ThesisPDF Available

Transmission and dynamics of VTEC O157:H7 A story about the complex associations between pathogen, host and environment

Authors:

Abstract and Figures

Verotoxin-producing Escherichia coli serotype O157:H7 (VTEC O157:H7), is a zoonotic pathogen often transmitted from cattle to humans. In Sweden, domestic transmission of a highly virulent subtype of VTEC O157:H7, originating in regional clusters of infected cattle farms, is increasing. To reduce the risk of transmission to humans a comprehensive picture of infection dynamics between and within farms are urgently needed. The aim of this thesis was to provide a holistic view of drivers of transmission and susceptibility from a regional, farm and animal perspective by combining epidemiology, microbiology, bioinformatics and animal welfare. The risk of introduction of VTEC O157:H7 on cattle farms was studied by collecting environmental samples in spring and fall from 80 farms. Information about farm characteristics, biosecurity and between farm contacts were collected by a questionnaire. On 4 farms, a more thorough environmental sampling with detailed analysis of strains was carried out during summer (between the spring and fall sampling). The results showed frequent transmission of VTEC O157:H7 between farms and that transmission occurs through human and animals contacts. To investigate drivers of colonisation and transmission on farm level, individual samples from calves on 12 dairy farms with VTEC O157:H7 (established through environmental sampling) were collected. In addition to collecting information about pen and calf environment, a novel approach, using indicators of animal welfare and behaviour to study individual differences, was used to explore differences between colonised and non-colonised calves. The results suggest that social and active individuals are more likely to be colonised by the pathogen while animals showing signs of poor health and welfare were less likely to be colonised. Colonised animals shedding high levels of the bacteria were important for transmission but environmental exposure also increased risk of transmission within pens.
Content may be subject to copyright.
!CTA5NIVERSITATIS!GRICULTURAE3UECIAE
Acta Universitatis Agriculturae Sueciae
Doctoral Thesis No. 2020:16
The aim of this thesis was to provide a holistic view of drivers of transmission
and susceptibility of the zoonotic pathogen VTEC O157:H7 from a regional,
farm and animal perspective. The results show frequent transmission between
farms in a cattle dense area and that transmission occurs through human
and animals contacts. A small proportion of colonised “super-shedders” were
important for transmission and colonisation was associated with being a social
and active calf.
Lena-Mari Tamminen received her postgraduate education at the
Department of Clinical Sciences. She obtained her degree in veterinary medicine
in 2013 at the Faculty of Veterinary Medicine and Animal Science, SLU, Uppsala.
Acta Universitatis Agriculturae Sueciae presents doctoral theses from the
Swedish University of Agricultural Sciences (SLU).
SLU generates knowledge for the sustainable use of biological natural
resources. Research, education, extension, as well as environmental monitoring
and assessment are used to achieve this goal.
Online publication of thesis summary: http://pub.epsilon.slu.se/
ISSN 1652-6880
ISBN (print version) 978-91-7760-550-8
ISBN (electronic version) 978-91-7760-551-5
Doctoral Thesis No. 2020:16 • Transmission and dynamics of VTEC O157:H7 • Lena-Mari Tamminen
Doctoral Thesis No. 2020:16
Faculty of Veterinary Medicine and Animal Science
Transmission and dynamics of VTEC
O157:H7
A story about the complex associations between
pathogen, host and environment
Lena-Mari Tamminen
Transmission and dynamics of VTEC
O157:H7
A story about the complex associations between
pathogen, host and environment
Lena-Mari Tamminen
Faculty of Veterinary Medicine and Animal Science
Department of Clinical Science
Uppsala
Doctoral thesis
Swedish University of Agricultural Sciences
Uppsala 2020
Acta Universitatis agriculturae Sueciae
2020:16
ISSN 1652-6880
ISBN (print version) 978-91-7760-550-8
ISBN (electronic version) 978-91-7760-551-5
© 2020 Lena-Mari Tamminen, Uppsala
Print: SLU Service/Repro, Uppsala 2020
Verotoxin-producing Escherichia coli serotype O157:H7 (VTEC O157:H7), is a
zoonotic pathogen often transmitted from cattle to humans. In Sweden, domestic
transmission of a highly virulent subtype of VTEC O157:H7, originating in regional
clusters of infected cattle farms, is increasing. To reduce the risk of transmission to
humans a comprehensive picture of infection dynamics between and within farms are
urgently needed.
The aim of this thesis was to provide a holistic view of drivers of transmission and
susceptibility from a regional, farm and animal perspective by combining epidemiology,
microbiology, bioinformatics and animal welfare.
The risk of introduction of VTEC O157:H7 on cattle farms was studied by collecting
environmental samples in spring and fall from 80 farms. Information about farm
characteristics, biosecurity and between farm contacts were collected by a questionnaire.
On 4 farms, a more thorough environmental sampling with detailed analysis of strains
was carried out during summer (between the spring and fall sampling). The results
showed frequent transmission of VTEC O157:H7 between farms and that transmission
occurs through human and animals contacts.
To investigate drivers of colonisation and transmission on farm level, individual
samples from calves on 12 dairy farms with VTEC O157:H7 (established through
environmental sampling) were collected. In addition to collecting information about pen
and calf environment, a novel approach, using indicators of animal welfare and behaviour
to study individual differences, was used to explore differences between colonised and
non-colonised calves. The results suggest that social and active individuals are more
likely to be colonised by the pathogen while animals showing signs of poor health and
welfare were less likely to be colonised. Colonised animals shedding high levels of the
bacteria were important for transmission but environmental exposure also increased risk
of transmission within pens.
Keywords: EHEC, verotoxin, shigatoxin, calves , cattle, on-farm measures, super-
shedder
Author’s address: Lena-Mari Tamminen, SLU, Department of Clinical Sciences,
P.O. Box 7054, 750 07 Uppsala, Sweden
Transmission and dynamics of VTEC O157:H7
- A story about the complex associations between pathogen, host
and environmen
t
Abstract
Verotoxin-producerande Escherichia coli av serotypen O157:H7 (VTEC O157:H7) är en
zoonos som ofta sprids från nötkreatur till människor. I Sverige ökar antalet fall av en
virulent typ av VTEC O157:H7som förekommer i hög grad bland gårdar i vissa områden.
Ökad förståelse för spridningen mellan och på gårdar behövs för att minska spridningen
till människor.
Målet med denna avhandling var att ge en övergripande bild av hur spridningsdynamik
och andra faktorer påverkar förekomst av bakterien på regional-, gård- och individnivå
genom att kombinera forskningsområdena epidemiologi, mikrobiologi, molekylära
metoder och djurvälfärd.
Spridning mellan gårdar studerades genom att samla in miljöprover från 80 gårdar
under vår och höst. Information om besättningarna samt kontakter och andra riskfaktorer
samlades in med en enkät och på 4 gårdar genomfördes ytterligare provtagningar under
sommaren. Resultaten visar tät smittspridningen i området och att smittan kan spridas
mellan gårdarna genom kontakter mellan både djur och människor.
För att undersöka risk-faktorer för kolonisering och spridning inom gårdar provtogs
kalvar från 12 besättningar där bakterien påvisats genom miljöprovtagning vid två
tillfällen. I tillägg till att samla in information om box och miljö användes indikatorer för
välfärd och observation av beteenden för att undersöka individuella skillnader mellan
koloniserade och icke koloniserade djur. Resultaten visar att kolonisering av VTEC
O157:H7 var vanligare bland sociala och aktiva djur medan tecken på nedsatt hälsa och
välfärd inte var kopplade till kolonisering. Djur som utsöndrade höga mängder bakterier
(så kallade super-utsöndrare) var viktiga för smittspridningen men även exponering från
miljön var en risk.
Keywords: EHEC, verotoxin, shigatoxin, kalv, nötkreatur, bekämpning, super-
utsöndring
Author’s address: Lena-Mari Tamminen, SLU, Department of Clinical Sciences,
P.O. Box 7054, 750 07 Uppsala, Sweden
Smittspridning av VTEC
O157:H7
en skildring av komplexa
samband mellan bakterie, värd och miljö
Abstract
To understand the whole it is necessary to understand the parts. To understand
the parts, it is necessary to understand the whole. Such is the circle of
understanding
Ken Wilbert, The Eye of Spirit
The more we learn and the more advanced our technologies become, the more
we realise that we are living in an increasingly complex reality. We are moving
from recognising infectious disease as simply the presence of a pathogenic
organism towards a much more intricate pattern where the need and importance
of considering the interactions between pathogen, host and the environment is
becoming increasingly clear. A bacteria that exemplifies complexity in multiple
ways is verotoxin-producing Escherichia coli serotype O157:H7 (VTEC
O157:H7). It is able to cause severe disease in humans, but not in everyone who
gets infected. It can persist and multiply in the environment, as well as establish
itself in the gastrointestinal tract of ruminants and in particular cattle. As
opposed to humans, cattle do not get sick, but there is unexplained variation in
which cattle are colonised and which are not. This thesis is an effort to explore
factors influencing VTEC O157:H7 on herd and individual level with the aim of
filling knowledge gaps in some of the parts, thereby increasing the
understanding of the whole.
Preface
To Kenzo.
For enduring the most boring weeks of his life during the work of this
thesis and for always showing an inspiring fighting spirit and admirable
integrity.
Dedication
List of publications 9
List of tables 11
List of figures 12
Abbreviations 13
1 Introduction 15
1.1 A story 15
1.2 The beginning of this story 16
1.3 Verotoxin-producing E. coli O157:H7 16
1.3.1 The rise of VTEC 17
1.3.2 The origin of VTEC O157 and non-VTEC O157 18
1.3.3 Within serotype variation 18
1.3.4 Classification, identification and characterisation of VTEC
O157:H7 20
1.4 VTEC and human disease 23
1.4.1 Pathogenesis in human disease 24
1.4.2 The link between cattle and human disease 25
1.4.3 Transmission of VTEC O157:H7 26
1.5 VTEC O157:H7 in the cattle population 28
1.5.1 Farm dynamics 28
1.5.2 Colonisation and shedding 30
1.5.3 Individual heterogeneity and similarity 31
2 Aims and objectives 33
3 Overview and comments on materials and methods 35
3.1 Study design 35
3.2 Detection and enrolment of participating farms 36
3.3 Sampling of VTEC O157:H7 38
3.3.1 Sampling to identify positive farms 38
3.3.2 Individual sampling 39
3.3.3 Motivation for targeted sampling 40
3.4 Microbial analysis of VTEC O157:H7 40
Contents
3.4.1 Detection in environmental samples 40
3.4.2 Detection in individual samples 41
3.4.3 Characterisation and subtyping 41
3.5 Risk factors 42
3.5.1 General overview 42
3.5.2 Farm characteristics and management 43
3.5.3 Pen characteristics 44
3.5.4 Individual assessment 44
3.5.5 Analysis of hair cortisol 46
3.6 Statistical analysis 48
4 Results and discussion 51
4.1 Between farm transmission on Öland 51
4.1.1 Is it important to differentiate between introduction and
persistence? 55
4.2 Within farm prevalence and transmission 55
4.2.1 Management and susceptibility closely connected potential
drivers of transmission 56
4.2.2 Importance of super-shedding for transmission 61
4.2.3 Transmission dynamics in poor and good hygiene conditions 62
4.3 Stress, colonisation and susceptibility 64
4.4 Validity, bias and methodological considerations 67
4.4.1 Study population and external validity 67
4.4.2 Choice of methods 69
4.4.3 Assessment of risk factors/determinants 70
4.4.4 Statistical methods 71
5 Conclusions, reflections and future perspectives 73
5.1 The end of this story 73
5.2 What can we tell the farmer in the beginning of our story? 74
5.3 The never ending story 75
References 77
Popular science summary 95
Populärvetenskaplig sammanfattning 97
Acknowledgements 99
9
This thesis is based on the work contained in the following papers, referred to
by Roman numerals in the text:
I Tamminen, L.M.*, Söderlund, R., Wilkinson, D. A., Torsein, M.,
Eriksson, E., Churakov, M., Dicksved, J., Keeling, L.J., Emanuelson, U.
(2019). Risk factors and dynamics of verotoxigenic Escherichia coli
O157:H7 on cattle farms: An observational study combining information
from questionnaires, spatial data and molecular analyses. Preventive
Veterinary Medicine, 170, pp. 104726.
II Tamminen, L.M., Dicksved. J., Eriksson, Keeling, L.J., E. Emanuelson, U.
Untangling the role of environmental and host-related determinants for on-
farm transmission of verotoxin-producing Escherichia coli O157:H7
(manuscript)
III Tamminen, L.M.*, Hranac, C.R., Dicksved. J., Eriksson, E. Emanuelson,
U., Keeling, L.J. (2020). Socially engaged calves are more likely to be
colonised by VTEC O157:H7 than individuals showing signs of poor
welfare. (submitted)
IV Tamminen, L.M., Keeling, L.J., Svensson, A., Briot, L., Emanuelson, U.
Considerations for using hair cortisol as an indicator of welfare in dairy
calves (manuscript)
Paper I is reproduced with the permission of the publishers.
* Corresponding author.
List of publications
10
I Involved in the designing the questionnaire. Organised and performed
most analysis of data (except spatial clustering). Drafted the manuscript
and finalised it together with co-authors. Corresponded with the journal.
II Involved in formulating the research idea, planning and organising the
study. Contacted farmers and coordinated (and performed part of)
environmental samplings of farms. Performed sampling and observation of
individual animals. Performed analysis of the data, drafted the manuscript
and finalised it with input from co-authors.
III Involved in formulating the research idea, planning and organising the
study. Contacted farmers and coordinated (and performed part of)
environmental samplings of farms. Performed sampling and observation of
individual animals. Performed analysis of the data (with input from co-
authors), drafted the manuscript and finalised it with input from co-
authors.
IV Actively involved in formulating the research idea and development of the
protocol for hair cortisol analysis. Performed sampling and observation of
individual animals as well as preparation of hair samples together with
master student. Performed analysis of the data, drafted the manuscript and
finalised it with input from co-authors.
The contribution of
Lena-Mari Tamminen
to the papers included in this thesis
was as follows:
11
Table 1. Overview of larger outbreaks of verotoxin-producing Escherichia coli in
Sweden during the course of this project. Source:
Folkhälsomyndigheten (2020). 24
Table 2. Characteristics of the farms included in the individual sampling (paper
II-IV). Farm size includes the total number of cattle on the farm.
Sampled animals is the number of animals sampled for verotoxin-
producing Escherichia coli O157:H7 and the proportion colonised as
determined by recto anal mucosal swabs. Information about calf
mangement of calves, cleaning routines and biosecurity measures
were collected in a structured interview. Last 13 rows reflect farmers
spontaneous answers to the questions “How is introduction of
infectious agents prevented” and “How is transmission of infectious
agents between animals prevented?”. 52
Table 3. Risk factors for colonisation by verotoxin-producing Escherichia coli
O157:H7 in dairy calves in the first sampling stratified by pen
hygiene. 63
List of tables
12
Figure 1. Important virulence factors of verotoxin-producing Escherichia coli
O157:H7 and their localisation in the bacterial genome. 20
Figure 2. An overview of four common molecular methods used to characterise
verotoxin-producing Escherichia coli O157:H7. 22
Figure 3. Infection of verotoxin-producing Escherichia coli O157:H7 in humans. 25
Figure 4. The multiple routes of transmission for verotoxin-producing Escherichia
coli O157:H7 between cattle and humans include contaminated products
as well as indirect transmission through the environment or direct contact
with cattle. 27
Figure 5. Overview of sampling performed within the project. 36
Figure 6. Overshoe samples used in the environmental sampling. 38
Figure 7. Description of the preparation and extraction of hair cortisol. 47
Figure 8. Pasture on Öland with the traditional stone walls separating pastures of
animals from different farms in the background. 54
Figure 9. Distribution of age of sampled calves and results of individual sampling for
verotoxin-producing Escherichia coli O157:H7. 57
Figure 10. Causal diagram describing causal assumptions of risk factors for
colonisation of verotoxin-producing Escherichia coli O157:H7 analysed in
paper II. 58
Figure 11. The association between age (x-axis) and stocking density (y-axis) and
colonisation/shedding of verotoxin-producing Escherichia coli O157:H7 of
the sampled dairy calves. Coloured dots indicate colonised calf (as
detected by recto anal mucosal swabs) and size of dots indicates
shedding level. 59
Figure 12. Hair cortisol concentration (pg/μl) of calves colonised by verotoxin-
producing Escherichia coli O157:H7 (as determined by recto-anal
mucosal swabs) and non-colonised (negative) calves. 65
List of figures
13
CI
CV
eae
EHEC
EPEC
Gam
Gb3
Glm
HPA
HUS
IMS
LEE
MLVA
mTSB
OR
PBS
PT
RAMS
SMAC
SNP
SVA
VTEC
vtx
wgs
WQ
Abbreviations
14
15
1.1 A story1
For children the opportunity to come out to the countryside and visit a dairy farm
is a great learning experience. Imagine a group of children, perhaps a pre-school
class, visiting a dairy farm and the farmer proudly showing them around. The
children are excited to learn about how milk and meat are produced and the
highlight of the visit is meeting the animals, particularly petting the cute calves.
The day is a success and after the visit the farmer is approached, in the local
supermarket and other public places, by grateful parents describing the
children’s joy after the visit. But suddenly, just a few days later, the tune changes
drastically. People suddenly avoid the farmer at the supermarket and other
planned farm visits are cancelled. The farmer hears that some of the visiting
children have become terribly sick, some are even hospitalised and in a critical
condition. The doctors are saying that they have caught a bacteria called
EHEC from the animals. A wave of guilt and worry washes over the farmer.
Have the children really become sick because of the farm visit? Does this mean
that children in the household and the staff are in danger? There is also fear of
what will happen with the animals now that public health agencies want to
investigate the farm. The animals are perfectly healthy and high-producing! In
fact, nothing has changed on the farm and plenty of previous visiting groups
have passed through without anyone getting sick before. Can it really be the farm
animals? If so, how did it go wrong this time? Where did the bacteria come from?
And most importantly, how do you deal with a problem when there are no
symptoms?
1 This story is purely fictional but inspired from meetings with farmers experiencing the
introduction of a highly virulent verotoxin-producing E. coli in their area or on their farms.
