Essays on the term structure of interest rates

This PhD thesis contains three main chapters on macro finance, with a focus on the term structure of interest rates and the applications of state-of-the-art Bayesian econometrics. Except for Chapter 1 and Chapter 5, which set out the general introduction and conclusion, each of the chapters can be considered as a standalone piece of work. In Chapter 2, we model and predict the term structure of US interest rates in a data rich environment. We allow the model dimension and parameters to change over time, accounting for model uncertainty and sudden structural changes. The proposed timevarying parameter Nelson-Siegel Dynamic Model Averaging (DMA) predicts yields better than standard benchmarks. DMA performs better since it incorporates more macro-finance information during recessions. The proposed method allows us to estimate plausible realtime term premia, whose countercyclicality weakened during the financial crisis. Chapter 3 investigates global term structure dynamics using a Bayesian hierarchical factor model augmented with macroeconomic fundamentals. More than half of the variation in the bond yields of seven advanced economies is due to global co-movement. Our results suggest that global inflation is the most important factor among global macro fundamentals. Non-fundamental factors are essential in driving global co-movements, and are closely related to sentiment and economic uncertainty. Lastly, we analyze asymmetric spillovers in global bond markets connected to diverging monetary policies. Chapter 4 proposes a no-arbitrage framework of term structure modeling with learning and model uncertainty. The representative agent considers parameter instability, as well as the uncertainty in learning speed and model restrictions. The empirical evidence shows that apart from observational variance, parameter instability is the dominant source of predictive variance when compared with uncertainty in learning speed or model restrictions. When accounting for ambiguity aversion, the out-of-sample predictability of excess returns implied by the learning model can be translated into significant and consistent economic gains over the Expectations Hypothesis benchmark.

Essays on the term structure of interest rates

This PhD thesis contains three main chapters on macro finance, with a focus on the term structure of interest rates and the applications of state-of-the-art Bayesian econometrics. Except for Chapter 1 and Chapter 5, which set out the general introduction and conclusion, each of the chapters can be considered as a standalone piece of work. In Chapter 2, we model and predict the term structure of US interest rates in a data rich environment. We allow the model dimension and parameters to change over time, accounting for model uncertainty and sudden structural changes. The proposed time-varying parameter Nelson-Siegel Dynamic Model Averaging (DMA) predicts yields better than standard benchmarks. DMA performs better since it incorporates more macro-finance information during recessions. The proposed method allows us to estimate plausible real-time term premia, whose countercyclicality weakened during the financial crisis. Chapter 3 investigates global term structure dynamics using a Bayesian hierarchical factor model augmented with macroeconomic fundamentals. More than half of the variation in the bond yields of seven advanced economies is due to global co-movement. Our results suggest that global inflation is the most important factor among global macro fundamentals. Non-fundamental factors are essential in driving global co-movements, and are closely related to sentiment and economic uncertainty. Lastly, we analyze asymmetric spillovers in global bond markets connected to diverging monetary policies. Chapter 4 proposes a no-arbitrage framework of term structure modeling with learning and model uncertainty. The representative agent considers parameter instability, as well as the uncertainty in learning speed and model restrictions. The empirical evidence shows that apart from observational variance, parameter instability is the dominant source of predictive variance when compared with uncertainty in learning speed or model restrictions. When accounting for ambiguity aversion, the out-of-sample predictability of excess returns implied by the learning model can be translated into significant and consistent economic gains over the Expectations Hypothesis benchmark.

