Routinely collected herd management data in a variety of formats were collated from 468 dairy herds, and novel objective measures of data recording quality were developed and applied. This revealed that there was a substantial amount of variation in data quality between herds, and the vast majority of herds failed to meet the threshold level for at least one of the data quality measures used. Analysis of trends in reproductive performance across the herds with good quality fertility event recording suggested that their fertility was generally declining through the first half of the 2000s, but there was some evidence that improvements in submission rate were beginning to reverse this decline in the later years studied (up to 2007). Associations between reproduction and two endemic diseases common in dairy cattle (mastitis and lameness) were explored using multilevel discrete time survival modelling, and probabilistic sensitivity analysis (PSA) used to contextualise and illustrate the results. In both cases, statistical modelling revealed significant and sizeable associations between disease events and reproductive outcomes at lactation level. However, simulation and application of PSA showed that a herd’s incidence rate of either disease was highly unlikely to influence its overall reproductive performance to a clinically relevant degree when other inputs to herd fertility were also considered. Factors associated with the proportion of serves leading to a pregnancy (pregnancy rate) were explored using multilevel logistic regression modelling. This revealed that relatively little of the variation in herd pregnancy rate is explainable by routinely recorded milk recording data (including constituent concentration in early lactation as well as daily and lactation yields). A large amount of the unexplained variation was revealed to be at herd level and very little at cow level, suggesting that investigation of herd management practices associated with pregnancy rate would be rewarding.

The small East African shorthorn zebu (EASZ) is an indigenous stable zebu x taurine admixed cattle population highly adapted to the local African environment, i.e. disease resistance, thermotolerance, sustain themselves on poor feed quality, in comparison to the exotic productive cattle. Understanding the genetic control of their adaptations will help sustainable breeding improvement programs, both within populations and through crossbreeding. In this thesis, genome-wide SNP data, from two commercially available SNP arrays; the Illumina BovineSNP50 BeadChip v.1 and the Illumina BovineHD BeadChip, and full genome sequence of 10 pooled EASZ samples were used to define candidate regions with signatures of positive selection in the genome of EASZ. In chapter 2, the low-density SNP chip has defined 24 candidate regions (21 in the autosomes and three in the sex chromosome X (BTA X)) by two Extended Haplotype Homozygosity (EHH)-based (intra-population iHS and inter-population Rsb) analyses and an inter-population FST approach. Characterising the EASZ genome with high-density genome-wide SNP analyses (meta-SS, Rsb, iHS and ΔAF) and pooled heterozygosity (Hp) sequence analysis, in chapters 3 and 4, identified a total of 266 autosomal and 20 BTA X candidate regions. Analysing genome-wide SNP data of zebu cattle populations from Uganda and Nigeria classified some of these regions as East African zebu-sharing and East and West African zebu-sharing candidate regions. Genes and quantitative trait loci (QTL) associated with adaptive traits (e.g., reproduction, immunity and heat stress) were found within the identified candidate regions. Moreover, Nine non-synonymous variants were found in five candidate genes and considered as possible causative mutations under selection. Finally, the mitochondrial DNA (mtDNA) of 13 EASZ were investigated and affiliated to T1a, T1b and T1b1 sub-haplogroups. No selective advantage was found to be associated with taurine mtDNA over the zebu type. Although purifying selection is the main selective factor influencing African cattle mtDNA, in Cox-2 gene of the T1b/T1b1 sub-haplogroups a non-synonymous mutation may confer a selective advantage. The results of this thesis are just the first milestone to define the valuable genetic material in EASZ, which should be conserved in future composite cattle upon informed selective crossbreeding and/or the recently introduced gene modification techniques (TALEN and CRISPER/Cas9).