1
Introduction
16
1.2 The beginning of this story
Although the farmer in the previous story is fictional, the zoonotic pathogen
verotoxin-producing Escherichia coli serotype O157:H7 (VTEC O157:H7) is
often found on farms in relation to outbreaks of disease among humans, often
involving children. During the last 40 years the pathogen has emerged as an
important risk to public health due to the severe disease and the high risk of life-
threatening complications. In Sweden, the number of cases of gastrointestinal
disease in humans due to VTEC O157:H7 attributed to the cattle population
remains high despite national efforts to control transmission. Despite extensive
research and many scientific publications, important gaps in our understanding,
and therefore our ability to implement an efficient control program, remain.
Within this project a multidisciplinary approach, combining epidemiology,
microbiology, ethology and bioinformatics, is used to fill some of these gaps.
We also combine studies of different levels, i.e. between farms, within farm as
well as between and within animal, to create a comprehensive picture of
pathogen dynamics on Swedish farms. Our approach enables new perspectives,
for example on the role of animal behaviour in disease transmission and
exposure to the pathogen, but also supports and increases the confidence in
previously suggested risk factors and theories. But let us start from the
beginning.
1.3 Verotoxin-producing E. coli O157:H7
The nomenclature used to describe verotoxin-producing Escherichia coli
(VTEC) and disease caused by it can easily confuse anyone. For example, the
acronyms VTEC, EHEC and STEC are used frequently to describe the same
pathogen and an understanding of the history, and relationship between VTECs
and other E. coli is required to understand the differences between the acronyms.
As the name suggests, VTEC of serotype O157:H7 is an E. coli that have
acquired the ability to produce a particular toxin that is interchangeably called
verotoxin or shigatoxin (more about this below). As all E. coli, it is a gram-
negative, rod-shaped, facultatively anaerobeic bacterium belonging to the family
Enterobacteriaceae (Gally & Stevens 2017). It is distinguished as serotype
O157:H7 by diversity of the O-antigens (part of a lipopolysaccharide in the
outer membrane) and the H-antigens (flagellar proteins), a method used to
differentiate between different types of E. coli since the 1940s (Kauffmann 1947;
Orskov et al. 1977). It is an important member of the group enterohemoragic E.
coli (EHEC), a group of pathogenic E. coli able to cause bloody diarrhea and
severe complications (Kaper & Nataro 1998).
17
1.3.1 The rise of VTEC
In 1982, two unusual outbreaks of gastrointestinal disease characterised by
severe abdominal pain, bloody diarrhoea and little or no fever occurred in the
states of Oregon and Michigan in the United States. At least 47 people became
ill and epidemiological investigations revealed that the illness was associated
with eating hamburgers at restaurants belonging to the same fast-food chain
(Riley et al. 1983). The ill persons were infected with a rare type of E. coli of
serotype O157:H7, which did not behave as previously recognised
enteropathogenic E. coli (EPEC), and it was suggested that a yet unknown type
of enterotoxin may have caused the serious illness (Riley et al. 1983).
Riley et al. (1983) were indeed correct about an enterotoxin causing the
disease but it was not completely unknown at the time of the outbreak. In fact,
two research groups, working in parallel, had already come across it. In 1977,
Konowalchuk et al. found that a group of EPEC produced an unknown toxin
with the ability to kill vero cells. Due to this ability, it was named verotoxin.
Around the same time, O’Brien and LaVeck (1983) also identified a new toxin
produced by an EPEC (of serotype O26). They found that this toxin was very
similar to the toxin produced by the bacteria Shigella dysenteriae and therefore
this toxin was called shiga-like toxin (O’Brien & LaVeck 1983). After the
outbreak in 1982 it became clear that these toxins were the same and that the
same toxin was produced by the E. coli causing the outbreak (Johnson et al.
1983; O’Brien et al. 1983). However, both names are still being used
interchangeably in literature today and agreement on which would be most
appropriate to use has caused debate (Calderwood et al. 1996; Karmali et al.
1996). In Sweden, the term verotoxin has been traditionally used within the
veterinary field while shigatoxin has been the preferred term in human medicine.
To keep to tradition the term verotoxin (vtx) will therefore be used in this thesis.
After the outbreaks of severe disease in Oregon and Michigan, the
importance of VTEC was further acknowledged when Karmali et al. (1983)
linked verotoxin to the severe complication haemolytic uremic syndrome
(HUS), a syndrome characterised by trombocytopenia, hemolytic anemia and
kidney failure. Since then the history of outbreaks, disease and public health
costs has only enhanced the importance of VTEC worldwide and prompted a
large research interest (Kaper & O’Brien 2014). Thus, our knowledge and
understanding of the group of VTEC has increased substantially since the
outbreak in 1982, but many questions remain.
18
1.3.2 The origin of VTEC O157 and non-VTEC O157
As O’Brien and LaVeck (1983) observed, serotype O157:H7 is just one of
multiple serotypes of E. coli with the ability to produce verotoxin. However,
when Karmali et al. (2003b) classified VTECs into five seropathotypes, based
on incidence, involvement in outbreaks and association with severe disease
serogroup O157 stood out. Due to the high incidence and common
occurrence in outbreaks serotypes O157:H7 and O157:NM were the only
serotypes classified as seropathotype A (the most important/severe).
Phylogenetic analysis has also suggested that VTEC O157:H7 stands out from
other verotoxin-producing serotypes (often referred to as non-O157) (Whittam
1998; Hazen et al. 2013). The serotype O157:H7 and its inferred ancestor
O155:H7 is categorised as EHEC1, a group of relatively closely related strains
that separated from a common ancestor as long as 4.5 million years ago (Reid et
al. 2000). The group EHEC2 contains other serotypes able to cause disease (like
O26, O103) and these are less closely related (Abu-Ali et al. 2009). Instead, it
appears that this group has acquired their virulence factors in different ways and
at different time points (Reid et al. 2000).
Horizontal gene transfer allows distantly related E. coli to exchange genes
(through plasmids, phages and pathogenicity islands) between each other driving
adaptation to new environmental challenges (reviewed by Lawrence 2002 and
Dobrindt 2005). Genes frequently exchanged between bacteria are part of the
accessory genes, while stable genes (within family, species or subtypes) make
up the core genome. The genome of E. coli contains between 4200-5500 genes
and of these ~2000 genes are core genes, i.e conserved among all strains (Rasko
et al. 2008; Touchon et al. 2009; Kaas et al. 2012). Thus, the variable, i.e. the
accessory, genome makes up more than half of the genome. Variation in this part
of the genome is huge since these genes come from genepool of more than
18 000 genes (Touchon et al. 2009; Kaas et al. 2012). An example of the
importance of this type of evolution is the large German outbreak where an
enteroaggregative E. coli of serotype O104:H4 was able to cause severe disease
and HUS by acquiring the ability to produce verotoxin type 2 (as reviewed by
Denamur 2011).
1.3.3 Within serotype variation
The flexible and adaptive capabilities of E. coli also means that there can be
significant variation also within serotypes. For example strains can carry genes
coding for different types of verotoxins (Scheutz et al. 2012). Toxins are grouped
into two branches; verotoxin type 1 (vtx1) and type 2 (vtx2) (Scheutz et al.
2012). Vtx1 is very similar to the toxin produced by S. dysenteriae and is
19
generally associated with milder disease and fewer cases of the complication
HUS than vtx2, although there are exceptions (EFSA 2013). Some strains
produce both vtx1 and vtx2 and there are also subcategories within vtx1 and vtx2
(also associated with differences in virulence) that appear in different
combinations (Scheutz 2014).
Virulence is, however, complicated and producing virulent verotoxins is not
enough if other important virulence mechanisms are lacking. For example,
missing the locus of enterocyte attachment (LEE), which enables attachment to
host cells, may lead to inability to attach to cells and cause disease (McDaniel et
al. 1995). Hence, not all VTECs fulfil the criteria for being defined as an EHEC
(able to cause enterohaemoragic disease in humans). There are also examples of
strains that have caused disease without LEE (Kaper et al. 2004) which
emphasises the importance of genome flexibility and the pathogens potential to
find new ways of causing disease.
Analysis of Dutch clinical isolates of varying serotypes has shown that toxin-
type does not cluster with core genome, indicating that vtx production was
highly influenced by horizontal gene transfer (Ferdous et al. 2016). Reid et al.
(2000) observed the same pattern for vtx while the virulence factor LEE did
cluster with the core genome. Hence, there appears to be differences in the
importance of horizontal gene transfer and the role of common ancestors for
acquisition of different genes (Gordienko et al. 2013).
Variation within O157:H7
An overview of the most well recognised virulence factors and localisation in
the genome of VTEC O157:H7 are presented in Figure 1. However, as in other
serotypes there is variation also within serotype O157:H7. By developing a
single-nucleotide polymorphism (SNP) typing system, Manning and colleagues
(2008) were able to distribute 519 VTEC O157:H7 strains into nine evolutionary
clades with different associations with human disease. Although many clades
were associated with human outbreaks, for example clade 3 was behind the first
outbreak in Michigan and Oregon, one stood out with higher rates of
hospitalisation and frequency of the complication HUS; the clade 8 lineage.
Analysis of isolates from human cases infected in Sweden, collected between
2008-2011, has similarly shown that a high proportion of the persons had been
infected with clade 8 and that 10 out of 11 cases of HUS were caused by clade
8 (Söderlund et al. 2014). The association between clade 8 and severe disease
has been suggested to be due to overexpression of vtx2 (Neupane et al. 2011).
Studies suggest that acquisition and loss of virulence genes is highly dynamic
within VTEC O157:H7, leading to development and regression of pathogenic
strains (Kyle et al. 2012; Dallman et al. 2015; Byrne et al. 2018).
20
1.3.4 Classification, identification and characterisation of VTEC
O157:H7
Much of the recently gained understanding of the importance of within and
between serotype variation has been possible thanks to more advanced analytical
methods. There are now multiple options for detection and characterisation
depending on which depth of information is desired. Generally, methods of
analysis can be divided into culture-based methods, immunological methods and
molecular methods.
Detection/culture based methods
Culture of bacteria has been historically important and remains the gold
standard for establishing presence of viable VTEC O157:H7 in a sample. VTEC
O157:H7 will, like other E. coli, grow on ordinary blood agar but when the goal
is to detect O157:H7, agars that use this strains unique biochemical properties
(inability to ferment sorbitol or produce β-glucuronidase) are helpful (Ojeda et
al. 1995; Kaper & Nataro 1998).
Figure 1. Important virulence factors of verotoxin-producing Escherichia coli O157:H7 and their
localisation in the bacterial genome. Source: Croxen & Finlay 2010; Mellies & Lorenzen 2014;
Gally & Stevens 2017. Illustration: Lena-Mari Tamminen.
21
One of the most common is Sorbitol MacConkey (SMAC) agar. This agar is
selective for the family Enterobacteriacae and the non-sorbitol fermenting
colonies of VTEC O157:H7 are distinguishable by their lack of colour compared
to other sorbitol fermenting E. coli. However, strains of VTEC O157:H7 able to
ferment sorbitol have been identified and there are VTECs of other serotypes
than O157:H7 that are not able to ferment sorbitol (Gunzer et al. 1992; Schmidt
et al. 1999). Adding cefixime and tellurite (CT-SMAC) inhibits growth of non-
vtx producing E. coli and increases rate of isolation of VTEC O157:H7 from
cattle, but there are also indications that it may inhibit the growth of some strains
of VTEC O157 (Zadik et al. 1993; Karch et al. 2005). Similarly, novobicin can
increase selection of O157:H7 (Okrend et al. 1990).
Another alternative is to use a chromogenic agar (like CHROMagar or
rainbow agar). This agar uses the inability of E. coli O157 to produce β-
glucuronidase. VTECs, including VTEC O157:H7, and non-vtx producing E.
coli can be differentiated by colour (vtx-positive isolates are mauve-coloured)
(Kaper & Nataro 1998; Hirvonen et al. 2012). However, sorbitol fermenting
strains of VTEC O157:H7 may not grow on this type of agar either (Hirvonen et
al. 2012). As there are no completely selective agars for O157:H7 it is
recommended that further confirmation of suspected colonies is carried out after
culture (Kaper & Nataro 1998).
Phage typing is a culture-based method that has been extensively used to
subtype VTEC O157:H7. Phage type (PT) of a strain is determined by culturing
the strain on an agar diffused with different lytic bacteriophages to produce a
susceptibility profile (Van der Merwe et al. 2014). Different PTs have been
associated with different virulence in VTEC O157:H7 (Lynn et al. 2005; Mora
et al. 2007).
Immunological methods
There is a variety of immunological methods that can be used in different steps
of analysis of VTEC O157:H7. For example, sensitivity in culture-based
methods can be further increased by using immunomagnetic separation.
Antibody coated paramagnetic beads are used to bind, pick up and separate
O157:H7 from other bacteria in a sample before plating on for example SMAC
agar (Karch et al. 1996). This concentrates the O157:H7 in the sample and
reduces the risk of other bacteria present in the sample outcompeting them.
For confirmation of cultured colonies multiple assays and commercial kits
are available, ranging from ELISAs, latex reagents and labelled antibodies, that
detect surface proteins (mainly O and H antigen) but also vero-toxins (Kaper &
Nataro 1998). Some of the ELISAs can be used directly on fecal samples and
thereby saves time (Dylla et al. 1995; Park et al. 1996). The downside is that
22
there is a risk for cross-reactivity with other closely related bacteria, like
Citrobacter freundii, Escherichia hermanni and Salmonella Urbana (Park et al.
1996). Bacterial vtx-production may on the other hand be influenced by for
example culture conditions (Boone et al. 2016).
Molecular methods for characterisation and subtyping
Molecular methods analysing bacterial DNA are, with decreasing prices and
increased availability, becoming the standard for detection and characterisation
of VTEC within research, diagnostics and outbreak situations (Newell & La
Ragione 2018). Already they have provided us with the increased understanding
of the variation within serotypes and virulence mechanisms described above as
well as possibilities for detailed investigation of outbreaks as well as bacterial
phylogeny (Eppinger & Cebula 2015; Land et al. 2015). Although a detailed
description of these methods is beyond the scope of this thesis, an overview of
some of the common molecular methods and level of detail they provide is
presented in Figure 2.
Figure 2. An overview of four common molecular methods used to characterise verotoxin-
producing Escherichia coli O157:H7. Source: Söderlund (2015). Illustration: Lena-Mari
Tamminen.
23
1.4 VTEC and human disease
VTEC has been estimated to cause 2 801 000 acute illnesses per year worldwide
but the estimated incidence between regions varied significantly (1.4 to 152
cases per 100 000 inhabitants) (Majowicz et al. 2014). The subregions (as
defined by the World Health Organization) with highest estimated incidence
were EMR B and EMR D2. Countries that have reported a high incidence, i.e.
cases per 100 000 inhabitants, in 2018 were for example New Zealand (18.9)
(ESR 2020) and Ireland (20.0) (ECDC 2020). The Swedish incidence of VTEC
during 2018 was 8.7, the third highest European notification rate reported to the
European Centre for Disease Control (ECDC). This continues the increasing
trend in incidence of cases that has been observed since 2006
(Folkhälsomyndigheten 2020). In addition, 40 cases of HUS (the highest number
of annual cases reported so far) were reported in 2018 and half of these occurred
in children less than 10 years of age (Folkhälsomyndigheten 2020). A study has
estimated that the economic burden (sum of direct and indirect costs) of VTEC
in Sweden during 2006 was 1.3 million euros for VTEC. Compared to the cost
of Campylobacter, a more common cause of gastrointestinal disease estimated
to cost 26.1 million euros, this is relatively small. However, although the total
public health burden is larger for Campylobacter, the burden per case, i.e. the
consequences for an individual infected as well as the cost per case, is much
higher for VTEC due to the potentially severe consequences of infection
(Toljander et al. 2012).
Although other serotypes can cause disease, VTEC O157:H7 is the serotype
most commonly associated with human disease and was the most commonly
reported serotype in Sweden during 2018 (EFSA 2013; Folkhälsomyndigheten
2020). In a Swedish study of VTEC in Jönköping county, O157:H7 was found
to be the dominating serotype causing bloody diarrhoea (Bai et al. 2018) and the
subtype clade 8 has been identified as the major cause of cases with the
complication HUS (Söderlund et al. 2014). In addition to sporadic cases, often
associated with farm visits or contact with cattle faeces, clade 8 has caused large
national outbreaks with high proportion of infected persons developing HUS
(Table 1).
2. EMR B: Bahrain, Cyprus, Iran (Islamic Republic of), Jordan, Kuwait, Lebanon, Libyan Arab
Jamahiriya, Oman, Qatar, Saudi Arabia, Syrian Arab Republic, Tunisia, United Arab Emirates;
EMRD: Afghanistan, Djibouti, Egypt, Iraq, Morocco, Pakistan, Somalia, Sudan, Yemen.
24
Table 1. Overview of larger outbreaks of verotoxin-producing Escherichia coli in Sweden during
the course of this project. Source: Folkhälsomyndigheten (2020).
Year
Number
of cases
HUS
cases
Type of VTEC
Source
2018
(July
-September)
116
14
(12 %)
O157:H7
Clade 8
(stx2a, stx2c, eae)
Unknown/Foodborne*
2016
2017
(Sept
ember-
February)
26
6
(23%)
O157:H7
Clade 8
(stx2a, stx2c, eae)
Meat (Cattle)
2016
(July
-September)
8
3
(38%)
O157:H7
Clade 8
(stx2a, stx2c, eae)
Farm contact (Cattle)
2015-2016
(November
May)
70
0
O103:H2
(
stx1, eae
)
Unknown/Foodborne
2015-2016
(September-April)
57
0
O26:H11
(stx1a, eae)
Unkown/Foodborne
*Outbreak also included person to person spread by e.g. recreational swimming
1.4.1 Pathogenesis in human disease
Only a small number of virulent VTEC is a sufficient infectious dose for disease
in humans (Griffin & Tauxe 1991; Newell & La Ragione 2018) and post-
outbreak calculation has suggested that even less than 50 bacteria may be enough
(Tilden et al. 1996). Symptoms vary between no signs of infection, abdominal
pain, mild or bloody diarrhoea to the severe complication HUS, a potentially
fatal syndrome including trombocytopenia, hemolytic anemia and kidney failure
(Karmali et al. 1983; Tarr et al. 2005).