Development of dendritic cells in the intestine

The intestinal tract is exposed to a large variety of antigens such as food proteins, commensal bacteria and pathogens and contains one of the largest arms of the immune system. The intestinal immune system has to discriminate between harmless and harmful antigens, inducing tolerance to harmless antigens and active immunity towards pathogens and other harmful materials. Dendritic cells (DC) in the mucosal lamina propria (LP) are central to this process, as they sample bacteria from the local environment and constitutively migrate to the draining mesenteric lymph nodes (MLN), where they present antigen to naïve T cells in order to direct an appropriate immune response. Despite their crucial role, understanding the function and phenotype of LP DC has been hampered by the fact that they share phenotypic markers with macrophages (mφ), which are the dominant population of mononuclear phagocyte (MP) in the LP. Recent work in our own and other laboratories has established gating strategies and phenotyping panels that allow precise discrimination between intestinal DC and mφ using the mφ specific markers CD64 and F4/80. In this way four bona fide DC subsets with distinct functions have been identified in adult LP based on their expression of CD11b and CD103 and a major aim of my project was to understand how these subsets might develop in the neonatal intestine. At the beginning of my PhD, the laboratory had used these new methods to show that signal regulatory protein α (SIRPα), an inhibitory receptor expressed by myeloid cells, was expressed by mφ and most DC in the intestine, except for those expressing CD103 alone. In addition, mice carrying a non-signalling mutation in SIRPα (SIRPα mt) had a selective reduction in CD103+CD11b+ DC, a subset which is unique to the intestinal LP. This was the basis for the initial experiments of my project, described in Chapter 3, where I investigated if the phenotype in SIRPα mt mice was intrinsic to haematopoietic cells or not. To explore this, I generated bone marrow (BM) chimeric mice by reconstituting irradiated WT mice with SIRPα mt BM, or SIRPα mt animals with WT BM. These experiments suggested that the defect in CD103+CD11b+ DC was not replicated in DC derived from BM of SIRPα origin. However as this seemed inconsistent with other data, I considered the possibility that 18 the phenotype may have been lost with age, as the BM chimeric mice were considerably older than those used in the original studies of SIRPα function. However a comparison of DC subsets in the intestine of WT and SIRPα mt mice as they aged provided no conclusive evidence to support this idea. As these experiments did show age-dependent effects on DC subsets, in Chapter 4, I went on to investigate how the DC populations appeared in the intestine and other tissues in the neonatal period. These experiments showed there were few CD103+CD11b+ DC present in the LP and migratory DC compartment of the MLN in the neonate and that as this population gradually increased in proportion with age, there was a reciprocal decrease in the relative proportion of CD103-CD11b+ DC. Interestingly, most of the changes in DC numbers in the intestine were found during the second or third week of life when the weaning process began. To validate my findings that there were few CD103+CD11b+ DC in the neonate and that this was not merely an absence of CD103 upregulation, I examined the expression of CD101 and Trem-1, markers that other work in the laboratory had suggested were specific to the CD103+CD11b+ DC lineage. My work showed that CD101 and Trem-1 were co- expressed by most CD103+CD11b+ DC in small intestine (SI) LP, as well as a small subset of CD103-CD11b+ DC in this tissue. Interestingly, Trem-1 was highly specific to the SI LP and migratory DC in the MLN, but absent from the colon and other tissues. CD101 expression was also only found on CD11b+ DC, but showed a less restricted pattern of distribution, being found in several tissues as well as the SI LP. The relative timing of their development suggested there might be a relationship between CD103+CD11b+ and CD103-CD11b+ DC and this was supported by microarray analysis. I hypothesised that the CD103-CD11b+ DC that co-expressed