tmed Fn Fs Pl Bp (M0) H 71 43 43 14 0 43 7 (n=14) K 71 21 29 14 0 50 7 (n=14) M* 58 33 0 0 8 17 0 (n=12) Total 68 32 25 10 2 38 5 (n=40) 1338 a Detectable by species-specific qPCR. * Farm had no reported active cases of DD during 1339 sampling. Farms H and K had active cases of DD during sampling period, but with an unknown 1340 prevalence. 1341

81 1342 1343 Figure S2.1: Analysis of Illumina MiSeq read counts in sequencing controls. A) Sequences 1344 for each group were counted after raw reads were processed with DADA2, and are displayed in 1345 mean read count ± SD. Labels represent, in order, healthy skin (M0), DD lesion samples (M1, 1346 M2, M3, M4, M4.1), negative control (NC) consisting of RNase/DNase-free water, and DNA 1347 extraction control (EC). B) Microbial composition of negative controls based on phylum level 1348 taxonomy. 1349

82 1350

1351 Figure S2.2: Pairwise associations between log copy numbers of species in DD lesions. 1352 Species copy numbers per mg biopsy tissue were transformed with natural logarithm after 1353 addition of a pseudocount of 1. Each dot represents bacterial copy number in separate samples, 1354 split into active (M1, M2, M4.1) and inactive (M3, M4) DD lesions. Spearman rank correlation 1355 coefficients measured degree of association between each species pair in all DD lesions. 1356

1357

83 1358 Chapter 3: Meta-analysis of bovine digital dermatitis microbiota reveals

1359 distinct microbial community structures associated with lesions

1360 Abstract

1361 Bovine digital dermatitis (DD) is a significant cause of infectious lameness and economic losses

1362 in cattle production across the world. There is a lack of a consensus across different 16S

1363 metagenomic studies on DD-associated bacteria that may be potential pathogens of the disease.

1364 The goal of this meta-analysis was to identify a consistent group of DD-associated bacteria in

1365 individual DD lesions across studies, regardless of experimental design choices including sample

1366 collection and preparation, hypervariable region sequenced, and sequencing platform. A total of

1367 6 studies were included in this meta-analysis. Raw sequences and metadata were identified on

1368 the NCBI sequence read archive and European nucleotide archive. Bacterial community

1369 structures were investigated between normal skin and DD skin samples. Random forest models

1370 were generated to classify DD status based on microbial composition, and to identify taxa that

1371 best differentiate DD status. Among all samples, members of Treponema, Mycoplasma,

1372 Porphyromonas, and Fusobacterium were consistently identified in the majority of DD lesions,

1373 and were the best genera at differentiating DD lesions from normal skin. Individual study and

1374 16S hypervariable region sequenced had significant influence on final DD lesion microbial

1375 composition (P < 0.05). These findings indicate that members of Treponema, Mycoplasma,

1376 Porphyromonas, and/or Fusobacterium may have significant roles in DD pathogenesis, and

1377 should be studied further in respect to developing more effective treatment and mitigation

1378 strategies.

1379

84 1380 Importance

1381 This study addresses a major gap in the understanding of digital dermatitis (DD) microbiology.

1382 Lesion-associated microbiota have been varied in their composition and degree of association

1383 across independent studies. We aimed to identify a core bacterial group within DD lesions across

1384 multiple studies and individual lesions. Regardless of study design choices and their influence on

1385 microbial composition, this meta-analysis identified a consistent bacterial group that

1386 differentiates DD lesions from normal skin. This study provides compelling evidence for the

1387 universal involvement of Treponema, Mycoplasma, Porphyromonas, and Fusobacterium in DD

1388 pathogenesis, and establishes these bacterial groups as targets in future studies that attempt to

1389 understand the etiopathogenesis of DD.

1390

1391 Introduction

1392 Bovine digital dermatitis (DD) is an infectious skin lesion that was first reported in Italy in 1974

1393 (1), and is a significant cause of lameness in cattle across the world (2). DD is a multifaceted

1394 disease, having both painful ulcerative lesions and chronic hyperkeratotic stages across the

1395 different morphological presentations of the lesions (3, 4). Treatment and production costs

1396 associated with DD lead to significant losses for both beef and dairy producers mainly due to

1397 either decreased average daily gain and decreased milk production (5–8). Prevalence of DD at

1398 the herd level is reported to be as high as 93% (9), but prevalence estimates vary based on a

1399 multitude of factors including parity, breed, housing type and hoof trimming frequency (9–12).

1400 Current treatment and prevention strategies rely on topical applications and footbaths of broad

1401 spectrum antimicrobial agents, most commonly tetracycline or copper sulfate, and are largely

85 1402 ineffective at returning lesions to normal skin and preventing recurrence of active lesions (13,

1403 14).

1404

1405 Early attempts to determine etiological agents of DD pointed to Treponema as the main pathogen

1406 of these lesions. More recent evidence shifts the focus to studying the roles of both Treponema

1407 and additional anaerobic bacteria in lesion formation and progression, leading to a polymicrobial

1408 hypothesis of DD causation (15, 16). The quantities of Treponema are dramatically higher in

1409 most DD lesions compared to healthy skin; however, there exists a wide variety of Treponema

1410 species and phylotypes across individual animals and between studies (17–19). Most commonly,

1411 Treponema phagedenis, T. medium, T. refringens, T. denticola, and T. pedis are identified in DD

1412 lesions, but presence and population dynamics of each species varies across studies (19–21).

1413

1414 Additional bacteria outside the genus Treponema likely play some role in DD pathogenesis

1415 based on their presence and abundance within DD lesions; however, the conclusions on which of

1416 these bacteria potentially are involved can vary dramatically across studies of DD microbiota. In

1417 addition to Treponema, many recent metagenomic studies associate Mycoplasma abundance with

1418 DD lesions (19, 22, 23). Inconsistencies most notably arise between studies when implicating

1419 other bacterial genera in DD pathogenesis. Porphyromonas (20, 22), Fusobacterium (22, 24),

1420 and Dichelobacter (19, 24), among a few other genera are inconsistently associated with DD

1421 lesion microbiota and pathogenesis. It is currently unknown if these major differences between

1422 metagenomic studies are due to differences in sequencing, bioinformatic processing, sample size

1423 or if DD lesion microbiota differs across beef and dairy cattle breeds.

1424

86 1425 Comparing conclusions between metagenomic studies can lead to erroneous interpretations, as

1426 variation in study objectives and analytical processing can influence findings and result in

1427 contradictory comparisons between study outcomes. Many metagenomic studies process

1428 individual sequencing reads into operational taxonomic units, which are not directly comparable

1429 across experiments and can be a source of variation when drawing comparisons between studies

1430 (25). Primer bias in deep amplicon sequencing can also have a significant impact on observed

1431 bacterial populations, as certain primers used in the amplification process before sequencing

1432 might not amplify all bacteria at equal rates, but the extent of this impact is not determined yet in

1433 DD-specific microbial communities. In addition, different study designs, experimental

1434 conditions, and statistical analyses can significantly alter the results and conclusions on

1435 microbial composition (26). Therefore, a meta-analysis is required in order to draw controlled

1436 and quantitative conclusions across these metagenomic DD studies.

1437

1438 The differences in findings between DD metagenomic studies leads to a lack of consensus on

1439 which bacteria may be potential pathogens, which is a limiting factor for future DD pathogenesis

1440 research. The goal of this meta-analysis was to use a consistent analytical approach on pooled

1441 sequences from individual DD metagenomic studies to identify a DD-associated microbiota that

1442 consistently and accurately differentiates DD lesions from normal skin. Identifying a consistent

1443 DD-associated microbiota, regardless of individual animal variation, 16S hypervariable region

1444 sequenced, cattle breed, and other study design choices would provide strong evidence for a

1445 potential core DD microbiota and initiate and accelerate future targeted research efforts toward

1446 determining etiopathogenesis mechanisms of DD.

1447

87 1448 Methods

1449 Data collection and selection criteria

1450 Searches for publicly available data were performed on the NCBI Sequence Read Archive (SRA)

1451 database and the European Nucleotide Archive using the search term “digital dermatitis”, and

1452 limiting search results by “BioProject”. BioProject accession numbers (study identifiers)

1453 containing high throughput sequencing reads and associated metadata were collected. For

1454 inclusion in the meta-analysis, all studies had to have publicly available sequencing reads before

1455 September 1, 2020. Studies were required to be a bovine DD metagenomic study, and have

1456 conducted deep amplicon sequencing of a 16S rRNA hypervariable region using universal

1457 bacterial primers on samples originating from bovine skin. Raw data (DNA sequences and

1458 associated quality scores) and sufficient metadata (containing information differentiating

1459 samples by DD diagnosis) had to be publicly available for inclusion in the meta-analysis.

1460 Additional information on sample collection was obtained for each study. All samples were

1461 collected with biopsy punches from feet cleaned with water or chlorhexidine of either live

1462 animals or at a slaughterhouse. Depth and exact location of skin lesion samples were not

1463 universally available across all studies. Other relevant details concerning experimental

1464 processing of samples before sequencing are described in Table 3.1. Fastq files were all

1465 downloaded from the NCBI SRA database (www.ncbi.nlm.nih.gov/sra).

1466

1467 Sequence processing

1468 Raw fastq files that contained reads from both the V1V2 and Treponema-specific V3V4 16S

1469 rRNA regions were processed with Bowtie2 v.2.3.5.1 (27) in order to separate reads by

1470 hypervariable region. Afterwards, sequences were processed using the DADA2 R package

88 1471 v.1.14.1 (28). Quality filtering and base call error models were conducted on each study

1472 independently, in order to account for differences in error rates between sequencing runs. Reads

1473 were truncated after the average Phred score dropped below 30, and any reads with ambiguous

1474 nucleotides were removed. The recommended DADA2 settings for each sequencing platform

1475 (Illumina and LS454) were used to infer amplicon sequence variants (ASVs). Taxonomy was

1476 assigned using the DADA2 naïve Bayesian classifier against the SILVA v.138 database (29).