While anyone can be infected by VTEC O157:H7 it is generally, as in the
introductory story, children and elderly that are most susceptible to the
complication HUS (Gould et al. 2009). The serious complication is a result of
the verotoxins entering the bloodstream and binding to the Gb3 receptor on the
surface of host cells, for example on platelets. This leads to cell damage,
secretion inflammatory chemokines and cytokines as release of thrombin which
activates thrombosis (Karpman & Ståhl 2014). In the kidney this causes
glomerular cell damage and in some cases the toxin can cause severe neurologic
dysfunction (Karpman & Ståhl 2014). A summarized description of the
pathogenesis, course of infection and treatment options is presented in Figure 3.
25
Figure 3. Infection of verotoxin-producing Escherichia coli O157:H7 in humans. Modified from
Karpman & Ståhl (2014), Tarr (2005) and Trachtman et al. (2012). Illustration: Lena-Mari
Tamminen.
1.4.2 The link between cattle and human disease
Cattle are considered the main reservoir of VTEC O157:H7 and a pattern where
isolates from cattle are closely related to clinical human isolates has been
observed in many studies (Zhang et al. 2007; Bono et al. 2012; Franz et al. 2012;
Jung et al. 2013; Strachan et al. 2015; Arimizu et al. 2019). Phylogeographic
analysis of bovine and human isolates from four continents suggests that the
common ancestor of O157:H7 arose in the Netherlands around 1980 and was
then spread around the world in a pattern that fits very well with trade routes of
cattle (Franz et al. 2019). According to the analysis by Franz et al. (2019), VTEC
O157:H7 from Europe was introduced in Sweden in 1982 and that the virulent
clade 8 type was introduced from the United States around 1990. This
corresponds relatively well with the first human case of VTEC O157:H7 in
Sweden, which occurred in 1988. The first case was followed by a small number
of sporadic cases until the first large outbreak occurred in 1995 (Ziese et al.
1996).
There is also a relationship between cattle density and cases of VTEC
O157:H7 in humans, and a higher risk of HUS in animal contact related
26
outbreaks has been observed in the United States (Kistemann et al. 2004; Frank
et al. 2008; Heiman et al. 2015). In addition to the examples presented in table
1, several large Swedish outbreaks have been associated with the Swedish cattle
population. For example, the so far largest outbreak of VTEC O157:H7 (135
cases, 11 HUS) was caused by lettuce contaminated by cattle pasturing upstream
of the water irrigation point and in 2002 an outbreak with a very high incidence
of HUS (12/39 cases) was cause by contaminated beef sausages (Sartz et al.
2008; Söderström et al. 2008). The role of controlling VTEC O157:H7 in cattle
for preventing human cases is emphasized in the national strategy for prevention
of EHEC signed by the Swedish Public Health Agency, the Board of Agriculture,
the National Food Agency, SVA, as well as the National Board of Health and
Welfare (Socialstyrelsen, 2014).
However, not all variants of O157:H7 present in the cattle population appear
to cause problems. Söderlund et al. (2014) observed more variation in isolates
from the cattle population compared to clinical cases. Other studies have also
observed lineages from cattle that do not appear to be associated with human
disease (Zhang et al. 2007; Bono et al. 2012; Franz et al. 2012). Analysis of
genomes of bovine and human isolates of O157:H7 has indicated that only 10 %
of bovine isolates have zoonotic potential (Lupolova et al. 2016).
There are also indications that exposure to cattle may have a protective effect
against disease caused by VTEC O157:H7. For example, increased immunity
and less clinical disease in farm resident children compared to non-farm resident
children has been observed in the United States (Belongia et al. 2003). A larger
proportion of rural populations have been shown to have antibodies against vtx
compared to urban populations but it is unknown to which extent these
antibodies represent exposure to pathogenic O157:H7 (Haack et al. 2003;
Karmali et al. 2003a). Haack et al. (2003) argued that the strain must be
pathogenic to evoke an antibody response, but considering the variation of
virulence within the VTEC it is possible that mild infections may induce
antibody protection that reduces the risk of severe infection. Indeed less
pathogenic lineages carrying vtx2c have been identified from both cattle and
healthy people (Kawano et al. 2012).
1.4.3 Transmission of VTEC O157:H7
From being considered a food borne pathogen, first associated with undercooked
meat, many pathways of transmission for VTEC O157:H7 have been recognised
(Figure 4). Large outbreaks are often foodborne and associated with cattle
products, like meat and unpasteurized milk, or contaminated vegetables, such as
salad or spinach (Michino et al. 1999; Howie et al. 2003; Grant et al. 2008;
27
Heiman et al. 2015). Sporadic cases often occur through contact with shedding
animals, an environment where animals have been or contact with cattle faeces
in other ways (Locking et al. 2001; Crump et al. 2002). A recent meta-analysis
focusing on sporadic cases identified raw/undercooked meat (population
attributable fraction 19%), person to person spread (15%), contact with animals
(14%) and visiting farms (12%) as the most important routes (Kintz et al. 2017).
Although food borne outbreaks tend to be dramatic and involve many cases, the
risk visiting a pasture has been estimated to be associated with a 100 times
greater risk compared to eating a burger (Strachan et al. 2006). Spread through
vegetables and green leafs have also been highlighted to be a particularly
important public health risk as these products are often consumed raw which
increases the number of live bacteria ingested (Griffin & Karmali 2017). Tarr et
al. (2018) found that some lineages of VTEC were more likely to spread through
raw milk while other types were more often associated with transmission
through vegetables and fruits. These differences may reflect bacterial ability to
survive in different types of environment and changes in the bacterial population
may lead to changes in the routes of transmission.
Figure 4. The multiple routes of transmission for verotoxin-producing Escherichia coli O157:H7
between cattle and humans include contaminated products as well as indirect transmission through
the environment or direct contact with cattle. Modified from Chapman et al. (2018).
Illustration: Lena-Mari Tamminen.
28
1.5 VTEC O157:H7 in the cattle population
As described in the pathogenesis section (p 24) vtx binding to Gb3 receptors are
responsible for causing symptoms in humans (Schüller et al. 2007). Cattle have
been suggested to have a different distribution of Gb3 receptors than humans
(Pruimboom-Brees et al. 2000) which may be the reason why colonisation rarely
causes symptoms in cattle (Kolenda et al. 2015). However, young calves
infected with high doses of VTEC O157:H7 can develop diarrhoea and
colonisation of the intestine induces a local inflammation and damage to
intestinal cells (Dean-Nystrom et al. 1997; Nart et al. 2008). Thus, the bacteria
should not be considered a commensal part of the microbiota of cattle.
Prevalence of VTEC O157:H7 among Swedish cattle has been investigated
several times by collection of faecal samples from cattle at slaughterhouses in
different regions. The first study, performed between 1996 and1997, found that
1.2 % of the slaughtered cattle were positive for the pathogen (Albihn et al.
2003). In samples collected between 1998 to 2000 a higher proportion, 8.3 % of
the sampled cattle were positive (Eriksson et al. 2005). Marked regional
differences were observed, especially the county of Halland stood out with a
prevalence of 23.3% combined with the highest incidence of human cases. A
following longitudinal study, comparing farms in four regions of Sweden
between 2009 and 2013, found regional differences in prevalence of clade 8
(only found in Falköping and Halland) (Widgren et al. 2015). Prevalence of
positive faecal samples at slaughter since 2000 has varied between 2.2-3.5%. In
the 2014-2015 and the 2017-2018 sampling, strains belonging to clade 8 were
only found in samples from the counties Öland and Skåne, indicating that the
virulent strain had moved to a new region (Erik Eriksson, Swedish National
Veterinary Institute).
1.5.1 Farm dynamics
Between farm transmission over large distances appear to be driven by trade
of cattle and, just as VTEC O157:H7 spread across the world through common
cattle trade routes, modelling suggests that it spread along trade routes within
Sweden (Widgren et al. 2016; Franz et al. 2019). In addition,
purchase/introduction of new animals has been identified as a risk factor for
establishing the pathogen on farms in Sweden and other countries (Schouten et
al. 2004; Herbert et al. 2014; Widgren et al. 2015). However, there are also signs
of local transmission, as infected neighbouring farms increases the risk of a farm
being positive and nearby farms often share related strains (Zhang et al. 2010;
Herbert et al. 2014; Widgren et al. 2015).
29
Prevalence of VTEC O157:H7 among cattle often follows a seasonal pattern.
In Sweden a peak during summer and fall (July-November) has been observed
in faecal sampling at slaughterhouses (Albihn et al. 2003). This corresponds well
to the seasonal pattern observed in Scotland and the Netherlands which are
countries with comparable climate to Sweden (Schouten et al. 2004; Gunn et al.
2007; Smith et al. 2016; Henry et al. 2019). Studies performed in the United
States and Australia have identified that climate related variables, such as
temperature, relative maximum soil temperature, wind speed, humidity and rain
also influences shedding and prevalence of VTEC O157:H7 (Williams et al.
2014; Benjamin et al. 2015; Lammers et al. 2015)
Although a majority of farms appear to clear infection in 4-6 months after
introduction, there are many examples of farms that remain positive for VTEC
O157:H7 over long periods (Hancock et al., 1997; Herbert et al., 2014; Rice et
al., 1999; Widgren et al., 2015). Strains of VTEC O157:H7 that are more likely
to cause human disease also appear more likely to persist on farms (Carlson et
al. 2009; Herbert et al. 2014). Dairy farms have been observed to have a higher
prevalence and remain positive for a longer time period compared to beef farms
(Cobbaut et al. 2009; Widgren et al. 2015; Smith et al. 2016).
Larger herds (total number of animals) also appear to be associated with risk
of infection according to several studies (Eriksson et al. 2005; Herbert et al.
2014; Benjamin et al. 2015). However, there are some studies that have shown
the opposite or no effect (Wilson et al. 1993; Cobbaut et al. 2009; Cho et al.
2013). Other examples of herd level risk factors that have been suggested are
wild bird density (starlings and geese), presence of pig on the farm and spreading
slurry on grazing lands (Synge et al. 2003; Eriksson et al. 2005; Gunn et al.
2007; Cernicchiaro et al. 2012).
On infected farms, VTEC is often found among younger animals (Eriksson
et al. 2005; Kuhnert et al. 2005). Contacts between adult animals and calves as
well as other groups of animals on farm is associated with increased risk and
keeping groups of animals together has been suggested to be the most cost
effective on-farm measure to reduce prevalence of the pathogen (Ellis-Iversen
et al. 2008; Cernicchiaro et al. 2012; Lyons et al. 2013). Other common on-farm
measures suggested to reduce prevalence of VTEC O157:H7 on infected farms
include improving hygiene, like maintaining dry bedding and reducing faecal
contamination of bedding and water troughs (Lejeune & Wetzel 2007; Ellis-
Iversen et al. 2008; Tamminen et al. 2018). A major cleaning of the barn has
been shown to decrease vtx2 found in milk filters on Finnish dairy farms
(Jaakkonen et al. 2019). However, it should also be noted that washing may also
spread the pathogen among animals inside the barn, for example flushing
30
alleyways with water has been associated with an increased risk (Garber et al.
1999).
Other interventions to reduce presence of VTEC O157:H7 on dairy farms
include changing dietary practices, adding feed additives (e.g. probiotics), phage
therapy and vaccination, but so far their impact remain limited (Besser et al.
2014).
1.5.2 Colonisation and shedding
As in humans VTEC O157:H7 colonises the large intestine of cattle. More
specifically the lymphoid dense tissue of the rectoanal-junction (Naylor et al.
2003). Thus, the bacteria has to survive through the passage of the
gastrointestinal tract, including the rumen, as well as compete with members of
the microbiota during passage and at the colonisation site (reviewed by
Ducarmon et al. 2019). This process is highly dependent on cues from the host
and other members of the microbiota as recently reviewed by (Pifer & Sperandio
2014).
Colonisation of the terminal rectum is associated with increased shedding
levels (Low et al. 2005; Davis et al. 2006) and colonised animals have been
suggested to be responsible for shedding a large proportion of all VTEC
O157:H7 shed into the environment. Both mathematical modelling of Scottish
data and a study of fecal shedding at slaughter have suggested that a small
proportion of colonised animals (<10 %) are responsible for 95-99% of all
VTEC O157:H7 shedd into the environment (Omisakin & MacRae 2003;
Matthews et al. 2006b). Shedding levels higher than 103 and 104 colony forming
units (cfu)/gram feces have been suggested to indicate “super-shedding” due to
colonisation (reviewed by Chase-Topping et al. 2008). The high number of
bacteria shed by super-shedders increases the risk of transmission of VTEC
O157:H7 to humans by increasing hide and carcass contamination of groups of
animals (Cobbold et al. 2007; Stephens et al. 2009). These high shedders have
also been suggested to drive transmission to other animals in the pen which leads
to new animals becoming colonised and keeping the pathogen circulating within
farms (Matthews et al. 2006a; Cobbold et al. 2007; Spencer et al. 2015; Widgren
et al. 2018).
In addition, colonisation has been associated with an increased duration of
shedding (Rice et al. 2003; Cobbold et al. 2007; Lim et al. 2007). However,
longitudinal studies with more frequent sampling have reported intermittent
shedding, with daily variation, of colonised animals (Robinson et al. 2004;
Lammers et al. 2016). A proposed reason for the intermittent shedding is that
colonisation is associated with formation of biofilm, which, when it has become
31
large enough, is released in chunks - so called biofilm sloughing (reviewed by
Munns et al. 2015). This would explain how cattle could shed high levels of the
bacteria on one sampling whereas later the same day it is not detected at all.
However, due to the short duration of super-shedding observed in some studies,
the role of colonised and super-shedding individuals in pathogen transmission
and persistence has been questioned (Munns et al. 2014; Williams et al. 2015).
It has been suggested that strains that are more virulent to humans are also
better equipped to colonise cattle intestines and are shed at higher levels (Chase-
Topping et al. 2007; Carlson et al. 2009).
1.5.3 Individual heterogeneity and similarity
Environmental exposure of VTEC O157:H7 influences the risk of colonisation
and shedding in cattle and studies have reported a synchronised increase of
shedding on group level, so called super-shedding events (Williams et al. 2014;
Lammers et al. 2015). However, there is considerable differences in shedding
levels and duration of shedding during these reported events and other studies
show similar heterogeneity (Robinson et al. 2004; Jonsson et al. 2009; Sheng et
al. 2016).
The heterogeneity between individuals exposed to the same environment and
same strain of VTEC O157:H7 indicates that there are host related differences
influencing colonisation (as reviewed by Munns et al. 2015). To some extent,
variation may be related to amount of bacteria the animal is exposed to as a
higher dose is associated with increased risk of colonisation (Sheng et al. 2016).
However, even a low dose can lead to high shedding levels in some individuals
(Besser et al. 2001). It also appear that the number of shedding periods vary. In
a longitudinal study of beef cattle followed for a year from 4-6 months of age,
82 % of the animals shed only once while other animals shed longer or multiple
periods (Rhades et al. 2019).
The colonisation process of VTEC O157:H7 is complex and dependent on
signals from the host and other member of the host microbiome (Pifer &
Sperandio 2014; Baümler & Sperandio 2016). Shedding and colonisation of
VTEC O157:H7 has been associated with lower diversity of gut microbiota and
differences in host gene expression in the terminal rectum which may explain
some individual variation observed (Xu et al. 2014; Mir et al. 2016; Wang et al.
2017). However, considering the close interactions between microbiota, VTEC
O157:H7 and enterocytes (including the immunomodulatory effects of the
pathogen), it cannot be excluded that these differences are in fact a result of
colonisation. Bacteriophages, i.e. viruses that infect bacteria, may also influence
32
shedding dynamics. Hallewell et al. (2014) sampled 6 super-shedders and 5 low-
shedders daily during 5 weeks and found that low-shedder had higher prevalence
of phages and more T4-like phages, which had strong lytic abilities against
O157:H7.
Modelling suggest that both between animal variability (a small proportion
of more susceptible animals) or within animal variability (all animals have
potential to shed but variability arises from transmission dynamics) can generate
similar patterns as observed in observational studies (Chen et al. 2013). If the
latter is true, all animals may have potential to become colonised or super-
shedders at some time or in some settings and instead of looking for the
subpopulation of colonised individuals focus should be on when animals become
colonised. It has also been proposed that stress hormones, like noradrenalin, may
have direct effects on the pathogen, promoting colonization and shedding (as
reviewed by Freestone et al. 2008). Increased risk of shedding has also been
observed following weaning, long haul transportation and feed deprivation
(Cray & Casey 1998; Rugbjerg et al. 2003; Bach et al. 2016). There are also
unexplored potential explanations for heterogeneity observed between animals.
Modelling has linked heterogeneity in social contacts between calves to
transmission dynamics (Turner et al. 2008). In addition, personality and animal
wellbeing has impacts on animal behaviour, and thereby how it handles changes
and stress induced by for example transportation, weaning or other factors
(Wiepkema et al. 1987; Sapolsky 1994; Lecorps et al. 2018; Neave et al. 2018).
.
33
The transmission and dynamics of VTEC O157:H7 in cattle are complex and
involve risk factors associated with host, pathogen and environment. The aim of
this thesis is to increase the understanding of this complexity and the role of
these factors in pathogen persistence, transmission and host susceptibility. The
overall goal is to synthesise the complexity and identify target areas for
preventive on farm measures to reduce prevalence of VTEC O157:H7.
The specific objectives were:
¾ To identify factors related to transmission of VTEC O157:H7 between farms
¾ To describe prevalence and dynamics of VTEC O157:H7 in dairy herds with
known presence of the pathogen
¾ To identify drivers of colonization of individual animals
¾ Explore the hypothesis that chronic stress increases susceptibility to
colonisation
2
Aims and objectives
34
35
This section will be used to provide an overview, as well as comments and
reflections on the methods used across all studies with references to the papers
where applicable. Detailed descriptions of materials and methods can be found
in the respective papers.
3.1 Study design
The focus of paper I was farm persistence and between farm transmission and
here 80 farms on the island of Öland, a region where VTEC O157:H7 had been
recently detected at the start of this project, were selected by convenience and
sampled twice. Information about herd characteristics and between sampling
activities were collected by postal questionnaires around the time of the second
sampling. For paper II-IV individual sampling of calves from 12 dairy farms
where presence VTEC O157:H7 had been established by environmental samples
was performed. In paper II, transmission dynamics between two sampling
occasions were evaluated in relation to pen-level risk factors, age and sex of
animals. Paper III focused on the first individual sampling and animal-based
indicators and behaviour to identify differences between calves colonised by
VTEC O157:H7 and those not. Non-colonised individuals housed with
colonised individuals were compared to control for environmental factors and
explore individual differences. In paper IV results of hair cortisol samples
collected from the animals were used to explore the associations between animal
based indicators and stress. Before the first individual sampling, a thorough
environmental sampling to identify groups of animals shedding VTEC O157:H7
was performed. The different parts of the study with references to the respective
papers are outlined in Figure 5.