Obesity is a growing problem for humans and their horses and ponies, yet emaciated animals still form an important part of the work of equine welfare charities. Non-invasive yet reliable methods of assessing equine body fat are required in order to promote management procedures to improve animal welfare. The overall objective of this work was to investigate the application of a horse-specific body condition scoring system in ponies in order to either validate or revise it, or even replace it with a novel system if necessary. Seasonal differences in appetite, body mass (BM) gain, body condition score (BCS) change and direct (ultrasound) and indirect (morphometry and D2O dilution) measures of body fat were explored in two relatively homogeneous groups of mature Welsh mountain pony mares, studied over summer (June –September 2007) and winter (January-April 2008). The ponies in each group were paired so that, at study outset, two ponies were ‘thin’ (BCS, 1-3/9); two were ‘moderate’ (BCS, 4-6/9); and two were ‘obese’ (BCS, 7-9/9). The greatest appetites (peak 4.6% BM as DMI), increases in body mass (~60kg) and in BCS (~3 points) were recorded for ponies of non-obese outset condition in summer (non-ObS, n=4). For ponies of non-obese outset condition in winter (non-ObW, n=3), appetites peaked at 3.5% BM as DMI, BM increased by a mean of 50kg and BCS increased by ~2 points over the 3 month study period. Appetites for all obese (Ob, n=4) ponies remained almost constant (~2% BM as DMI; peak 2.3% BM as DMI) and minimal changes in BM (n=3) and BCS (n=4) were recorded, regardless of season. All measures of body fat increased for non-Ob ponies (non-ObS>non-ObW). An exponential relationship was determined between body fat content and BCS and for values > 6, BCS was not a useful predictor of actual body fat content. The endogenous circannual mechanisms to encourage winter weight loss were insufficient to prevent the development of obesity in ad libitum fed ponies. The effects of dietary restriction to 1% BM as DMI were studied in a group of 5 overweight or obese mature pony mares (BCS 5.6-8/9). Those measures outlined above were likewise recorded. All ponies remained healthy throughout the 12 week trial. Overall, BM reduced by 1% of outset BM per week. Approximately half the lost BM comprised fat, but fatter animals lost relatively more fat. Despite an average loss of ~30 kg BM, BCS did not change appreciably suggesting that BCS was a relatively poor indicator of early weight/fat loss in obese ponies. The relationships between BCS, direct (ultrasonic) and indirect (morphometric and D2O dilution derived) measures of body fat and actual body fat content determined by both physical dissection and chemical cadaver analysis were explored using 7 donated mature Welsh pony mares (BCS 1.25 to 7/9). Body ‘fat’ content (dissected white adipose tissue or chemically-extracted lipid fractions) was the most variable constituent of the cadavers (up to 1/3rd body mass), and was non-linearly related to BCS. From these studies, it was also possible to validate the D2O dilution technique for the measurement of total body water and fat in ponies. Contemporaneously gathered data for BCS and body fat (D2O dilution) from 48 separate observations were explored statistically. A non-linear association between body fat content and BCS was confirmed, with a cut off value of BCS 7/9, above which BCS was less useful for determining body fat content. A novel BCS system was created and is undergoing field trials.

Equine laminitis manifests as a painful condition of the foot and is characterised by disruption of the dermo-epidermal lamellar bond within the equine hoof. Chronic pain in laminitis leads to debilitation and if unchecked euthanasia of the horse and is therefore a significant welfare issue for horses worlwide. P2X receptors are shown to have a pivotal role in pain pathways as well as other tissue functions making them attractive candidates for research in naturally occurring diseases such as equine laminitis. In this study, the presence and distribution of P2X receptors in tissues of the equine digit and its associated vasculature and innervation was investigated. The investigation was extended to tissues from horses suffering from naturally occurring chronic active lamintis in an attempt to identify changes in distribution or expression of the receptors in these tissues. Finally, the functionality of the P2X receptors expressed in equine nervous tissue was investigated. Gene expression (mRNA) for all seven receptors was found in the studied tissues and protein expression for P2X1-3 and 7 receptor subtype proteins in these tissues was confirmed with Western blot analysis. Immunohistochemical analysis of P2X1-3 and 7 receptor subtypes distribution showed differential cellular distribution within normal equine tissue and differential distribution and expression between normal and diseased tissues too. In particular, in a horse with shorter duration of chronic laminitis (1 month), P2X3 receptor subtype immunostaining in DRG neurons presented a marked increase in large neurons which coincided with a decrease in the proportion of these neurons in the DRG. Also, horses suffering from chronic laminitis displayed expression of P2X3 receptor subtype in smooth muscle cells of the palmar digital artery which was absent in normal horses. In chronic laminitic hooves, P2X7 receptor subtype showed particularly strong staining in partially keratinised cells. Finally, intracellular calcium measurements provided evidence for the presence of functional purinergic and P2X receptors in equine spinal cord and DRG cells, respectively. The expression of P2X receptor subtypes in key equine tissues for laminitis and chronic pain in laminitis and their differential expression in laminitic tissues highlights a potential role in chronic laminitis molecular mechanisms. Further research in this area may provide insights into putative targets for disease modulation and pain management in chronic laminitis and other conditions of the horse associated with chronic pain.

