influence of bone-associated and cardiovascular biomarkers
TRANSCRIPT
Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine 1785
Influence of bone-associated and cardiovascular biomarkers on vascular events and mortality in relation to renal dysfunction
PING-HSUN WU
ACTA UNIVERSITATIS
UPSALIENSIS UPPSALA
2021
ISSN 1651-6206 ISBN 978-91-513-1336-8 URN urn:nbn:se:uu:diva-456907
Dissertation presented at Uppsala University to be publicly examined in Enghoffsalen, Ing 50 bv, Akademiska sjukhuset, Uppsala, Wednesday, 15 December 2021 at 09:00 for the degree of Doctor of Philosophy (Faculty of Medicine). The examination will be conducted in English. Faculty examiner: Associate Professor Jonas Spaak (Cardiology at the Department of Clinical Sciences, Danderyd Hospital, Karolinska Institutet).
Abstract Wu, P.-H. 2021. Influence of bone-associated and cardiovascular biomarkers on vascular events and mortality in relation to renal dysfunction. Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine 1785. 63 pp. Uppsala: Acta Universitatis Upsaliensis. ISBN 978-91-513-1336-8.
Biomarkers can help physicians identify subjects with an increased cardiovascular risk. Apart from the clinical factors, some biomarkers have been recognized as important predictors and risk factors for cardiovascular disease in renal disease. The applicability of biomarkers may be limited in patients with kidney disease due to the complex etiology of cardiovascular disease, which warrants separate evaluations, including established and novel biomarkers. The overall aim of the thesis was to investigate the association between bone-associated markers and cardiovascular proteins on death and vascular events in the elderly male population and patients with kidney disease.
Study I included 3,014 participants in Swedish multicenter prospective Osteoporotic Fractures in Men (MrOS) cohort and investigated the associations between Klotho single-nucleotide polymorphism and mortality. Two potentially damaging single-nucleotide polymorphisms (rs9536314 and rs9527025) in the Klotho gene were not associated with mortality.
Study II investigated the association between mineral bone markers and all-cause mortality / cardiovascular mortality. The composite evaluation of elevated fibroblast growth factor-23 levels, vitamin D deficiency, and renal impairment was associated with mortality.
Study III evaluated the bone-associated proteins and mortality/composite vascular events in the 331 Demark hemodialysis patients. Osteoprotegerin, as one of the most promising bone-related proteins, was associated with composite vascular events independent of cytokine.
Study IV investigated the association between 92 proteins measured by proximity extension assay and mortality/composite vascular events in hemodialysis patients. A higher level of Interleukin-8, T-cell immunoglobulin and mucin domain 1, C-C motif chemokine 20, and lower level of stem cell factor and galanin peptides were associated with poor outcomes.
This thesis addressed the issue of bone-vascular axis and cardiovascular disease. We evaluated from gene levels to circulating protein levels and from the general population to patients with kidney disease. Based on our research findings, more evidence was linked between bone and vascular complications. We also identified several cardiovascular proteins considered potentially important predictors for cardiovascular disease in patients with renal failure, especially hemodialysis patients.
Keywords: biomarkers, bone markers, cardiovascular markers, proteomics, vascular event, mortality, renal function, hemodialysis
Ping-Hsun Wu, Department of Medical Sciences, Endocrinology and mineral metabolism, Akademiska sjukhuset, Uppsala University, SE-751 85 Uppsala, Sweden.
© Ping-Hsun Wu 2021
ISSN 1651-6206 ISBN 978-91-513-1336-8 URN urn:nbn:se:uu:diva-456907 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-456907)
To my lovely family…... you give me the roots to stand tall and strong
List of Papers
This thesis is based on the following papers, which are referred to in the text
by their Roman numerals.
I Ping-Hsun Wu, Per-Anton Westerberg, Andreas Kindmark, Åsa
Tivesten, Magnus K Karlsson, Dan Mellström, Claes Ohlsson,
Bengt Fellström, Torbjörn Linde, Östen Ljunggren. (2020)
The association between Single Nucleotide Polymorphisms of
Klotho Gene and Mortality in Elderly Men: The MrOS Sweden
Study. Sci Rep, Jun 24;10(1):10243; doi: 10.1038/s41598-020-
66517-5
II Ping-Hsun Wu, Per-Anton Westerberg, Magnus K. Karlsson,
Dan Mellström, Torbjörn Linde, Bengt Fellström, Östen Ljung-
gren. The effect of fibroblast growth factor 23, vitamin D, and
renal function on all-cause and cardiovascular mortality: The
MrOS Sweden Study. (Manuscript)
III Ping-Hsun Wu, Rie Io Glerup, Jeppe Hagstrup Christensen, My
Hanna Sofia Svensson, Torbjörn Linde, Östen Ljunggren, Bengt
Fellström.(2021) Osteoprotegerin predicts cardiovascular events
in patients treated with hemodialysis. Nephrol Dial Transplant,
Jun 4;gfab192. doi: 10.1093/ndt/gfab192.
IV Ping-Hsun Wu*, Rie Io Glerup*, My Hanna Sofia Svensson,
Niclas Eriksson, Jeppe Hagstrup Christensen, Philip de Laval,
Inga Soveri, Magnus Westerlund, Torbjörn Linde, Östen Ljung-
gren, Bengt Fellström. Novel biomarkers detected by proteomics
predict death and cardiovascular events in hemodialysis patients.
(Submitted)
*Authors contributed equally to this work
Reprints were made with permission from the respective publishers.
Contents
Introduction ................................................................................................... 11 Cardiovascular disease in patients with chronic kidney disease .............. 11 Chronic kidney disease and mineral bone disease ................................... 12 Biomarkers in patients with chronic kidney disease ................................ 12 Vascular calcification and cardiovascular disease ................................... 13 Bone-vascular axis ................................................................................... 13 Bone markers in patients with kidney disease .......................................... 15 Protein biomarkers and cardiac structure and function ............................ 15
Proteomics technology .................................................................................. 17 Brief introduction of proteomics .............................................................. 17 Biotechnology of clinical proteomics using proximity extension assays . 18
Aims .............................................................................................................. 19 General aim .............................................................................................. 19 Specific aims of the studies ...................................................................... 19
Study samples ............................................................................................... 21 The Swedish part of the Osteoporotic Fractures in Men Study (MrOS) .. 21 Denmark Århus Hemodialysis Cohort ..................................................... 21
Methods ........................................................................................................ 22 Study designs and methods ...................................................................... 22 Biomarker measurements ......................................................................... 27 Outcomes definitions and mortality assessment ....................................... 29 Statistical analysis .................................................................................... 30
Main results ................................................................................................... 33 Study I ...................................................................................................... 33 Study II ..................................................................................................... 33 Study III ................................................................................................... 35 Study IV ................................................................................................... 36
Additional Preliminary Study ....................................................................... 37 Method and Statistical analysis of the additional preliminary study ........ 37 Preliminary results of additional study ..................................................... 38
Discussion ..................................................................................................... 42
Klotho gene SNPs and all-cause or CV mortality .................................... 42 The mineral bone factors and all-cause or CV mortality ......................... 43 Bone-associated markers and mortality or CVE in hemodialysis
patients ..................................................................................................... 45 CV-associated markers for mortality or CVE in hemodialysis patients ... 46 Mortality differences among hemodialysis patients between Asian and
Caucasian ................................................................................................. 48 Strengths and limitations .......................................................................... 50
Conclusion and future perspectives .............................................................. 51
Summary in Swedish (sammanfattning på svenska) ..................................... 53
Acknowledgments......................................................................................... 54
References ..................................................................................................... 55
Abbreviations
AMI Acute myocardial infarction
CAD Coronary artery disease
CCL20 C-C motif chemokine 20
CKD Chronic kidney disease
CKD-BMD Chronic kidney disease and mineral bone disease
CRP C-reactive protein
CTSD Cathepsin D
CTSL1 Cathepsin L1
CV Cardiovascular
CVD Cardiovascular disease
CVE Composite vascular event
DKK-1 Dickkopf-related protein 1
eGFR Estimated glomerular filtration rate
ESKD End-stage kidney disease
FGF23 Fibroblast growth factor 23
GAL Galanin peptides
HGF Hepatocyte growth factor
IL-8 Interleukin-8
IDI Integrated discrimination improvement
NRI Net reclassification improvement
OPG Osteoprotegerin
PAD Peripheral artery disease
PCR Polymerase chain reaction
PEA Proximity extension assay
PTH Parathyroid hormone
RANKL Receptor activator of nuclear factor κB ligand
RF Random forest
SCF Stem cell factor
SNP Single nucleotide polymorphism
TIA Transient ischemic attack
TIM-1 T-cell immunoglobulin and mucin domain 1
TRAIL TNF-related apoptosis-inducing ligand
TRAIL-R2 TNF-related apoptosis-inducing ligand receptor 2
VSMC Vascular smooth muscle cells
10
11
Introduction
Cardiovascular disease in patients with chronic kidney
disease
Chronic kidney disease (CKD) is a global public health problem associated
with high morbidity and mortality. Cardiovascular disease (CVD) is the major
cause of premature death in patients with CKD.1, 2 Death from CVD is far
more common in patients with CKD than progression to end-stage kidney dis-
ease (ESKD).3 Chronic hemodialysis treatment is associated with a risk of
premature mortality that is 10–50 times higher than that in the age-matched
general population, with CVD being a leading cause of death.4 Large prospec-
tive cohort studies have documented significant associations of worse kidney
function and proteinuria with CVD and mortality that are independent of tra-
ditional risk factors.2, 5, 6 CKD has been recognized as an independent risk fac-
tor for CVD and has now been recognized as a coronary artery disease (CAD)
risk equivalent,7 similar to diabetes mellitus, suggesting that the risk from
CKD is equivalent to that from established CAD. Population-based studies in
Asia have also established CKD as a substantial risk factor for CVD and all-
cause mortality, suggesting CKD is a serious public health issue that trans-
cends ethnicity.8-10 The prevalence of CVD in patients 66 years of age and
older who have CKD is 68.8% compared with 34.1% in those who do not have
CKD. The complex relationship between CVD and kidney disease is thought
to be attributable to shared traditional risk factors (e.g., diabetes mellitus, hy-
pertension, physical inactivity, left ventricular hypertrophy, smoking, family
history, and dyslipidemia) as well as to the influence of nontraditional risk
factors in the presence of CKD (e.g., endothelial dysfunction; vascular medial
hyperplasia, sclerosis, and calcification; volume overload; abnormalities in
mineral metabolism; anemia and malnutrition; inflammation; oxidative stress;
hyperparathyroidism, hyperhomocysteinemia, and autonomic imbalance).