1477 Prior to analysis, all ASVs classified as Eukaryota, Mitochondria, or Chloroplast were removed.

1478

1479 Data analysis

1480 Due to the lack of consistent information across studies on individual lesion stages, all samples

1481 included in analysis were categorized as “DD positive” or “DD negative”. DD positive samples

1482 were defined as any sample from DD lesion skin at any stage of the disease. DD negative

1483 samples were classified as any sample from skin with no visible DD lesion present. Any samples

1484 that were obtained from treated lesions were removed from downstream analysis. For diversity

1485 analyses, samples with less than 1500 reads were removed, and the remaining samples were

1486 rarefied down to the minimum sequencing depth. Observed ASV count, Chao1 richness estimate,

1487 Fisher’s alpha and Shannon’s diversity index were used to evaluate diversity (richness and

1488 evenness) within DD positive and negative samples. Analysis of variance (ANOVA) followed by

1489 post hoc Tukey’s Honest Significant Difference (TukeyHSD) test determined any significant

1490 differences of diversity between DD statuses. For beta diversity analyses, ASVs were grouped at

1491 phylum, family, and genus taxonomic levels, and Bray-Curtis dissimilarities between samples

1492 were explored using Principal Coordinates (PCoA) analysis. Permutational ANOVA

1493 (PERMANOVA) with 999 permutations, followed by a test for multivariate homogeneity of

89 1494 group dispersions (BETADISP) was performed to identify differences in microbial composition

1495 between DD negative and DD positive samples.

1496

1497 The influence of different study variables and experimental designs on microbial composition

1498 was measured exclusively on DD positive samples by repeating the diversity analysis methods

1499 described above. Alpha diversity and relative abundance plots evaluated the effect of different

1500 16S rRNA hypervariable regions sequenced, and beta diversity measured the variation in

1501 microbial composition by individual study, 16S rRNA hypervariable region, and sequencing

1502 platform. All diversity metrics and statistical analyses were conducted using the vegan package

1503 (v. 2.5.6) in R (30).

1504

1505 Random forest modeling was used to predict DD status based on microbial composition, and

1506 identify the microbiota that best differentiates DD positive from DD negative samples. Prior to

1507 modeling, ASVs were grouped at phylum, family, and genus ranks and rare taxa (present in less

1508 than 1% of samples) were removed. Sample ASV counts were normalized using a centered log

1509 ratio transformation, with a pseudocount of 1 applied to all ASVs. Randomly generated training

1510 sample sets were made of 70% of each DD negative and positive samples, and testing sample

1511 sets contained the remaining 30% of samples in each group. Random forest models containing

1512 1000 trees were trained using 100 repeats of 10-fold cross validation, and the number of

1513 variables sampled at each node was optimized as part of the caret R package v.6.0.86 (31).

1514 Model performance was evaluated based on accuracy of classification and kappa score, which

1515 was determined by constructing a confusion matrix based on the testing sample set. Relative

1516 variable importance along with relative abundance of each taxa was used to identify the

90 1517 microbiota that differentiate DD positive from DD negative samples. A P value of less than 0.05

1518 was considered statistically significant for all comparisons. All analysis and figures were

1519 completed in R version 3.5.3.

1520

1521 Results

1522 Study selection and characteristics

1523 A total of 24 BioProject IDs were identified upon the initial search. Next, 12 studies were

1524 removed because they did not conduct deep amplicon sequencing of a 16S rRNA hypervariable

1525 region. Four studies were excluded because they were not from bovine skin/lesion samples, or

1526 did not have any raw fastq files associated with the BioProject ID. One study was removed

1527 because it did not use universal bacterial primers for sequencing. Finally, one study was removed

1528 due to a lack of metadata to differentiate samples by DD status. Included in the meta-analysis

1529 were 6 studies (Table 3.1), which all provided sufficient metadata to differentiate the DD status

1530 (DD negative samples, n = 37; DD positive samples, n = 190). The total number of raw

1531 sequences available for meta-analysis was 16,659,674, and after quality filtering, denoising, and

1532 merging, 8,827,942 reads were taxonomically classified. These samples were further categorized

1533 by sequencing platform, cow breed (if known), 16S rRNA hypervariable region, DNA extraction

1534 kit, country of sample origin, and if the sample was known to come from an animal treated for

1535 DD. Samples that came from treated lesions (n = 16) were removed from the meta-analysis in

1536 order to focus solely on the differences in microbial populations between DD positive and DD

1537 negative skin.

1538

91 1539 Characterization of microbiota in DD positive and DD negative skin

1540 For all measures of alpha diversity tested, DD positive skin microbiota had significantly lower (P

1541 < 0.05) diversity, richness, and evenness when compared to DD negative skin microbiota (Table

1542 3.2). Bray-Curtis distances between samples showed a significant difference (P < 0.05) in

1543 microbial composition between DD negative and DD positive skin (Fig. 3.1, Table 3.3),

1544 regardless of the taxonomic rank tested. However, BETADISP was significant (P < 0.05) at the

1545 phylum level (Table 3.3). For these Bray-Curtis distances, the percent of variation explained by

1546 DD status increased as taxonomic rank lowered from phylum to genus (Table 3.3). When ASVs

1547 were grouped at the genus level, approximately 20.5% of the variation in microbial composition

1548 could be explained by DD status (Table 3.3). DD positive samples cluster in two major groups

1549 based on Bray-Curtis dissimilarities, and are visually separated by high and low Spirochaetaceae

1550 relative abundance (Fig. S3.1).

1551

1552 Distinct bacterial populations were found when DD negative skin microbiota was compared to

1553 DD positive skin microbiota. At the phylum level, Spirochaetota relative abundance was

1554 noticeably higher in DD positive skin, which was present at roughly 3% in DD negative skin to

1555 35% relative abundance in DD positive skin (Fig. 3.2A). There were also higher relative

1556 abundances of Fusobacteriota, Bacteroidota, and Campilobacterota in DD positive skin (Fig.

1557 3.2A). DD negative skin was dominated primarily by members of Firmicutes, Actinobacteriota,

1558 and Proteobacteria (Fig. 3.2A). Additional trends emerged when increasing the taxonomic

1559 resolution to family and genus levels, as shown in Fig. 3.2B and Fig. 3.2C. The families

1560 Spirochaetaceae and Fusobacteriaceae were higher in relative abundance in DD positive skin,

1561 similarly to their phylum level equivalents. In addition, families Mycoplasmataceae and

92 1562 Porphyromonadaceae were higher in relative abundance in DD positive skin compared to DD

1563 negative skin. At genus level grouping, Treponema dominated the DD positive skin microbiota;

1564 however, there were 4 other groups that also showed considerably higher relative abundance in

1565 DD positive skin compared to DD negative skin. Fusobacterium and Peptoanaerobacter were

1566 both relatively absent in DD negative skin, but were present at roughly 3% and 5% relative

1567 abundance in DD positive skin, respectively. Porphyromonas and Mycoplasma were present in

1568 DD negative skin at roughly 3% and 1%, respectively, but both had higher relative abundance in

1569 DD positive skin at greater than 6% relative abundance. In terms of bacterial prevalence,

1570 Treponema, Porphyromonas, Mycoplasma, Fusobacterium, and Peptoanaerobacter were present

1571 in the majority of DD positive skin samples, and were more prevalent in DD positive samples

1572 relative to DD negative skin samples (Table S3.1). Mycoplasma, Treponema, and

1573 Porphyromonas were all present in at least 94% of DD positive samples; whereas,

1574 Peptoanaerobacter and Fusobacterium were present in 81% and 63% of DD positive samples,

1575 respectively (Table S3.1). However, Treponema were also present in approximately 86% of DD

1576 negative samples, along with Porphyromonas and Peptoanaerobacter also present in more than

1577 60% of DD negative samples (Table S3.1). Amnipila, Ezakiella, and Campylobacter, are genera

1578 that had low relative abundances at less than 1% in DD positive skin (Fig. 3.2C), but were all

1579 present in greater than 75% of DD positive samples, while absent in the majority of DD negative

1580 skin samples (Table S3.1).

1581

1582 Random forests classifiers built at each taxonomic rank all had at least 90% accuracy in

1583 classifying DD status based on microbial composition (Table 3.4). When bacterial taxonomy was

1584 grouped at the genus level, models had the highest accuracy at 97.06% and a kappa score of 0.82

93 1585 (Table 3.4). Of the top 30 important genera in the random forest model, 22 were associated with

1586 DD negative skin (Fig. 3.3A, B, and C). Mycoplasma was the third ranked genus in the model,

1587 and was the first ranked DD positive-associated bacteria in the model. Seven additional genera

1588 were designated as important markers of DD positive skin microbiota, of these, only Treponema,

1589 Fusobacterium, and Porphyromonas had relative abundances of greater than 3% in DD lesions,

1590 whereas the remaining 4 genera had a relative abundance of roughly 0.1% or less in DD positive

1591 skin.

1592

1593 Study variables influencing the microbiota of DD positive skin

1594 Permutational analysis of variance on Bray-Curtis distances identified BioProject (individual

1595 studies) as the largest source of variation (20%) in microbial composition between DD positive

1596 samples, regardless of taxonomic levels tested (Fig. 3.4; Table 3.5). The average relative

1597 abundances of non-rare taxa were relatively similar across studies, although one study detected

1598 distinctly lower relative abundance of Spirochaetaceae in DD positive samples (Fig. S3.2).

1599 There were 3 different 16S rRNA hypervariable regions sequenced, and was another significant

1600 variable influencing the microbiota, accounting for roughly 11% of the differences between

1601 samples (Table 3.5). Sequencing platform, either Illumina or Life Sciences 454 pyrosequencing

1602 explained the lowest amount of variation of all categories tested (Table 3.5). All study variables,

1603 except for BioProject at phylum level grouping, displayed significant heterogenous dispersion

1604 across groups (P < 0.05) as identified in Table 3.6.