3
Overview and comments
on materials
and methods
36
Figure 5. Overview of sampling performed within the project.
3.2 Detection and enrolment of participating farms
Identification of farms with VTEC O157:H7 is difficult, as animals do not show
any symptoms of carrying the pathogen nor is there any routine sampling of the
bacteria on farms, except in association with epidemiological investigations of
human cases suspected to be connected to specific farms. We were also
interested in focusing on strains of O157:H7 able to cause disease in humans
(like clade 8) and not the less virulent strains that appear in the cattle population.
As a previous study has shown that farms often clear strains of O157:H7 over
time (although some farms remain positive over longer periods) (Widgren et al.
2015), detecting and enrolling farms with pathogenic strains was a significant
challenge.
37
Thanks to the national slaughter prevalence study performed by the Swedish
Veterinary Institute (SVA) around the start of the project we knew that clade 8
had recently been detected on the island of Öland. There was also some
additional information available about clade 8 prevalence on Öland as the
company Farm and Animal Health (Sv. Gård och Djurhälsan) had performed
environmental sampling of 80 farms in the spring of 2014 for initiating a study
on vaccine efficiency. However, soon after initiation, the vaccine trial was
cancelled due to high mortality in the first vaccinated calves. At this point we
were invited to collaborate and together with the Farm and Animal Health the
study on transmission dynamics presented in paper I was developed.
As described in Paper I, the 80 farms in the study were recruited by the local
livestock association (VÄXA). The staff visited farms across the island and
combined sampling in the project with routine visits. They continued recruiting
until 80 farms were reached and these farms were sampled in spring and fall
2014. The help from the local staff in encouraging farmer participation was
invaluable as there was some hesitance to participate due to the risk of being
associated with having VTEC O157:H7.
The close collaboration with Farm and Animal Health was also crucial for
recruiting farms to the studies in paper II-IV. During the time of the project Farm
and Animal Health conducted sampling of farms in the regions Öland, Skåne
and Blekinge as well as of farms associated with human cases in other regions.
Through these samplings, farms where the pathogen had already been identified
could be enrolled in the project. To increase the number of farms in the project
we also contacted farmers on Öland in areas where Farm and Animal Health was
not monitoring the infection status as well as an area in Falköping where several
farms positive for clade 8 had been previously identified (Widgren et al. 2015).
Farms were continuously recruited and sampling carried out between fall 2015
and spring 2017. Farmers were first contacted by phone (after permission to
share contact details had been procured by Farm and Animal Health or VÄXA)
and informed about the project. If environmental sampling had not already been
performed in other projects this was scheduled. To save time the initial
environmental sampling was in many cases performed by staff from VÄXA
according to standard protocol for identifying farms connected with human
outbreaks provided by SVA.
38
3.3 Sampling of VTEC O157:H7
3.3.1 Sampling to identify positive farms
Environmental sampling in paper I, and the initial environmental sampling of
farms in paper II-IV to establish presence of VTEC O157:H7, consisted of
pooled pat samples and overshoe samples from young stock (6-12 months) and
calves (<6 months). Sampling was done as previously described and validated
by Widgren et al. (2013). In short, overshoe samples were collected by walking
around the pen area with a gauze soaked with phosphate buffered saline (PBS)
fitted over plastic overshoes (Figure 6). While walking around the pen a pooled
pat sample was collected by sampling from 15-20 fresh faecal pats around the
pen. Samples were sent to SVA by postal service.
Figure 6. Overshoe samples used in the environmental sampling.
On farms positive for VTEC O157:H7 that agreed to take part in the individual
study (paper II-IV) a more thorough sampling was performed in the pens of all
groups of animals to identify in which buildings/pens the pathogen was present.
At minimum, separate overshoe samples from the pens of non-weaned calves,
weaned calves, young stock and adult animals were taken. If any of these groups
consisted of more than 50 animals housed in different pens or groups of animals
39
housed in separate buildings additional samples were taken. Overshoe samples
were sent to SVA by postal service or kept on ice and transported by car.
3.3.2 Individual sampling
After results from the second environmental sampling were analysed, the farm
was visited for sampling of individual animals. Up to 30 animals were sampled
on this visit and pens from which positive environmental samples had been
acquired were targeted. A sample size calculation based on results from pilot
sampling on a positive farm, which indicated that the prevalence of colonised
animals was 15 %, specified that minimum 21 animals should be sampled
(assuming test sensitivity 0.9) to identify at least one colonized animal and 20
animals was defined as a minimum number of animals sampled in a group.
However, on some farms the number of animals in pens indicated as positive by
environmental sampling was smaller than 20. In these cases all animals were
sampled. Sampling in larger groups was systematically randomised. On arrival
the calf standing nearest to the observer was selected and then every second or
third calf (depending on total number of animals in pen) was selected for
sampling.
The aim was to sample animals from all pens that had been positive in the
environmental sampling. However, there were practical constraints that
prevented sampling from being carried out as planned on some farms. Generally
animals were restrained either individually (using feeders and other structures in
the pen) or in groups. If available on the farm temporary fencing panels were
brought into the pens. However, on some farms the pens had no practical or safe
way of restricting animal movement for sampling. This problem mostly occurred
for older and larger animals housed in large groups in pens without possibility
to reduce the accessible pen area. In one case sampling of younger animals could
not be performed as animals managed to jump out of the pen. When it was not
possible to approach and restrain animals without risk of injury to the animals
and samplers, sampling in the pen was aborted/not performed.
First a fecal sample was collected from the rectum of the calf and placed in a
plastic jar. Following this, a foam coated cotton swab was used to swab the recto-
anal junction approximately 2-5 cm from the rectum. The area was swabbed
for 1 minute before the recto anal mucosal swab (RAMS) was put into a Falcon
tube with 2.5 ml of sterile phosphate buffer. Samples were then stored in a cooler
and either sent by postal service or transported in the cooler to the SVA for
analysis the following day. Sampling was started in the younger groups of
animals and plastic gloves were changed between each calf.
40
Follow up sampling
Animals that were positive for VTEC O157:H7 in the first individual sampling
were resampled again after about 5 weeks. On this occasion, 2-3 controls
(previously negative individuals) in the same age as the positive animals and
housed with a colonised animal were also sampled in the same way as described
for the first sampling.
3.3.3 Motivation for targeted sampling
The motivation for using a targeted sampling regime, both to identify positive
farms as well as identify positive groups of animals was dual. The obvious
reason was to improve cost efficiency and increase the probability of identifying
positive farms as well as colonised and shedding animals. In addition, the idea
was to enable comparison between farms and animals that were at risk of being
infected (and avoid introducing noise by including farms and individuals that
had not exposed to the pathogen). For Paper I this was achieved by limiting the
study population to farms on Öland where the pathogen was present and
transmission ongoing. Similarly, to avoid inclusion of unexposed individuals the
study population in the second part of paper II and paper III was narrowed down
to only include animals that were housed together with colonised animals.
3.4 Microbial analysis of VTEC O157:H7
3.4.1 Detection in environmental samples
Environmental samples (pair of overshoe samples or pooled faecal samples)
were pre-enriched in modified tryptic soy broth (mTSB) supplemented with
novbiocin (20 mg/ml) for 18-24 hours in 41.5 °C ± 0.5 °C before
immunomagnetic separation (IMS) (Dynabeads anti-E. coli O157;
Dynal/Thermo Fisher) was performed. Paramagnetic beads were then plated on
CT-SMAC agar (0.05 mg/l cefixime and 2.5 mg/l of potassium tellurite) and
incubated in 37 °C for 18-24 hours. Analysis started within 2 days of sampling.
Due to logistic reasons, there were farms where the collection of the first
environmental sampling and the more thorough sampling was performed at the
same time. In these cases the additional samples were kept in 2°C while the first
samples were analysed and only analysed if the first samples were positive for
VTEC O157:H7.
41
3.4.2 Detection in individual samples
The RAMS and fecal samples collected in the study were handled differently.
RAMS were vortexed for 1 minute and 20 ml of mTSB supplemented with 20
mg/l novobiocin was added to 2 ml of the sample. After pre-enrichment (for 18-
24 hours in 41.5°C ± 0.5C°) IMS using paramagnetic beads (Dynabeads anti-E.
coli O157; Thermo Fisher) was performed and the beads were plated on CT-
SMAC agar (0.05 mg/l cefixime and 2.5 mg/l potassium tellurite) and incubated
for 18-24 hours in 37°C. Confirmation on two isolates per farm was performed
using latex agglutination and PCR as described above.
Fecal samples on the other hand were stored in 2°C during analysis of the
RAMS (approximately 2-3 days). Samples from RAMS positive calves were
analysed by direct plating and to enable quantification of shedding levels a
tenfold dilution of 10 grams of feces was made and plated on CT-SMAC agar.
Plates were then incubated for 18-24 hours in 37°C.
The combination of pre-enrichment and IMS for the RAMS was used to
achieve a high sensitivity and reduce the risk of false negative calves (Rice et al.
2003; Davis et al. 2006). As RAMS positivity has been shown to be correlated
with positive fecal samples (Rice et al. 2003; Greenquist et al. 2005; Davis et al.
2006) and a pilot sampling on two farms, where only RAMS positive animals
were found to be shedding, it was decided to only analyse the fecal samples from
RAMS positive calves. Another motivation for this decision was that we were
interested in the shedding of colonised animals and not VTEC O157:H7 passing
through the gastrointestinal tract without colonisation. As the interest was the
actual shedding levels of the individuals, enrichment was avoided although this
results in lower sensitivity (Davis et al. 2006).
3.4.3 Characterisation and subtyping
All samples from which non-sorbitol fermenting E. coli was cultured were
confirmed by agglutination of 5 suspected colonies using a latex kit (DR 622;
Oxoid). In paper I, positive colonies were also tested biochemically using the
API 20 E system (bioMérieux) and analysed with PCR to identify the presence
of genes coding for verotoxin 1 and 2 (vtx1 and vtx2) and intimin (eaeA) (Gannon
et al. 1997; Paton & Paton 1998). In paper II and III, two positive isolates (as
determined by agglutination) from each sampling occasion were analysed with
PCR to detect presence of genes coding for O157, vtx1, vtx2 and eaeA (Nielsen
& Andersen 2003; Perelle et al. 2004).
In paper I, additional characterisation was also performed. Belonging to clade 8
was determined by real-time PCR and multi-locus variable number tandem
repeat analysis typing (MLVA) was performed as previously described by
42
Söderlund et al. (2014). Whole genome sequencing (wgs) was performed on 30
isolates of clade 8 recovered from 4 farms during 2014. DNA was extracted with
DNeasy Blood & Tissue kit (Qiagen), sequencing libraries were prepared using
the Nextera XT kit and the sequencing were done on the Illumina MiSeq system
with 2 x 250 bp paired-end reads. Detailed description of analysis and processing
of raw reads is found in paper I.
3.5 Risk factors
3.5.1 General overview
Risk factors analysed in this thesis are related to between farm, within farm as
well as between and within animal dynamics. As a first step (Paper I) we
explored local transmission of VTEC O157:H7 on the island of Öland. Although
studies have shown that local transmission occurs (Herbert et al. 2014; Widgren
et al. 2015), how it occurs remains unexplored. Öland is one of Swedens most
cattle dense regions and, considering the association between high number of
cases and high cattle density (Innocent et al. 2005), understanding of dynamics
within such an area is of public health importance.
Many studies have investigated risk factors on farm and pen level and while
similar results are common for some factors there are also conflicting results as
well as risk factors that are unconfirmed by additional studies. Farm, pen and
management related factors are also complexly interrelated and, when all are not
accounted for, correlations and confounding may affect estimates of the included
variables. In paper II we included pen-level variables suggested to be important
for prevalence and transmission in previous studies and analysed their impact on
colonisation together to account for confounding and possible correlation. We
also follow up on the role of suggested risk factors for new infections after five
weeks (in pens where colonised individuals were identified) to validate the
repeatability of the results.
The last part of the project (including Paper III and IV) focused on animal
level determinants to address the individual heterogeneity observed in
colonisation of VTEC O157:H7. As in paper II, a targeted sampling design and
selection of cases and controls from the same pens was used to enable
comparison of individual differences. By combining observations of animal-
based welfare measures and behaviour, we explored possible drivers of
transmission and host factors related to increased susceptibility or resistance
(Paper III) and provided new perspectives on animal-level risk factors associated
with colonisation. In addition, hair cortisol analysis was used to explore the often
43
proposed, but poorly established, association between stress and colonisation.
Although hair cortisol has been suggested to be a promising objective way to
measure stress we evaluate methodological aspects and cortisol levels in relation
to welfare parameters (Paper IV) and discuss our results in relation to these
findings to clarify the difficulties of making inference about animal stress.
3.5.2 Farm characteristics and management
In Paper I, coordinates representing the farm building were retrieved from
national registry for productions sites (Swedish Board of Agriculture through the
national database Geodata). Information about risk factors was collected through
questionnaires sent to farmers by postal service around the time of the second
sampling. The questionnaire was developed together with representatives from
Farm and Animal Health. The first edition was reviewed by a veterinarian
specialized in cattle medicine and herd health and some questions were
rephrased before it was sent out. It contained mostly closed questions, with space
for free comments, about farm characteristics (number of animals, type of
production etc.), contacts with other farms (on pasture, co-use of agricultural
machines) as well as visits to and on other farms. Documents for retrieving a
small economical compensation for participating in the project was also sent
together with the questionnaire with the hope that farmers would complete both
and return them together. If no reply had come about a month after the
documents had been sent, farmers were contacted by phone to remind them of
the questionnaire and possibility to get financial compensation.
During this phone call, farmers were also asked if they needed some
clarifications about the questionnaire. It appeared that farmers had not had
problems answering questions but they found the questions about which other
farms they or their animals were in contact with time consuming to fill in. This
was due to many farms having a large number of contacts but also that they had
to actively ask their neighbours about their farm identification number (Farm
ID) or write their names. The issue of this also became clear when looking at the
returned questionnaires. On these particular questions there were multiple
occasions where one farmer had filled in contact with another farm while this
farm had not mentioned the other farm. There was also farmers that had written
down the names of the farmer his animals had contact with instead of farm ID.
In most cases we were able to tie the names to farm ID but there were occasions
when this could not be done. Thus, these questions could only be used for
descriptive purposes.
On the farms included in paper II-IV a structured interview based on a
questionnaire was performed to provide background information about the
44
farms. This included questions about number of animals on the farm,
management, feeding and cleaning routines different groups of animals and
farmer perception of health and welfare.
3.5.3 Pen characteristics
On the 12 farms positive for VTEC O157:H7 (included in paper II-IV),
characteristics of pens indicated to be positive in the environmental sampling
were collected on the day of the individual sampling. The size of the pen was
measured with a laser telemeter and the number of animals in the pen counted.
The number of drinkers per pen and the cleanliness of them was assessed based
criteria from the Welfare Quality protocol (1 = drinker and water clean to 3 =
water and drinker dirty) (Welfare Quality® 2009). The bedding material (fecal
contamination and wetness) was visually assessed and type of pen was also
noted. On the first visit wetness of bedding and fecal contamination were scored
together in a single measure (clean: limited faeces visible, dry bedding; some
dirt: faecal contamination of bedding material clearly visible and/or bedding wet
in part of the pen; very dirty: faecal contamination visible and/or bedding wet in
the whole pen). During the sampling of the first visits, it was noted that although
bedding sometimes appeared clean and dry from a distance it could be very wet
under the surface. Thus, for the follow up sampling, this measure was expanded
and cleanliness and wetness assessed separately.
Initially animals per square meter was used as a measure for stocking density.
However, during analysis of the data it was suspected that number of animals
did not describe the stocking density very well. As animal size increases with
age number of animals per square meter does not mean the same for young
animals as for older animals. Instead, a stocking density measure that reflected
change of weight as animals get older was calculated. First, an average number
of kilograms within the pen was created by multiplying average age (in days) of
calves within the pen with the average daily weight gain (estimated to 0.81 kg)
and the total number of animals in the pen. This number was then divided with
the area of the pen (m2). Although this measure was not an exact measure of the
kilograms within pen it represents at least a more meaningful estimation of
stocking density than the first measure.
3.5.4 Individual assessment
The heterogeneity in colonisation indicates that there are host differences, either
intrinsic or extrinsic that influence the susceptibility or exposure. Within the
field of animal welfare, it has been well documented that individuals differ in
45
their behaviour as well as coping, and by using validated animal-based measures
developed to assess welfare these individual differences can be studied (Broom
1986; Wiepkema et al. 1987; Duncan 2005; Lecorps et al. 2018). As welfare and
stress are closely linked (although poor welfare does not necessarily mean high
stress or vice versa)(Veissier & Boissy 2007), considering the association
between welfare and colonisation may also add some clarity to the connection
between colonisation and stress.
The protocol used for individual observations used in paper III and IV was
developed based on existing protocols for assessing health and welfare in dairy
cows and calves. The basis was protocols and background material developed
within the Welfare Quality (WQ) project (Welfare Quality® 2009). However,
due to the wide scope of the existing WQ protocol it was trimmed and simplified
for easier and faster scoring. Part of this development was inspired by observing
a certified assessor perform a herd health and welfare assessment within VÄXAs
welfare scheme “Ask the cow” (Louise Winblad von Walter, VÄXA, personal
communication). The VÄXA protocol contains similar measures but is designed
to be performed faster than the WQ assessment. In addition, previous studies on
dairy calf behaviour and welfare were used to select and define behaviours as
well as some measures. Descriptions of the different measures and their origins
are presented in Paper III, supplementary material S1). All observations were
carried out by the author with support from students and on some occasions staff
from VÄXA.
Performing the welfare and behaviour assessment
Individual assessment in the form of undisturbed behavioural assessment was
started as soon as the animals to be sampled in the positive pens had been
identified and their ID-number noted. Before the visit farmers had been asked
what time they would start activity in the barn where the animals to be sampled
were kept and the start of the visits were planned to coincide with this time so
animals would be active (Bokkers & Koene 2001). The aim was to start
observations when the farmer started work in the barn (normally feeding the
calves) which was communicated to the farmer during the phone call. However,
on some occasions the farmer had started feeding the calves earlier with the
intention of being helpful and not being in the way of the observations. Hence,
some farms were observed post-feeding. Still, low activity was rarely a problem
in the groups that were fed during or before the visit. Instead low levels of
activity were seen more often in groups that had continuous access to feed
throughout the night and did not appear as excited about new food being
provided.