Cardio-renal syndrome continues to pose both diagnostic and therapeutic chal-
lenges for those with heart failure.11 Thus, CVD is an important comorbidity
for patients with CKD.
12
Chronic kidney disease and mineral bone disease
CKD is characterized by mineral and bone disorders associated with ex-
traosseous and cardiovascular (CV) calcifications. Experimental studies and
clinical observations in the general population and in patients with CKD show
an inverse relationship between the extent of CV calcifications and bone min-
eral density or bone metabolic activity. Hyperphosphatemia, reduced Klotho
expression and impaired soft-tissue calcification defenses are key pathophys-
iologic components of chronic kidney disease and mineral bone disease
(CKD–MBD). In patients receiving dialysis, the coronary artery calcification
score was found to be inversely correlated with vertebral bone mass.12 A high
systemic calcification score combined with bone histomorphometry sugges-
tive of low bone activity was observed in patients receiving hemodialysis.13
Disturbances of the fibroblast growth factor-23 (FGF23)/alpha-Klotho
(Klotho) axis in CKD are related to CKD-associated vascular calcification.
The decrease in Klotho expression in CKD might provide several explanations
for the link between the uremic state and vascular calcification. Thus, admin-
istration of protein-bound uremic toxin to mice induced not only vascular cal-
cification,14 but also a decrease in tubular Klotho expression subsequent to
Klotho gene hypermethylation (“epigenetic modification”).15
Biomarkers in patients with chronic kidney disease
Biomarkers that will help to improve the identification of patients at risk of
CV events has been at the core of extensive research in the general population
and in patients with CKD.16 The predicate of that approach is that an accurate
early assessment of CV risk will facilitate more aggressive and focused treat-
ment of those in greater need of preventive measures to reduce event rates.
Several important clinical CVD risk factors for patients with CKD have been
established.17 In addition, other biomarkers representing potential risk factors
for CVD in patients with CKD have also been identified: high-sensitivity C-
reactive protein (hsCRP),18 fibroblast growth factor 23 (FGF23),19, 20 and N-
terminal pro-brain natriuretic peptide (NT-proBNP),21 among others. Cardiac
troponin T is cardiac-specific and is not present in serum after non-myocardial
muscle damage. Detecting asymptomatic CAD, especially multi-vessel dis-
ease, as a predictor of mortality in CKD patients could be helpful. However,
as in the case of several other markers, variability of the test for troponin T
both day-to-day and from one laboratory to another could limit the possibility
of issuing widely applicable guidelines for using this biomarker in CKD.22
Other biomarkers, such as calciprotein particles, sclerostin or Dickkopf-re-
lated protein 1 (DKK-1), Neutrophil Gelatinase-Associated Lipocalin, plasma
growth differentiation factor-15, asymmetric dimethylarginine require more
13
study before their roles as key mediators of CV damage and potential thera-
peutic targets in patients with CKD can be elucidated.22 Given the extremely
high CV risk and complex routes of CVD development in patients with CKD,
new biomarkers and prediction based on composite biomarkers are worthy of
investigation.
Vascular calcification and cardiovascular disease
Individuals with vascular calcification present the most significant risk of pro-
gressive CVD. The dramatically increased CV risk in patients with CKD is
directly related to vascular calcification.23 Calcification of the intima is a part
of atherosclerosis, while medial calcification is the hallmark of arteriosclero-
sis. Vascular calcification is a multi-step process that includes supersaturated
phosphate and calcium precipitation in the extracellular milieu and cell-medi-
ated processes, such as apoptosis, osteochondrogenic differentiation, and elas-
tin degradation.24 This phenotypic conversion of vascular smooth muscle cells
into an osteoblast-like phenotype is induced by osteoblast transcriptional ac-
tivation of osteocyte-specific proteins, such as FGF23 and sclerostin through
a Wnt-dependent mechanism.25 In vascular smooth muscle cells, transdiffer-
entiation and osteogenic lineage allocation of multipotent mesenchymal pro-
genitors can generate osteogenic progenitors. Metabolic and inflammatory in-
sults affect vascular function, causing vascular mesenchymal cells to develop
into osteogenic cells, and impair normal vascular smooth muscle cell func-
tion.26
Animal studies demonstrated that Klotho protects the vasculature against cal-
cification in CKD.27 Disturbances of the FGF23/ Klotho axis are related to
CKD-associated vascular calcification. In addition to Klotho, several other in-
hibitors of vascular calcification have been identified: inorganic substances
such as pyrophosphate and magnesium, and peptidic substances such as fe-
tuin-A, osteoprotegerin (OPG), and bone morphogenetic protein 7.28
Bone-vascular axis
Clinical and experimental studies suggested a bone–vascular axis because of
a link between bone and artery remodeling and alterations.26 Although the
physiologic link between vascular illness and bone alterations is undoubtedly
part of the aging process, several studies found that bone–vessel associations
remained significant after adjusting for age, suggesting that the causative re-
lationship is age-independent. Nevertheless, the factors or mechanisms under-
lying those associations are not well understood. They might reflect common
14
mechanisms acting on both systems (e.g., inflammation, hyperlipidemia, dia-
betes, smoking, or generalized arterial disease per se). A bidirectional endo-
crine connection between bone and the vasculature mutually benefits bone and
vascular health.26 The kidney plays a crucial role in the process by regulating
phosphate excretion and Klotho expression. Hyperphosphatemia, reduced
Klotho expression, and impaired soft tissue calcification defenses are key
pathophysiologic components of chronic kidney disease and mineral bone dis-
ease (CKD-MBD).29 Other essential factors in bone and vascular metabolism
include proteins such as the bone morphogenetic proteins, receptor activator
of nuclear factor κB ligand [RANKL], OPG, and matrix Gla protein; parathy-
roid hormone; phosphate; oxidized lipids; and vitamins D and K.
Numerous clinical investigations have now proved the link between osteopo-
rosis and vascular calcification, and the increased risk of CVD in patients with
osteoporosis (Figure 1). However, compared with placebo in the FREEDOM
(Fracture Reduction Evaluation of Denosumab in Osteoporosis every 6
Months) trial, denosumab (a monoclonal antibody against RANKL that im-
proves bone marrow density and reduces fracture risk) did not affect the pro-
gression of aortic calcification or the incidence of CV adverse events.30 Alt-
hough the relationship between OPG/RANK/RANKL and vascular calcifica-
tion is not yet fully elucidated, further studies focus on RANKL inhibitor and
CV outcomes.
Figure 1. The Bone-vascular axis and cardiovascular complications
15
Bone markers in patients with kidney disease
The human body remodels (turns over) bone at both cortical and cancellous
sites. Bone remodeling is regulated both by systemic factors (e.g., parathyroid
hormone [PTH], phosphorus, 1,25-dihydroxyvitamin D, circulating scle-
rostin, and perhaps FGF23) and by local microenvironmental factors
(RANKL, OPG, and sclerostin). Several biochemical markers of bone turno-
ver can be measured in serum, including bone resorption and bone formation
markers. Data suggest that serum levels of specific bone turnover markers
might help discriminate the various forms of renal bone disease. Those serum
marker levels are valuable in assessing systemic bone turnover in osteoporosis
and in assessing the body’s response to antiresorptive agents (which inhibit
bone turnover) and anabolic agents (which stimulate bone turnover).31-33 Bio-
chemical markers of bone turnover—in particular, serum PTH and bone-spe-
cific alkaline phosphatase—might be able to differentiate between biopsy-
proven adynamic renal bone disease, hyperparathyroid bone disease, and os-
teomalacia.
The baseline bone turnover markers sclerostin and tartrate-resistant acid phos-
phatase 5b are independent predictors of bone loss in patients receiving dial-
ysis.34 However, investigations are still required to use noninvasive markers
of bone resorption in patients with uremia. Moreover, the dynamic marker of
bone mass density requires an observation period of 6 months to 1 year be-
tween measurements. In contrast, metabolic markers accurately reflect the
state of bone metabolism at the time of measurement. . When there is doubt
about choosing a drug for pharmacotherapy, metabolic markers can help to
suggest the appropriate selection. Furthermore, in evaluating the effects of
drug therapy on disease improvement, an assessment of bone metabolism at
the time of diagnosis is recommended whenever possible.
Recently, the clinical application of bone metabolic markers has progressed
significantly, and measurements of the accepted indices provide a better un-
derstanding of pathologic osteoporosis. Guidance for the application and as-
sessment of bone metabolic markers in clinical practice is therefore crucial.
Protein biomarkers and cardiac structure and function
There is evidence that impairment of left ventricular diastolic relaxation de-
velops in concert with activation of circulating plasminogen activator inhibi-
tor-1.35 Myocardial fibrosis is also a significant determinant of diastolic func-
tion. The plasma level of the carboxy-terminal peptide of procollagen type I
has thus been suggested as a valuable marker for determining the level of fi-
16
brotic activity36, 37 and diastolic dysfunction.38 Hyperphosphatemia can de-
plete Klotho levels and increase FGF-23, which can directly contribute to car-
diac hypertrophy and fibrosis, and vascular calcification.11 Moreover, in ani-
mal models, the soluble inflammation and fibrosis biomarker galectin-3 has
been shown to have a causative role in renal and cardiac fibrosis, potentially
representing common pathophysiology for progression toward ESKD and car-
diomyopathy.39 Galectin-3 is implicated in the pathogenesis of fibrosis in the
heart, but it also increases with normal aging and renal impairment.40 Galec-
tin-3 has also been shown to promote transforming growth factor-β–mediated
activation of fibroblasts in the kidney.41 Thus, a renal–cardiac interaction
could be observed.
Accumulating evidence has proved that several noninvasive markers are prog-
nostic indicators in patients with CKD. But head-to-head comparison studies
between echocardiography and other noninvasive markers in patients with
CKD are limited. Biomarkers cannot replace echocardiography, but might
play a complementary role in evaluating CV risk in CKD. Exploring the link
between proteomics and cardiac structure and function is an excellent way to
begin to understand the pathophysiology of cardiac remodeling in uremia-as-
sociated cardiomyopathy.