Canine infections with antimicrobial resistant (AMR), particularly multi-drug resistant (MDR) bacteria are increasing, severely limiting therapeutic options, and representing an animal health issue. In addition, with potential transfer of AMR bacteria between dogs, their environment, humans and other animals, there may also be a public health risk. Commensal isolates can be a source of clinical infections and studies reporting the prevalence of AMR and risk factors for such isolates are important. Furthermore, one of the most significant impacts upon commensal bacterial populations is antimicrobial therapy that may select for pre-existing AMR organisms or transmission of resistance determinants. The aim of this work was to investigate AMR amongst canine commensal bacterial populations and the effects of five different antimicrobials, authorised to treat dogs in the UK, on these populations both during and after therapy. Three groups of dogs were enrolled: healthy non-antimicrobial treated, non-vet visiting dogs (n = 28), to investigate longitudinal carriage of faecal E. coli; healthy non-antimicrobial treated, non-vet visiting, dogs (n = 73) and antimicrobial treated, non-hospitalised dogs (n = 127) to investigate longitudinal carriage of mucosal staphylococci and faecal E. coli. Staphylococci and E. coli isolated from swabs (nose/perineum) and faecal samples respectively, were tested for phenotypic AMR and carriage of resistance genes by PCR assay. Staphylococci were assigned to species by PCR assay (nuc gene), MALDI-TOF-MS and sequencing (tuf gene). Healthy dog E. coli underwent phylo-typing, and a selection of longitudinal healthy dog E. coli isolates were genotyped. Questionnaire data were used to formulate independent variables. Statistical analysis included Pearson’s Chi-square, survival analysis and multivariable logistic regression; multilevel for clustered data. The prevalence of meticillin-resistant (MR; 42%) and MDR staphylococci (resistant to ≥ 3 antimicrobial classes; 34%) was high amongst healthy dogs, however MR-coagulase positive staphylococci were not detected. The most common species detected was S. epidermidis (52% of dogs), followed by S. pseudintermedius (44%). S. aureus was only detected in a small number of dogs (8%). Faecal E. coli with AMR to at least one tested drug (63%), MDR (30%) and AmpC-production (16%) were prevalent in healthy dogs, however ESBL-producers (1%) were rare. Healthy dogs carried a predominance of phylogenetic group B1; group B2 E. coli isolates were less likely to have AMR while group D isolates were more likely. Carriage of E. coli with AMR to at least one tested drug was common and persistent, whereas MDR, AmpC- and ESBL-types were intermittent or transient. Genotyping revealed high intra-dog diversity with frequent new genotypes and resistance phenotypes detected over time. AMR staphylococci and E. coli were detected in more dogs following antimicrobial treatment than baseline, but generally returned to pre-treatment levels within three months. Eating raw meat/animal stools, living with other dogs/in-contact humans or pets that had been hospitalised/in-contact humans working with farm animals were associated with the detection of AMR canine commensal bacteria. In particular, following treatment with beta-lactams or fluoroquinolones there was a significant increase in the detection of MDR and AmpC-producing E. coli or MR- and MDR staphylococci, respectively. However significant differences were not detected at one month after the end of treatment. Antimicrobial therapy is a risk factor for antimicrobial resistant commensal bacteria in dogs and recovery may take up to three months after the end of treatment. This highlights the importance of prudent antimicrobial use and prescribing guidelines. However other factors, such as diet, in-contacts, co-selection and bacterial fitness may be involved in the carriage of resistant bacteria and should be considered.