46
Undisturbed behavioural observations were carried out for a total of 20
minutes per pen divided into 5 minute intervals. During this time the observers
were careful about standing still, not making noise and standing preferably at
least 1 m from the pen fence. Movement between pens was done slowly and as
quietly as possible. The observed animals appeared to take notice of the persons
watching them in the beginning of the first observation period but generally lost
interest when nothing interesting happened. The difficult situations arose when
it was not possible to stand at least 1 m away from the pen or the observer had
to move closer to them to move between pens as this peaked their interest. For
pens where it was not possible to see all calves all the time, the pen was observed
from another location for another 20 minutes.
It is obvious that observing a pen for 20 minutes does not represent all
behaviours calves perform during a day. It is also uncertain how repeatable this
behaviour would be if the observations wold have been performed several days
in a row. However, due to the time consuming sampling this was the maximum
time available. (Visits still lasted between 5 a.m to 8 p.m on occassions). The
time for observation in the WQ is 10 minutes for a segment of 25 animals but
this time was increased to enable observation of more behaviours. Nevertheless,
the observations should be considered as a snapshot taken on one day during an
active period. The limitation in observed time should mainly be associated with
a risk of type II errors (failure to observe a difference when it is there) as a lower
frequency of observed behaviours decreases the power of the study. Also, if
calves have very different behavioural patterns, for example individual
differences in when during the day grooming behaviour is performed, this would
similarly decrease the ability to differentiate between grooming and not
grooming calves and make an association more difficult to identify.
3.5.5 Analysis of hair cortisol
Using cortisol to measure the activity of the hypothalamicpituitary
adrenocortical (HPA) axis in itself is nothing new. Concentration of the hormone
has been analysed in blood, saliva, faeces and urine for many years in studies of
welfare and acute as well as chronic tress (as reviewed by Mormède et al. 2007).
Analysing hair cortisol content offers a non-invasive possibility to measure
retrospective levels of circulating cortisol, as cortisol is incorporated in the hair
while it grows, which is not influenced by daily fluctuations and stress around
sampling (Lee et al. 2015; Burnard et al. 2017). When the project began hair had
been used to analyse cortisol from cattle but in studies with a smaller number of
animals (González-de-la-Vara et al. 2011; Moya et al. 2013; Burnett et al. 2014).
47
The concentration of cortisol was determined using an ELISA kit designed
for salivary cortisol (Salimetrics Europe Ltd, Art 1-3002) according to the
manufacturer’s instructions (validated by Moya et al. 2013). Final cortisol
content in hair was calculated using the formula suggested by Meyer et al. (2014)
and inter- and intra-assay coefficients of variation (CV) calculated according to
manufacturer instructions (Salimetrics 2018).
A detailed description of the hair cortisol extraction is presented in Paper IV
and the steps visualised in Figure 7. To enable extraction of a larger number of
samples the protocols suggested in previous studies on cattle required some
changes. To be able to pulverise the large number of samples in a standardised
way, freezing of hair followed by bead beating with three chrome steel (3.2 mm
in diameter, BioSpec Products, Cat. No. 11079132c) in the tubes was done. The
freezing step was added to achieve a more homogenous pulverisation thanks to
input from researchers working with dog hair at Linköping University (Roth et
al. 2016).
During the processing of samples, it became clear that the washing procedure
did not remove all dirt and hair with severe faecal contamination also appeared
to have changed structure. As the washing procedure is known to impact cortisol
extraction (Davenport et al. 2006) and it was suspected that the changes in
structure might also have an impact it was decided to score the samples of
remaining dirt to enable control for this in the analysis. As dirt on dark hair
Figure 7. Description of the preparation and extraction of hair cortisol.
48
would potentially be more difficult to spot the dirt score was correlated with
discoloration of the final solution of extracted cortisol. The impact of dirt is
discussed in detail in paper IV.
After drying of the first batch (samples 1-41) it was also clear that one
centrifugation did not remove all the hair particles and some were transferred
with the supernatant to the final sample. To ensure this did not happen an
additional centrifugation was added. The extra centrifugation did not impact
cortisol content in the final sample (Paper IV). In the initial protocol, 200 μl of
PBS was used to dissolve the extracted cortisol. However, the dried cortisol was
hard to dissolve with this amount of PBS. When the volume was decreased
slightly (to 150 μl) for a test batch the process was easier. As CV values of this
batch were smaller the protocol was changed to 150 μl PBS.
3.6 Statistical analysis
Data was entered in Excel and all statistical analysis performed in the statistical
software R (R Core Team 2018). Several different approaches were used and
details of each analysis, as well as R-packages used, are outlined in the papers I-
IV.
To summarise, Fisher’s exact test, Wilcoxon rank test or Chi-square tests were
used in the univariable analyses in paper I, II and IV. Paper I and II included
generalised linear mixed-effects models with a logit link (glm), to assess risk
factors for presence of VTEC O157:H7 on farms and risk factors for
colonisation. Paper I also included analyses of spatial clustering (using Cuzick-
Edwards’ kNN (k nearest neighbours) and Ripley’s K function tests). In paper
II, the glm was complemented with generalised additive models (gam) to
investigate non-linear associations. In paper III cluster analysis, elastic net
regression and principal component regression were combined to study
individual risk factors. These methods were used to enable analysis without
reducing the data. This was done to provide a holistic perspective where
associations between the variables could be used to enable interpretation of
possible underlying meanings. The different methods also have different
strengths as they are not dependent on the same assumptions. In paper IV, gam
and elastic net regression were used to analyse how hair cortisol concentrations
were influenced by methodological changes, age and welfare indicators.
There were a small proportion of missing values in the welfare observations
(for more details see paper III and IV). In most cases observations of one or two
variables of an individual were missing. As the number of animals in the study
was limited, and one missing value would mean that the individual could not be
included in the elastic net regression, imputation of the missing values was
49
performed using non-parametric random forest imputation as described in Paper
III and IV.
Some additional analyses (not presented in the papers) are included in this
thesis. Causal pathways to visualise assumptions of causality in paper II were
created using DAGitty v3.0 (www.dagitty.net). Stratified analysis of risk factors
from paper II was performed using generalised mixed models with a logit link
in the package lme4 (Bates et al. 2015). Analysis of the association between hair
cortisol and colonisation was performed using generalised additive models in the
package mgcv (Wood 2004, 2011). Figures were created using the packages
ggplot2 and ggeffects (Wickham 2016; Lüdecke 2018).
50
51
4.1 Between farm transmission on Öland
The results from paper I indicate that local transmission on the island of Öland
was common and strains were frequently exchanged between farms. One of the
risk-factors for introduction of the infection was purchase of animals, which is a
recognised risk factor in previous studies and the suggested underlying driver
for transmission over large distances (Widgren et al. 2015, 2016; Franz et al.
2019). Trade of animals is also a well-known risk for introduction of other
infectious diseases on dairy farms and avoiding purchase of animals was the
measure most commonly mentioned when the farmers participating in Paper II-
IV were asked how they protect animals on their farm from infectious disease
(Table 2).
Still, the results from paper I indicate that avoiding purchase of animals will
not be enough if the farm is located in an area with high cattle density where
VTEC O157:H7 circulates. The analysis of risk factors as well as more detailed
analysis of isolates using wgs, point to human activities (visitors travelling
between farms) being responsible for introducing the pathogen on farms. A
previous study of small scale dairy farms in Mexico have correspondingly
observed a genetic pattern that matched shared forage storage and milking staff
(Rosales-Castillo et al. 2011).
Still, humans may not literally have to carry the pathogen between barns. As
sharing of agricultural machines was a risk factor for being positive in the fall
sampling, moving vehicles between farms may be enough. For example flies,
known to be able to spread VTEC O157:H7 (Ahmad et al. 2007) and found to
carry VTEC O157:H7 in paper I, may pick up the pathogen from faecal
contamination on vehicles or travel in vehicles between farms.
4
Results and discussion
52
Table 2. Characteristics of the farms included in the individual sampling (paper II-IV). Farm size includes the total number of cattle on the farm. Sampled animals
is the number of animals sampled for verotoxin-producing Escherichia coli O157:H7 and the proportion colonised as determined by recto anal mucosal swabs.
Information about calf mangement of calves, cleaning routines and biosecurity measures were collected in a structured interview. Last 13 rows reflect farmers
spontaneous answers to the questions “How is introduction of infectious agents prevented” and “How is transmission of infectious agents between animals
prevented?”. (Yes) = mentioned but applied with exceptions.
Farm
F1
F2
F3
F4
F5
F6
F7
visit 1
F7
visit 2
F8
F9
visit 1
F9
visit 2
F10
F11
F12
Farm size
330
375
650
300
422
380
360
350
700
135
130
250
Sampled animals
24
26
20
20
28
26
17
20
25
29
20
25
20
18
Proportion colonised
25%
8%
5%
10%
11%
39%
29%
20%
12%
17%
15%
12%
30%
17%
Calves in single pens
(weeks)
8-12
1
~2
0
8
~2
2-3 days
2-4
6
~2
2-3
4
Weaning age (weeks)
8-12
14
8
8-10
8
10-12
12
8
8
8-10
12-16
12
Yearly cleaning
Yes
Yes
Yes
No
Every
second
year
Yes
Yes
Yes
"Most years"
Yes
Yes
Some
buildings
Regular use of
disinfectant
No
Slaked
lime
Slaked
lime
Slaked
lime
Slaked
lime
No
Slaked lime
Yes +
Slaked
lime
Yes +
Slaked lime
No
No
Yes
How is introduction of infection to the herd prevented:
Avoid purchase of
animals
Yes
Yes
Yes
(Yes)
(Yes)
Yes
Yes
(Yes)
Protective clothes for
visitors
(Yes)
Yes
Yes
Yes
Yes
Yes
Yes
(Yes)
Yes
53
Farm
F1
F2
F3
F4
F5
F6
F7
visit 1
F7
visit 2
F8
F9
visit 1
F9
visit 2
F10
F11
F12
No contact with other
farms on pasture
Yes
How is transmission between groups of animals prevented:
Rinse boots between
groups
Yes
Yes
Yes
(Yes)
Yes
Yes
(Yes)
Yes
(Yes)
Yes
Desifection at entrance
Yes
Clean feed through
Yes
Yes
Yes
Control rats and
wildlife
Yes
Desinfect between
groups of animals
Yes
Dont move sick
animals
Yes
Wash hands after
handling sick animals
Yes
Avoid contact between
young and old animals
Yes
54
Animal contacts (nose-nose contact on pasture) were also an important risk
factor. However, it should be kept in mind that farms on Öland differ from farms
in other regions of Sweden. On Öland, pastures are often adjacent to pastures of
other farms and animals are often transported from the main farm to dierent
pastures during summer. There are often only simple fences (traditional stone
walls) around pastures separating the animals (Figure 8). This means that there
is an unusually high rate of contacts between farms and the extent of animal to
animal contacts between farms located on different parts of the island becomes
very clear in the questionnaire study (See Paper I, Figure 4). The high contact
rate likely explains how the strain of clade 8 spread so efficiently across the
island and why the proportion of farms with clade 8 is high compared to other
national studies including other regions (Eriksson et al. 2005; Söderlund et al.
2014; Widgren et al. 2015).
Another interesting observation from paper I was animals picking up strains
of clade 8 previously found on neighbouring farms on pasture and introducing it
on farms. It is possible that strains from neighbouring farms were acquired
through animal contacts on pasture but another explanation could be contact with
a common environmental reservoir like flies, birds and wild game as has been
discussed previously. Differentiating whether environmental presence of the
same strains found in cattle is a reservoir of infection or a spill-over from
circulation among cattle is beyond this study to ascertain. However, considering
that farm and cattle density appear to be of importance for transmission the
results of this thesis supports the latter.
Figure 8. Pasture on Öland with the traditional stone walls separating pastures of animals from
different farms in the background. (Photo: Lena-Mari Tamminen)
55
4.1.1 Is it important to differentiate between introduction and
persistence?
In paper I it was found that contact with a known positive farm (positive for
VTEC O157:H7 in the spring sampling) was a risk factor for persistence
(positive on both sampling occasions) as well as new infection (a previously
negative farm becoming positive in the fall sampling). It was also the risk factor
most strongly associated with being positive in the fall sampling regardless of
previous status in the spring sampling (OR 6.8, CI 1.6-32.3). This suggests that
some of the farms that were positive in both the spring and the fall sampling
were infected with a new strain during summer and hence were not really farms
where infection persisted. The results from the whole genome sequencing of
strains from four of the farms also exemplifies that different strains were
circulating among the farms between the spring and fall sampling (Paper I,
Figure 5). However, being positive in both the spring and fall sampling was also
associated with farm size (large farms more likely to be positive) and combining
milk and meat production. These risk factors are not clearly related to
introduction of new strains. In addition, the farms positive in both samplings had
significantly fewer neighbours within 5 km compared to the farms that cleared
infection. There are several examples of farms that have remained positive over
time despite a turnover of animals (Lahti et al., 2003; Tamminen et al., 2018) as
well as farms where circulation of the same strain over time has been confirmed
using PFGE (Joris et al. 2013; Herbert et al. 2014).
If the underlying pattern behind observed persistence is in fact frequent
introductions to a farm, on-farm measures applied to prevent persistence will be
an unnecessary cost to the farmer as the farm would likely clear the infection if
new introductions stopped. However, on a farm where an environmental
reservoir or circulation of infection within groups of animals is occurring,
external biosecurity measures will not reduce the prevalence. Not separating
these two scenarios also risks introducing noise to studies of transmission and
prevalence on farms, just as we are likely observing in paper I.
4.2 Within farm prevalence and transmission
The separate environmental samples collected from young calves, weaned
calves, young stock and dairy cows in this study showed a large variation
between farms. On the 12 farms where thorough environmental sampling was
performed, only 1 had positive samples from all groups of animals. On all farms,
the pathogen was found among calves between 2-6 months of age. On 6 farms,
groups including animals up to 12 months were also positive in the
environmental sampling. However, these groups were often difficult to
56
distinguish as animals were housed in overlapping age constellations. Dairy
cows and non-weaned calves were positive only on part of the farms (n = 2 and
5 respectively). Low prevalence among dairy cows and non-weaned calves is
consistent with previous studies (Mechie et al. 1997; Rugbjerg et al. 2003; Gunn
et al. 2007; Cho et al. 2009). In addition to the variation between groups
observed in the environmental sampling, the individual sampling showed
variation in colonised animals between pens. Colonised and shedding animals
were generally not found in all pens on a farm and there were several examples
of farms where differences in prevalence were observed within the same age
groups if animals were housed in separate buildings. Thus, it appears possible to
keep transmission from occurring between groups of animals despite most of the
farms not having strategies for preventing disease spread between groups (Table
2). Avoiding transmission between groups of animals on farms by keeping
groups together has previously been suggested as a cost effective measure to
reduce prevalence of VTEC O157:H7 on farms (Ellis-Iversen et al. 2008; Lyons
et al. 2013). However, there are indications that other vectors may participate in
spreading the bacteria between groups on a farm. For example, a cat on the farm,
identified as a potential risk factor for being positive for VTEC O157:H7 in
paper I (although only significant with 90 % confidence), may both bring the
pathogen to the farm and circulate it within the farm as cats tend to move freely
among groups of animals. As mentioned above, flies and birds may also play a
possible role in transmission on farms. But, since VTEC O157:H7 was generally
associated with a subset of animals on the farm this does not appear to be a huge
problem.
4.2.1 Management and susceptibility closely connected potential
drivers of transmission
Interdependencies between predictors
The risk factors included in paper II were selected based on associations with
colonisation and shedding in previous literature. Analysis of risk factors on pen
level from the first individual sampling showed an interaction between stocking
density and age, suggesting that with low stocking density the risk of
colonisation increased with age while in high stocking density the risk of
colonisation decreased with increasing age (Paper II, Figure 1). The effect was
most notable as a large difference in risk of colonisation in young animals
housed in high density compared to low density. However, in the follow up
sampling (5 weeks after the initial sampling) this association was no longer
observed. The average age of animals in the second sampling was higher (130
57
days compared to 122) but most importantly there was a smaller proportion
young animals (below 100 days) compared to the first sampling (Figure 9 A).
Thus, the second sampling may not have had the power to detect an increased
risk at young age. Stocking density, on the other hand, was observed to be an
important risk factor for colonisation, both in the first and the second sampling
but the effect was estimated to be slightly larger in the first sampling compared
to the second (OR 1.99 and 1.31 respectively). A potential reason for this
difference can be found when considering a causal network of the variables
included in the two models (Figure 10).
Figure 9. Distribution of age of sampled calves and results of individual sampling for verotoxin-
producing Escherichia coli O157:H7. A: Number of sampled calves and results from recto anal
mucosal swab (colonised/non-colonised). B: Number of calves and results of individual sampling
including faecal shedding levels. C: Proportion colonised and shedding within age groups.
58
Figure 10. Causal diagram describing causal assumptions of risk factors for colonisation of
verotoxin-producing Escherichia coli O157:H7 analysed in paper II. (m2=Pen size)
As seen by the arrows in the figure, associations between the risk factors; age,
stocking density, pen hygiene as well as water to cattle ratio and the presence of
super-shedders were assumed. (Based on the assumption that these variables
influence the risk of individual colonisation, they likely influence the risk of
colonisation of peers and through this effect the presence of super-shedders). If
part or all of the effects of these variables on colonisation is mediated through
increased presence of super-shedders their effect will be reduced or blocked
when the presence of super-shedder is included in the model (Dohoo et al. 2014).
Thus, the estimate of stocking density in the second model in paper II likely is a
better estimation of the direct effect on the risk of colonisation of an individual
(as the effect of stocking density on other calves in the pen was accounted for).