17
Proteomics technology
Brief introduction of proteomics
All cells, organs, and organisms contain proteins, which are polymers of
amino acids. All proteins that can be measured in a biological sample are con-
sidered part of the proteome.42 Proteomics is a technique for identifying pro-
teins and determining their quantity, modifications, and interactions. Prote-
omics data usually reveal information about the biological function, which is
crucial in metabolic processes and illnesses.43 Proteins may be identified and
quantified using various analytical methods, including liquid chromatog-
raphy-tandem mass spectrometry, isotope-coded protein labeling, and protein
microarrays. Mass spectrometry has been regarded as a "Gold Standard"
method for identifying and analyzing individual proteins in expression prote-
omics research. However, mass spectrometry is time-consuming and expen-
sive compared to immunoaffinity assay. The choice of an affinity-based ap-
proach or mass spectrometry for a given experiment is influenced by several
factors. Affinity-based approaches have the following advantages: (a) high
sensitivity for numerous low-abundance proteins, (b) high sample throughput,
(c) ease of use, (d) modest instrument investment, and (e) ability to target spe-
cific proteins of biological or clinical relevance that have been defined a priori.
However, limitations include (a) the inability to identify proteins not targeted
by the assay, (b) the inability to distinguish posttranslational modifications
and isoforms, (c) the potential influence of coding DNA variants on epitope
structures and reagent affinity, (d) batch effects, which can cause significant
variability in reagents and irreproducibility, and (e) the lack of open science
platforms for several of the available methods.44
Using immunoaffinity assays, such as proximity extension assays (PEA), we
may now investigate many proteins simultaneously. Cross-reactivity, which
prevents multiplexing, specificity difficulties, and inadequate sensitivity to
identify proteins in the lower abundance spectrum, has been criticized in im-
munoaffinity tests.45 Proximity extension assays are nucleotide-labeled immu-
noassays that have been proven to identify proteins with excellent specificity
in small sample quantities.45 Due to the requirement of dual recognition of a
target protein by a matched pair of DNA-conjugated antibodies, the proximity
extension test has high specificity.
18
Biotechnology of clinical proteomics using proximity extension assays
A PEA is a 96-plex immunoassay for high-throughput detection of protein
biomarkers in liquid samples. A matched pair of antibodies linked to unique
oligonucleotides (“proximity probes”) binds to the respective protein target
for each biomarker. In the original assay design, one PEA probe consisted of
a double-stranded oligonucleotide attached to the antibody at its 3′ end, with
a 9 or so nucleotide 3′ overhang at the 5′ end. That overhang was complemen-
tary to the 3′ end of the oligonucleotide bound to the other antibody partner.
After incubation of the proximity probes with a sample containing the antigen
recognized by the probes, the overhanging 3′ end could hybridize to the 5′
oligonucleotide, and after the addition of a DNA polymerase, the free 3′ OH
was extended in the 5′–3′ direction toward the attachment site of the 5′ oligo-
nucleotide thereby generating a full-length amplicon and hybridization site for
the upstream primer. Thus, amplification and detection of the target antigen
by polymerase chain reaction (PCR) were facilitated. A modification has now
replaced that arrangement.
Each of the 2 single-stranded oligonucleotides contains a complementary site
for pair-wise annealing with the other oligonucleotide, allowing for an exten-
sion by a DNA polymerase and eliminating the requirement for a double-
stranded oligonucleotide with a 3′ overhang. Furthermore, a reasonable choice
of DNA polymerase minimizes background noise and improves the assay’s
sensitivity. Each of the 96 oligonucleotide antibody pairs contains unique
DNA sequences, allowing for a hybridization only to each other. Subsequent
proximity extension will create 96 unique DNA reporter sequences, which are
amplified by real-time PCR.
A limiting factor of multiplexed immunoassays is antibody cross-reactivity,
which restricts the degree of multiplexing of most assays to below 10-plex.
Olink panels will not detect cross-reactive events, given that only matched
DNA reporter pairs are amplified with real-time PCR. The new approach al-
lows for scalable multiplexing without loss of specificity and sensitivity. The
addition of a DNA polymerase, therefore, leads to an extension of the hybrid-
izing oligo, bound to one of the probes, creating a DNA amplicon that can
subsequently be detected and quantified by quantitative real-time PCR.45, 46
19
Aims
General aim
The overall aim of this thesis was to investigate the association of bone-asso-
ciated biomarkers and CV proteins with death and vascular events in the gen-
eral population and patients with kidney disease (Figure 2).
Specific aims of the studies
The aim of Study I was to investigate the associations between Klotho single-
nucleotide polymorphisms (SNPs) and all-cause mortality or CV mortality in
Sweden multicenter prospective Osteoporotic Fractures in Men (MrOS) co-
hort.
The aim of Study II was to investigate abnormal levels of mineral bone mark-
ers (FGF23, vitamin D, and PTH) and all-cause mortality or CV mortality in
the MrOS cohort.
The aim of Study III was to evaluate bone-associated proteins and clinical
outcomes (mortality or composite vascular events [CVEs]) in Demark’s Århus
HD cohort. In addition, the effects of calcium, phosphate, PTH, and inflam-
mation cytokine on bone-associated protein were also investigated to evaluate
how they might improve the prediction of CVEs.
The aim of Study IV was to explore the associations of 92 protein biomarkers
(measured by PEA) with death and CVEs in the Denmark Århus HD cohort.
20
Figure 2. The overall and specific aim of the thesis: Bone-associated and CV markers– target on cardiovascular disease
21
Study samples
The Swedish part of the Osteoporotic Fractures in Men
Study (MrOS)
The prospective international MrOS study of osteoporosis and fracture risk
enrolled elderly male participants in the United States, Sweden, and Hong
Kong. We included Sweden’s MrOS participants 69–81 years of age. The par-
ticipants (n = 3014) were randomly selected from a national population data-
base, constituting three sub-cohorts in Uppsala (n = 999), Göteborg (n =
1010), and Malmö (n = 1005). Ethics committees in Sweden approved the
MrOS study at Uppsala (Ups 01-057), Göteborg (Gbg M 014-01), and Lund
(LU 693-00). The study was performed in accordance with the Declaration of
Helsinki. Informed consent was obtained from all study participants.
Denmark Århus Hemodialysis Cohort
The prospective multicenter Århus HD cohort study protocol was approved
by the Regional Research Ethics Committee of The North Jutland Region
(protocol N-20100041), and the study complied with the Declaration of Hel-
sinki. Patients receiving HD at 5 dialysis facilities in Jutland, Denmark, were
enrolled between December 2010 and March 2011, and follow-up continued
for 5 years. Patients receiving chronic HD who were less than 18 years of age
or who had acute kidney injury or an inability to understand informed consent
were excluded. Serum samples were collected immediately before a dialysis
session and were stored at –80°C for later analysis. One physician reviewed
medical records for all participants at study entry and during follow-up. Clin-
ical events recorded during follow-up included cause of death and CVEs (a
composite endpoint of acute myocardial infarction [AMI], unstable angina,
ischemic stroke, transient ischemic attack, hemorrhagic stroke, and peripheral
artery disease [PAD]). All patients were followed for up to 5 years, until death,
loss to follow-up, dialysis modality change, kidney transplantation, or HD
treatment cessation.
22
Methods
Study designs and methods
An overview of each of the Studies I–IV is presented in Table 1.
Table 1. Overview of the designs and methods for studies I–IV
Study I II III IV
Design Cohort study Cohort study Cohort study Cohort study
Subjects Total of 2921
male subjects
with Klotho
gene SNP and
follow-up rec-
ords in Sweden
MrOs cohort
Total of 3014
male subjects
with FGF23,
PTH, vitamin
D, and renal
function as
well as follow-
up records in
Sweden MrOs
cohort
Total of 331 he-
modialysis pa-
tients with 9
bone associated
protein markers
in Denmark År-
hus hemodialysis
cohort
Total of 331 he-
modialysis pa-
tients with 92
CV protein
markers in Den-
mark Århus he-
modialysis co-
hort
Expo-
sure
Klotho Gene
SNP
Mineral bone
markers
Bone markers CV biomarkers
Out-
come
All-cause mor-
tality
CV mortality
All-cause
mortality
CV mortality
All-cause mor-
tality
CV mortality
CVE
All-cause mor-
tality
CV mortality
CVE
Analysis Logistic regres-
sion
Cox regression
Cox regres-
sion
Cox regression Cox regression
23
The detailed study flowchart of each of the Studies I–IV is presented
below
Study I47
24
Study II
25
Study III48
26
Study IV
27
Biomarker measurements
Genotyping of the Klotho gene (Study I)
Using standard methods, DNA was extracted from the whole blood of 3014
study participants. Using the HaploView 4.2 tagger algorithm (Mark Daly lab,
Broad Institute, Cambridge, MA, U.S.A.), SNPs covering the klotho gene and
the flanking 5′ and 3′ regions were tagged.49 Genotyping was performed using
matrix-assisted laser desorption/ionization time-of-flight mass spectrometry
on a Sequenom MassARRAY system (Sequenom Inc., Newton, MA, U.S.A.)
with iPLEX assay. The primers were designed using the MassARRAY Assay
Design software (version 3). PCR was performed according to the standard
iPLEX methodology.
Genotyping quality control was completed by excluding from the analysis in-
dividual samples or SNPs with genotype call rates below 95%, SNP assays
with poor-quality spectra or cluster plots, and SNPs that significantly deviated
from the Hardy–Weinberg equilibrium (p < 0.05). Genotype data for the SNPs
were pairwise calculated with respect to linkage disequilibrium and identifi-
cation of SNPs with the highest haplotype predictability. Successful genotyp-
ing was obtained for 23 SNPs, with an overall call rate of 98.9%. Allele fre-
quencies for those SNPs were calculated, and the Hardy–Weinberg equilib-
rium was tested for each SNP using chi-square goodness-of-fit statistics for
all but 2 SNPs (rs1888057 and rs2283368) in the cohort. The latter 2 SNPs
were subsequently excluded from further analyses.
HaploView 4.2 was accessed to generate haplotype blocks, linkage disequi-
librium values and diagrams, and SNPs tagged using the tagger algorithm.49
The data for chromosome location, minor genotype frequency, SNP function
and regulation of the human klotho gene, and wild or mutated residue of the
nonsynonymous SNPs were used according to program requirements.