Feline calicivirus (FCV) is an important veterinary pathogen of domestic cats. It is highly variable and rapidly evolving such that many different strains of varying pathogenicity exist from avirulent, to mildly virulent and hypervirulent. FCV is one of only a small number of viruses within the Caliciviridae to propagate readily in vitro. This property has made FCV a model for the study of the viruses of the calicivirus family. However, relatively little remains known about the way in which these viruses affect host cellular mechanisms. Despite the increasing use of post-genomic technologies in other areas of science, what is known about the effects of calicivirus replication upon host cells has generally relied on relatively conventional technologies to target known defined pathways. In order to investigate FCV modulation of the host cell proteome a 2D-DIGE experiment was performed upon a single viral strain (FCV-F9) with four time points analysed. The results of this analysis showed that 30 host proteins were differentially expressed in the cells infected with FCV-F9 and of these 30 proteins, 14 were successfully identified using mass spectrometry and 11 of these had cytoskeletal proteins within their identifications. To further evaluate these host proteome changes and to explore whether variations in strain pathogenicity were associated with varying affects in the host proteome, a second 2D-DIGE experiment was performed upon four strains of FCV isolated from cats exhibiting a range of clinical signs. As in the first experiment, host cytoskeletal proteins appeared to be targeted for modulation during infection. Overall, this experiment uncovered 131 modulated host proteins with 61 proteins successfully identified by mass spectrometry and 16 of these had cytoskeletal proteins within their identifications. Also found within this experiment was evidence of the putative virulent FCV strain UKOS-A having a different affect upon 14 of the differentially expressed proteins to the other three FCV strains tested. UKOS-A was found to down-regulate these proteins whereas the other strains had an up-regulatory effect, these proteins could be a potential marker of hypervirulence but would need further study to confirm their validity. To follow the modulation of the identified host cytoskeletal proteins, immunofluorescent staining coupled with confocal microscopy was used to track FCV capsid protein and two cytoskeletal proteins vimentin and tubulin. Results showed that during infection with the four FCV strains tested both vimentin and tubulin were both significantly altered during infection with the intensity of fluorescent staining increasing in all cases. To complement the proteomic analyses, an experiment designed to discover differentially expressed gene transcripts was conducted using RNA-seq technology. Overall, there were 354 genes found to be differentially transcribed within the infected cells at both four and seven hrs post infection (p.i.) with the majority of modulated genes experiencing down-regulation at four hrs p.i. and up-regulation at seven hrs p.i.. Once again, genes within the cytoskeletal regulation pathway were up-regulated alongside genes from the apoptosis regulatory pathway and immune system response pathways. To further characterise the putative virulent strain used in earlier studies, attempts were made to sequence the rest of this virus using polymerase chain reaction and conventional Sanger sequencing technologies. Although almost 50% of the UKOS-A strain was sequenced, the sequence obtained was no more similar or different to other virulent systemic disease (VSD) causing strains. Overall, the most marked effects of FCV infection upon the host were found amongst the cytoskeletal genes and proteins which have been demonstrated to undergo extensive modification. This is a completely novel finding in this family of viruses. Further studies will be needed to identify the consequences of the identified cytoskeletal changes to both the host and the virus.