Considering this there may be an alternative explanation for why the
interaction between age and stocking density does not influence colonisation
when presence of super-shedders are accounted for. It is possible that the effect
of the interaction is related to the presence of shedders, i.e. that the presence of
59
shedders is more common in high stocking density and young animals. The
presence of shedders was closely connected to the outcome (colonisation) in the
first individual sampling, since analysis of shedding was only performed on
samples from colonised individuals. Thereby it was not independent from the
response variable and not included in the risk factors for colonisation analysed
in the first sampling. In Figure 11, the association between colonisation,
shedding levels, age and stocking density from the first and second sampling are
visualized. In sampling 1 there does appear to be a group of animals around 100
days shedding higher levels although there were also older animals shedding. In
sampling 2 high shedding also occurred around 100 days and it is noteworthy
that the highest shedding animals are housed in higher stocking density. The
shedding pattern of calves in relation to their age is presented in Figure 9 (B &
C). Out of the RAMS positive calves a low frequency of animals were shedding
(19 individuals in the first sampling and 17 individuals in the second). It appear
as if the proportion of RAMS positive calves shedding high levels of bacteria
was higher in the age group 50-100 days in sampling 1 and 50-150 days in
sampling 2. Thus, there are signs of an association between young age and
increased shedding but this should be interpreted with caution due to the small
number of shedding individuals.
Figure 11. The association between age (x-axis) and stocking density (y-axis) and
colonisation/shedding of verotoxin-producing Escherichia coli O157:H7 of the sampled dairy
calves. Coloured dots indicate colonised calf (as detected by recto anal mucosal swabs) and size of
dots indicates shedding level.
60
Age and weaning may influence exposure in different ways
Age was found to have a small negative coefficient in the elastic net regression
(Paper III) which supports a decreased risk of colonisation with increasing age.
This is in agreement with many studies where prevalence of VTEC O157:H7
has been observed to decrease with age (Kuhnert et al. 2005; Gunn et al. 2007;
Mir et al. 2015). However, as mentioned above, non-weaned calves have been
associated with a lower prevalence (Garber et al. 1995; Hancock et al. 1997;
Rugbjerg et al. 2003). This is consistent with the result that VTEC O157:H7 was
found in the environment of non-weaned calves on only 5 of the sampled farms.
However, analysis of individual risk factors for colonisation in the first sampling
(paper III) and transmission among positive groups of animals (Paper II, part 2)
did not indicate a protective effect of drinking milk or an effect of weaning. Most
of the farms included in this study kept calves in single crates for around 2 weeks
before group housing (Table 2) which is common practice in Sweden (although
exceptions occur). Common practice in many other countries, like the United
States, is to keep calves in single crates until weaning around 2 months of age.
As grouping of calves before weaning has been identified as a risk factor
regardless of weaning age (Garber et al. 1995), the practice of early group-
housing may explain the discrepancy between our study and previous studies.
This indicates that other factors, sometimes associated with weaning, such as
changes in the management of young animals and being introduced to other
animals, influence the risk of shedding and colonisation. This is consistent with
studies in the UK where the effect of age and weaning was not significant after
accounting for management related factors (Synge et al. 2003; Smith et al. 2016)
and that VTEC O157:H7 has been identified in high prevalence in very young,
group housed calves on New Zealand (Browne et al. 2018).
The first months of the life of a dairy calf is a dynamic period with changes
in feeding, housing and social contacts. Social grooming behaviour, one of the
important risk factors on individual level (Paper III), increases rapidly during
first week calves are housed together (Abdelfattah et al. 2018; Horvath & Miller-
Cushon 2018) and it is important for maintaining social contacts (Færevik et al.
2007). Mixing of calves has been observed to lead to a marked increase in
behaviours directed towards other calves but the effect disappeared with
increased number of regroupings (Veissier et al. 2001). In another study,
regrouping has been shown to be associated with increased grooming mainly of
familiar calves (Horvath & Miller-Cushon 2018). Thus, a reason for the
decreasing proportion of colonised animals observed with increasing age, in this
and other studies, may be related to reduced transmission as groups and social
contacts stabilise with age.
61
4.2.2 Importance of super-shedding for transmission
The second part of paper II emphasises the importance of super-shedders as the
risk of colonisation in the follow up sampling was much larger in animals housed
with a super-shedder compared to being housed with a non-shedding colonised
animal (OR 9.8, CI 3.9-50.6). This finding supports the importance of super-
shedders on transmission of VTEC O157:H7 proposed in previous studies
(reviewed by Chase-Topping et al. 2008). The selection of animals in the follow
up sampling included only animals housed with a colonised individual in the
first sampling, which means that all animals should have had the opportunity to
be exposed to some level of the pathogen. Despite that it has been shown that
low doses are enough to induce colonisation and shedding (Besser et al. 2001)
the presence of a super-shedder played an important role in the dynamics.
However, we cannot, based on these results, identify whether the presence of a
shedder was related to individual characteristics of the particular individual or
group level characteristics. The presence of a super-shedder may indicate a pen
where some necessary, unmeasured, requirements for transmission are fulfilled,
like contacts within pen (as suggested by Turner et al. 2008) or environmental
contamination (Gautam et al. 2015). Considering that the individual risk factors
for colonisation in these pens were related to social interactions as well oral
exposure, through for example grooming, supports a combination of both
explanations (Paper III).
It is well known that high shedding can lead to large and unpredictable
fluctuations in environmental prevalence of the pathogen and that other animals
become infected with strains that are being shed in high levels by pen mates
(Chase-Topping et al. 2007; Cobbold et al. 2007; Stephens et al. 2009; Henry et
al. 2019). However, recent longitudinal studies questioned the role of super-
shedding individuals as periods of high shedding were rare and occurred during
short periods (Munns et al. 2014; Lammers et al. 2015). Instead of individual
super-shedders they suggest that synchronised shedding of many animals is
driving transmission. In this study, the proportion of colonised animals was
relatively stable, few animals per farm were found to be colonised and even
fewer super-shedding (1-2 individuals). Although we cannot know the dynamics
between the two sampling occasions, it is clear that the presence of a super-
shedder has an important role in the dissemination of the pathogen.
62
4.2.3 Transmission dynamics in poor and good hygiene conditions
In our study, pen hygiene was not a significant risk factor for presence of
colonised individuals in the first individual sampling and, while moisture was
indicated to be important in the follow-up sampling, faecal contamination was
negatively associated with colonisation (Paper II). In addition, analysis of risk
factors on individual level (Paper III) did not suggest associations between poor
hygiene and colonisation. Instead, calves with poor cleanliness scores (below
hocks and body) were less likely to be colonised by VTEC O157:H7. This
contradicts recent suggestions that contamination of the environment and
ingestion of faecal material is the major driver for transmission (Gautam et al.
2015; Spencer et al. 2015). Variables most strongly associated with colonisation
were self-licking and licking other calves which may be related to direct
transmission between animals as VTEC O157:H7 is commonly found on the
hides of animals housed with a super-shedder (Arthur et al. 2009; Stephens et
al. 2009).
Still, the most important route of transmission may depend on context. This
was illustrated when the animals in the first individual sampling in paper II were
stratified by pen hygiene. The calves housed in clean pens differed slightly from
calves in pens with poor hygiene as animals were housed in lower stocking
density (average 24 kg/m2 compared to 30 kg/m2) and were slightly younger.
Reanalysis of risk factors for the separate groups (using univariable analysis and
multivariable analysis as performed in Paper II) revealed interesting differences
(Table 3). For example, the interaction between age and stocking density was
only significant for calves housed in clean pens and not for calves in pens with
poor hygiene. As explained in paper II, the interaction means that in clean pens
young animals housed in high stocking densities were more likely to be
colonised by the pathogen but that the risk decreased with increasing age. The
stratified analysis suggests that age and stocking density did not influence
individual susceptibility to colonisation in a dirty pen. This could be interpreted
as direct transmission between individuals being less important in pens with high
environmental exposure. Thus, in a dirty environment calves are probably
infected through environmental exposure and direct contacts while in a clean
pen the only way to become colonised is through contact with another individual.
The reduced risk with increasing age in the clean pens support the suggestion by
Gautam et al. (2015) that direct transmission is less effective than from the
environment.
63
Table 3. Risk factors for colonisation by verotoxin -producing Escherichia coli O157:H7 in dairy calves in the first sampling stratified by pen hygiene.
Colonisation of VTEC O157:H7* in:
Clean pens:
Uni-
Multi-
Dirty pens:
Uni-
Multi-
(Score 1)
variable analysis
(Score 2-3)
variable analysis
RAMS-
RAMS+
p
β
p
RAMS-
RAMS+
p
β
p
Number of calves
112
18
149
38
Shedders >103 cfu/g
3
1
Shedders >104 cfu/g
4
9
Sex = M (%)
40 (35, 7)
2 (11,1)
0,06
-1,42
0,10
50 (33,6)
7 (18,4)
0,08
-0,70
0,18
Age (months) (median [IQR])
3,82
[1,80, 6,11]
2,93
[2,35, 6,08]
0,97
1,15
0,02
4,10
[3,23, 5,07]
3,47
[2,81, 4,07]
0,01
-0,02
0,95
Stocking density (10 kg/m2)
(median [IQR])
2,43
[2,09, 2,92]
2,43
[2,36, 3,7
3]
0,15
1,52
0,01
2,92
[2,33, 4,04]
3,34
[2,92, 3,41]
0,04
0,50
0,21
Animals in pen
(median [IQR])
11,00
[6,00, 18,25]
7,00
[4,00, 16,50]
0,26
0,00
0,97
9,00
[7,00, 13,00]
8,00
[7,00, 18,50]
0,66
-0,03
0,64
Pentype (%)
0,56
0,06
Straw
64 (57,1)
13 (72,2)
54 (36,2)
22 (57,9)
Deepstraw
35 (31,2)
4 (22,2)
-1,55
0,22
79 (53,0)
14 (36,8)
0,13
0,85
Loose housing
13 (11,6)
1 ( 5,6)
-0,94
0,74
0
0
Slatted floor
0
0
16 (10,7)
2 ( 5,3)
-1,10
0,58
Water/cattle ratio
(median [IQR])
0,09
[0,06, 0,17]
0,17
[0,06, 0,25]
0,45
0,11
0,98
0,11
[0,10, 0,14]
0,14
[0,11, 0,20]
0,00
10,24
0,07
Stocking density:Age (months)
-0,42
0,01
-0,06
0,29
*As determined by detection of bacteria in recto-anal mucosal swabs (RAMS); Fisher exact test or Wilcoxon rank sum test; Generalised linear mixed model
with a logit link.
64
In the analysis of risk factors for colonisation in the follow up sampling
(Paper II part 2), being housed in a pen with faecal contamination of bedding
reduced risk while a wet bedding was associated with increased risk. It is known
that moisture and faecal contamination influences persistence and growth of
VTEC O157:H7 in the environment (as discussed in paper II). But poor hygiene
will not only influence VTEC O157:H7 it will also have an impact on the
health, well-being and behaviour of animals in the pen, for example cleaning
frequency of pens has been associated with calf diarrhoea (Klein-Jöbstl et al.
2014). Sick individuals will modify their behaviour to overcome disease (Hart
1988) which may impact exposure to agents like VTEC O157:H7. Behavioural
changes in calves include reduced self-grooming, feeding and social interactions
and have been even observed in early and mild stages of respiratory disease
(Borderas et al. 2008; Cramer et al. 2016; Hixson et al. 2018). Sick calves are
also less likely to approach novel objects or a stationary human (Cramer &
Stanton 2015), possibly indicating a less exploratory behaviour. This fits very
well with the finding that individual risk factors not associated with colonisation
in paper III were associated with diarrhoea, coughing and nasal discharge (signs
of respiratory disease) as well as other indicators of poor welfare.
4.3 Stress, colonisation and susceptibility
The hypothesis that stress is related to colonisation of VTEC O157:H7 has been
suggested in several studies and reviews (Cray & Casey 1998; Chase-Topping
et al. 2007; Rostagno 2009; Munns et al. 2015). However, few studies have
actually explored the association deeper than connecting changes in
management associated with stress (such as dietary or heat stress, weaning,
movement and transport) and suggesting that increased stress makes calves more
susceptible to colonisation and shedding (Cray & Casey 1998; Chase-Topping
et al. 2007; Bach et al. 2016; Stenkamp-Strahm et al. 2018). But, as described
for weaning above, it is not necessarily the increased susceptibility of the host,
but the increased exposure to the pathogen due to changes in environmental
exposure, behaviour or social contacts, which is the actual risk.
In this study, three approaches to investigate stress were used. Firstly, the
associations between colonisation and indicators of poor welfare were explored.
Although stress and welfare are not the same, they are closely linked (Veissier
& Boissy 2007), and animals experiencing poor welfare should be more likely
to also experience stress. As mentioned above, colonisation was associated with
social and active calves that were grooming themselves and others, while
animals that were showing signs of poor health and welfare were less likely to
be colonised (Paper III).
65
Secondly, comparing hair cortisol concentrations of the colonised and non-
colonised groups (Figure 12) using a generalised additive model (including the
variables amount of buffer, faecal contamination of hair as well as age identified
as important influencers in paper IV) showed no significant association. The
model indicated an interaction between age and hair cortisol, suggesting that
high hair cortisol and increasing age was connected with colonisation, but this
was not statistically significant (p = 0.12). This could indicate an association
between colonisation and hair cortisol in older animals but there were too few
old animals in the study to explore this. Considering that previous studies on
cows have showed associations between increased hair cortisol and clinical
disease (Comin et al. 2013; Burnett et al. 2015) and increased risk of shedding
of VTEC O157:H7 in downer cows (Byrne et al. 2003), supports the possibility
that hair cortisol may be associated with colonisation in older animals.
Figure 12. Hair cortisol concentration (pg/μl) of calves colonised by verotoxin-producing
Escherichia coli O157:H7 (as determined by recto-anal mucosal swabs) and non-colonised
(negative) calves.
66
Thirdly, individual reactivity and fearfulness, indicators of coping styles that
are linked to vulnerability to stress (Koolhaas et al. 1999), were assessed in paper
III. A previous study looking into temperament and shedding of VTEC O157:H7
in calves compared excitable, intermediate and calm calves (classified by a
temperament index), found that calm animals were more likely to shed VTEC
O157:H7 than other calves (Schuehle Pfeiffer et al. 2009). In our study, a similar
association between reduced risk of colonisation and high reactivity was
observed (Paper III), further supporting that personality and/or coping style
influences colonisation.
Schuehle Pfeiffer et al. (2009) also analysed serum cortisol levels but could
not find an association between shedding and non-shedding animals as there was
no difference in cortisol levels between the intermediate and calm groups of
animals. However, the excitable group had higher serum cortisol levels. This is
inconsistent with the results of the hair cortisol analysis and welfare measures in
this study (paper IV) where high reactivity was associated with low hair cortisol.
This may be a result of the different methods used for cortisol analysis. Serum
cortisol represents immediate changes while hair cortisol represents average hair
cortisol levels during hair growth (Lee et al. 2015). There are also other
important regulators of coping, like serotonin, that are unaccounted for in both
studies which may explain the discrepancies observed (Koolhaas et al. 2007). In
addition, as discussed in paper IV, the cortisol levels of a calf in a poor non-
stimulating environment may be difficult to differentiate from an individual in a
good environment exposed to acceptable, stimulating challenges (Korte et al.
2007).
Combining the results of the three approaches there is no clear indication that
stress is related to increased risk of colonisation. This is consistent studies
focusing on increased susceptibility colonisation due to heat and handling stress,
where no effect of either were observed (Brown-Brandl et al. 2009; Sheng et al.
2016). Instead results indicate that animals showing signs of coping well are
more likely to be colonised while animals showing signs of poor welfare and
disease were less likely to be colonised. However, there are signs of interesting
differences in personality and behaviour between colonised and non-colonised
calves and we propose that these are related to different exposure to the
pathogen. An additional important aspect of the identified risk factors is that the
risk factors associated with colonisation, i.e. being a socially engaged and active
calf, may be associated with a more efficient dissemination of an infectious
agent. A super-spreader is a description of an individual who has more
opportunities to infect others, through for example through a high number of
contacts (Chase-Topping et al. 2008). Although super-shedding and super-
spreading by definition are independent traits (the first referring to interactions
67
between host and pathogen while the latter refers to interactions between hosts),
risk factors identified for colonisation in this study suggest that colonisation, and
thereby super-shedding, and super-spreading of VTEC O157:H7 are associated.
4.4 Validity, bias and methodological considerations
4.4.1 Study population and external validity
Identification and selection of farms
Almost from the beginning of the project a major concern and limitation has
been obvious identification of farms positive for VTEC O157:H7 to enrol in
the study. As animals do not show any symptoms and Sweden does not have an
active national surveillance, apart from a slaughter prevalence study including a
small proportion of slaughtered animals every second year, finding farms
infected with VTEC O157:H7 was difficult. Previous studies had shown a
regionally high prevalence in Falköping (Widgren et al. 2015) and the slaughter
prevalence study performed by SVA indicated that the pathogen was present on
the island of Öland and the counties Skåne and Blekinge (Erik Eriksson, SVA,
personal communication) which is why these areas were targeted for
environmental sampling. Environmental samplings of farms around Falköping
(guided by the results from Widgren et al. 2015) were all negative, which
supports the national slaughter prevalence study suggesting that clade 8 in this
areas has decreased.
Thanks to the collaboration with Farm and Animal Health, a number of farms
where VTEC O157:H7 were found in a parallel research project as well in
association with human disease farms were enrolled. Most of these farms were
located on Öland but one was located in Småland county and would not have
been identified in any other way. The samples collected from Skåne and
Blekinge were also guided by results from Farm and Animal Health and
performed in an area where several farms had been positive just weeks before
our sampling. However, most farms had cleared the pathogen. This region also
included the only farm that cleared the infection during the time between
environmental sampling and individual sampling. This may indicate that there
were differences between Öland and Skåne/Blekinge, either in the circulating
strains ability to persist on farms or in transmission between farms.
Regarding the effect of potential differences between circulating strains there
is much to learn about survival and persistence of different types of VTEC
O157:H7. Considering the flexible genome of VTEC O157:H7, strain
68
differences are expected and can be present even in closely related strains.
However, we still lack the knowledge to fully understand the impact of these
differences. It may be a strength to study farms in an area where one dominating
strain is present (like on Öland) as this decreases variation and noise introduced
by strain differences. In addition, all strains in our study belonged to a virulent
subtype and understanding how these subtypes behave is a priority compared to
other, less virulent, subtypes.
While the majority of farms were located on the island of Öland, and thus
somewhat similar, the farms included a variety of farm sizes, housing systems
and management and should therefore include a representative selection of
Swedish dairy herds. However, the unique regional characteristics, like high rate
of contacts between farms and high cattle density, raises the question of how
relevant the risk factors identified in this study are for other regions. We argue
that although all routes of transmission that were identified in paper I may not
apply to all regions, it remains relevant to identify them. The unique
characteristics and frequent contacts between farms on Öland may have given
us the opportunity to discover routes that would have been difficult to identify
in other regions with fewer contacts.