FGF23 and renal function measurement (Study II)
Serum samples were collected and stored at -80° C. The serum concentration
of intact FGF23 was analyzed using ELISA (Kainos Laboratories Interna-
tional; Tokyo, Japan).50 The estimated glomerular filtration rate (eGFR) based
on cystatin C (Cystatin C Immunoparticles; Dako A/S, Glostrup, Denmark)
was calculated using the formula: eGFR cystatin C=79.901*(Cyst C [mg/L])-
1.4389. This proxy for eGFR shows a strong correlation with the iohexol clear-
ance rate (R2=0.956).51, 52
28
Proximity extension assay method (Study III and Study IV)
Bone-associated proteins and inflammatory cytokine proteins were assessed
using the Proseek Multiplex 96×96 PEA with the CVD I panel (Olink Biosci-
ence, Uppsala, Sweden). The assay simultaneously measures proteins using 2
particular antibodies for each protein, linking a PCR reporter sequence. Each
sample includes 2 incubations, 1 extension, and 1 detection control to deter-
mine the lower detection limit and to normalize the measurements on the Flu-
idigm BioMark HD real-time PCR platform. The 9 bone-associated proteins
investigated included cathepsin D (CTSD), cathepsin L1 (CTSL1), Dickkopf-
related protein 1, FGF23, leptin, OPG, RANKL, TNF-related apoptosis-in-
ducing ligand (TRAIL), and TNF-related apoptosis-inducing ligand receptor 2
(TRAIL-R2). In addition, all 92 CV proteins were evaluated for their associa-
tions with mortality and CV outcomes.
29
Outcomes definitions and mortality assessment
In Studies I and II, mortality was retrieved from Statistics Sweden. Follow-
up duration was recorded from the baseline visit (2001–2004) to the date of
death or last mortality data collection (March 1, 2008). Cause-of-death data
were recorded using International Classification of Diseases codes based on
death certificates obtained from the Swedish Cause of Death Register. For this
thesis, CV death was defined using codes I00–I99 from the International Clas-
sification of Diseases, 10th revision.
In Studies III and IV, medical records for all participants were reviewed by
1 physician at study entry and during follow-up. Clinical events recorded dur-
ing follow-up included cause of death and CVEs (a composite endpoint of
AMI, unstable angina, ischemic stroke, transient ischemic attack, hemorrhagic
stroke, and PAD). Individual CVEs were described as follows:
a. AMI was diagnosed when at least 2 of these 3 criteria were met: Pres-
ence of cardiac ischemic chest pain; electrocardiography findings
meeting the criteria for a definitive diagnosis during an event (new or
presumed new significant ST-segment–T wave changes, or new left
bundle branch block, or development of pathologic Q waves); ele-
vated cardiac enzymes measured during an event.
b. Unstable angina requiring hospitalization was defined as symptoms
of myocardial ischemia at rest (chest pain or equivalent) or an accel-
erating pattern of angina with frequent episodes associated with pro-
gressively decreased exercise capacity with no evidence of acute
AMI.
c. Ischemic stroke was defined as an acute episode of neurologic dys-
function caused by vascular injury lasting 24 hours or more.
d. A transient ischemic attack was defined as sudden onset of neurologic
symptoms, presumed to be ischemic, resolving in less than 24 hours,
clearly attributable to focal involvement of the central nervous system
(or of the eye) with no signs of a corresponding recent cerebral infarc-
tion on brain imaging.
e. Hemorrhagic stroke was defined as an acute episode of focal neuro-
logic symptoms with the presence of cerebral hemorrhage in the ap-
propriate territory on brain imaging (computed tomography or mag-
netic resonance imaging) caused by a nontraumatic intraparenchymal,
intraventricular, or subarachnoid hemorrhage.
f. PAD was defined as limb ischemia with major amputation or urgent
peripheral revascularization for ischemia.
30
Statistical analysis
In Study I, logistic regression was first used to analyze all associations of
klotho gene SNPs with mortality outcomes in an initial filtering process. Sub-
sequently, Cox regression was fitted to SNPs associated with a p value below
the 0.05 threshold, representing final results for those SNPs.53 The associa-
tions between genotypes for the SNPs and all-cause or CV mortality during
follow-up were evaluated using odds ratios and 95% confidence intervals
(CIs) with the homozygous major allele as reference in additive models. At
baseline, the models were tested unadjusted and adjusted for age, body mass
index, smoking, comorbidities (hypertension, diabetes mellitus, CAD, stroke,
cancer), eGFR, phosphate, and FGF23. Dominant and recessive genetics mod-
els were also analyzed. Cox regression models with unadjusted hazard ratios
(HRs) and 95% CIs were used to confirm findings. Tests for associations and
estimates for HRs were computed with adjustment for age, body mass index,
and smoking. Because eGFR and FGF23 were strongly correlated with mor-
tality, subgroup analyses on selected SNPs for mortality stratified by eGFR
(≥60 mL/min/1.73 m2, <60 mL/min/1.73 m2) and FGF23 (≥60 pg/mL,
<60 pg/mL). Kaplan–Meier curves with log-rank tests were used to examine
survival by klotho variant.
In Study II, cut-off values for analyses of mineral bone factors were based on
a literature review of associations of those factors with all-cause or CV mor-
tality. The selected cut-off values were 60 pg/mL for FGF23,54 60 ml/min for
eGFR,55, 56 50 nmol/l for vitamin D,57, 58 and 4.5 pmol/l for PTH.59-61 Kaplan–
Meier curves were generated to show the cumulative probabilities for all-
cause and CV mortality over 6 years’ observation time, and differences were
evaluated using a log-rank test. In addition, unadjusted associations of contin-
uous FGF23, PTH, eGFR, and vitamin D levels with all outcomes were as-
sessed and presented as restricted cubic splines with 3 knots to allow for non-
linear associations. Given that circulating FGF23 was highly correlated with
renal function, restricted cubic splines were also applied to demonstrate
FGF23-related outcomes associations stratified by renal function (normal re-
nal function, CKD stage 3A, and CKD stage 3B or less). Because of potential
nonlinear effects and interactions between mineral metabolic factors, the Ran-
dom Forest (RF) method was used to identify and rank relationships between
the predictive factors and the outcomes.62 The Gini index was applied to meas-
ure mineral metabolic factors and renal function importance in the RF algo-
rithm.63 The more important predictive factors had larger Gini scores.
To evaluate the interplay of bone and mineral metabolic factors with aspects
of kidney function related to the risk of all-cause and CV mortality, partici-
pants were allocated to these 3 groups: (1) Participants with normal eGFR (≧
31
60 ml/min), a normal vitamin D level, and FGF23 below 60 pg/ml. (2) Partic-
ipants with any single abnormality in these parameters. (3) Participants with
any 2 abnormalities in these parameters. After ensuring fulfillment of the pro-
portional hazards assumption by Schoenfeld residuals trend tests, multivaria-
ble Cox proportional hazards regressions were performed to find any associa-
tions of all-cause and CV mortality with mineral metabolic factors and abnor-
mal renal function components, with adjustment for age, baseline characteris-
tics (body mass index, smoking), and comorbidities (diabetes mellitus,
hypertension, stroke, and CAD). To determine the independent effect of
FGF23 and vitamin D on all-cause and CV mortality, adjusted analyses of the
associations were performed stepwise using Cox regression models with ad-
justment for factors including age, body mass index, smoking, comorbidities,
eGFR, and clinical laboratory data.
In Study III, the bone-associated proteins were elaborated using the RF
method in a multistage decision process that identifies and ranks relationships
between predictive and response variables. The RF algorithm constructs a
classification tree, randomly sampling the predictors, choosing the best split-
ting variables, and predicting new data by combining the predictions from all
trees to estimate the error rate and, in this case, to list the bone-associated
proteins that provide the highest predictive performance for all-cause and CV
mortality, and CVEs. The RF process62 can detect both linear and nonlinear
effects and potential protein-protein interactions, thereby identifying the pro-
tein that best differentiates groups. The Gini index was then applied to meas-
ure the importance of the bone-associated protein.63 The larger the Gini index
(range: 1–100), the more important the variable. Multivariable Cox regression
models were also used to determine the relationships between bone-associated
proteins (as continuous variables) and clinical outcomes of interest, adjusting
for confounders. All-cause mortality and CVE rates were displayed as
Kaplan–Meier curves, with the bone-associated proteins divided into tertiles.
Outcomes related to bone-associated proteins in their tertile groups and as
continuous variables were evaluated in 4 different models, with adjustment
for several confounders (Model 1: adjusted for age and sex; Model 2:
Model 1, plus blood pressure and comorbidities; Model 3: Model 2 plus la-
boratory data [albumin, CRP, low-density and high-density lipoprotein cho-
lesterol, phosphate] and dialysis duration; Model 4: Model 3 plus calcium and
PTH). The associations of the bone-associated proteins (as continuous varia-
bles) with the outcomes were further adjusted for single cytokine inflamma-
tory biomarkers (interleukins IL-6, IL-8, IL-16, IL-18, IL-27, IL-1RA, and
IL6-RA) separately.
To assess each biomarker’s incremental benefit, the C statistic was compared
using models with traditional risk factors with and without the addition of the
32
biomarker. The integrated discrimination improvement (IDI) and the continu-
ous (category free) net reclassification improvement (NRI) 64, 65 were esti-
mated to quantify the degree of correct reclassification as a result of adding
biomarkers to the clinical information risk model and the AURORA (A Study
to Evaluate the Use of Rosuvastatin in Subjects on Regular Hemodialysis: An
Assessment of Survival and Cardiovascular Events) risk model 66. The AU-
RORA risk prediction model contains age, albumin, CRP, history of CV dis-
ease, and diabetes.66
In study IV, univariable and multivariable Cox regression analyses were per-
formed to examine the risk for the clinical outcomes (all-cause and CV mor-
tality, and CVEs) in relation to PEA-selected CV protein biomarkers. That
analysis used continuous models (HR per standard deviation increase in nor-
malized protein expression). Multivariable Cox proportional hazards models
were adjusted for age, sex, smoking status, cause of ESKD, dialysis vintage,
dialysis treatment time per week, diabetes mellitus, previous myocardial in-
farction, previous unstable angina, previous cerebrovascular disease, previous
PAD, and blood levels of albumin, phosphate, and CRP.
33
Main results
Study I
In study I, polymorphisms in the klotho gene were analyzed using computa-
tional tools to predict coding and non-coding SNPs. SNPs (rs9536314,
rs9527025, rs9527026, rs564481) from the coding region (exon) were selected
for further bioinformatic analysis using SIFT, PolyPhen-2, PROVEAN,
SNPs3D, LS-SNPs, and MutPred. SNP rs9536314 was predicted to be “dam-
aging” in SIFT, “probably damaging” in PolyPhen-2, “deleterious” in
PROVEAN, and “deleterious” in SNPs3D and MutPred. SNP rs9527025 was
predicted to be “damaging” in SNPs3D, LS-SNPs, and MutPred.