1605

1606 There were significant differences (P < 0.05) in richness, evenness, and diversity within DD

1607 positive samples from the V3V4 region compared to both V1V2 and V1V3 for all diversity

94 1608 measures evaluated (Table 3.7). No significant differences in any alpha diversity measurement

1609 existed between samples from V1V2 and V1V3 hypervariable regions (Table 3.7). In DD lesions

1610 from V1V3 amplification, there was a notable absence of Fusobacterium compared to the other

1611 two hypervariable regions, which show a Fusobacterium relative abundance of 2-4% in DD

1612 positive skin (Fig. 3.5A and B). Although the proportions varied across sequencing regions, DD

1613 positive skin appeared to be dominated by Treponema, regardless of 16S rRNA hypervariable

1614 region (Fig. 3.5B). Other bacterial genera with a relative abundance higher than 2% that were

1615 common across all sequencing regions in DD positive skin were Mycoplasma,

1616 Peptoanaerobacter, and Porphyromonas (Fig. 3.5B). Although no DD negative skin samples

1617 were acquired from the V1V3 region, microbiota of DD negative skin from V1V2 and V3V4

1618 sequencing both contained a community dominated by members of the phyla Firmicutes,

1619 Actinobacteriota, Bacteroidota, and Proteobacteria (Fig. 3.5A).

1620

1621 Discussion

1622 In this meta-analysis, Treponema, Mycoplasma, Porphyromonas, and Fusobacterium were the

1623 genera that best differentiated DD positive skin from DD negative skin, indicating their possible

1624 role in DD pathogenesis. These genera were present in the majority of DD affected animals, and

1625 were relatively more abundant in DD positive skin microbiota compared to DD negative skin.

1626 Whereas DD-associated microbiota and conclusions on which bacteria may have a role in DD

1627 etiology is varied across literature, the findings of this study identified a relatively consistent

1628 bacterial consortium in DD affected skin, regardless of study variables and experimental design

1629 choices. We also identify a consistent DD-associated microbiota across dairy and beef cattle

1630 breeds, providing justification for extrapolating dairy DD knowledge to beef cattle DD.

95 1631

1632 Although it was not possible to identify any taxa in 100% of animals with DD, there are strong

1633 associations of Treponema, Mycoplasma, Fusobacterium, and Porphyromonas with DD positive

1634 skin. This core group of genera we identified agrees best with the outcome of two previous

1635 metagenomic studies (22; B. Caddey, K. Orsel, S. Naushad, H. Derakhshani, J. De Buck,

1636 submitted for publication), in which they identify the same microbiota that best differentiates DD

1637 lesions from healthy skin. We identified other bacterial groups that displayed strong associations

1638 with DD skin, mainly members of Firmicutes, but these were left out of the primary core group

1639 because of their relatively low rank in the random forest model and low relative abundance in

1640 DD positive skin. However, low abundant taxa can have the capability to modulate community

1641 structures as keystone taxa (32), and ignoring taxa because of their low abundance may result in

1642 an incomplete picture of DD pathogenesis. Although it is unknown if low abundance taxa play a

1643 role in DD microbial community structure formation, we know Dichelobacter nodosus drives

1644 pathogenesis of ovine foot rot and only has a relative abundance of less than 2% in diseased

1645 tissue (33), so it is possible that low abundant taxa may contribute to DD pathogenesis.

1646

1647 Treponema has been the most common genus implied as an etiologic agent of DD. Consistent

1648 with this study, virtually every other metagenomic study of DD observes a higher relative

1649 abundance of Treponema in most DD lesions compared to normal skin (19, 20, 22, 23, 34).

1650 Conversely, it is not uncommon to see some healthy skin samples to also have high proportions

1651 of at least one species of Treponema (35; B. Caddey, K. Orsel, S. Naushad, H. Derakhshani, J.

1652 De Buck, submitted for publication). Currently, there is little knowledge of pathogenesis

1653 mechanisms of individual Treponema spp., but T. pedis, T. medium, and T. phagedenis are some

96 1654 of the most commonly identified across DD literature (18, 35), and may have pathogenic

1655 potential as identified in murine infection models (36, 37).

1656

1657 Mycoplasma are more recently implicated in DD etiology (19, 22), and they have been identified

1658 in this meta-analysis as the best DD-associated genus at differentiating disease status. There is a

1659 lack of understanding of which Mycoplasma spp. are important to DD development, due to a

1660 lack of culture isolates or species-specific identification methods, thus we only have short 16S

1661 rRNA amplicon-based identifications which do not give a definitive species identification. The

1662 major focus in literature tends to be on M. fermentans in DD pathogenesis, whose presence has

1663 been determined by PCR (22) and by bacterial culture from DD (38). No DD studies have

1664 identified an abundance or presence of M. bovis, a major pathogen of bovine respiratory disease

1665 (39). The lack of type cultures of Mycoplasma spp. isolated from DD lesions impairs our ability

1666 to further understand their roles in DD pathogenesis.

1667

1668 There is previous literature showing Fusobacterium as differentiating DD lesions from healthy

1669 skin (22, 24, 40), but is suggested to have a larger role in chronic lesions due to it being more

1670 prevalent in those lesions (23). We did not identify Fusobacterium in any of the samples from

1671 V1V3 amplification. This could suggest that Fusobacterium is not important for lesion

1672 formation, and perhaps impacts severity of lesions as a secondary invader. However, there could

1673 be potential confounding effects that may have caused Fusobacterium to be absent in V1V3

1674 reads, such as that only active lesion stages were sequenced, or there could be potential primer

1675 mismatches with DD-specific strains resulting in poor amplification efficiency. Two different

1676 Fusobacterium spp. have been identified and are both present in the majority of DD lesions of

97 1677 beef cattle (B. Caddey, K. Orsel, S. Naushad, H. Derakhshani, J. De Buck, submitted for

1678 publication); therefore, interest should remain on Fusobacterium as a potential pathogenic agent

1679 of DD. Porphyromonas is also a possible contributor to DD pathogenesis that is detectable in all

1680 active lesions in one study (B. Caddey, K. Orsel, S. Naushad, H. Derakhshani, J. De Buck,

1681 submitted for publication), but along with Fusobacterium, its superficial location in the dermis

1682 has led some studies to limit their conclusions on its potential involvement (24). However, in

1683 periodontal disease, which has similar higher level microbial community structure to DD,

1684 Porphyromonas plays a significant role in influencing the metabolic activity of Treponema

1685 species (41). Additionally, Porphyromonas and Fusobacterium species generate mixed-species

1686 biofilms as a mechanism to impair the bovine innate immune system (42). While it is not evident

1687 that the same microbial strategies are involved in DD pathogenesis, it provides significant

1688 motivation to identify interactions between DD-associated species to determine a framework of

1689 the synergistic mechanisms used as pathogenesis mechanisms.

1690

1691 From the key group of DD-associated bacteria we mentioned in this study, multiple species

1692 within Treponema, Porphyromonas, and Mycoplasma have been identified within DD

1693 microbiota (20, 22, 35). Due to the unreliability of species level classifications of the 16S rRNA

1694 hypervariable regions sequenced in this meta-analysis, comparisons at the species level were not

1695 performed (43). Methods to reliably quantify species-level dynamics across DD lesions is

1696 essential to validating these metagenomic results, and account for differences in 16S rRNA copy

1697 number and primer specificity between individual taxa.

1698

98 1699 One of the more common species implicated in DD lesion formation, D. nodosus, the etiologic

1700 agent of ovine foot rot, were present in less than 30% of DD samples in this meta-analysis (19,

1701 24, 44). Insufficient amplification of D. nodosus isolates through one pair of universal 16S rRNA

1702 primers has been shown (45), and could be a potential reason for D. nodosus absence in the

1703 majority of samples in this study. Dichelobacter nodosus is primarily suggested to have a

1704 potential role in early establishment of DD lesions, and then appears relatively rare in later lesion

1705 stages (19, 44), which could explain our inability to identify D. nodosus as clinically relevant.

1706 Further investigations of D. nodosus populations is warranted for early DD lesions, in order to

1707 characterize its potential role in epithelial infiltration and facilitation of skin colonization. Some

1708 studies not included in this meta-analysis have identified bacteria that show strong associations

1709 to DD lesion microbiota, such as Candidatus Aemobophilus asiaticus, which was present in large

1710 abundance in two studies (20, 21), but not detected in this meta-analysis. Similarly,

1711 Guggenheimella bovis has been implicated in a potential DD pathogenesis role (46) but also was

1712 not identifiable in this meta-analysis.

1713

1714 Machine learning classifiers can help shed light on the complexity of polymicrobial infections

1715 that have major individual variation. These models appraise individual bacterial population

1716 dynamics with respect to overall community structure instead of the traditional statistical

1717 approach which primarily study individual taxa in isolation (47). Random forest models provide

1718 a relative variable importance ranking, and in this meta-analysis, Mycoplasma was the best DD-

1719 associated bacteria at differentiating DD status, instead of Treponema, which is historically the

1720 most often group implicated in DD etiology. The relative variable importance is not a function of

1721 biological relevance of each organism, but rather scores variables based on their reduction of

99 1722 randomness in model prediction, and therefore we cannot conclude a relative ranking among

1723 DD-associated bacteria in their importance to DD pathogenesis.

1724

1725 Each of the studies included in this meta-analysis had significant differences in the microbial

1726 compositions of their samples based on PERMANOVA analysis. However, this could be due to

1727 the fact there was significant heterogenous dispersion between groups, so we cannot conclude

1728 that there are significant differences in the microbial composition between studies and study

1729 variables. The skin microbial composition and diversity of DD lesions is quite variable across

1730 individuals (24), which is one of the reasons why it has been difficult to pin down a causative

1731 agent. This individual animal variation may be causing the heterogenous dispersions among

1732 groups tested for study bias.

1733

1734 In our meta-analysis, there were 2 studies that contained two-thirds of the total samples. These

1735 large studies may have skewed our results, but that is unlikely, given that our principal

1736 coordinates analysis appeared to show more diversity within studies than between studies, and

1737 it’s likely that the massive variation in Treponema abundance between samples was the source of

1738 this variation within studies. Observing the bacterial population differences between the 16S

1739 rRNA hypervariable regions identified minimal differences in the core bacterial group identified

1740 as potential DD pathogens, furthering evidence suggesting that these two large studies did not

1741 skew our results. An additional concern in study bias includes the significantly greater diversity

1742 in samples from V3V4 16S rRNA hypervariable regions compared to both V1V2 and V1V3.

1743 There are at least two potential reasons for this: 1) the samples that used V3V4 sequencing were

1744 also the only samples to include skin samples from beef cattle breeds rather than just dairy

100 1745 breeds; 2) V3V4 primers may amplify a wider range of bovine DD skin microbiota. DD in beef

1746 cattle is a recent field of study, as DD is emerging in those populations (48), and since dairy and

1747 beef animals have separate housing environments and genetics, there is potential for their skin

1748 microbiota to have significant differences.