CD101 and Trem-1 may be the cells that developed into CD103+CD11b+ DC. To investigate this I analysed how CD101 and Trem-1 expression changed with age amongst the DC subsets in SI LP, colonic LP (CLP) and MLN. The proportion of CD101+Trem-1+ cells increased amongst CD103+CD11b+ DC in the SI LP and MLN with age, while amongst CD103+CD11b+ DC in the CLP this decreased. This was not the same in CD103-CD11b+ DC, where CD101 and Trem-1 expression was more varied with age in all tissues. CD101 and Trem-1 were not expressed to any great extent on CD103+CD11b- or CD103-CD11b- DC. The phenotypic development of the 19 intestinal DC subsets was paralleled by the gradual upregulation of CD103 expression, while the production of retinoic acid (RA), as assessed by the AldefluorTM assay, was low early in life and did not attain adult levels until after weaning. Thus DC in the neonatal intestine take some time to acquire the adult pattern of phenotypic subsets and are functionally immature compared with their adult counterparts. In Chapter 5, I used CD101 and Trem-1 to explore the ontogeny of intestinal DC subsets in CCR2-/- and SIRPα mt mice, both of which have selective defects in one particular group of DC. The selective defect seen amongst CD103+CD11b+ DC in adult SIRPα mt mice was more profound in mice at D7 and D14 of age, indicating that it may be intrinsic to this population and not highly dependent on environmental factors that change after birth. The expression of CD101 and Trem-1 by both CD103+CD11b+ and CD103-CD11b+ DC was reduced in SIRPα mt mice, again indicating that this entire lineage was affected by the lack of SIRPα signalling. However there was also a generalised defect in the numbers of all DC subsets in many tissues from early in life, suggesting there was compromised development, recruitment or survival of DC in the absence of SIRPα signalling. In contrast to the findings in SIRPα mt mice, more CD103+CD11b+ DC co-expressed CD101 and Trem-1 in CCR2-/- mice, while there were no differences in the expression of these molecules amongst CD103-CD11b+ DC. This may suggest that CCR2+ CD103-CD11b+ DC are not the cells that express CD101 and Trem-1 that are predicted to be the direct precursors of CD103+CD11b+ DC. I also examined the expression of DC growth factor receptors on DC subsets from mice of different ages, but no clear age or subset- related patterns of the expression of mRNA for Csf2ra, Irf4, Tgfbr1 and Rara could be observed. Next, I investigated whether Trem-1 played any role in DC development. Preliminary experiments in Trem-1-/- mice show no differences between any of the DC subsets, nor were there any selective effects on individual subsets when DC development from Trem-1-/- KO and WT BM was compared in competitive chimeras. However these experiments were difficult to interpret due to viability problems and because I found an unexpected defect in the ability of Trem-1-/- BM to generate all DC, irrespective of whether they expressed Trem-1 or not. 20 The final experiments I carried out were to examine the role of the microbiota in driving the differentiation of intestinal DC subsets, based on the hypothesis that this could be one of the environmental factors that might influence events in the developing intestine. To this end I performed experiments in both antibiotic treated and germ free adult mice, both of which showed no significant phenotypic differences amongst any of the DC subsets. However the study of germ free mice was compromised by recent contamination of the colony and may not be the conclusive answer. Together the data in this thesis have shown that the population of CD103+CD11b+ DC, which is unique to the intestine, is not present at birth. These cells gradually increase in frequency over time and as this occurs there is a reciprocal decrease in the frequency of CD103-CD11b+ DC. Along with other results, this leads to the idea that there may be a linear developmental pathway from CD103-CD11b+ DC to CD103+CD11b+ DC that is driven by non-microbial factors that are located preferentially in the small intestine. My project indicates that markers such as CD101 and Trem-1 may assist the dissection of this process and highlights the importance of the neonatal period for these events.