Our study design was cross-sectional and thus we do not know how long the
pathogen had been circulating on each farm. As discussed in Paper II animals
infected with VTEC O157:H7 develop an immune response, although there
appear to be strains that can overcome these responses partly or completely
(Hoffman et al. 2006; Corbishley et al. 2014). Depending on how long the
pathogen had been circulating on each study farm, the dynamics may be
different. For example, sampling a farm just as infection hade been introduced
in a naïve population there could be more colonised animals and increased levels
of shedding leading to different risk factors than entering a farm with a long-
lasting presence. Thus, there is a risk that the study design biased the results.
Selection of animals for individual sampling
Farms in papers II-IV represent a selection of farms where VTEC O157:H7 had
been identified and environmental sampling was used further target groups of
positive animals on the farm. The targeted sampling approach was used to avoid
sampling and analysing animals not exposed to the pathogen and increasing the
number of colonised animals included in the study. While previous large scale
studies have struggled with identifying enough calves, e.g. only 34 calves out of
1324 calves were positive in a recent large study from the United States
(Stenkamp-Strahm et al. 2018), we identified at least one colonised individual
on all sampled farms except one.
69
In addition to using environmental sampling to narrow the individual
sampling to a population exposed to the bacteria, calves for follow-up sampling
in paper II and analysis in paper III were selected based on a case-control
approach where only individuals housed with a colonised individual were
included in the analysis. This enabled us to control for environmental exposure
and decreased the risk of introducing noise by including animals that had not
been exposed to the pathogen. However, there is also a risk that this approach
may have led to over-matching and that some determinants associated with
negative pens were not discovered. For example, the environmental sampling
directed the individual sampling to groups of young animals and only sampling
among these animals may be a reason why age was not a strong determinant in
our study as in other studies sampling a wider variety of age groups on farms or
following animals over time (Nielsen et al. 2002; Mir et al. 2015). As the aim
was to understand transmission and risk for colonisation in the presence of
VTEC O157:H7, this risk was accepted.
There were also some practical constraints during sampling and there were
cases when pens with older calves and young stock indicated as positive by
environmental sampling could not be sampled. These cases were commonly a
large pen where animals had space to run that lacked of restraining facilities. To
avoid injury of both humans and animals sampling was not performed in these
situations. However, this may have skewed sampling so older animals sampled
in our study represent animals kept in smaller pens than young stock generally
are. Still, this should mainly be a potential bias in the first part of paper II where
it may explain why the effect of high stocking density was only observed in
young animals.
4.4.2 Choice of methods
There is a large variation in methods that have been used to study colonisation
and super-shedding. Some studies uses faecal counts, others RAMS (with or
without enumeration), and enrichment, culture and confirmation protocols vary.
The most sensitive method for detecting colonisation has been suggested to be
RAMS (Cobbold et al. 2007), which were used to analyse all individual samples
in this study. Spencer et al. (2015) estimated a median sensitivity of 0.78 (95%
CI 0.730.82) for the RAMS and 0.46 (95% CI 0.420.51) for the faecal test.
Spencer et al. (2015) did not use immunomagnetic separation during analysis of
the RAMS, as in this study, and we therefore expect an even higher sensitivity.
For the analysis in paper II and III we also chose to focus on colonised
individuals and not shedding levels as studies have shown that shedding patterns
can be intermittent and RAMS positive animals may have shed high levels just
70
before the sampling. However, there is a risk that a RAMS may have been
contaminated by VTEC O157:H7 that is just passing through the gastrointestinal
system without colonisation. This would lead to wrongly classifying a non-
colonised individual as colonised and reduce the specificity of the test.
Enumeration of faecal shedding was only done for RAMS positive calves as
RAMS are considered to indicate colonisation while low fecal shedding has been
observed to occur also in non-colonised animals (Davis et al. 2006; Cobbold et
al. 2007). In addition, in a pilot study, including 40 animals from two farms
where all faecal samples were enumerated, VTEC O157:H7 was only found in
faecal samples from animals with positive RAMS. Faecal samples were stored
in 2°C during analysis of the RAMS. As it was a concern that this might lead to
reduced levels of VTEC O157:H7 in the samples, we performed a pilot study
where levels of VTEC O157:H7 where analysed before and after storage in the
fridge. There were no indications of reductions in levels of VTEC O157:H7,
instead, a small increase was noticed in some samples. This may have been due
to VTEC O157:H7 increasing in numbers but more likely due to an uneven
distribution of VTEC O157:H7 in the sample or a reduction of other bacteria that
may have competed with VTEC O157:H7 when plated on the CT-SMAC agar.
Super-shedding has been proposed to be shedding more than 104 cfu/g faeces
(Chase-Topping et al. 2008) but a recent study has proposed that 103 may be
sufficient to influence transmission (Spencer et al. 2015). In paper II we chose
the latter definition but also explored the effect of animals shedding more than
104 cfu/g faeces. Using 103 better explained the risk of new infections, compared
to using the higher number 104 (AUC 81% compared to 77%), which supports
the suggestion by Spencer et al. (2015). However, as only 3 animals shed
between 103 and 104 more studies confirming this is needed.
4.4.3 Assessment of risk factors/determinants
There are also some methodological aspects regarding the determinants included
that warrant attention. Collection of information in paper I was done by a postal
questionnaire and there is a risk that not all farmers perceived the questions in
the same way. In addition, the formulation of some questions was not optimal.
For example, in the question about farm contacts farmers were asked to write
down contact farms. When going through the contact patterns it became clear
that there were several cases where one farmer had indicated contact with
another farm which had not stated that farm as a contact. These discrepancies
indicate that farmers varied in how thoroughly they filled in the answers to these
questions. It might have been useful to ask farmers to estimate the number of
contact farms first, before asking for detailed contacts.
71
The animal-based assessments in paper III and IV were all performed by one
person and consequently inter-observer reliability is not an issue in this study.
However, although assessments were practiced before the study began, there is
a risk that intra-observer reliability changed over the two years the study was
performed. Still, considering that colonised and non-colonised animals were
compared within pens (and assessed on the same day) the differences observed
within groups should be relevant and not due to a possible drift in the assessment.
The behavioural observations were performed for 20 minutes during an
active time of the day (when activity in the barn started in the morning). This
means that the behaviours observed reflect a snapshot in time and are not
representative of all behaviours performed during one day. However, this
snapshot represents individual differences within the pen. Longer observation
time had likely increased the frequency of observed behaviour and potentially
added associations that we did not have the power to detect.
Finally, one of the aims was to explore the hypothesis that chronic stress
increases the risk of colonisation. Due to the difficulties of objectively
measuring stress, we combined three different approaches including welfare
assessment, fearfulness and reactivity as well as hair cortisol. However, neither
of these are perfect measures of chronic stress and only provide indications about
stress experienced by the calves. With increased understanding of the HPA-axis
and coping in relation to calves personalities, it may become possible to better
interpret the meaning of these findings.
4.4.4 Statistical methods
A variety of different models and analyses have been used within this project
and especially linear and logistic regression. One important assumption for these
types of models are independence of the predictors (Dohoo et al. 2014). In all
papers there were dependencies between observations and these were generally
handled by including random effects to control for clustering. In paper I, this
was done by including farm in the multivariable analysis, where two sampling
occasions per farm were included in the model, and pen was included as a
random effect to control for clustering in samplings on individual level.
However, random effects was not possible to include in the elastic net regression
(paper III and IV) and in these models pen (paper III) and herd (paper IV) were
included as a fixed effects. There are some recognised drawbacks with
controlling for clustering using fixed effects first of all it is not possible to
include other pen/herd level predictors, which limited the analysis to within-
pen/farm level variables. In addition, inference made from such a model is
specific to the actual pens and not the general population and having pen as a
72
fixed effect increases the number of predictors. The elastic net regression
technique can handle the large number of predictors and as a clear aim of paper
III and IV was to explore individual differences within pen/farm accounting for
this variation was considered appropriate. However, this means that
interpretation of estimates is slightly different in this model compared to the
others. It should also be remembered that the estimates in an elastic net
regression are biased (shrunk by a penalty-term to control variance inflation) and
not directly translatable to OR (James et al. 2013).
The assumption of linearity is another important consideration in regression
analysis and in paper II and IV the usefulness of gams to identify non-linear
associations is clear. Especially the interaction observed in paper II was
interesting and provided important insights and this model might have been
useful when exploring risk factors using multivariable analysis in paper I as well.
Different approaches were used for model building in the different papers.
For the first part of paper I, the glm was reduced using stepwise backwards
selection using Akaike Information Criterion (AIC) while in paper II model
building was guided by a causal diagram. The model in the second part of paper
II was reduced using likelihood ratio test. In both studies all removed variables
were reintroduced one by one to control for confounding by studying the change
of estimates. Although stepwise selection using AIC is often used to reduce
models (Dohoo et al. 2014), and convenient due to the relatively large number
of variables in paper I, the structured approach in paper II is preferred as more
care can be taken to intervening, mediating and confounding variables.
In paper III, we wanted to study the combined effects of the highly
interdependent variables. Adding the principal component regression was an
alternative that was not constrained by correlation between variables or the
assumption of normal distribution in the creation of the components and made it
possible to explore how variables may have had context dependent meaning. It
was also interesting to combine principle component regression, driven by
variance of the observations, with a clustering method, which identifies
homogenous subgroups among observations, to visualise the associations from
different perspectives. There are of course alternatives that could have been
used. For example network analysis and exploratory factor analysis would have
been interesting options (James et al. 2013). In paper IV the main question was
related to the association between hair cortisol to identify indicators of welfare
and interdependencies of variables were not explored further. However, adding
analyses, like principle component regression and cluster analysis, could
potentially provide deeper understanding of hair cortisol concentration and, for
example, personalities of calves.
73
5.1 The end of this story
In the work of this thesis, new insights into the complex associations involved
in transmission and persistence of the pathogen between and within farms have
been gained and possible target areas for on farm measures to reduce prevalence
have been identified. The main conclusions are:
¾ Frequent transmission of virulent strains between nearby farms occur in
cattle dense areas with frequent contacts. Many neighboring farms increase
the risk of infection on a farm and the routes of transmission are related to
both human and animal contacts between farms.
¾ On infected farms, VTEC O157:H7 is most often found among calves and
young stock and only occasionally among dairy cows. Transmission
dynamics within farms is influenced by direct contacts between animals,
presence of super-shedding animals as well as pen hygiene. The most
influential driver is context dependent, which means that different farms may
require different measures.
¾ Drivers of colonisation of individual animals include social and active
behaviour, related to increased exposure of the bacteria, while indicators of
poor health and welfare decrease the risk. This indicates that the variation in
colonisation observed within groups is related to different levels of exposure.
¾ Although individual differences and personality appears to influence risk,
there are no signs of an association between chronic stress and increased
susceptibility to colonisation of VTEC O157:H7.
5
Conclusions, reflections and future
perspectives
74
5.2 What can we tell the farmer in the beginning of our
story?
Although VTEC O157:H7 is a public health issue and not a pathogen of cattle,
farmers are the key when it comes to reducing farm prevalence and reducing
transmission from animals to humans. Farmers are the ones best placed for
preventing sporadic human cases caused by direct contacts with animals by
informing visitors, especially children, to wash their hands after contact with the
animals. Thus, it is important that farmers are aware of this pathogen and that it
can suddenly appear on the farms without animals showing any symptoms of
infection. Perhaps it can also be comforting to know that the farm is probably
not the only farm in the area that is affected, and that on most farms the pathogen
will disappear almost as suddenly as it came. In addition, there are concrete
measures that can help the farm clear the infection faster, like making sure that
bedding is dry and trying to reduce stocking density in animal groups.
Due to the many potential routes of transmission, it may not be possible to
remove the risk of introduction of VTEC O157:H7 on farms. However,
implementation of biosecurity measures reducing human and animal contacts
between other farms will decrease the risk. Such measures could include
providing protective clothing for visitors, especially the ones travelling between
farms, and avoiding pastures where animals can have contact with animals from
other farms. If the latter is not possible, separating animals returning from
pasture, giving them time to clear the pathogen before being reintroduced among
animals on the farm, may be an alternative. Testing would be required to be
certain that a group has cleared the pathogen (as the pathogen does not lead to
symptoms of disease), but it is questionable if this additional cost to the farmer
can be motivated.
Knowing that active and social animals may be the ones most likely to be
colonised emphasises that the farmer has not overlooked signs of poor health in
the animals spreading the bacteria. This also means that cleaning pens between
groups of animals is important, even when groups of animals do not show signs
of disease. Still, cleaning does not prevent the already colonised animals from
shedding and hygiene efforts need to be complemented with measures that
prevent transmission between groups of animals, like the well-known action of
avoiding mixing of groups and preventing the pathogen from moving between
pens by dirty boots, flies and other potential vectors.
Although VTEC O157:H7 is not really a problem for the animals, these
measures will have the added benefit of reducing transmission of other infections
and improve general health of animals. Thus, the added costs of handling this
zoonotic pathogen should be associated with other benefits than reducing the
risk of transmission to humans.
75
5.3 The never ending story
The increasing trend of human cases of VTEC O157:H7 and the common
association with cattle, highlights the importance of reducing the prevalence of
VTEC O157:H7 on cattle farms. During the work of this thesis we have provided
new insights but the results also lead to new questions and hypothesises that
should be explored in future studies.
For handling of infected farms these include:
¾ On-farm persistence is linked with both a contaminated environment and
presence of shedding animals. Thus, in addition to measures targeting
environmental presence of the pathogen in pens, measures that reduce the
shedding of individuals are needed. There is ongoing research of vaccines,
phage-therapy and dietary supplements but maybe it is also time to consider
how management and housing can influence oral exposure to the pathogen.
For example, keeping cow and calf together, providing hay, providing milk
through a teat, using a more gradual weaning process and to provide access
to pasture have been observed to reduce non-nutritional oral activities of
animals, and such management actions may thereby reduce exposure to
VTEC O157:H7.
¾ Despite many suggestions of drivers of transmission and possible on-farm
measures, few intervention studies where the effect of measures are studied
have been performed. To enable estimations of effectiveness, costs and
benefits such studies are needed.
On animal level many questions about susceptibility and resistance to
colonisation, which could help development of on-farm measures remain to be
explored.
For example:
¾ It is clear that super-shedding matters and that there are individual drivers
influencing colonisation but to which extent does individual characteristics
(in behaviour, immunity and/or susceptibility) influence the risk of
colonisation compared to increased environmental exposure?
76
¾ Are shedding levels really higher in younger animals or are they simply
higher in naïve animals encountering VTEC O157:H7 for the first time? To
which extent can previous exposure to less virulent strains circulating in the
cattle population influence colonisation of VTEC O157:H7?
¾ Are there different drivers of colonisation in young and old animals? If young
animals are more susceptible to infection, but increased colonisation
resistance develops over time, it is important to identify the source of this
resistance. Is it related to changes in behaviour, microbiome, stress or
immunity? Can we enhance this resistance in some way?
In paper I the added benefit of whole genome sequencing for understanding
transmission is clear. Thanks to decreasing costs and increased availability of
molecular methods it will be possible to look into:
¾ If on-farm transmission occurs and is driven by colonised animals or an
environmental/other on-farm reservoir?
¾ The differences between on-farm persistence and reoccurrence of the
bacteria. To which extent does on-farm persistence actually represent new
introductions?
¾ What intrinsic factors are important for VTEC O157:H7 and enhances strains
capability to persist in environment and to colonise cattle and humans?
¾ Is VTEC O157:H7 dependent or inhibited by other bacteria and are there
other, less virulent, bacteria that can reduce VTEC O157:H7?
77
Abdelfattah, E.M., Karousa, M.M., Lay, D.C., Marchant-Forde, J.N. & Eicher, S.D. (2018). Short
communication: Effect of age at group housing on behavior, cortisol, health, and leukocyte
differential counts of neonatal bull dairy calves. Journal of Dairy Science, vol. 101 (1), pp. 596
602. DOI: 10.3168/jds.2017-12632
Abu-Ali, G.S., Lacher, D.W., Wick, L.M., Qi, W. & Whittam, T.S. (2009). Genomic diversity of
pathogenic Escherichia coli of the EHEC 2 clonal complex. BMC Genomics, vol. 10, pp. 116
Ahmad, A., Nagaraja, T.G. & Zurek, L. (2007). Transmission of Escherichia coli O157:H7 to cattle by
house flies. Preventive Veterinary Medicine, vol. 80 (1), pp. 7481
Albihn, A., Eriksson, E., Walle, C. & Aspán, A. (2003). Verotoxinogenic Escherichia coli ( VTEC )
O157:H7 - A Nationwide Swedish Survey of Bovine Faeces. Acta veterinaria Scandinavica, vol.
44 (1), pp. 4352
Arimizu, Y., Kirino, Y., Sato, M.P., Uno, K., Sato, T., Gotoh, Y., Auvray, F., Brugere, H., Oswald, E.,
Mainil, J.G., Anklam, K.S., Döpfer, D., Yoshino, S., Ooka, T., Tanizawa, Y., Nakamura, Y.,
Iguchi, A., Morita-Ishihara, T., Ohnishi, M., Akashi, K., Hayashi, T. & Ogura, Y. (2019). Large-
scale genome analysis of bovine commensal Escherichia coli reveals that bovine-adapted E. Coli
lineages are serving as evolutionary sources of the emergence of human intestinal pathogenic
strains. Genome Research, vol. 29 (9), pp. 14951505
Arthur, T.M., Keen, J.E., Bosilevac, J.M., Brichta-Harhay, D.M., Kalchayanand, N., Shackelford, S.D.,
Wheeler, T.L., Nou, X. & Koohmaraie, M. (2009). Longitudinal study of Escherichia coli
O157:H7 in a beef cattle feedlot and role of high-level shedders in hide contamination. Applied
and Environmental Microbiology, vol. 75 (20), pp. 65156523
Bach, S.J., Mcallister, T.A., Mears, G.J. & Schwartzkopf-genswein, K.S. (2016). Long-Haul Transport
and Lack of Preconditioning Increases Fecal Shedding of Escherichia coli and Escherichia coli
O157:H7 by Calves. Journal of Food Protection, vol. 67 (4), pp. 672678
Bai, X., Mernelius, S., Jernberg, C., Einemo, I.M., Monecke, S., Ehricht, R., Löfgren, S. & Matussek,
A. (2018). Shiga toxin-producing Escherichia coli infection in Jönköping county, Sweden:
Occurrence and molecular characteristics in correlation with clinical symptoms and duration of
stx shedding. Frontiers in Cellular and Infection Microbiology, vol. 8 (MAY) DOI:
10.3389/fcimb.2018.00125
Bates, D., Mächler, M., Bolker, B. & Walker, S. (2015). Fitting Linear Mixed-Effects Models using
lme4. Journal of Statistical Software, vol. 67 (1), pp. 148
Baümler, A.J. & Sperandio, V. (2016). Interactions between the microbiota and pathogenic bacteria in
References
78
the gut. Nature, vol. 535 (7610), pp. 8593
Belongia, E.A., Chyou, P., Greenlee, R.T., PerezPerez, G., Bibb, W.F. & DeVries, E.O. (2003).