In the additive univariate logistic regression model, 3 SNPs were associated
with increased risk of all-cause mortality, including the TT genotype of
rs9536282, the GG genotype of rs9536314, and the CC genotype of
rs9527025. However, none of the 18 SNPs were statistically significantly as-
sociated with the risk for CV mortality. In adjusted analyses controlling for
age, body mass index, smoking, comorbidities (hypertension, diabetes melli-
tus, CAD, stroke, cancer), eGFR, phosphate, and FGF23, the rs9536282,
rs9536314, and rs9527025 genotypes were not associated with all-cause or
CV mortality. In the analyses of dominant and recessive models, an associa-
tion of rs9527025 and rs9527026 with all-cause mortality was observed under
the recessive model. A Cox proportional hazards analysis to determine the
relationships of the rs9536314 and rs9527025 genotypes with mortality found
no statistical significance in either the unadjusted or adjusted model. Subgroup
analyses of SNPs rs9536314 and rs9527025 in the additive univariate Cox
regression models demonstrated no significant risk for all-cause or CV mor-
tality when stratified by eGFR or FGF23.
Study II
In study II, Kaplan–Meier curves showed greater all-cause mortality associ-
ated with higher FGF23 (≥60 pg/mL vs. <60 pg/mL), renal function impair-
ment (eGFR <60 mL/min vs. ≥60 mL/min), and vitamin deficiency (vita-
min D <50 nmol/L vs. ≥50 nmol/L)vitamin D. However, no difference in all-
cause mortality was found between the PTH groups (<4.5 pmol/L vs.
34
≥4.5 pmol/L). Similarly, in the survival analysis, participants having higher
FGF23 levels or renal function impairment compared with those having lower
FGF23 levels or normal renal function were also observed to have a higher
CV mortality rate. In the vitamin D and PTH subgroups, CV mortality was
not different.
Evaluating the nonlinear associations of mineral metabolic factors and renal
function with outcomes, the cubic spline demonstrated an increased risk for
all-cause mortality related to FGF23 level, renal function, and vitamin D level.
However, spline curves were found to be flat for the association of PTH level
with all-cause mortality and for that of vitamin D level with CV mortality.
Given that FGF23 was strongly correlated with renal function, the association
of all-cause mortality with FGF23 was stratified by renal function (normal
renal function, CKD stage 3A, and CKD stage 3B or lower). An increased HR
for all-cause and CV mortality was observed for an increased level of FGF23
in CKD stage 3B or lower. In addition, the Gini index was applied to the RF
analysis to compare the importance of FGF23, PTH, vitamin D, and renal
function. The highest-ranking factor for both all-cause and CV mortality was
FGF23.
Given the complex interaction between FGF23, PTH, vitamin D, and renal
function, we evaluated the associations between those factors in composite
and mortality risk. However, because the survival difference at the PTH cutoff
value was found to be insignificant, only the abnormal components of FGF23,
vitamin D, and renal function were evaluated in the composite analysis. Com-
pared with participants having a normal renal function (eGFR > 60 mL/min),
a normal vitamin D level, and low FGF23 (<60 pg/mL), those having renal
impairment (eGFR < 60 mL/min), vitamin D deficiency (<50 nmol/L), and
high FGF23 (≥60 pg/mL) had an increased risk for all-cause and CV mortal-
ity, after adjustment for confounders.
To evaluate the influence of confounding factors on FGF23 and vitamin D,
Cox regression models were used to investigate the associations of FGF23 or
vitamin D with outcomes, using stepwise adjustment for various factors. In
the unadjusted analyses, FGF23 and vitamin D were both associated with an
increased risk of all-cause mortality. However, the increased risk of all-cause
mortality in the high FGF23 group was attenuated after controlling eGFR.
Vitamin D deficiency was independently associated with all-cause mortality
even after controlling for all confounders. In contrast, vitamin D deficiency
was not associated with CV mortality. The increased risk of CV mortality in
the high FGF23 group was attenuated after controlling for other comorbidities
(diabetes, hypertension, stroke, and CAD). Longer follow-up on the clinical
outcome will be analyzed prior to the final sub-mission of this paper.
35
Study III
An importance plot of the relationships of bone-associated proteins with all-
cause mortality, CV mortality, and CVEs was produced by applying the Gini
index to RF analyses. The highest-ranking bone markers were TRAIL-R2 for
all-cause death, OPG for CV death, and OPG and TRAIL-R2 for CVEs. In
multivariable Cox regression analysis, only CTSL1 (HR: 1.62; 95% CI: 1.17
to 2.23) was associated with an increased risk for all-cause death. Four indi-
vidual bone markers were positively associated with CVEs: OPG (HR: 2.16;
95% CI: 1.44 to 3.26), TRAIL-R2 (HR: 1.66; 95% CI: 1.22 to 2.25), CTSD
(HR: 1.88; 95% CI: 1.24 to 2.85), and CTSL1 (HR: 1.63; 95% CI: 1.14 to
2.34).
The Kaplan-Meier plots of the associations between the bone marker tertiles
and the risk of all-cause death show a graded increase in risk in the higher
tertile groups of CTSL1, TRAIL-R2, and OPG. Higher concentrations of
CTSD, OPG, TRAIL-R2, and lower concentrations of RANKL were associ-
ated with increased risk for CVEs. Results for the relationships between the
bone markers and outcomes (all-cause death and CVEs) were similar in the
crude and adjusted analyses whether the tertile subgroup or continuous ap-
proach was used. After multivariable adjustment for age, sex, blood pressure,
comorbidities, laboratory data, HD treatment duration, and calcium and PTH
levels, the level of CTSL-1 was positively associated with an increased risk of
all-cause death. In the adjusted model for CVEs, 3 individual bone markers
(CTSD, OPG, TRAIL-R2) were associated with an increased risk for CVEs.
Because bone markers are moderately to strongly correlated with inflamma-
tory cytokines and considered to be potential confounders, the associations of
bone markers with outcomes (all-cause death and CVEs) were further adjusted
for cytokines individually. The association between all-cause mortality and
CTSL1 remained significant after adjustment for IL-6, IL-16, IL-18, IL-27,
IL-1RA, and IL6-RA. However, adjustment for IL-8 attenuated the associa-
tion between CTSL1 and risk for all-cause mortality (HR: 1.43; 95% CI: 0.98
to 2.08; p = 0.067). Adjustment for IL-8 also attenuated the associations with
CVEs of TRAIL-R2 (HR: 1.37; 95% CI: 0.96 to 1.96; p = 0.083), CTSD (HR:
1.38; 95% CI: 0.84 to 2.27; p = 0.2), and CTSL1 (HR: 1.25; 95% CI: 0.83 to
1.90; p = 0.29). Interestingly, the risk for CVEs of OPG remained significant
after adjustment for the various individual cytokines. For each unit increase in
OPG, the HR for CVEs after adjustment for IL-8 was 1.71 (95% CI: 1.06 to
2.78; p = 0.029).
OPG was the only bone-associated biomarker that improved the predictive
performance of our clinical risk model and the AURORA risk model. The
addition of OPG to a clinical risk model predicting CVEs showed a significant
36
NRI (0.32; 95% CI: 0.04 to 0.44; p = 0.03) and IDI (0.06; 95% CI: 0.01 to
0.10; p = 0.01). Adding OPG to the AURORA risk prediction model also re-
sulted in a NRI (0.21; 95% CI: 0.02 to 0.36; p = 0.02) and an IDI (0.04; 95%
CI: 0.004 to 0.09; p < 0.001).
Study IV
In multivariable-adjusted Cox models in study IV, 15 proteins were associated
with increased risk for all-cause death; 5, with increased risk for CV death;
and 17, with increased risk for CVEs. The adjusted Cox models also found 5
proteins related to a decreased risk for all-cause death and 2 related to a de-
creased risk for CV death.
A Venn diagram summarizing the PEA proteins associated with outcomes in
the adjusted multivariable model found that IL-8, T cell immunoglobulin and
mucin domain 1 (TIM-1), and C-C motif chemokine 20 (CCL20) were bi-
omarkers for all-cause and CV mortality, and CVEs. The HRs of IL-8 in the
adjusted multivariable model were 1.29 (95% CI: 1.11 to 1.51; p = 0.001) for
all-cause death, 1.34 (95% CI: 1.02 to 1.76; p = 0.036) for CV death, and 1.33
(95% CI: 1.11 to 1.59; p = 0.002) for CVEs. The HRs of TIM-1 in the adjusted
multivariable model were 1.26 (95% CI: 1.01 to 1.57; p = 0.036) for all-cause
death, 1.68 (95% CI: 1.15 to 2.45; p = 0.007) for CV death, and 1.34 (95% CI:
1.05 to 1.72; p = 0.017) for CVEs. The HRs of CCL20 in the adjusted multi-
variable model were 1.30 (95% CI: 1.11 to 1.52; p = 0.001) for all-cause death,
1.39 (95% CI: 1.08 to 1.79; p = 0.010) for CV death, and 1.27 (95% CI: 1.08
to 1.51; p = 0.005) for CVEs. Interestingly, the stem cell factor (SCF) and
galanin peptides (GAL) were both associated with reduced risk of all-cause
and CV mortality. The corresponding HRs in the adjusted multivariable mod-
els were, for SCF, 0.82 for all-cause death (95% CI: 0.69 to 0.96; p = 0.016)
and 0.72 for CV death (95% CI: 0.56 to 0.94; p = 0.015), and for GAL, 0.75
for all-cause death (95% CI: 0.63 to 0.90; p = 0.002) and 0.72 for CV death
(95% CI: 0.53 to 0.97; p = 0.033).
37
Additional Preliminary Study
Owning to another prospective hemodialysis cohort was conducted in Taiwan
Kaohsiung Medical University Hospital (KMUH), we launched a new study
to compare the hemodialysis patients between Denmark and Taiwan using in-
dividual data. The aim of the additional preliminary study was to compare
mortality and CVE differences for patients receiving HD in cohorts from Den-
mark (Århus HD cohort) and Taiwan (KMUH HD cohort). (Figure 3).