Oxidative stress and its haemodynamic consequences in chronic kidney disease

Chronic kidney disease (CKD) is associated with increased cardiovascular risk in comparison with the general population. This can be observed even in the early stages of CKD, and rises in proportion to the degree of renal impairment. Not only is cardiovascular disease (CVD) more prevalent in CKD, but its nature differs too, with an excess of morbidity and mortality associated with congestive cardiac failure, arrhythmia and sudden death, as well as the accelerated atherosclerosis which is also observed. Conventional cardiovascular risk factors such as hypertension, dyslipidaemia, obesity, glycaemia and smoking, are highly prevalent amongst patients with CKD, although in many of these examples the interaction between risk factor and disease differs from that which exists in normal renal function. Nevertheless, the extent of CVD cannot be fully explained by these conventional risk factors, and non-conventional factors specific to CKD are now recognised to contribute to the burden of CVD. Oxidative stress is a state characterised by excessive production of reactive oxygen species (ROS) and other radical species, a reduction in the capacity of antioxidant systems, and disturbance in normal redox homeostasis with depletion of protective vascular signalling molecules such as nitric oxide (NO). This results in oxidative damage to macromolecules such as lipids, proteins and DNA which can alter their functionality. Moreover, many enzymes are sensitive to redox regulation such that oxidative modification to cysteine thiol groups results in activation of signalling cascades which result in adverse cardiovascular effects such as vascular and endothelial dysfunction. Endothelial dysfunction and oxidative stress are present in association with many conventional cardiovascular risk factors, and can be observed even prior to the development of overt, clinical, vascular pathology, suggesting that these phenomena represent the earliest stages of CVD. In the presence of CKD, there is increased ROS production due to upregulated NADPH oxidase (NOX), increase in a circulating asymmetric dimethylarginine (ADMA), uncoupling of endothelial nitric oxide synthase (eNOS) as well as other mechanisms. There is also depletion in exogenous antioxidants such as ascorbic acid and tocopherol, and a reduction in activity of endogenous antioxidant systems regulated by the master gene regulator Nrf-2. In previous studies, circulating markers of oxidative stress have been shown to be increased in CKD, together with a reduction in endothelial function in a stepwise fashion relating to the severity of renal impairment. Not only is CVD linked to oxidative stress, but the progression of CKD itself is also in part dependent on redox sensitive mechanisms. For example, administration of the ROS scavenger tempol attenuates renal injury and reduces renal fibrosis seen on biopsy in a mouse model of CKD, whilst conversely, supplementation with the NOS inhibitor L-NAME causes proteinuria and renal impairment. Previous human studies examining the effect of antioxidant administration on vascular and renal function have been conflicting however. The work contained in this thesis therefore examines the effect of antioxidant administration on vascular and endothelial function in CKD. Firstly, 30 patients with CKD stages 3 – 5, and 20 matched hypertensive controls were recruited. Participants with CKD had lower ascorbic acid, higher TAP and ADMA, together with higher augmentation index and pulse wave velocity. There was no difference in baseline flow mediated dilatation (FMD) between groups. Intravenous ascorbic acid increased TAP and O2-, and reduced central BP and augmentation index in both groups, and lowered ADMA in the CKD group only. No effect on FMD was observed. The effects of ascorbic acid on kidney function was then investigated, however this was hindered by the inherent drawbacks of existing methods of non-invasively measuring kidney function. Arterial spin labelling MRI is an emerging imaging technique which allows measurement of renal perfusion without administration of an exogenous contrast agent. The technique relies upon application of an inversion pulse to blood within the vasculature proximal to the kidneys, which magnetically labels protons allowing measurement upon transit to the kidney. At the outset of this project local experience using ASL MRI was limited and there ensued a prolonged pre-clinical phase of testing with the aim of optimising imaging strategy. A study was then designed to investigate the repeatability of ASL MRI in a group of 12 healthy volunteers with normal renal function. The measured T1 longitudinal relaxation times and ASL MRI perfusion values were in keeping with those found in the literature; T1 time was 1376 ms in the cortex and 1491 ms in the whole kidney ROI, whilst perfusion was 321 mL/min/100g in the cortex, and 228 mL/min/100g in the whole kidney ROI. There was good reproducibility demonstrated on Bland Altman analysis, with a CVws was 9.2% for cortical perfusion and 7.1% for whole kidney perfusion. Subsequently, in a study of 17 patients with CKD and 24 healthy volunteers, the effects of ascorbic acid on renal perfusion was investigated. Although no change in renal perfusion was found following ascorbic acid, it was found that ASL MRI demonstrated significant differences between those with normal renal function and participants with CKD stages 3 – 5, with increased cortical and whole kidney T1, and reduced cortical and whole kidney perfusion. Interestingly, absolute perfusion showed a weak but significant correlation with progression of kidney disease over the preceding year. Ascorbic acid was therefore shown to have a significant effect on vascular biology both in CKD and in those with normal renal function, and to reduce ADMA only in patients with CKD. ASL MRI has shown promise as a non-invasive investigation of renal function and as a biomarker to identify individuals at high risk of progressive renal impairment.