Anterior cruciate ligament rupture (ACL) is a major source of morbidity in the dog, leading to severe osteoarthritis of the knee joint and marked lameness. Following rupture, the ACL will not heal and in the dog, ACL rupture is thought to be the end stage of degenerative ligament disease (non-contact ACL injury). The extracellular matrix (ECM) of CLs has been extensively studied but little is known of the role of elastic fibres in the physiology of the ECM, the mechanics of CL function and in CL degeneration. Elastic fibres include polymers of fibrillins (microfibrils), bundles of microfibrils (oxytalan fibres) and elastin fibres (bundles of microfibrils with an elastin core). The hypothesis of this thesis is that elastin has a mechanical and a biological role in the canine cruciate ligament complex. It is further hypothesised that the distribution and function of elastic fibres will vary between three breeds of dog with differing risk of ACL rupture are: the greyhound with a low risk, the beagle with a low-to-moderate risk and the Labrador retriever with a high risk. The distribution of elastic fibres, fibrillins and cells was investigated throughout the CL complex using a combination of histochemical staining and immunofluorescence. CL microanatomy was studied using Nomarski differential interference microscopy. Elastin was measured biochemically and compared to histologic assessment of tissue architecture, elastic fibre staining and other biochemical parameters. The biological effect of elastin degradation products (EDPs) was assessed in an in vitro ACL cell culture model. A low risk dog breed to ACL rupture (greyhound) was used in all investigations and comparisons were made with other breeds with regard to cellular and elastic fibre anatomy. Differences in cell morphology between breeds with differing risk of ACL rupture may reflect fundamental differences in CL physiology possibly through altered cell-to-cell communication. Cellular and matrix changes, considered degenerative, were seen throughout the CL complex and may reflect adaptation rather than degeneration in certain dog breeds such as the greyhound. Elastin content ranged from 5.9 to 19.4% of ligament dry weight. This was a far greater proportion of canine CLs than previously. Elastin fibres may have a mechanical role in bundle reorganization following ligament deformation. The distribution of fibrillins 1 and 2 was different from the pattern previously reported in tendon and may represent a fundamental difference between ligament and tendon. In the greyhound CL there was a significant proportional increase in oxytalan fibre staining with advancing CL degeneration. This response was seen also in the Labrador retriever and the beagle but the increase in oxytalan fibre staining was less marked with advancing degeneration. Therefore production of oxytalan fibres may reflect a healing response in damaged CL tissue in breeds at a low risk of ligament rupture. Fragments of elastin containing the VGVAPG motif affect canine ACL cells in vitro resulting in increased transcription of fibrillin 2 mRNA. Additionally, there was synergism with TGF-β1 resulting in upregulation of the elastin laminin receptor 1, through which EDPs are transduced. EDPs may thus have a role in response to injury in the CL.

The racing Greyhound presents us with an opportunity to study the characteristics of a successful athlete and the costs and risks such specialisation entails. This thesis investigates the nature of the injuries suffered by racing Greyhounds and how adaptation of the musculoskeletal system to the unique pattern of stresses encountered during racing and training might impact upon the risk of injury. Racing Greyhounds sustain a number of musculoskeletal injuries. Several of these, notably fatigue fractures of distal limb bones, are very similar to those seen in human athletes and military recruits (Armstrong et al. 2004; Beck et al. 2000; Brukner et al. 1996; Kowal 1980; Matheson et al. 1987). The most common, often leading to the dog being euthanatised, is fracture of the right tarsus. Evaluation of tarsal fractures via radiography alone frequently resulted in an underestimation of the severity of the injuries, whereas the use of computed tomography provided a more detailed, accurate assessment. Evidence of asymmetric bone remodelling was found in the distal limb bones of racing Greyhounds. Rail‐side bones had significantly higher bone density and increased levels of bone resorption and formation markers compared to contralateral bones. Greyhound bones also have regional differences in trabecular architecture. In contrast, Staffordshire Bull Terrier (SBT) bones did not show these differences. Additionally, Greyhound distal limb tendons appear well adapted to withstand the high stresses of racing; they are stronger, stiffer, and in the pelvic limbs, return more elastic strain energy than the corresponding SBT tendons. Greyhounds had left‐to‐right asymmetries in the tensile properties of their pelvic limb tendons, which SBTs did not. SBTs are not bred for racing and are unlikely to encounter asymmetric stresses. Therefore, the adaptive changes observed in the Greyhound bones and tendons appear to result from the asymmetric stresses encountered by the Greyhounds during racing around ovoid tracks.