Figure 3. The study was designed to compare clinical outcomes in hemodialysis patients among Demark Århus cohort and the Taiwan KMUH cohort
Method and Statistical analysis of the additional preliminary study
Participants from the dialysis units of Kaohsiung Medical University Hospital
(KMUH) and Kaohsiung Municipal Hsiao-Kang Hospital were enrolled be-
tween August 2016 and January 2017. The enrolled participants were more
than 18 years of age and had received HD for more than 90 days. All partici-
pants received high-efficiency dialyzers therapy 3 times per week, with a tar-
get for adequate dialysis of Kt/V > 1.2. The HD parameters were 250–
300 mL/min blood flow, 500 mL/min dialysate flow, and 3.5–4 hours per di-
alysis session. The baseline characteristics of the patients (age, sex, dialysis
38
vintage, arteriovenous access type (fistula or graft), primary cause of kidney
failure [hypertension, diabetes, glomerulonephritis, or others], comorbidities,
medications, and biochemical data) were obtained from electronic health care
record systems. Hypertension was defined as blood pressure exceeding
140/90 mmHg or the use of blood-pressure-lowering drugs. Diabetes mellitus
was defined as HbA1C exceeding 6.5% or the use of glucose-lowering drugs.
The outcomes and mortality assessment were assessed as the previous defini-
tion (study III and study IV) by physician diagnosis in Demark Århus cohort
and Taiwan KMUH cohort. Kaplan–Meier curves were generated to show the
cumulative probabilities of all-cause mortality and CVEs over 4 years of ob-
servation time, and differences were examined with a log-rank test. Further,
the differences between the two regions in all-cause mortality and CVEs were
analyzed in 4 different Cox regression models, with adjustment for confound-
ers. (Model 1: age, sex, body mass index, systolic and diastolic blood pressure,
cause of ESKD, and HD vintage; Model 2: Model 1 plus comorbidities [dia-
betes mellitus, myocardial infarction, ischemic stroke]; Model 3: Model 2
plus clinical laboratory data [hemoglobin, albumin, low-density lipoprotein
cholesterol, CRP, total calcium, phosphate, PTH) and propensity score;
Model 4: weighted by stabilized inverse probability of treatment weighting).
The association between circulating high-sensitivity troponin T (hsTNT) and
outcomes (all-cause mortality and CVEs) in two cohorts was also evaluated
to illustrate the predictive effect across geographic regions.
Preliminary results of additional study
In the additional study, Demark’s Århus cohort and Taiwan KMUH cohort
were compared for differences in baseline characteristics, all-cause mortality,
and CVEs. Compared with patients in the KMUH cohort, patients in the Århus
cohort were older and more likely to be male. They also had a shorter HD
duration, a proportionally greater history of ischemic stroke, and a lesser his-
tory of diabetes mellitus. Patients in the Århus cohort also had a higher body
mass index; higher CRP, total calcium, phosphate, and hsTNT; and lower sys-
tolic and diastolic blood pressure, and PTH (Table 2).
Kaplan–Meier curves showed greater all-cause mortality and CVEs in the År-
hus cohort than in the KMUH cohort (Figure 4). After controlling for con-
founders, including in adjusted multivariable models and stabilized inverse
probability of treatment weighting models, Cox regression analysis found that
all-cause mortality and CVEs were greater in the Århus cohort than in the
KMUH cohort (Table 3).
39
Given that hsTNT was an excellent biomarker for predicting CVD, we further
evaluated its relationships with the high mortality risk cohort (Århus) and the
low mortality risk cohort (KMUH). Importantly, Kaplan–Meier curves for
hsTNT tertiles demonstrated that a higher level of hsTNT was associated with
higher risk for all-cause mortality and CVEs in patients at high CV risk (Århus
cohort) and in patients at low CV risk (KMUH cohort) (Figure 5 and 6).
Table 2. Baseline characteristics of participants in Taiwan KMUH hemodialysis co-hort and Demark Århus hemodialysis cohort
Taiwan KMUH
cohort
(N=347)
Demark Århus
cohort
(N=331)
P
Age, years 60.0 (51.0, 67.0) 67.1 (57.0, 76.5) <0.001
Male gender 187 (53.9%) 211 (63.7%) 0.011
Body mass index, kg/m2 23.6 (21.0, 25.8) 25.2 (22.2, 29.2) <0.001
Systolic blood pressure, mmHg 147.0 (128.5, 163.0) 141.0 (122, 158) 0.011
Diastolic blood pressure, mmHg 80.0 (70.0, 89.0) 69.5 (59.0, 79.0) <0.001
Cause of ESKD <0.001
Hypertension 37 (10.7%) 45 (13.6%)
Diabetes Mellitus 121 (34.9%) 61 (18.4%)
Glomerulonephritis 125 (36.0%) 39 (11.8%)
Others* 64 (18.4%) 186 (56.2%)
Hemodialysis vintage, years 5.0 (2.0, 12.0) 2.5 (0.9, 5.3) <0.001
Comorbidities
Diabetes mellitus 148 (42.7%) 90 (27.2%) <0.001
Myocardial infarction 59 (17.1%) 61 (18.4%) 0.713
Ischemic stroke 23 (6.6%) 61 (18.5%) <0.001
Clinical laboratory data
Hemoglobin, g/dL 10.7 ± 1.2 11.6 ± 1.3 <0.001
Albumin, g/dL 3.9 ± 0.3 3.9 ± 0.4 0.457
Low-density lipoprotein, mg/dl 93.0 (71.0, 116.5) 88.8 (65.6, 123.6) 0.816
C-reactive protein, mg/L 0.8 (0.2, 2.9) 5.7 (2.3, 15.0) <0.001
Total Calcium, mg/dL 9.3 ± 0.9 9.5 ± 0.8 <0.001
Phosphate, mg/dL 4.6 (3.9, 5.3) 6.3 (5.4, 7.4) <0.001
Parathyroid hormone, pg/ml 336.8 (149.6, 589.5) 198.1 (98.1, 333.0) <0.001
Cardiovascular biomarker
hsTNT, ng/l 52.0 (32.0, 81.0) 63.1 (40.5, 109.0) <0.001
Log-transformed hsTNT 1.7 ± 0.3 1.8 ± 0.3 <0.001
Abbreviation: ESKD, end-stage kidney disease; hsNTN, high sensitivity troponin T
Continuous variables as mean and standard deviation or median and interquartile
range; categorical variables as n (%).
*Other causes of end-stage renal disease include polycystic kidney disease, tumor,
systemic lupus erythematosus, gout, interstitial nephritis
40
Figure 4. Kaplan–Meier curves for (A) all-cause mortality and (B) composite vascu-lar events in patients of Demark Århus hemodialysis cohort and Taiwan KMUH he-
modialysis cohort
Figure 5. Kaplan-Meier curves of hsTNT tertiles for all-cause mortality in patients of Demark Århus hemodialysis cohort, Taiwan KMUH hemodialysis cohort, and in
both cohorts combined.
Figure 6. Kaplan-Meier curves of hsTNT tertiles for CVEs in patients
of Demark Århus hemodialysis cohort, Taiwan KMUH hemodialysis cohort, and in both cohorts combined
41
Table 3. The comparison of all-cause mortality, composite vascular events, and single vascular event (myocardial infarction, ischemic stroke, peripheral artery disease) in hemodialysis patients between Taiwan KMUH cohort and Demark Århus cohort
Outcomes Hazard Ratio (95% Confidence Interval)
Unadjusted Model 1a Model 2b Model 3c Model 4d
All-cause mortality
Taiwan KMUH cohort 0.22 (0.16 - 0.31)* 0.20 (0.13 - 0.31)* 0.20 (0.13 - 0.31)* 0.11 (0.02 - 0.54)* 0.14 (0.09 - 0.22)*
Demark Århus cohort 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference]
CVE
Taiwan KMUH cohort 0.52 (0.38 - 0.71)* 0.48 (0.32 - 0.72)* 0.50 (0.33 - 0.75)* 0.54 (0.32 - 0.92)* 0.58 (0.35 - 0.95)*
Demark Århus cohort 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference]
Myocardial infarction
Taiwan KMUH cohort 1.18 (0.74 - 1.90) 0.89 (0.49 - 1.64) 0.87 (0.46 - 1.63) 0.76 (0.33 - 1.73) 1.23 (0.52 - 2.88)
Demark Århus cohort 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference]
Ischemic stroke
Taiwan KMUH cohort 0.45 (0.25 - 0.82)* 0.46 (0.20 - 1.04) 0.49 (0.21 - 1.11) 0.71 (0.24 - 2.11) 0.73 (0.26 - 2.07)
Demark Århus cohort 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference]
Peripheral artery disease
Taiwan KMUH cohort 0.28 (0.17 - 0.47)* 0.28 (0.15 - 0.52)* 0.29 (0.16 - 0.55)* 0.36 (0.16 - 0.84)* 0.14 (0.07 - 0.28)*
Demark Århus cohort 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference] 1 [Reference] aModel 1: adjusted for age, sex, body mass index, systolic blood pressure, diastolic blood pressure, cause of ESKD, and hemodialysis vintage. bModel 2: model 1 with further adjusted for comorbidities (Diabetes mellitus, Myocardial infarction, Ischemic stroke). cModel 3: model 2 with further adjusted for clinical laboratory data (Hemoglobin, Albumin, Low-density lipoprotein, C-reactive protein, Total
Calcium, Phosphate, Parathyroid hormone), and propensity score. dModel 4: Confounders control using weighted by stabilized inverse probability of treatment weighting (stabilized IPTW)
42
Discussion
Klotho gene SNPs and all-cause or CV mortality
In study I, SNPs in the klotho gene were evaluated by computational pro-
grams and databases that use various methods to predict SNP damage. The
nonsynonymous SNPs rs9536314 and rs9527025 were classified as the most
damaging. In the clinical study, unadjusted logistic regression analysis using
an additive model found an association of all-cause mortality risk with the TT
genotype of rs9536282, the GG genotype of rs9536314, and the CC genotype
of rs9527025. However, no statistically significant associations between pol-
ymorphisms in klotho and mortality risk were observed in unadjusted Cox
regression or adjusted logistic regression models. Results from the dominant
and recessive models were similar. Subgroup analyses of rs9536314 and
rs9527025 stratified by eGFR or FGF23 also produced negative results for the
MrOS cohort.