The role of non-coding RNA in the development of pulmonary arterial hypertension

Pulmonary arterial hypertension (PAH) is a progressive disease of the small pulmonary arteries, characterised by pulmonary vascular remodelling due to excessive proliferation and resistance to apoptosis of pulmonary artery endothelial cells (PAECs) and pulmonary artery smooth muscle cells (PASMCs). The increased pulmonary vascular resistance and elevated pulmonary artery pressures result in right heart failure and premature death. Germline mutations of the bone morphogenetic protein receptor-2 (bmpr2) gene, a receptor of the transforming growth factor beta (TGF-β) superfamily, account for approximately 75%-80% of the cases of heritable form of PAH (HPAH) and 20% of sporadic cases or idiopathic PAH (IPAH). IPAH patients without known bmpr2 mutations show reduced expression of BMPR2. However only ~ 20% of bmpr2-mutation carriers will develop the disease, due to an incomplete penetrance, thus the need for a ‘second hit’ including other genetic and/or environmental factors is accepted. Diagnosis of PAH occurs most frequently when patients have reached an advanced stage of disease. Although modern PAH therapies can markedly improve a patient’s symptoms and slow the rate of clinical deterioration, the mortality rate from PAH remains unacceptably high. Therefore, the development of novel therapeutic approaches is required for the treatment of this multifaceted disease. Noncoding RNAs (ncRNAs) include microRNAs (miRNAs) and long noncoding RNAs (lncRNAs). MiRNAs are ~ 22 nucleotide long and act as negative regulators of gene ex-pression via degradation or translational inhibition of their target mRNAs. Previous studies showed extensive evidence for the role of miRNAs in the development of PAH. LncRNAs are transcribed RNA molecules greater than 200 nucleotides in length. Similar to classical mRNA, lncRNAs are translated by RNA polymerase II and are generally alternatively spliced and polyadenylated. LncRNAs are highly versatile and function to regulate gene expression by diverse mechanisms. Unlike miRNAs, which exhibit well-defined actions in negatively regulating gene expression via the 3’-UTR of mRNAs, lncRNAs play more diverse and unpredictable regulatory roles. Although a number of lncRNAs have been intensively investigated in the cancer field, studies of the role of lncRNAs in vascular diseases such as PAH are still at a very early stage. The aim of this study was to investigate the involvement of specific ncRNAs in the development of PAH using experimental animal models and cell culture. The first ncRNA we focused on was miR-143, which is up-regulated in the lung and right ventricle tissues of various animal models of PH, as well as in the lungs and PASMCs of PAH patients. We show that genetic ablation of miR-143 is protective against the development of chronic hypoxia induced PH in mice, assessed via measurement of right ventricular systolic pressure (RVSP), right ventricular hypertrophy (RVH) and pulmonary vascular remodelling. We further report that knockdown of miR-143-3p in WT mice via anti-miR-143-3p administration prior to exposure of mice to chronic hypoxia significantly decreases certain indices of PH (RVSP) although no significant changes in RVH and pulmo-nary vascular remodelling were observed. However, a reversal study using antimiR-143-3p treatment to modulate miR-143-3p demonstrated a protective effect on RVSP, RVH, and muscularisation of pulmonary arteries in the mouse chronic hypoxia induced PH model. In vitro experiments showed that miR-143-3p overexpression promotes PASMC migration and inhibits PASMC apoptosis, while knockdown miR-143-3p elicits the opposite effect, with no effects observed on cellular