Diarrhea Incidence and FarmRelated Risk Factors for Escherichia coli O157:H7 and
Campylobacter jejuni Antibodies among Rural Children . The Journal of Infectious Diseases,
vol. 187 (9), pp. 14601468
Benjamin, L.A., Jay-Russell, M.T., Atwill, E.R., Cooley, M.B., Carychao, D., Larsen, R.E. & Mandrell,
R.E. (2015). Risk factors for Escherichia coli O157 on beef cattle ranches located near a major
produce production region. Epidemiology and Infection, vol. 143 (1), pp. 8193
Besser, T.E., Richards, B.L., Rice, D.H. & Hancock, D.D. (2001). Escherichia coli O157:H7 infection
of calves: infectious dose and direct contact transmission. Epidemiology and Infection, vol. 127,
pp. 555560
Besser, T.E., Schmidt, C.E., Shah, D.H. & Shringi, S. (2014). “Preharvest” Food Safety for Escherichia
coli O157 and Other Pathogenic Shiga Toxin-Producing Strains. Microbiology Spectrum, vol. 2
(5), pp. 421431
Bokkers, E.A.M. & Koene, P. (2001). Activity, oral behaviour and slaughter data as welfare indicators
in veal calves: A comparison of three housing systems. Applied Animal Behaviour Science, vol.
75 (1), pp. 115
Bono, J.L., Smith, T.P.L., Keen, J.E., Harhay, G.P., McDaneld, T.G., Mandrell, R.E., Jung, W.K.,
Besser, T.E., Gerner-Smidt, P., Bielaszewska, M., Karch, H. & Clawson, M.L. (2012).
Phylogeny of shiga toxin-producing escherichia coli o157 isolated from cattle and clinically ill
humans. Molecular Biology and Evolution, vol. 29 (8), pp. 20472062
Boone, J.T., Campbell, D.E., Dandro, A.S., Chen, L. & Herbein, J.F. (2016). A rapid immunoassay for
detection of Shiga toxin-producing Escherichia coli directly from Human Fecal samples and its
performance in detection of toxin subtypes. Journal of Clinical Microbiology, vol. 54 (12), pp.
30563063
Borderas, T.F., De Passillé, A.M. & Rushen, J. (2008). Behavior of dairy calves after a low dose of
bacterial endotoxin. Journal of Animal Science, vol. 86 (11), pp. 29202927
Broom, D.M. (1986). Indicators of poor welfare. British Veterinary Journal, vol. 142 (6), pp. 524526
Brown-Brandl, T.M., Berry, E.D., Wells, J.E., Arthur, T.M. & Nienaber, J.A. (2009). Impacts of
individual animal response to heat and handling stresses on escherichia coli and E. coli O157:H7
fecal shedding by feedlot cattle. Foodborne Pathogens and Disease, vol. 6 (7), pp. 855864
Browne, A.S., Midwinter, A.C., Withers, H., Cookson, A.L., Biggs, P.J., Marshall, J.C., Benschop, J.,
Hathaway, S., Haack, N.A., Akhter, R.N. & French, N.P. (2018). Molecular epidemiology of
Shiga toxinproducing Escherichia coli (STEC) on New Zealand dairy farms: Application of a
culture-independent assay and whole-genome sequencing. Applied and Environmental
Microbiology, vol. 84 (14) e00481-18. DOI: 10.1128/AEM.00481-18
Burnard, C., Ralph, C., Hynd, P., Edwards, J.H. & Tilbrook, A. (2017). Hair cortisol and its potential
value as a physiological measure of stress response in human and non-human animals. Animal
Production Science, vol. 57, pp. 401414
Burnett, T.A., Madureira, A.M.L., Silper, B.F., Nadalin, A., Tahmasbi, A., Veira, D.M. & Cerri, R.L.A.
(2014). Short communication: Factors affecting hair cortisol concentrations in lactating dairy
cows. Journal of Dairy Science, vol. 97 (12), pp. 76857690. DOI: 10.3168/jds.2014-8444
Burnett, T.A., Madureira, A.M.L., Silper, B.F., Tahmasbi, A., Nadalin, A., Veira, D.M. & Cerri, R.L.A.
(2015). Relationship of concentrations of cortisol in hair with health, biomarkers in blood, and
reproductive status in dairy cows. Journal of Dairy Science, vol. 98 (7), pp. 44144426. DOI:
79
10.3168/jds.2014-8871
Byrne, C.M., Erol, I., Call, J.E., Kaspar, C.W., Buege, D.R., Hiemke, C.J., Fedorka-Cray, P.J., Benson,
A.K., Wallace, F.M. & Luchansky, J.B. (2003). Characterization of Escherichia coli O157:H7
from downer and healthy dairy cattle in the upper midwest region of the United States. Applied
and Environmental Microbiology, vol. 69 (8), pp. 46834688
Byrne, L., Dallman, T.J., Adams, N., Mikhail, A.F.W., McCarthy, N. & Jenkins, C. (2018). Highly
pathogenic clone of shiga toxinproducing escherichia coli O157:H7, England and Wales.
Emerging Infectious Diseases, vol. 24 (12), pp. 23032308
Calderwood, S.B., Akeson, D.W.K., Keusch, G.T., Barrett, T.J. & Griffin, P.M. (1996). Proposed New
nomenclature for SLT (VT) family. ASM News., vol. 62, pp. 118119
Carlson, B.A., Nightingale, K.K., Mason, G.L., Ruby, J.R., Choat, W.T., Loneragan, G.H., Smith, G.C.,
Sofos, J.N. & Belk, K.E. (2009). Escherichia coli O157:H7 strains that persist in feedlot cattle
are genetically related and demonstrate an enhanced ability to adhere to intestinal epithelial
cells. Applied and Environmental Microbiology, vol. 75 (18), pp. 59275937
Cernicchiaro, N., Pearl, D.L., McEwen, S. a, Harpster, L., Homan, H.J., Linz, G.M. & Lejeune, J.T.
(2012). Association of wild bird density and farm management factors with the prevalence of E.
coli O157 in dairy herds in Ohio (2007-2009). Zoonoses and public health, vol. 59 (5), pp. 320
9. DOI: 10.1111/j.1863-2378.2012.01457.x
Chase-Topping, M.E., Gally, D., Low, C., Matthews, L. & Woolhouse, M. (2008). Super-shedding and
the link between human infection and livestock carriage of Escherichia coli O157. Nature
reviews. Microbiology, vol. 6 (12), pp. 90412. DOI: 10.1038/nrmicro2029
Chase-Topping, M.E., McKendrick, I.J., Pearce, M.C., MacDonald, P., Matthews, L., Halliday, J.,
Allison, L., Fenlon, D., Low, J.C., Gunn, G. & Woolhouse, M.E.J. (2007). Risk factors for the
presence of high-level shedders of Escherichia coli O157 on Scottish farms. Journal of clinical
microbiology, vol. 45 (5), pp. 1594603. DOI: 10.1128/JCM.01690-06
Chen, S., Sanderson, M. & Lanzas, C. (2013). Investigating effects of between- and within-host
variability on Escherichia coli O157 shedding pattern and transmission. Preventive Veterinary
Medicine, vol. 109 (12), pp. 4757
Cho, S., Fossler, C.P., Diez-Gonzalez, F., Wells, S.J., Hedberg, C.W., Kaneene, J.B., Ruegg, P.L.,
Warnick, L.D. & Bender, J.B. (2009). Cattle-level risk factors associated with fecal shedding of
Shiga toxin-encoding bacteria on dairy farms in Minnesota, USA. Canadian Veterinary Journal,
vol. 73, pp. 151156
Cho, S., Fossler, C.P., Diez-Gonzalez, F., Wells, S.J., Hedberg, C.W., Kaneene, J.B., Ruegg, P.L.,
Warnick, L.D. & Bender, J.B. (2013). Herd-level risk factors associated with fecal shedding of
Shiga toxin-encoding bacteria on dairy farms in Minnesota, USA. Canadian Veterinary Journal,
vol. 54 (7), pp. 693697
Cobbaut, K., Berkvens, D., Houf, K., De Deken, R. & De Zutter, L. (2009). Escherichia coli O157
prevalence in different cattle farm types and identification of potential risk factors. Journal of
Food Protection, vol. 72 (9), pp. 18481853. DOI: 10.4315/0362-028X-72.9.1848
Cobbold, R.N., Hancock, D.D., Rice, D.H., Berg, J., Stilborn, R., Hovde, C.J. & Besser, T.E. (2007).
Rectoanal Junction Colonization of Feedlot Cattle by Escherichia coli O157: H7 and Its
Association with Supershedders and Excretion Dynamics. Applied and environmental
microbiology, vol. 73 (5), pp. 15631568. DOI: 10.1128/AEM.01742-06
Comin, A., Peric, T., Corazzin, M., Veronesi, M.C., Meloni, T., Zufferli, V., Cornacchia, G. & Prandi,
A. (2013). Hair cortisol as a marker of hypothalamic-pituitary-adrenal axis activation in Friesian
80
dairy cows clinically or physiologically compromised. Livestock Science, vol. 152 (1), pp. 3641
Corbishley, A., Ahmad, N.I., Hughes, K., Hutchings, M.R., McAteer, S.P., Connelley, T.K., Brown, H.,
Gally, D.L. & McNeilly, T.N. (2014). Strain-dependent cellular immune responses in cattle
following Escherichia coli O157:H7 colonization. Infection and Immunity, vol. 82 (12), pp.
51175131
Cramer, M.C., Ollivett, T.L. & Stanton, A.L. (2016). Associations of behavior-based measurements and
clinical disease in preweaned, group-housed dairy calves. Journal of Dairy Science, vol. 99 (9),
pp. 74347443. DOI: 10.3168/jds.2015-10207
Cramer, M.C. & Stanton, A.L. (2015). Associations between health status and the probability of
approaching a novel object or stationary human in preweaned group-housed dairy calves.
Journal of Dairy Science, vol. 98 (10), pp. 72987308. DOI: 10.3168/jds.2015-9534
Cray, W.C. & Casey, T. (1998). Effect of Dietary Stress on Fecal Shedding of Escherichia coli O157:
H7 in Calves. Applied and Environmental Microbiology, vol. 64 (5), pp. 16. DOI:
10.1128/aem.64.5.1975-1979.1998
Crump, J.A., Sulka, A.C., Langer, A.J., Schaben, C., Crielly, A.S., Gage, R., Baysinger, M., Moll, M.,
Withers, G., Toney, D.M., Hunter, S.B., Hoekstra, R.M., Wong, S.K., Griffin, P.M. & Van
Gilder, T.J. (2002). An outbreak of Escherichia coli O157:H7 infections among visitors to a
dairy farm. The New England journal of medicine, vol. 347 (8), pp. 555560
Dallman, T.J., Ashton, P.M., Byrne, L., Perry, N.T., Petrovska, L., Ellis, R., Allison, L., Hanson, M.,
Holmes, A., Gunn, G.J., Chase-Topping, M.E., Woolhouse, M.E.J., Grant, K.A., Gally, D.L.,
Wain, J. & Jenkins, C. (2015). Applying phylogenomics to understand the emergence of Shiga-
toxin-producing Escherichia coli O157:H7 strains causing severe human disease in the UK.
Microbial Genomics, vol. 1 (3), pp. 113. DOI: 10.1099/mgen.0.000029
Davenport, M.D., Tiefenbacher, S., Lutz, C.K., Novak, M.A. & Meyer, J.S. (2006). Analysis of
endogenous cortisol concentrations in the hair of rhesus macaques. General and Comparative
Endocrinology, vol. 147 (3), pp. 255261
Davis, M.A., Rice, D.H., Sheng, H., Dale, D., Besser, T.E., Cobbold, R., Hovde, J., Hancock, D.D. &
Hovde, C.J. (2006). Comparison of Cultures from Rectoanal-Junction Mucosal Swabs and Feces
for Detection of Escherichia coli O157 in Dairy Heifers. Applied and environmental
microbiology, vol. 72 (5), pp. 37663770
Dean-Nystrom, E.A., Bosworth, B.T., Cray, W.C. & Moon, H.W. (1997). Pathogenicity of Escherichia
coli O157:H7 in the intestines of neonatal calves. Infection and Immunity, vol. 65 (5), pp. 1842
1848
Denamur, E. (2011). The 2011 Shiga toxin-producing Escherichia coli O104:H4 German outbreak: A
lesson in genomic plasticity. Clinical Microbiology and Infection, vol. 17 (8), pp. 11241125
European Society of Clinical Microbiology and Infectious Diseases. DOI: 10.1111/j.1469-
0691.2011.03620.x
Dobrindt, U. (2005). (Patho-)Genomics of Escherichia coli. International Journal of Medical
Microbiology, vol. 295 (67), pp. 357371
Dohoo, I., Martin, W. & Stryhn, H. (2014). Veterinary epidemiologic research. 2nd. ed. Charlottetown:
VER inc.
Ducarmon, Q.R., Zwittink, R.D., Hornung, B.V.H., van Schaik, W., Young, V.B. & Kuijper, E.J.
(2019). Gut Microbiota and Colonization Resistance against Bacterial Enteric Infection.
Microbiology and Molecular Biology Reviews, vol. 83 (3), pp. 129
Duncan, I.J.H. (2005). Science-based assessment of animal welfare: farm animals. Revue Scientifique et
81
Technique Office International des Epizooties, vol. 24 (2), pp. 483492.
Dylla, B.L., Vetter, E.A., Hughes, J.G. & Cockerill, F.R. (1995). Evaluation of an immunoassay for
direct detection of Escherichia coli O157 in stool specimens. Journal of Clinical Microbiology ,
vol. 33 (1), pp. 222224
ECDC (2020) Disease data from ECDC Surveillance Atlas - Escherichia coli infection, Availible at:
https://www.ecdc.europa.eu/en/escherichia-coli-ecoli/surveillance/atlas [2020-03-02]
EFSA (2013). Scientific Opinion on VTEC-seropathotype and scientific criteria regarding pathogenicity
assessment. EFSA Journal, vol. 11 (4), 3138. DOI: 10.2903/j.efsa.2013.3138
ESR (Institute of Environmental Science and Research Ltd) (2020) Notifiable Diseases Intelligence
Dashboard, Availible at: https://www.esr.cri.nz/our-services/consultancy/public-
health/notifiable-diseases-intelligence-dashboard/ [2020-03-02]
Ellis-Iversen, J., Smith, R.P., Van Winden, S., Paiba, G., Watson, E., Snow, L.C., Cook, Alasdair, J.C.
& Escherichia, V. (2008). Farm practices to control E. coli O157 in young cattle-A randomised
controlled trial. Veterinary Research, vol. 39 (3), DOI: 10.1051/vetres:2007041
Eppinger, M. & Cebula, T.A. (2015). Future perspectives, applications and challenges of genomic
epidemiology studies for food-borne pathogens: A case study of Enterohemorrhagic Escherichia
coli (EHEC) of the O157:H7 serotype. Gut Microbes, vol. 6 (3), pp. 194201
Eriksson, E., Aspan, A., Gunnarsson, A. & Vågsholm, I. (2005). Prevalence of verotoxin-producing
Escherichia coli (VTEC) O157 in Swedish dairy herds. Epidemiology and Infection, vol. 133
(2), pp. 349358. DOI: 10.1017/S0950268804003371
Færevik, G., Andersen, I.L., Jensen, M.B. & Bøe, K.E. (2007). Increased group size reduces conflicts
and strengthens the preference for familiar group mates after regrouping of weaned dairy calves
(Bos taurus). Applied Animal Behaviour Science, vol. 108 (34), pp. 215228
Ferdous, M., Friedrich, A.W., Grundmann, H., de Boer, R.F., Croughs, P.D., Islam, M.A., Kluytmans-
van den Bergh, M.F.Q., Kooistra-Smid, A.M.D. & Rossen, J.W.A. (2016). Molecular
characterization and phylogeny of Shiga toxinproducing Escherichia coli isolates obtained
from two Dutch regions using whole genome sequencing. Clinical Microbiology and Infection,
vol. 22 (7), pp. 642.e1-642.e9 DOI: 10.1016/j.cmi.2016.03.028
Folkhälsomyndigheten (2020). Enterohemorragisk E. coli infektion (EHEC). Available at:
http://www.folkhalsomyndigheten.se/amnesomraden/statistik-och-
undersokningar/sjukdomsstatistik/enterohemorragisk-e-coli-infektion-ehec/ [2020-03-02]
Frank, C., Kapfhammer, S., Werber, D., Stark, K. & Held, L. (2008). Cattle Density and Shiga Toxin-
Producing Escherichia coli Infection in Germany: Increased Risk for Most but Not All
Serogroups. Vector-Borne and Zoonotic Diseases, vol. 8 (5), pp. 635644. DOI:
10.1089/vbz.2007.0237
Franz, E., Van Hoek, A.H.A.M., Van Der Wal, F.J., De Boer, A., Zwartkruis-Nahuis, A., Van Der
Zwaluw, K., Aarts, H.J.M. & Heuvelinkd, A.E. (2012). Genetic features differentiating bovine,
food, and human isolates of Shiga toxin-producing Escherichia coli O157 in The Netherlands.
Journal of Clinical Microbiology, vol. 50 (3), pp. 772780
Franz, E., Rotariu, O., Lopes, B.S., Macrae, M., Bono, J.L., Laing, C., Gannon, V., Söderlund, R.,
Hoek, A.H.A.M. Van, Friesema, I., French, N.P., George, T., Biggs, P.J., Jaros, P., Rivas, M.,
Chinen, I., Campos, J., Jernberg, C., Gobius, K., Mellor, G.E., Chandry, P.S. & Perez-reche, F.
(2019). Phylogeographic analysis reveals multiple international transmission events have driven
the global emergence of Escherichia coli O157:H7. Clinical Infectious Diseases