The aim of the present study was to investigate the hormone, sexual behaviour and neuropeptide profiles of follicular phase ewes and examine alterations after the application of acute stressors. In study 1, follicular phases of intact ewes were synchronised with progesterone vaginal pessaries. Ewes then received saline vehicle, insulin (4 iu/kg) or endotoxin (LPS; 100ng/kg) at 28h after progesterone withdrawal (PW; time zero). In study 2, this protocol was repeated, but animals were killed at 0h, 16h, 31h and 40h after PW and brain tissue retrieved. In study 1, there was a delay of 17.6h and 7.2h (P<0.05), respectively, in half the insulin-treated animals (‘insulin-delayed’) but not in the other half; and a delay of 22.5h and 20.7h (P<0.001), respectively, in all LPS-treated animals. Plasma oestradiol concentrations decreased after both stressors (P<0.001) and cortisol increased in all groups (P<0.05); whereas progesterone increased in the insulin-delayed and LPS groups only (P<0.05). In study 2, immunohistochemistry was used to examine transcriptional activation (co-expression with c-Fos) of various neuropeptides in the hypothalamus and preoptic area. In control ewes, the maximum percentage of dynorphin cells co-localising c-Fos (i.e., activated) was observed at 31h after PW (52%; P<0.05), whereas maximum activated kisspeptin and neurokinin B cells occurred at 40h after PW (49 and 42%, respectively; P<0.05). The percentage of activated dopamine cells decreased before the onset of sexual behaviour (from 70 to 26%; P<0.05) whereas β-endorphin activation was lower during the LH surge (from 41 to 10%; P<0.05). In contrast, neuropeptide Y and somatostatin activation was higher during the surge (from 21 to 36%; P<0.08; and from 14 and 9% to 47 and 73%, respectively; P<0.05). However, LPS decreased the percentage of activated dynorphin cells (to 11%; P<0.05) and kisspeptin cells (to 22%; P<0.05). On the contrary, insulin decreased the percentage of activated dynorphin cells (to 27%) in two of the insulin-treated animals (insulin-responders) but not in the other two; whereas the percentage of activated kisspeptin cells increased in all insulin-treated animals (52%; P<0.05). Neurokinin B was not altered by either treatment. Furthermore, insulin increased the percentage of activated β-endorphin, neuropeptide Y and somatostatin cells in the ARC (to 71, 72 and 63%, respectively, P<0.05) but LPS did not have the same effect. In the VMN, activation of somatostatin cells was greater in all LPS treated animals (from 8 to 27%; P<0.05) but only in two of the insulin-treated animals (to 55 and 76%; insulin-responders) but not in the other two (to 5 and 6%; insulin-non-responders). These results indicate that there is a specific hormonal, behavioural and neuropeptide pattern during the follicular phase of intact ewes and this is disturbed by acute LPS or insulin administration in the late follicular phase, leading to the disruption of the LH surge.

The work in this thesis falls into three parts. The first part relates to the time spent with the industry as part of this CASE Studentship, whilst the second and third parts relate to stochastic transmission models and the analysis of interventions imposed upon these models. The second and third parts are linked by a common aim, which is to develop models to understand the dynamics of Salmonella transmission on a pig farm and thus identify key drivers of Salmonella. The thesis begins with an assessment and analysis of a Farm Tool Questionnaire that was developed by the industry. A total of 28 farms were visited, had pooled faecal samples taken and completed the Farm Tool Questionnaire. The main aim of this study was to pilot the developed tool and identify any areas that could be modified in order to enhance its usability. Furthermore, the results from the study were used in an attempt to highlight any possible areas of farm management that differ between Platinum farms and non-Platinum farms. It was shown that Platinum farms were likely to adopt a subset of biosecurity practices, which should consequently encourage farms to adopt a range of biosecurity practices rather than focusing on one aspect of biosecurity. The thesis then turns to the development of mathematical models in order to try and understand how the components of the system interact by using both numerical simulation and mathematical analysis. As farming methods differ considerably between farms, two key forms of unit structure were analysed: a fully slatted unit and a solid floored unit. The models were developed using a semi-stochastic transmission model similar to Xiao et al. [2006] (Y. Xiao, D. Clancy, N. P. French & R. G. Bowers. A semi-stochastic model for Salmonella infection in a multi-group herd. Mathematical Biosciences, 200(2):214-233, 2006). These were then used to assess any differences in dynamics as a result of farm structure. Finally, both sets of models were analysed in order to identify any possible interventions that could have some form of control on Salmonella prevalence at slaughter. The models showed that the key drivers of Salmonella transmission were the amount of bacteria shed and the probability of infection after exposure. As such, interventions focusing on these aspects should be implemented in order to see the most beneficial results. The rate at which infection was able to spread when shedding was high was found to be of great importance within the various models; indicating that solid flooring is a potential risk factor. Furthermore, as infection was able to spread quickly within the solid-floored unit, the time interval at which cleaning and disinfection were carried out could be of importance. However, this would require further investigation.