proliferation. Interestingly, miR-143-3p-enriched exosomes derived from PASMCs mediated cell-to-cell communication between PASMCs and PAECs, contributing to the pro-migratory and pro-angiogenic phenotype of PAECs that underlies the pathogenesis of PAH. Previous work has shown that miR-145-5p expression is upregulated in the chronic hypoxia induced mouse model of PH, as well as in PAH patients. Genetic ablation and pharmacological inhibition (subcutaneous injection) of miR-145-5p exert a protective against the de-velopment of PAH. In order to explore the potential for alternative, more lung targeted delivery strategies, miR-145-5p expression was inhibited in WT mice using intranasal-delivered antimiR-145-5p both prior to and post exposure to chronic hypoxia. The decreased expression of miR-145-5p in lung showed no beneficial effect on the development of PH compared with control antimiRNA treated mice exposed to chronic hypoxia. Thus, miR-143-3p modulated both cellular and exosome-mediated responses in pulmonary vascular cells, while the inhibition of miR-143-3p prevented the development of experimental pulmonary hypertension. We focused on two lncRNAs in this project: Myocardin-induced Smooth Muscle Long noncoding RNA, Inducer of Differentiation (MYOSLID) and non-annotated Myolnc16, which were identified from RNA sequencing studies in human coronary artery smooth muscle cells (HCASMCs) that overexpress myocardin. MYOSLID was significantly in-creased in PASMCs from patients with IPAH compared to healthy controls and increased in circulating endothelial progenitor cells (EPCs) from bmpr2 mutant PAH patients. Exposure of PASMCs to hypoxia in vitro led to a significant upregulation in MYOSLID expres-sion. MYOSLID expression was also induced by treatment of PASMC with BMP4, TGF-β and PDGF, which are known to be triggers of PAH in vitro. Small interfering RNA (siR-NA)-mediated knockdown MYOSLID inhibited migration and induced cell apoptosis without affecting cell proliferation and upregulated several genes in the BMP pathway in-cluding bmpr1α, bmpr2, id1, and id3. Modulation of MYOSLID also affected expression of BMPR2 at the protein level. In addition, MYOSLID knockdown affected the BMP-Smad and BMP-non-Smad signalling pathways in PASMCs assessed by phosphorylation of Smad1/5/9 and ERK1/2, respectively. In PAECs, MYOSLID expression was also induced by hypoxia exposure, VEGF and FGF2 treatment. In addition, MYOSLID knockdown sig-nificantly decreased the proliferation of PAECs. Thus, MYOSLID may be a novel modulator in pulmonary vascular cell functions, likely through the BMP-Smad and –non-Smad pathways. Treatment of PASMCs with inflammatory cytokines (IL-1 and TNF-α) significantly in-duced the expression of Myolnc16 at a very early time point. Knockdown of Myolnc16 in vitro decreased the expression of il-6, and upregulated the expression of il-1 and il-8 in PASMCs. Moreover, the expression levels of chemokines (cxcl1, cxcl6 and cxcl8) were sig-nificantly decreased with Myolnc16 knockdown. In addition, Myolnc16 knockdown decreased the MAP kinase signalling pathway assessed by phosphorylation of ERK1/2 and p38 MAPK and inhibited cell migration and proliferation in PASMCs. Thus, Myolnc16 may a novel modulator of PASMCs functions through anti-inflammatory signalling pathways. In summary, in this thesis we have demonstrated how miR-143-3p plays a protective role in the development of PH both in vivo animal models and patients, as well as in vitro cell cul-ture. Moreover, we have showed the role of two novel lncRNAs in pulmonary vascular cells. These ncRNAs represent potential novel therapeutic targets for the treatment of PAH with further work addressing to investigate the target genes, and the pathways modulated by these ncRNAs during the development of PAH.