1749

1750 Conclusion

1751 This meta-analysis has identified, through a consistent analytical approach of skin microbiota

1752 across multiple studies, that Treponema, Mycoplasma, Porphyromonas, and Fusobacterium were

1753 key genera that different DD lesions from normal skin. These genera are consistently associated

1754 with the majority of skin samples of DD lesions. Based on abundance data, Mycoplasma best

1755 differentiates DD lesions from normal skin and should be a priority focus in future research on

1756 DD pathogenesis. Treponema are the most abundant bacteria in DD, but have also been

1757 identified in low amounts in most DD negative skin samples. Further analysis on the individual

1758 species of Treponema, as well as species of the other potential DD pathogens, is warranted to

1759 further understand their roles in lesion formation and development. This study is an

1760 accumulation of current understanding of DD microbiology, and provides strong evidence to

1761 standardize future research to focus primarily on the potential DD pathogens identified in this

1762 study.

1763

1764 Acknowledgements

1765 Funding was supplied by Alberta Agriculture and Forestry, the Canadian Dairy Commission

1766 under the Workplace Development Initiative, and the Simpson Ranch Grant at the University of

1767 Calgary Faculty of Veterinary Medicine.

101 1768

1769 References

1770 1. Cheli R, Mortellaro C. 1974. La dermatite digitale del bovino, p. 208–213. In Proceedings

1771 of the 8th International Conference on Diseases of Cattle.

1772 2. Read DH, Walker RL. 1998. Papillomatous digital dermatitis (footwarts) in California

1773 dairy cattle: Clinical and gross pathologic findings. J Vet Diagnostic Investig 10:67–76.

1774 3. Döpfer D, Koopmans A, Meijer FA, Szakáll I, Schukken YH, Klee W, Bosma RB,

1775 Cornelisse JL, Van Asten AJAM, Ter Huurne AAHM. 1997. Histological and

1776 bacteriological evaluation of digital dermatitis in cattle, with special reference to

1777 spirochaetes and Campylobacter faecalis. Vet Rec 140:620–623.

1778 4. Berry SL, Read DH, Famula TR, Mongini A, Döpfer D. 2012. Long-term observations on

1779 the dynamics of bovine digital dermatitis lesions on a California dairy after topical

1780 treatment with lincomycin HCl. Vet J 193:654–658.

1781 5. Dolecheck KA, Overton MW, Mark TB, Bewley JM. 2019. Use of a stochastic simulation

1782 model to estimate the cost per case of digital dermatitis, sole ulcer, and white line disease

1783 by parity group and incidence timing. J Dairy Sci 102:715–730.

1784 6. Gomez A, Cook NB, Socha MT, Döpfer D. 2015. First-lactation performance in cows

1785 affected by digital dermatitis during the rearing period. J Dairy Sci 98:4487–4498.

1786 7. Davis-Unger J, Pajor E, Schwartzkopf-Genswein K, Marti S, Dorin C, Spackman E, Orsel

1787 K. 2017. Economic impacts of lameness in feedlot cattle. Transl Anim Sci 1:467–479.

1788 8. Kulow M, Merkatoris P, Anklam KS, Rieman J, Larson C, Branine M, Döpfer D. 2017.

1789 Evaluation of the prevalence of digital dermatitis and the effects on performance in beef

1790 feedlot cattle under organic trace mineral supplementation. J Anim Sci 95:3435–3444.

102 1791 9. Solano L, Barkema HW, Mason S, Pajor EA, LeBlanc SJ, Orsel K. 2016. Prevalence and

1792 distribution of foot lesions in dairy cattle in Alberta, Canada. J Dairy Sci 99:6828–6841.

1793 10. Yang DA, Johnson WO, Müller KR, Gates MC, Laven RA. 2019. Estimating the herd and

1794 cow level prevalence of bovine digital dermatitis on New Zealand dairy farms: A

1795 Bayesian superpopulation approach. Prev Vet Med 165:76–84.

1796 11. Holzhauer M, Hardenberg C, Bartels CJM, Frankena K. 2006. Herd- and cow-level

1797 prevalence of digital dermatitis in the Netherlands and associated risk factors. J Dairy Sci

1798 89:580–588.

1799 12. Bran JA, Daros RR, von Keyserlingk MAG, LeBlanc SJ, Hötzel MJ. 2018. Cow- and

1800 herd-level factors associated with lameness in small-scale grazing dairy herds in Brazil.

1801 Prev Vet Med 151:79–86.

1802 13. Jacobs C, Orsel K, Mason S, Barkema HW. 2018. Comparison of effects of routine topical

1803 treatments in the milking parlor on digital dermatitis lesions. J Dairy Sci 101:5255–5266.

1804 14. Krull AC, Shearer JK, Gorden PJ, Scott HM, Plummer PJ. 2016. Digital dermatitis:

1805 Natural lesion progression and regression in Holstein dairy cattle over 3 years. J Dairy Sci

1806 99:3718–3731.

1807 15. Yano T, Moe KK, Yamazaki K, Ooka T, Hayashi T, Misawa N. 2010. Identification of

1808 candidate pathogens of papillomatous digital dermatitis in dairy cattle from quantitative

1809 16S rRNA clonal analysis. Vet Microbiol 143:352–362.

1810 16. Santos TMA, Pereira R V., Caixeta LS, Guard CL, Bicalho RC. 2012. Microbial diversity

1811 in bovine Papillomatous digital dermatitis in Holstein dairy cows from upstate New York.

1812 FEMS Microbiol Ecol 79:518–529.

1813 17. Klitgaard K, Boye M, Capion N, Jensen TK. 2008. Evidence of multiple Treponema

103 1814 phylotypes involved in bovine digital dermatitis as shown by 16S rRNA gene analysis and

1815 fluorescence in situ hybridization. J Clin Microbiol 46:3012–3020.

1816 18. Klitgaard K, Bretó AF, Boye M, Jensen TK. 2013. Targeting the treponemal microbiome

1817 of digital dermatitis infections by high-resolution phylogenetic analyses and comparison

1818 with fluorescent in situ hybridization. J Clin Microbiol 51:2212–2219.

1819 19. Krull AC, Shearer JK, Gorden PJ, Cooper VL, Phillips GJ, Plummera PJ. 2014. Deep

1820 sequencing analysis reveals temporal microbiota changes associated with development of

1821 bovine digital dermatitis. Infect Immun 82:3359–3373.

1822 20. Zinicola M, Lima F, Lima S, Machado V, Gomez M, Döpfer D, Guard C, Bicalho R.

1823 2015. Altered microbiomes in bovine digital dermatitis lesions, and the gut as a pathogen

1824 reservoir. PLoS One 10:e0120504.

1825 21. Zinicola M, Higgins H, Lima S, Machado V, Guard C, Bicalho R. 2015. Shotgun

1826 metagenomic sequencing reveals functional genes and microbiome associated with bovine

1827 digital dermatitis. PLoS One 10:e0133674.

1828 22. Nielsen MW, Strube ML, Isbrand A, Al-Medrasi WDHM, Boye M, Jensen TK, Klitgaard

1829 K. 2016. Potential bacterial core species associated with digital dermatitis in cattle herds

1830 identified by molecular profiling of interdigital skin samples. Vet Microbiol 186:139–149.

1831 23. Hesseling J, Legione AR, Stevenson MA, McCowan CI, Pyman MF, Finochio C, Nguyen

1832 D, Roic CL, Thiris OL, Zhang AJ, van Schaik G, Coombe JE. 2019. Bovine digital

1833 dermatitis in Victoria, Australia. Aust Vet J 97:404–413.

1834 24. Moreira TF, Facury Filho EJ, Carvalho AU, Strube ML, Nielsen MW, Klitgaard K, Jensen

1835 TK. 2018. Pathology and bacteria related to digital dermatitis in dairy cattle in all year

1836 round grazing system in Brazil. PLoS One 13:e0193870.

104 1837 25. Callahan BJ, McMurdie PJ, Holmes SP. 2017. Exact sequence variants should replace

1838 operational taxonomic units in marker-gene data analysis. ISME J 11:2639–2643.

1839 26. Goodrich JK, Di Rienzi SC, Poole AC, Koren O, Walters WA, Caporaso JG, Knight R,

1840 Ley RE. 2014. Conducting a microbiome study. Cell 158:250–262.

1841 27. Langmead B, Salzberg SL. 2012. Fast gapped-read alignment with Bowtie 2. Nat Methods

1842 9:357–359.

1843 28. Callahan BJ, Mcmurdie PJ, Rosen MJ, Han AW, Johnson AJA, Holmes SP. 2016.

1844 DADA2: High-resolution sample inference from Illumina amplicon data. Nat Methods

1845 13:581–583.

1846 29. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, Peplies J, Glöckner FO.

1847 2013. The SILVA ribosomal RNA gene database project: Improved data processing and

1848 web-based tools. Nucleic Acids Res 41:590–596.

1849 30. Dixon P. 2003. VEGAN, a package of R functions for community ecology. J Veg Sci

1850 14:927–930.

1851 31. Kuhn M. 2008. Building predictive models in R using the caret package. J Stat Softw

1852 28:1–26.

1853 32. Banerjee S, Schlaeppi K, van der Heijden MGA. 2018. Keystone taxa as drivers of

1854 microbiome structure and functioning. Nat Rev Microbiol 16:567–576.

1855 33. Maboni G, Blanchard A, Frosth S, Stewart C, Emes R, Tötemeyer S. 2017. A distinct

1856 bacterial dysbiosis associated skin inflammation in ovine footrot. Sci Rep 7:1–11.

1857 34. Marcatili P, Nielsen MW, Sicheritz-Pontén T, Jensen TK, Schafer-Nielsen C, Boye M,

1858 Nielsen M, Klitgaard K. 2016. A novel approach to probe host-pathogen interactions of

1859 bovine digital dermatitis, a model of a complex polymicrobial infection. BMC Genomics

105 1860 17:987.

1861 35. Beninger C, Naqvi SA, Naushad S, Orsel K, Luby C, Derakhshani H, Khafipour E, De

1862 Buck J. 2018. Associations between digital dermatitis lesion grades and the quantities of

1863 four Treponema species. Vet Res 49:111.