A nonsignificant trend of association was observed for the nonsynonymous
SNPs rs9536314 and rs9527025 with all-cause mortality, but not CV mortal-
ity. The SNPs rs9536314 (F352V) and rs9527025 (C370S) have been reported
to be related to CV diseases.67, 68 However, a meta-analysis of rs9536314
(F352V) and rs9527025 (C370S) with respect to CV disease failed to indicate
any statistically significant association in the additive genetic model, the dom-
inant genetic model, or the recessive genetic model.69 The minor allele fre-
quency differences for rs9536314 and rs9527025 in Sweden’s MrOs data and
other databases (the 1000 Genomes EUR Population, the HapMap CEU Pop-
ulation, and the Genome Aggregation Database European Population) might
partly explain the outcomes difference. Although computational prediction
signaled a “damaging” effect of the klotho KL-VS genotype in our study, no
statistically significant differences were observed in the cohort, possibly be-
cause the cohorts were different in age, sex, ethnicity, and comorbidities. The
functional klotho KL-VS variant results in amino acid substitutions that alter
secretion, catalytic activity, and functionality of the αkl protein.70, 71 However,
a study has also reported that substitution SNPs rs9536314 (F352V) and
rs9527025 (C370S) alter amino acid-related shedding and trafficking. Still,
the effects are small, and intragenic complementation could further minimize
the effects in vivo.72
43
Another klotho gene synonymous variant, rs564481 (C1818T), located in the
fourth exon, showed no association with all-cause or CV mortality in our
study, according to earlier studies that showed no CVD correlation for this
polymorphism.68, 69, 73 Although rs564481 is a synonymous variant, it is likely
to be functionally relevant, given that reports have associated it with CV risk
factors (blood pressure, glucose metabolism, and lipid levels)74, 75 and CAD76
in Asian populations. The SNP rs577912 was previously related to higher
mortality risk in patients receiving hemodialysis77 but our clinical study re-
sulted in negative findings. Interestingly, the computational tool FATHMM
predicts that the rs577912 genotype could be a damaging intron non-coding
region; however, no association with mortality outcomes was found in our
study.
The mineral bone factors and all-cause or CV mortality
In Study II, we demonstrated that high FGF23 was associated with all-cause
and CV mortality, especially in individuals with moderate to severe CKD
(CKD 3B or worse). Moreoever, in an RF analysis, compared with vitamin D
or PTH, FGF23 was found to be the mineral metabolic factor most important
to all-cause mortality and CV mortality. However, that association was atten-
uated after controlling for renal function. Conversely, vitamin D deficiency
was independently associated with all-cause mortality, but not CV mortality.
High and low PTH showed no relationship with mortality in our study. In a
combined evaluation, as the values for FGF23, vitamin D, and renal function
became more abnormal, all-cause and CV mortality were found to rise. Be-
cause compensation effects between these mineral bone factors are known,
FGF23 excess, vitamin D deficiency, and renal impairment together imply
loss of the compensating effect and might predict poor outcomes. Our findings
tease out the complexity of these mineral metabolic factors, renal function,
and mortality relationships.
In our study, the association of FGF23, but not vitamin D, with the risk for
all-cause mortality attenuated after adjustment for eGFR, indicating that renal
function remains the central factor. In contrast, the association of vitamin D
deficiency with mortality risk was independent of known mineral metabolism
and renal function. As reported, vitamin D deficiency was found in several
meta-analytic studies to be associated with mortality.78-80 However, vitamin D
deficiency can result from comorbidities and nutrient deficiencies, which
could themselves explain the observed links with mortality.81, 82 Currently, no
conclusive evidence shows that vitamin D supplementation provides a sur-
vival benefit in adults.83 In fact, vitamin D supplementation increases FGF23
concentrations84, which might counter the benefits of vitamin D supplements.
44
FGF23 can suppress the production of active vitamin D metabolites. Any det-
rimental effect of FGF23 might be influenced by reduced vitamin D activity.85
In our study, vitamin D deficiency was found to be a factor independently as-
sociated with all-cause mortality beyond FGF23. Other reports support our
finding that the highest risk for all-cause mortality was found in hemodialysis
patients with a low vitamin D level and with a high FGF23 level.19
High serum FGF23 has been observed as a physiologic reaction to impaired
renal function when renal phosphate excretion is compromised.86 The health
impact of these mineral bone factors should therefore be evaluated simultane-
ously. During the early stages of CKD, expression of the renal FGF23 co-
receptor, αkl, progressively declines in response to kidney damage and de-
clines along with the loss of functional nephrons are lost, promoting partial
resistance to FGF23’s physiologic actions.87-89 As a compensatory mecha-
nism, FGF23 rises above normal values by a factor of 1000 to maintain a neu-
tral phosphate distribution.87, 90, 91 The compensatory increase in FGF23 pro-
motes the suppression of 1-25-dihydroxyvitamin D production, which in turn
promotes the elevation of PTH, causing secondary hyperparathyroidism.92
Thus, renal dysfunction is the major contributing factor to increased circulat-
ing FGF23. As demonstrated in our study, high FGF23 was not associated
with risk for all-cause mortality after adjustment for eGFR. Thus, high FGF23
is a consequence, but not a cause, of risk for mortality.
FGF23 has been associated with non-CV causes of death,93 reflecting a lack
of specificity between a rise in FGF23 concentration and disease risk. In our
study, a higher level of FGF23 was not associated with increased CV mortality
after adjustment for diabetes, hypertension, stroke, and CAD. Theoretically,
FGF23 rises before any other marker of mineral bone disorder, and so it tem-
porally mirrors the rise in CV risk as kidney disease progresses. Many condi-
tions can cause that rise, and its presence and severity generally reflect disease
activity. Nevertheless, elevated FGF23 has still been associated with mortality
across a spectrum of CKD.91, 92 In our study, lower CKD status, particularly
eGFR less than 45 ml/min, produced higher mortality spline plots. Because
the kidney is the main source of systemic αKlotho, any reduction in nephron
numbers and in eGFR is followed by a progressive decline in circulating
αKlotho.94 That the protection conferred by αkl is greater than the deleterious
effect of FGF23 has been demonstrated in clinical95 and animal studies.96 This
raises the possibility that αKlotho deficiency may be the primary alteration in
MBD.97 Thus, we found that renal function is the crucial factor contributing
to the risk association between FGF23 and mortality.
45
Bone-associated markers and mortality or CVE in hemodialysis patients
In Study III, we evaluated the relationships of 9 bone-associated biomarkers
with all-cause and CV mortality, and CVEs in 331 patients receiving HD. Us-
ing an RF approach, the importance of the bone biomarkers with respect to
outcomes was evaluated, and the top biomarkers were TRAIL-R2 for all-cause
death, OPG for CV death, and both OPG and TRAIL-R2 for CVEs. However,
only OPG was independently associated with CVEs in a cytokine adjustment
model. OPG could potentially provide clinical information beyond that cur-
rently available for the prediction of CVEs. Relationships of the bone-associ-
ated proteins, such as OPG, TRAIL-R2, CTSD, and CTSL1, with CV disease
have previously been reported.98-105 Our study also showed a relationship be-
tween TRAIL-R2 and OPG106 with higher CV risk. However, in a departure
from previous reports,104, 107 we found no correlation between low TRAIL and
poor outcomes.
In our study, OPG was the strongest predictor of CVEs. Elevated OPG has
previously been associated with aortic stiffness and coronary artery calcifica-
tion,108, 109 as well as CV outcomes.110 Interestingly, the RF approach was re-
cently used to show an association of OPG with CVEs in non-dialysis CKD.
Forne et al. investigated 19 biomarkers in a subcohort of the NEFRONA study
(Observatorio Nacional de Atherosclerosis en NEFrologia).111 They used var-
iable importance from the RF analysis for CV risk in the first step and then
evaluated the most promising variables for CVE in a Fine and Gray competing
risks model. OPG was a potential predictor of CV risk in the RF analysis and
also a statistically significant predictor of increased CVEs in the competing
risks model.111 Elevated OPG had already been associated with increased CV
mortality112-114 in patients treated with hemodialysis. Compared with our
study, the study by Sigrist et al. enrolled patients receiving hemodialysis who
were younger and had lower mortality rates109 ; patients receiving hemodialy-
sis who were enrolled in other studies by Morena et al.112 and Winther et al.114
had a higher prevalence of baseline CV diseases. In contrast to these previous
studies,112-114 OPG was not an independent risk marker for all-cause mortality
in our study. The results of discrepancies may exist because of different con-
founders' adjustments across studies.
Although a positive association between baseline OPG concentration and risk
for CV disease has been shown, some analytic issues should be discussed. No
study has evaluated the independent association between OPG and CVEs in a
model adjusted for inflammatory cytokines.108, 109, 112, 113, 115 Inflammation is
well known to be a significant risk factor for mortality and CV events in pa-
tients with CKD.116 Inflammatory cytokines influence the production of OPG,
suggesting that the OPG/RANKL/RANK system plays a modulatory role in
46
vascular injury and inflammation.117 One study suggested that OPG levels are
significantly correlated with inflammatory markers, such as IL-6.118 As
demonstrated in our study, OPG was positively associated with inflammatory
cytokines, such as IL-6, IL-8, IL-16, IL-18, IL-27, IL-1RA, and IL-6RA.
However, we found no significant correlation between OPG and CRP, in line
with previous studies in patients with CKD and those receiving hemodialy-
sis.109, 112 It has been suggested that OPG alone or in combination with elevated
CRP is associated with worse outcomes in such patients.109, 112 In addition,
after adjustment for confounders (including CRP), OPG was found to be a
significant independent predictor of PAD in patients receiving peritoneal di-
alysis.119 In our adjusted multivariable model, we also demonstrated that OPG
predicts CVEs even after adjustment for CRP. Furthermore, we demonstrated
that the association between OPG and CVEs was independent of cytokine ac-
tivity or calcium/phosphate/PTH parameters. More studies are needed to con-
firm our findings.
CV-associated markers for mortality or CVE in hemodialysis patients
In Study IV, a recent PEA-based proteomics study in the Mapping of Inflam-
matory Markers in Chronic Kidney disease (MIMICK) cohort, TIM-1, matrix
metalloproteinase (MMP)-7, tumor necrosis factor TNF receptor 2, IL-6,
MMP-1, brain natriuretic peptide, ST2, hepatocyte growth factor (HGF),
TRAIL-R2, spondin-1, and FGF23 showed significant associations with CV
death after adjustment for age and sex in 183 patients receiving HD.120 Several
of the proteins reported in the MIMICK study were also correlated with CVEs
in our study—specifically, TIM-1, MMP-7, IL-6, brain natriuretic peptide,
HGF, and TRAIL-R2. However, some proteins associated with CV death in
our study (e.g., tissue plasminogen activator [tPA], CCL20, GAL, IL-8, and
SCF) were not associated with CV death in the MIMICK cohort. Importantly,
TIM-1, also known as kidney injury molecule-1, was the most important risk
marker for CV death in the MIMICK cohort,120 a result confirmed in our study.