1864 36. Arrazuria R, Knight CG, Lahiri P, Cobo ER, Barkema HW, De Buck J. 2020. Treponema

1865 spp. Isolated from Bovine Digital Dermatitis Display Different Pathogenicity in a Murine

1866 Abscess Model. Microorganisms 8:1507.

1867 37. Elliott MK, Alt DP, Zuerner RL. 2007. Lesion formation and antibody response induced

1868 by papillomatous digital dermatitis-associated spirochetes in a murine abscess model.

1869 Infect Immun 75:4400–4408.

1870 38. Wyss C, Dewhirst FE, Paster BJ, Thurnheer T, Luginbühl A. 2005. Guggenheimella bovis

1871 gen. nov., sp. nov., isolated from lesions of bovine dermatitis digitalis. Int J Syst Evol

1872 Microbiol 55:667–671.

1873 39. Griffin D, Chengappa MM, Kuszak J, McVey DS. 2010. Bacterial Pathogens of the

1874 Bovine Respiratory Disease Complex. Vet Clin North Am Food Anim Pract 26:381–394.

1875 40. Sullivan LE, Evans NJ, Blowey RW, Grove-White DH, Clegg SR, Duncan JS, Carter SD.

1876 2015. A molecular epidemiology of treponemes in beef cattle digital dermatitis lesions

1877 and comparative analyses with sheep contagious ovine digital dermatitis and dairy cattle

1878 digital dermatitis lesions. Vet Microbiol 178:77–87.

1879 41. Tan KH, Seers CA, Dashper SG, Mitchell HL, Pyke JS, Meuric V, Slakeski N, Cleal SM,

1880 Chambers JL, McConville MJ, Reynolds EC. 2014. Porphyromonas gingivalis and

1881 Treponema denticola Exhibit Metabolic Symbioses. PLoS Pathog 10:e1003955.

1882 42. Lockhart JS, Buret AG, Ceri H, Storey DG, Anderson SJ, Morck DW. 2017. Mixed

106 1883 species biofilms of Fusobacterium necrophorum and Porphyromonas levii impair the

1884 oxidative response of bovine neutrophils in vitro. Anaerobe 47:157–164.

1885 43. Johnson JS, Spakowicz DJ, Hong BY, Petersen LM, Demkowicz P, Chen L, Leopold SR,

1886 Hanson BM, Agresta HO, Gerstein M, Sodergren E, Weinstock GM. 2019. Evaluation of

1887 16S rRNA gene sequencing for species and strain-level microbiome analysis. Nat

1888 Commun 10:5029.

1889 44. Rasmussen M, Capion N, Klitgaard K, Rogdo T, Fjeldaas T, Boye M, Jensen TK. 2012.

1890 Bovine digital dermatitis: Possible pathogenic consortium consisting of Dichelobacter

1891 nodosus and multiple Treponema species. Vet Microbiol 160:151–161.

1892 45. Calvo-Bado LA, Oakley BB, Dowd SE, Green LE, Medley GF, Ul-Hassan A, Bateman V,

1893 Gaze W, Witcomb L, Grogono-Thomas R, Kaler J, Russell CL, Wellington EMH. 2011.

1894 Ovine pedomics: The first study of the ovine foot 16S rRNA-based microbiome. ISME J

1895 5:1426–1437.

1896 46. Schlafer S, Nordhoff M, Wyss C, Strub S, Hübner J, Gescher DM, Petrich A, Göbel UB,

1897 Moter A. 2008. Involvement of Guggenheimella bovis in digital dermatitis lesions of dairy

1898 cows. Vet Microbiol 128:118–125.

1899 47. Topçuoğlu BD, Lesniak NA, Ruffin MT, Wiens J, Schloss PD. 2020. A Framework for

1900 Effective Application of Machine Learning to Microbiome-Based Classification

1901 Problems. MBio 11:e00434-20.

1902 48. Orsel K, Plummer P, Shearer J, De Buck J, Carter SD, Guatteo R, Barkema HW. 2018.

1903 Missing pieces of the puzzle to effectively control digital dermatitis. Transbound Emerg

1904 Dis 65:186–198.

1905 49. Gotoh Y, Chiba K, Sekiyama Y, Okada K, Hayashi T, Misawa N. 2020. 16S rRNA-based

107 1906 amplicon analysis of changes in the bacterial population in the lesions of papillomatous

1907 digital dermatitis in dairy cattle after topical treatment with allyl isothiocyanate. Microbiol

1908 Immunol 64:416–423.

1909

108 1910 Tables

1911 Table 3.1: Characteristics of studies and sequencing experiments included in meta-analysis Study Objectives Conclusions 16S DNA extraction Platform Sample # Readsb Breed / BioProject ID rRNA kit size / Country region Lesion stages Gotoh et Compare Treponema V1V3 Dneasy Blood and LS454 32 / 411,834 Dairy / PRJDB5495 al. 2020 microbiota relative Tissue Kit Active Japan (49) before and abundance is after reduced upon treatments. treatment. Nielsen et Determine Treponema, T- AllPrep Illumina 68e / 2,428,737 Dairy / PRJNA300499 al. 2016 what Mycoplasma, V3V4a DNA/RNA/miRNA Healthy, (V1V2) Denmark (22) microbiota Fusobacterium & Kit Inactive is and V1V2 1,613,962 positively Porphyromonas (T- associated are positively V3V4a) with DD. associated with DD. Hesseling Determine DD prevalence V3V4 N/Ac Illumina 18e / 2,714,524 Mixedd / PRJNA429866 et al. 2019 the is significantly Healthy, Australia (23) prevalence higher in dairy Active, of DD and breeds than beef Inactive identify cattle. which Treponema are bacteria are significantly consistent more abundant in DD in DD compared lesions. to normal skin. Beninger Quantify Treponema V1V2 Dneasy Blood and Illumina 16f / N/Ac Dairy / PRJNA478809 et al. 2018 the species Tissue Kit Active Canada (35) abundances composition and of the four quantities Treponema correlate with

109 species in DD lesion DD lesions. stages. Moreira et Describe DD lesion T- Dneasy Blood and Illumina 92 / 4,958,950 Dairy / PRJNA369034 al. 2018 DD progression is V3V4 Tissue Kit Healthy, Brazil (24) microbiota associated with & DDg in year- Treponema V1V2 round species and grazing Dichelobacter dairy cattle. nodosus. Caddey et Identify DD lesions were V3V4 Dneasy Blood and Illumina 98 / 4,860,255 Beef / PRJNA664530 al., and associated with Tissue Kit Healthy, Canada submitted describe species of Active, for bacterial Treponema, Inactive publication populations Mycoplasma, associated Porphyromonas, with DD Fusobacterium, lesions. and Bacteroides 1912 a T-V3V4 refers to primers selectively amplifying Treponema within the V3V4 hypervariable region 1913 b Number of reads represents the published total number of sequencing reads after quality filtering 1914 c Not reported in BioProject description or associated publication 1915 d Skin samples came from both dairy and beef breeds. Metadata was not sufficient to link breed to individual samples 1916 e Skin samples came from slaughterhouse animals instead of on farm 1917 f Skin samples were disinfected with Chlorhexidine prior to sampling 1918 g Individual lesion stages were not reported 1919

110 1920 Table 3.2: Alpha diversity estimatesa of DD negative and DD positive skin microbiota Observed Chao1 Shannon Fisher ASVs DD 384.5 ± 566.5 ± 5.2 ± 0.7 189.0 ± negative 167.7 354.0 137.6 DD 197.4 ± 244.6 ± 4.1± 1.1 74.6 ± 85.5 positive 144.5 197.4 P value 4.3x10-11 3.8x10-13 1.9x10-8 3.3.x10-10

1921 a Alpha diversity estimates shown are mean plus/minus standard deviation

1922

1923 Table 3.3: Permutation analysis of variance and analysis of multivariate homogeneity of 1924 group dispersions on Bray-Curtis distances of DD negative and DD positive skin 1925 microbiota grouped at different taxonomic levels PERMANOVAa BETADISPERb Pseudo-F R2 P value F value P value ratio Phylum 48.534 0.17874 0.001 12.384 0.001 Family 54.471 0.19631 0.001 0.1454 0.712 Genus 57.519 0.20504 0.001 2.653 0.114 1926 a Permutational analysis of variance 1927 b Analysis of multivariate homogeneity of group dispersions 1928

1929 Table 3.4: Evaluation of random forest model performance on classifying DD status from 1930 microbial composition at three taxonomic levels Phylum Family Genus

Accuracy 0.90 0.91 0.97 Kappa 0.61 0.52 0.82 1931

1932

111

1933 Table 3.5: Study variable PERMANOVA analysis on Bray-Curtis distances of DD positive 1934 skin microbiota grouped at different taxonomic levels Phylum Family Genus Study Pseudo- R2 P Pseudo-F R2 P Pseudo- R2 P variable F ratio value ratio value F ratio value BioProject (n=6) 9.5837 0.208 0.001 9.395 0.204 0.001 9.504 0.206 0.001 Amplicon (n=3) 12.466 0.118 0.001 11.852 0.113 0.001 12.23 0.116 0.001 Platform (n=2) 6.776 0.035 0.003 4.769 0.025 0.002 5.201 0.027 0.005 1935

1936

1937 Table 3.6: Study variable BETADISP analysis of Bray-Curtis distances between DD 1938 positive skin microbiota grouped at different taxonomic levels Study Phylum Family Genus variable F value P value F value P value F value P value BioProject 1.969 0.086 5.225 0.002 4.829 0.001 Country 2.615 0.036 5.485 0.001 4.754 0.004 Amplicon 7.973 0.001 10.665 0.001 9.758 0.001 Platform 6.219 0.011 11.791 0.003 9.945 0.002 1939

1940

1941 Table 3.7: Alpha diversity estimates* of DD positive skin microbiota across different 16S 1942 rRNA hypervariable regions** Observed Chao1 Shannon Fisher V1V2 135.2 ± 163.0 ± 3.3 ± 0.6a 32.6 ± 132.5a 198.2a 58.3a V1V3 98.4 ± 105.5 ± 3.1 ± 0.7a 19.5 ± 51.2a 56.5a 12.0a V3V4 368.6 ± 399.0 ± 5.2 ± 0.7b 109.3 ± 157.3b 172.2b 70.3b 1943 * Alpha diversity estimates shown are mean plus/minus standard deviation 1944 ** Different letters within a column indicate significant difference (p < 0.05) 1945

112

1946 Figures

1947

1948

1949 Figure 3.1: Principal coordinates analysis of Bray-Curtis distances of the microbiota in DD 1950 negative and DD positive skin. Bray-Curtis distances were calculated on samples with ASVs 1951 grouped at three different taxonomic levels: A) Phylum, B) Family, C) Genus. 1952

1953

113

1954

1955 Figure 3.2: Mean percent relative abundances of microbiota in DD negative and DD 1956 positive skin. A) Taxa were grouped at the phylum level, and all taxa that had less than 1% 1957 relative abundance were grouped. Taxa were grouped at the B) family and C) genus level, and all 1958 taxa that had less than 2.5% and 2% relative abundance, respectively, were grouped. 1959

1960

114

1961

1962

1963 Figure 3.3: Relative importance ranking of random forest classifier at genus level 1964 taxonomy. A) Percent relative importance of top 30 genera to the random forest classifier. 1965 Genera are coloured red if they have higher relative abundance in DD negative skin, and blue if 1966 they have a higher relative abundance in DD positive skin. B) Square-root transformed relative 1967 abundances in DD positive and negative skin. Relative abundances were square-root transformed 1968 for easier visualization of large differences in relative abundance between genera. C) Proportion 1969 of samples containing at least one sequence count of each genera. 1970

115

1971

1972

1973 Figure 3.4: Principal coordinates analysis on Bray-Curtis distances of DD positive skin 1974 microbiota. Bray-Curtis distances were calculated on DD positive skin samples in which ASVs 1975 were grouped at A) phylum, B) family, and C) genus level classifications. 1976

1977

1978

1979

1980

1981

1982 116

1983

1984 Figure 3.5: Mean percent relative abundance of DD negative skin and DD lesions from 16S 1985 rRNA hypervariable regions V1V2, V1V3, and V3V4. A) Taxa were grouped at the phylum 1986 level, and all taxa that had less than 1% relative abundance were grouped. Taxa were grouped at 1987 the B) genus level, and all taxa that had less than 2% relative abundance were grouped. 1988

1989

117

1990 Supplemental Material

1991 Table S3.1: Proportion of DD negative and DD positive skin samples containing at least one 1992 read of each genus Genus DD negative DD positive Relative proportion in DD skin skin lesionsa (DD positive – DD negative) Amnipila 0.19 0.83 0.64 Fretibacterium 0 0.49 0.49 Mycoplasma 0.49 0.94 0.45 Catonella 0.11 0.56 0.45 Ezakiella 0.41 0.82 0.41 Campylobacter 0.43 0.75 0.32 S5-A14a 0.16 0.47 0.31 Filifactor 0.11 0.41 0.30 Zymophilus 0 0.30 0.30 Porphyromonas 0.70 0.98 0.28 Fusobacterium 0.38 0.63 0.25 Negativicoccus 0.05 0.3 0.25 Centipeda 0.08 0.31 0.23 Anaerococcus 0.51 0.73 0.22 Parvimonas 0.11 0.32 0.21 Peptoanaerobacter 0.62 0.81 0.19 Peptococcus 0.30 0.48 0.18 Schwartzia 0 0.15 0.15 Gallicola 0.62 0.75 0.13 Peptoniphilus 0.65 0.76 0.11 Treponema 0.86 0.96 0.10 Desulfoplanes 0 0.09 0.09 1993 a top 30 genera ranked by difference in proportions of DD positive and DD negative skin samples 1994 are shown 1995

1996

118

1997

1998 Figure S3.1: Relative abundance heatmap of bacterial families on each sample included in 1999 the meta-analysis. Rare families present in less than 10% of samples were removed from this 2000 analysis. Color key values represent percent relative abundance of each family within a sample. 2001 Dendrograms were built using hierarchical clustering on Bray-Curtis dissimilarities. Vertical 2002 color coding of samples represents DD status (Pink: DD negative; Light Blue: DD positive). 2003

119

2004

2005 Figure S3.2: Heatmap of average relative abundances of DD positive microbiota within 2006 each BioProject. Only DD positive samples were included in the average relative abundance 2007 calculation within BioProjects. Taxa were grouped at family level, and rare families present in 2008 less than 10% of samples were removed from this analysis. Color key values represent average 2009 percent relative abundance of each family. Dendrograms were built using hierarchical clustering 2010 on Bray-Curtis dissimilarities. 2011

2012

2013

120

2014 Chapter 4: Summarizing Discussion

2015 DD lesion microbiota: feedlot beef cattle and beyond

2016 In this thesis, we provide a thorough characterization of DD lesion microbiota. We identified that

2017 Treponema, Mycoplasma, Porphyromonas, and Fusobacterium are relatively more abundant in

2018 DD lesions compared to healthy skin. Bacteroides pyogenes, P. levii, F. necrophorum and an

2019 uncharacterized Fusobacterium sp. were isolated from only DD lesion cultures. A novel

2020 multiplex qPCR targeting P. levii, B. pyogenes, and Fusobacterium sp. was developed, while

2021 existing qPCRs successfully quantified T. pedis, T. phagedenis, T. medium and F. necrophorum

2022 throughout lesion stages. All bacterial species tested were associated with DD lesions compared

2023 to healthy skin. Of particular interest, P. levii and T. phagedenis were detected in all M1 lesions.

2024

2025 A meta-analysis allowed for quantitative and controlled comparisons across DD metagenomic

2026 studies. A core group of DD-associated microbiota consisting of Treponema, Mycoplasma,

2027 Fusobacterium, and Porphyromonas was identified irrespective of individual study variations.

2028 Multiple low-abundant genera within Firmicutes were also associated with DD lesions. No major

2029 differences were evident in primary DD-associated microbiota between beef cattle and dairy

2030 cattle lesions. We were also able to identify Treponema, Porphyromonas, and Mycoplasma in

2031 greater than 90% of DD samples.

2032

2033 The meta-analysis identified that the key DD-associated taxa are the same in both beef and dairy

2034 breeds. This provides evidence that the key bacterial species we quantified using qPCR in beef

2035 DD lesions may similarly be associated in dairy DD lesions. Treponema, Fusobacterium,

2036 Porphyromonas, and Mycoplasma were the genera that best differentiated DD lesions from

121

2037 normal skin microbiota in both studies of this thesis. The primary species within the key DD-

2038 associated genera were P. levii, F. necrophorum, Fusobacterium sp., T. phagedenis, T. medium,

2039 and T. pedis. In addition, we strongly associated the presence of B. pyogenes with DD lesions

2040 through qPCR quantification, even though Bacteroides was not identified as a primary DD-

2041 associated genus in our meta-analysis. Although we have characterized the suspected major

2042 species present in DD for Porphyromonas and Fusobacterium, there are likely additional

2043 Treponema spp. we did not target, and there is a notable gap in our understanding of

2044 Mycoplasma spp. dynamics throughout lesion progression. As a result of the work in this thesis,

2045 we are within reach of a complete picture of the core taxa driving DD progression, and are

2046 striving towards elucidating etiopathogenesis mechanisms.

2047

2048 Key DD-associated microbiota

2049 Treponema, Porphyromonas, Fusobacterium, and Mycoplasma best differentiate DD lesions

2050 from normal skin, regardless of individual study and cattle breed. We identified Treponema as

2051 the predominant member of DD microbiota, confirming previous observations (1–3). Concurring

2052 with Beninger et al. 2018 (4), T. phagedenis was the most prevalent Treponema spp. in lesions

2053 and healthy skin. Prevalence of T. medium was highest in M1 lesions, but overall was not present

2054 in the majority of DD lesions in contrast to previous studies (4–7). We also identify T. pedis

2055 consistently throughout lesion development corresponding to previous data (4, 5, 8).

2056

2057 Key non-Treponema members of the DD microbiota were Mycoplasma, Porphyromonas, and

2058 Fusobacterium. We identified Mycoplasma populations increasing systematically towards

2059 chronic stages of DD, as shown in previous work (9). Mycoplasma were detectable in almost all

122

2060 DD lesions in our work and others (3); therefore, they potentially play crucial roles in DD

2061 development, and further characterization of the species involved can improve our understanding

2062 of Mycoplasma involvement in DD. Porphyromonas levii was identified in 100% active DD

2063 lesions. Other Porphyromonas spp. in periodontitis are capable of modulating expression

2064 patterns of other taxa to increase virulence (10–12), and a similar keystone role for P. levii is

2065 possible based on its high expression levels in DD lesions (13). Through combined use of

2066 culture, deep sequencing, and qPCR, we identified the presence of F. necrophorum and an

2067 uncharacterized Fusobacterium sp. in DD lesions. Whereas F. necrophorum is frequently

2068 identified in DD lesions (3, 8, 14), we strongly associated DD lesions with the presence of both

2069 Fusobacterium sp. and F. necrophorum. In addition, we were able to identify Bacteroides

2070 pyogenes associated with DD lesions for the first time through culture isolation and qPCR

2071 quantification.

2072

2073 Insights into DD lesion establishment and progression

2074 Based on abundance data presented in this dissertation, we increased our understanding of

2075 potential etiopathogenesis mechanisms of DD. Our data suggests that DD lesion formation is the

2076 result of P. levii colonization along with Fusobacterium and some combination of Treponema

2077 such as T. medium, T. phagedenis, or T. pedis. Secondary colonization and rapid expansion of

2078 DD-associated microbiota may result in progression through different lesion stages. Identifying

2079 polymicrobial interactions in lesions may reveal P. levii as a keystone member of DD lesions,

2080 which is an argument strengthened by our findings detecting P. levii in 100% of active stage

2081 lesions. There are additional potential roles for Fusobacterium spp. and B. pyogenes in the

2082 facilitation of anaerobic conditions or possible immune evasion strategies to aid in lesion

123

2083 progression and development (15–19). There are likely several mechanisms of polymicrobial

2084 synergy at work in DD lesion microbiota, such as metabolic interactions to increase virulence

2085 (10) and polymicrobial biofilms (11, 15). Transition from ulcerative lesions to chronic

2086 hyperkeratotic lesions is strongly associated with increased Mycoplasma abundance, and

2087 suggests that Mycoplasma may facilitate morphological change to a chronic and recurring lesion

2088 state. This is further supported by repeated demonstration of Mycoplasma spp. involved in many

2089 chronic inflammatory human diseases, and association of Mycoplasma with host cellular

2090 transformations (including cancer) through its influence on host gene expression (20). However,

2091 as a polymicrobial disease, there are likely several variations in the overall microbial

2092 composition that contributes to the progression and severity of DD cases in cattle. The

2093 hypothetical model of infection dynamics stated above are visualized in Appendix A.

2094

2095 As a direct result of extrapolating on abundance data, the above roles of potential DD pathogens

2096 are simply conjecture, albeit based on consistent observations across hundreds of lesion samples

2097 and multiple studies, providing significant evidence to warrant future research. Relative

2098 abundance data from deep sequencing is inherently biased alone, but with qPCR validation,

2099 becomes strong evidence of true community structure and bacterial population dynamics (21).

2100 However, there are still major limitations to validated abundance data in this context, because

2101 bacterial copy numbers do not translate to functional activity. In this work, we identify bacterial

2102 population trends across DD lesions, and work under the assumption that rapid increase or

2103 decrease in abundance across a lesion stage likely determines relative impact to disease status.

2104 Functional annotation can be achieved through metagenome assembly or other predictive

2105 methods on 16S metagenomics, but are only predictive of putative factors and depend on

124

2106 availability of suitable references for assembly (22–24). Nevertheless, we have completed the

2107 necessary foundational work to establish a narrowed subset of species and genera to focus on for

2108 future DD research into pathogenesis mechanisms.

2109

2110 Insights into DD transmission and infection dynamics

2111 We identified a significantly lower absolute abundance of all the species tested in our qPCR,

2112 except for Fusobacterium sp., in a feedlot without DD compared to feedlots with DD. In

2113 addition, T. medium, F. necrophorum, and B. pyogenes were undetectable in skin from the

2114 feedlot without any DD cases at the time of sampling. This may suggest that this group of

2115 bacteria are potential infectious agents and play a role in lesion formation. However, there are

2116 substantial limitations to these conclusions, including limited sample size, misdiagnosing DD,

2117 and varying management practices between farms; therefore, more farms and increased sample

2118 size are required to validate these findings before making further conclusions. From this work,

2119 we can generate hypotheses on the mechanisms of DD spread and infectious reservoirs, in which

2120 studies on this topic have mostly been limited to Treponema (25, 26).

2121

2122 Slurry or wet conditions likely aid in lesion formation, as these conditions are associated with

2123 increased DD prevalence and increased skin permeability, potentially increasing chances of early

2124 lesion pathogens colonizing skin (27, 28). However, the infection reservoirs for DD pathogens

2125 beyond Treponema are not well characterized. Treponema have been identified in many

2126 locations such as cattle mouth, rumen, gut, rectum, feces, and slurry (25, 29–32). Furthermore,

2127 Porphyromonas is present in higher abundance in slurry samples from farms with DD than farms

2128 without (31). More work is required to elucidate infection reservoirs and transmission routes,

125

2129 especially for non-Treponema DD microbiota using the qPCR assays developed and identified as

2130 part of this thesis.

2131

2132 Insights into future treatment and therapeutics

2133 The unsatisfactory treatment success rate of DD is a major issue for producers, and is

2134 contributing to overuse of antimicrobials leading to antibiotic resistance and environmental

2135 contamination from footbath components (33). Our results are encouraging for future

2136 development of more selective therapeutic strategies that could target key DD pathogenic

2137 causative agents or interrupt required bacterial synergy. Effective and efficient preventative

2138 measures are desired for the control of DD, and vaccine development has been discussed but

2139 attempts immunizing against Treponema so far have been unsuccessful in reducing DD

2140 prevalence (34). Vaccines protective against DD pathogens are a possible future, but researchers

2141 must consider the polymicrobial nature of DD in order to develop effective therapeutics.

2142 Successful vaccines will likely target several key pathogens, and must consider antigenic

2143 variation and strain variation within lesions. In periodontal disease mice models, immunizations

2144 against the keystone taxa P. gingivalis resulted in significant reductions in periodontitis and

2145 associated bone loss (35). This provides evidence that vaccines targeting key taxa in

2146 polymicrobial diseases can reduce prevalence, therefore, P. levii, Treponema spp.,

2147 Fusobacterium spp., and Mycoplasma spp. may be adequate targets for future DD therapeutics.

2148 In sheep, ovine foot rot vaccination against multiple D. nodosus serotypes reduces the odds of

2149 acquiring CODD (36). Because we associated the bovine foot rot causative agent, F.

2150 necrophorum, with bovine DD lesions, there are added implications on potential treatment

2151 connections between bovine DD and other hoof diseases. However, there are significant

126

2152 challenges with generating vaccines compensating for antigenic variation, and phenotypic

2153 variation of bacteria between monomicrobial and polymicrobial settings (37). Characterization of

2154 bacterial metabolic phenotypes throughout DD progression is required before vaccine

2155 development to ensure any antigens targeted are ubiquitously expressed, immunogenic, and

2156 accessible in biofilms (37).

2157

2158 Tools we have developed

2159 Through our work, we developed and validated a novel multiplex species-specific qPCR to

2160 quantify DD-associated bacteria. Combined with an existing Treponema species-specific

2161 multiplex qPCR (4), and a F. necrophorum qPCR (38), we were able to validate and characterize

2162 the species-level abundances between lesion stages. These tools are sensitive and reproducible

2163 methods that will henceforth increase efficiency of DD microbiology research. Studies on

2164 environmental reservoirs, potential future vaccine trials, culture optimizations, and many other

2165 applications can all equally benefit from these qPCR tools and accelerate progress in DD

2166 research.

2167

2168 Combining metagenomics, culturing, and qPCR provided comprehensive microbiological

2169 characterization of DD lesions, and validated species-level population dynamics throughout

2170 lesion stages. There are limitations to each of these methods alone, which are minimized when

2171 combined together. The inability of our metagenomic approach to designate reliable species-

2172 level taxonomy was compensated by full length 16S rRNA sequencing of culture isolates and

2173 species-specific qPCR development. Our qPCR accurately and reliably quantified individual

2174 species populations in skin samples, but we did not target all potential DD pathogens. Our

127

2175 metagenomic approach provides an estimate to the number of additional species that requires

2176 further validation and characterization in DD lesions. It is important to note that the species-

2177 specific multiplex qPCRs in this study were developed on publicly available genomes—most of

2178 which are not isolates from hoof lesions—and locally derived strains; therefore, these qPCRs

2179 may not amplify all global strains as efficiently and require further validation when more isolates

2180 become sequenced.

2181

2182 Next steps

2183 As previously mentioned, the implications on DD etiopathogenesis from this thesis are derived

2184 from abundance data, which have several limitations. In order to make accurate conclusions on

2185 the microbiota involved in DD causation, future studies are necessary. Continued validation of

2186 additional species-level dynamics in DD lesions must continue to fully understand the molecular

2187 epidemiology of DD. Meanwhile, low-abundant taxa should not be ignored in their potential

2188 involvement in DD pathogenesis. Furthermore, unique species pathogenic mechanisms and

2189 synergistic interactions can be understood through meta-transcriptomic/metabolomic/proteomic

2190 approaches. This can identify and validate keystone species of infections, and provide strong

2191 evidence of synergism and metabolic pathway interactions between species (12). Studies can be

2192 performed on co-cultures of DD-associated species and lesion biopsies to understand the full

2193 framework of pathogenesis in the context of DD lesion formation and progression.

2194 Understanding the species involved in DD pathogenesis and mechanisms of lesion development

2195 is key in developing more successful prevention and treatment strategies.

2196

128

2197 Significance for producers and researchers

2198 The findings presented in this thesis deliver a significant gain in knowledge to the research

2199 community and producers alike. This work contains many firsts in the field of DD microbiology,

2200 and will facilitate critical changes in our thinking of DD disease dynamics.

2201

2202 To the producers, our work provides a greater understanding of the bacteriological factors behind

2203 DD lesion formation and development. We identify potential causative agents of DD, and some

2204 bacterial species we targeted are already known to producers as pathogens of other hoof diseases,

2205 such as the foot rot agent F. necrophorum. Our results provide novel insights into potential

2206 transmission dynamics of DD, and the optimism for future therapeutic development, sparking

2207 discussion on the importance of on-farm biosecurity protocols and the need for more effective

2208 treatment and prevention strategies.

2209

2210 For the research community, our work establishes a consistent DD-associated microbiota, and

2211 provides a systematic framework of future research avenues to elucidate DD etiopathogenesis.

2212 We support a polymicrobial causation of DD, consisting of microbial constituents belonging to

2213 Treponema, Mycoplasma, Porphyromonas, and Fusobacterium. In previous studies of DD

2214 microbiota, species-level taxonomy is usually not validated, and in our work we perform

2215 accurate species classification while showcasing the critical importance of studying species

2216 population dynamics. We also provide crucial qPCR tools for efficient and reproducible

2217 quantification of potential DD pathogens, aiding in the acceleration of future research on DD and

2218 related diseases.

2219

129

2220 References

2221 1. Krull AC, Shearer JK, Gorden PJ, Cooper VL, Phillips GJ, Plummera PJ. 2014. Deep

2222 sequencing analysis reveals temporal microbiota changes associated with development of

2223 bovine digital dermatitis. Infect Immun 82:3359–3373.

2224 2. Nordhoff M, Moter A, Schrank K, Wieler LH. 2008. High prevalence of treponemes in

2225 bovine digital dermatitis-A molecular epidemiology. Vet Microbiol 131:293–300.

2226 3. Nielsen MW, Strube ML, Isbrand A, Al-Medrasi WDHM, Boye M, Jensen TK, Klitgaard

2227 K. 2016. Potential bacterial core species associated with digital dermatitis in cattle herds

2228 identified by molecular profiling of interdigital skin samples. Vet Microbiol 186:139–149.

2229 4. Beninger C, Naqvi SA, Naushad S, Orsel K, Luby C, Derakhshani H, Khafipour E, De

2230 Buck J. 2018. Associations between digital dermatitis lesion grades and the quantities of

2231 four Treponema species. Vet Res 49:111.

2232 5. Sullivan LE, Evans NJ, Blowey RW, Grove-White DH, Clegg SR, Duncan JS, Carter SD.

2233 2015. A molecular epidemiology of treponemes in beef cattle digital dermatitis lesions

2234 and comparative analyses with sheep contagious ovine digital dermatitis and dairy cattle

2235 digital dermatitis lesions. Vet Microbiol 178:77–87.

2236 6. Klitgaard K, Bretó AF, Boye M, Jensen TK. 2013. Targeting the treponemal microbiome

2237 of digital dermatitis infections by high-resolution phylogenetic analyses and comparison

2238 with fluorescent in situ hybridization. J Clin Microbiol 51:2212–2219.

2239 7. Nascimento LV, Mauerwerk MT, Dos Santos CL, De Barros Filho IR, Birgel EH,

2240 Sotomaior CS, Madeira HMF, Ollhoff RD. 2015. Treponemes detected in digital

2241 dermatitis lesions in Brazilian dairy cattle and possible host reservoirs of infection. J Clin

2242 Microbiol 53:1935–1937.

130

2243 8. Moreira TF, Facury Filho EJ, Carvalho AU, Strube ML, Nielsen MW, Klitgaard K, Jensen

2244 TK. 2018. Pathology and bacteria related to digital dermatitis in dairy cattle in all year

2245 round grazing system in Brazil. PLoS One 13:e0193870.

2246 9. Hesseling J, Legione AR, Stevenson MA, McCowan CI, Pyman MF, Finochio C, Nguyen

2247 D, Roic CL, Thiris OL, Zhang AJ, van Schaik G, Coombe JE. 2019. Bovine digital

2248 dermatitis in Victoria, Australia. Aust Vet J 97:404–413.

2249 10. Tan KH, Seers CA, Dashper SG, Mitchell HL, Pyke JS, Meuric V, Slakeski N, Cleal SM,

2250 Chambers JL, McConville MJ, Reynolds EC. 2014. Porphyromonas gingivalis and

2251 Treponema denticola exhibit metabolic symbioses. PLoS Pathog 10:e1003955.

2252 11. Zhu Y, Dashper SG, Chen YY, Crawford S, Slakeski N, Reynolds EC. 2013.

2253 Porphyromonas gingivalis and Treponema denticola synergistic polymicrobial biofilm

2254 development. PLoS One 8:e71727.

2255 12. Banerjee S, Schlaeppi K, van der Heijden MGA. 2018. Keystone taxa as drivers of

2256 microbiome structure and functioning. Nat Rev Microbiol 16:567–576.

2257 13. Marcatili P, Nielsen MW, Sicheritz-Pontén T, Jensen TK, Schafer-Nielsen C, Boye M,

2258 Nielsen M, Klitgaard K. 2016. A novel approach to probe host-pathogen interactions of

2259 bovine digital dermatitis, a model of a complex polymicrobial infection. BMC Genomics

2260 17:987.

2261 14. Staton GJ, Sullivan LE, Blowey RW, Carter SD, Evans NJ. 2020. Surveying bovine

2262 digital dermatitis and non-healing bovine foot lesions for the presence of Fusobacterium

2263 necrophorum, Porphyromonas endodontalis and Treponema pallidum. Vet Rec 186:450.

2264 15. Lockhart JS, Buret AG, Ceri H, Storey DG, Anderson SJ, Morck DW. 2017. Mixed

2265 species biofilms of Fusobacterium necrophorum and Porphyromonas levii impair the

131

2266 oxidative response of bovine neutrophils in vitro. Anaerobe 47:157–164.

2267 16. Tan ZL, Nagaraja TG, Chengappa MM. 1996. Fusobacterium necrophorum infections:

2268 virulence factors, pathogenic mechanism and control measures. Vet Res Commun 20:113–

2269 140.

2270 17. Narayanan S, Stewart GC, Chengappa MM, Willard L, Shuman W, Wilkerson M,

2271 Nagaraja TG. 2002. Fusobacterium necrophorum leukotoxin induces activation and

2272 apoptosis of bovine leukocytes. Infect Immun 70:4609–4620.

2273 18. Cunha F, Jeon SJ, Daetz R, Vieira-Neto A, Laporta J, Jeong KC, Barbet AF, Risco CA,

2274 Galvão KN. 2018. Quantifying known and emerging uterine pathogens, and evaluating

2275 their association with metritis and fever in dairy cows. Theriogenology 114:25–33.

2276 19. Plummer PJ, Krull A. 2017. Clinical perspectives of digital dermatitis in dairy and beef

2277 cattle. Vet Clin North Am - Food Anim Pract 33:165–181.

2278 20. Benedetti F, Curreli S, Zella D. 2020. Mycoplasmas–host interaction: Mechanisms of

2279 inflammation and association with cellular transformation. Microorganisms 8:1351.

2280 21. Jian C, Luukkonen P, Yki-Järvinen H, Salonen A, Korpela K. 2020. Quantitative PCR

2281 provides a simple and accessible method for quantitative microbiota profiling. PLoS One

2282 15:e0227285.

2283 22. Langille MGI, Zaneveld J, Caporaso JG, McDonald D, Knights D, Reyes JA, Clemente

2284 JC, Burkepile DE, Vega Thurber RL, Knight R, Beiko RG, Huttenhower C. 2013.

2285 Predictive functional profiling of microbial communities using 16S rRNA marker gene

2286 sequences. Nat Biotechnol 31:814–821.

2287 23. Sun S, Jones RB, Fodor AA. 2020. Inference-based accuracy of metagenome prediction

2288 tools varies across sample types and functional categories. Microbiome 8:46.

132

2289 24. Wilkinson TJ, Huws SA, Edwards JE, Kingston-Smith AH, Siu-Ting K, Hughes M,

2290 Rubino F, Friedersdorff M, Creevey CJ. 2018. CowPI: A rumen microbiome focussed

2291 version of the PICRUSt functional inference software. Front Microbiol 9:1095.

2292 25. Klitgaard K, Nielsen MW, Ingerslev HC, Boye M, Jensen TK. 2014. Discovery of bovine

2293 digital dermatitis-associated Treponema spp. in the dairy herd environment by a targeted

2294 deep-sequencing approach. Appl Environ Microbiol 80:4427–4432.

2295 26. Gillespie A V., Carter SD, Blowey RW, Staton GJ, Evans NJ. 2020. Removal of bovine

2296 digital dermatitis-associated treponemes from hoof knives after foot-trimming: A

2297 disinfection field study. BMC Vet Res 16:330.

2298 27. Wells SJ, Garber LP, Wagner BA. 1999. Papillomatous digital dermatitis and associated

2299 risk factors in US dairy herds. Prev Vet Med 38:11–24.

2300 28. Palmer MA, Donnelly RF, Garland MJ, Majithiya R, O’Connell NE. 2013. The effect of

2301 slurry on skin permeability to methylene blue dye in dairy cows with and without a history

2302 of digital dermatitis. Animal 7:1731–1737.

2303 29. Evans NJ, Timofte D, Isherwood DR, Brown JM, Williams JM, Sherlock K, Lehane MJ,

2304 Murray RD, Birtles RJ, Anthony Hart C, Carter SD. 2012. Host and environmental

2305 reservoirs of infection for bovine digital dermatitis treponemes. Vet Microbiol 156:102–

2306 109.

2307 30. Sullivan LE, Carter SD, Duncan JS, Grove-White DH, Angell JW, Evans NJ. 2015. The

2308 gastrointestinal tract as a potential infection reservoir of digital dermatitis-associated

2309 treponemes in beef cattle and sheep. Appl Environ Microbiol 81:7460–7469.

2310 31. Klitgaard K, Strube ML, Isbrand A, Nielsen MW, Jensen TK. 2017. Microbiota analysis

2311 of an environmental slurry and its potential role as a reservoir of bovine digital dermatitis

133

2312 pathogens. Appl Environ Microbiol 83:e00244-17.

2313 32. Zinicola M, Lima F, Lima S, Machado V, Gomez M, Döpfer D, Guard C, Bicalho R.

2314 2015. Altered microbiomes in bovine digital dermatitis lesions, and the gut as a pathogen

2315 reservoir. PLoS One 10:e0120504.

2316 33. Cook NB. 2017. A review of the design and management of footbaths for dairy cattle. Vet

2317 Clin North Am - Food Anim Pract 33:195–225.

2318 34. Ertze RA, Read DH, Hird DW, Berry SL. 2006. Field evaluation of prophylactic and

2319 therapeutic effects of a vaccine against (papillomatous) digital dermatitis in dairy cattle on

2320 two California dairies. Bov Pract 40:76–82.

2321 35. Darveau RP. 2010. Periodontitis: A polymicrobial disruption of host homeostasis. Nat Rev

2322 Microbiol 8:481–490.

2323 36. Duncan JS, Grove-White D, Moks E, Carroll D, Oultram JW, Phythian CJ, Williams HW.

2324 2012. Impact of footrot vaccination and antibiotic therapy on footrot and contagious ovine

2325 digital dermatitis. Vet Rec 170:462.

2326 37. Harro JM, Peters BM, O’May GA, Archer N, Kerns P, Prabhakara R, Shirtliff ME. 2010.

2327 Vaccine development in Staphylococcus aureus: Taking the biofilm phenotype into

2328 consideration. FEMS Immunol Med Microbiol 59:306–323.

2329 38. Witcomb LA, Green LE, Kaler J, Ul-Hassan A, Calvo-Bado LA, Medley GF, Grogono-

2330 Thomas R, Wellington EMH. 2014. A longitudinal study of the role of Dichelobacter

2331 nodosus and Fusobacterium necrophorum load in initiation and severity of footrot in

2332 sheep. Prev Vet Med 115:48–55.

2333

2334 2335 134

2336 Appendices

2337 Appendix A: Hypothetical representation of DD infection development as a function of

2338 associated microbiota dynamics. Microbial population distributions throughout lesion stages

2339 were used to estimate potential conditions and mechanisms for infection and pathogenesis of

2340 DD. Each colour represents DD stages. Lesion progression is not linear throughout M stages,

2341 thus the infection representation below can be read in multiple directions.

2342

135