Importantly, TIM-1, also known as kidney injury molecule-1, was the most
important risk marker for CV death in the MIMICK cohort.120 There could be
several reasons for the differences in the two studies, including a smaller sam-
ple size in MIMICK (n=183), the longer dialysis vintage in our study, and
adjustment for different confounders. In addition, the definition of CV death
was validated by a physician in our study, while the MIMICK cohort used
International Classification of Diseases diagnostic codes. Moreover, in a dis-
covery-validation study of individuals with CKD not receiving dialysis,
MMP-12 was significantly associated with an increased risk of major adverse
CV events (fatal or nonfatal myocardial infarction or fatal and nonfatal
47
stroke).121 That research also agrees with our finding that MMP-12 is an inde-
pendent biomarker predicting CVEs in patients receiving HD.
Chronic inflammation is an essential CV risk factor in patients with kidney
disease, characterized by enhanced production of CRP and other inflamma-
tory mediators, including IL-6 and IL-8. Those cytokines and chemokines
were previously associated with all-cause and CV mortality in patients receiv-
ing hemodialysis.122-125 In line with previous studies in such patients,126 EN-
RAGE, another inflammatory biomarker, was also positively associated with
CVEs. Furthermore, we found that TRAIL-R2 (member protein of the TNF-
receptor superfamily) and OPG, reflecting the bone-vascular axis, were asso-
ciated with a higher risk of CVEs. OPG has previously been described as a
CV marker for all-cause death in hemodialysis patients,112-114 and TRAIL-R2
is present in human atherosclerosis lesions, with higher expression in vulner-
able plaques than in stable ones.106
Potential mechanisms and clinical studies of novel biomarkers
Several novel biomarkers, such as HGF, CCL20, tPA, GAL, and SCF, were
related to death or CV outcomes.
HGF is a pleiotropic cytokine involved in regulating several biologic pro-
cesses, such as cardiometabolic activity, inflammation, angiogenesis, and tis-
sue repair.127 HGF is associated with CAD128 and ischemic stroke129 in patients
without CKD. In hemodialysis patients, higher levels of HGF are associated
with concentric left ventricular geometry130 and cerebral infarction.131 Those
associations might be related to leukocyte activation during hemodialysis.132
Our study is the first to demonstrate that HGF is associated with an increased
risk of both all-cause mortality and CVEs in hemodialysis patients.
Previous studies have shown that CCL20 excretion is increased in patients
with stage 5 CKD compared with healthy individuals and patients with
stages 1–3 CKD.133 Furthermore, patients with ischemic heart disease have
higher levels of circulating CCL20,134 and CCL20 is expressed in atheroscle-
rotic plaques.135 A recent study demonstrated that CCL20 is associated with
increased CVEs in patients with stage 3-5 CKD.121 CCL20 contributes to vas-
cular endothelial inflammation136 and triggers pathways similar to those acti-
vated by low-density lipoprotein cholesterol.135 The IL-6, IL-8, and CCL20
cytokines or chemokines have been associated with the Th17 CD4 lympho-
cyte.137 Several studies have shown that Th17 cells play a critical role in arte-
riosclerosis development.138 Genetic deletion of CCL20 receptors in Apoe−/−
mice decreases atherogenesis and endothelial inflammation.139 We, therefore,
speculate that the accumulation of CCL20 might significantly affect the risk
for CV disease in patients receiving hemodialysis.
48
The tPA glycoprotein is involved in coronary plaque rupture140 and is consid-
ered a marker for regulating endogenous fibrinolysis.141 In the general popu-
lation without kidney disease, there is an association between elevated circu-
lating levels of tPA and subsequent coronary heart disease.141 In general, cir-
culating levels of tPA are elevated in patients receiving dialysis compared
with healthy individuals,142 and circulating levels of tPA in hemodialysis pa-
tients are positively associated with several CV risk factors, such as age,
smoking, blood pressure, and CRP.142
GAL is an endocrine hormone of the central and peripheral nervous systems
involved in central CV regulation, affecting heart rate and blood pressure.143
However, functional properties of the galaninergic system are not fully eluci-
dated for cardiac diseases. In an animal model, GAL can limit myocardial in-
farction size and improve post-ischemic cardiac function recovery.144 It also
suppresses myocardial apoptosis and mitochondrial oxidative stress in cardiac
hypertrophic remodeling.145 These basic studies could support our results of
the negative association between GAL and CV mortality in patients receiving
hemodialysis.
The association between high levels of SCF and lower all-cause or CV mor-
tality was a novel finding of our study. Similar results were described in the
general population (Malmö Diet and Cancer study)146 and in patients with sta-
ble CAD.147 SCF is involved in vasculogenesis and cardiac repair by stimulat-
ing the recruitment and activation of bone marrow-derived stem cells and tis-
sue-resident progenitors.148 An increase in SCF expression occurs naturally in
response to myocardial infarction, which mediates the migration of c-kit+ car-
diac and bone marrow cells to the injured area for cardiac remodeling.149 We
found that circulating SCF was positively correlated with albumin. Albumin
is well known to be a predictive marker in patients receiving hemodialysis.
We speculate that SCF might play a protective role in vascular injury and
could reflect the clinical nutrition status in patients receiving hemodialysis.
Mortality differences among hemodialysis patients between Asian and Caucasian
In the additional preliminary study, we found that patients in Demark’s År-
hus HD cohort had a higher risk for all-cause mortality and CVEs than patients
in Taiwan’s KMUH HD cohort. Interestingly, racial and ethnic differences in
mortality for patients receiving HD have been reported in several studies.150-
153 Some reports suggested a 25% to 35% lower mortality rate in Asian pa-
tients than in White patients.154, 155 The causes of these ethnic disparities re-
main largely unknown. Factors that might explain the mortality differences
49
include genetic and environmental factor differences151 and racial disparities
in diet, nutrition, and inflammation.156 This recent comparison between hemo-
dialysis patients in the two nations will be developed further into a full manu-
script and hopefully published in the near future.
50
Strengths and limitations
Strengths of Studies I and II include the large sample size of the MrOS cohort
of community-dwelling men with nearly complete follow-up of surviving co-
hort participants, and outcome measures that were well-validated because of
high-quality national registers. We used 14 different computational tools in a
bioinformatics approach, plus a clinical study, to perform a complete evalua-
tion of klotho gene polymorphism. However, several limitations must still be
addressed. First, self-reported questionnaires were used at baseline visits, and
we cannot exclude the possibility of underestimation of smoking or disease
prevalence. Second, the study was observational, and so causality was not pos-
sible to cannot be determined. Third, the generalizability of our study is lim-
ited to healthy community-dwelling Swedish males. Fourth, the minor allele
frequency of rs9536314 homozygotes could be different in elderly and young
subjects.70 The latter limitation might also affect our study result because of
the enrolled elderly population in the MrOS cohort (Study I). Fifth, FGF23,
vitamin D, and PTH were measured at a single point, so we could not evaluate
those mineral metabolic factors as time-varying covariates (Study II). Sixth,
residual and unmeasured confounders might remain despite the attempt to ad-
just for potential confounders such as genetics and other CV biomarkers
(hsTNT) (Study II). Lastly, we could not evaluate other FGF23 regulators or
mediators (i.e., iron metabolism,157, 158 inflammatory proteins,158, 159 or 1-25-
dihydroxyvitamin D158) because of unavailable data in the Sweden MrOs co-
hort (Study II).
In Studies III and IV, the strengths of our investigation include a longitudinal
study design, proteomics chip for simultaneous measurement of multiple
bone-associated proteins, and exploration analysis using the RF approach.
Limitations include the observational design, including prevalent patients with
varying disease durations and low statistical power in some of our analyses.
The study focused mainly on White patients receiving HD therapy in Scandi-
navia, so extrapolations to non-White patients receiving HD or patients with
CKD not yet on dialysis should be made cautiously. Furthermore, the scale
for PEA-based protein levels is not an absolute concentration, so the perfor-
mance of the protein biomarkers and definition of risk thresholds in a clinical
setting warrants further study or the use of methods yielding absolute values,
possibly by calibrating the PEA values. The study also used single assess-
ments of the proteomic assay, so misclassification and short-term variability
are potential issues. Finally, the number of statistical tests performed for this
manuscript increases the risk of spurious findings; hence, novel findings not
previously published have to be replicated in an independent cohort of pa-
tients.
51
Conclusion and future perspectives
This thesis contributes to the identification of biomarkers related to all-cause
mortality, CV mortality, and CVEs in patients in general and in those with
ESKD, but especially markers targeting the bone–vascular axis. In Sweden’s
MrOs elderly community cohort, two potentially damaging SNPs (rs9536314
and rs9527025) in the klotho gene were not associated with all-cause mortality
or CV mortality (Study I). However, composite evaluation of elevated
FGF23, vitamin D deficiency, and renal impairment was associated with all-
cause and CV mortality. Among the bone-related factors, elevated FGF23 was
the most important marker for predicting mortality, but it was clearly influ-
enced by renal function. Long-term follow-up data on the outcome will be
analyzed prior to submission of the study (Study II). In Denmark Århus co-
hort of individuals receiving hemodialysis, one of the most promising bone-
related proteins, OPG, was associated with CV events independent of cytokine
activity (Study III). Using the PEA proteomics approach, we identified
higher IL-8, TIM-1, and CCL20, as well as lower SCF and GAL, as being
associated with poor outcomes (Study IV). We addressed the issue of the
bone–vascular axis and CVD. We evaluated from gene levels to circulating
protein levels and from the general population to patients with CKD. Based
on our research findings, more evidence was generated to link bone and vas-
cular complications. We also identified several CV proteins considered to be
potential biomarkers in patients receiving HD. Because we had 2 independent
prospective HD cohorts (Demark Århus cohort and Taiwan KMUH cohort), a
direct comparison of mortality and CVEs between the two geographic regions
was intriguing. We found a higher risk for all-cause mortality and CVEs in the
patients of the Århus cohort than in the patients of the KMUH cohort. Inter-
estingly, hsTNT can adequately predict outcomes in both cohorts. This will
all be analyzed further and developed into a separate communication (Addi-
tional preliminary study).
For the future, a simultaneous evaluation of PEA-based CV proteomics in the
Århus and KMUH cohorts will be fascinating. Given the risk for death or
CVEs triggered by uremic toxins in patients with kidney disease, an investi-
gation of molecules associated or interacting with uremic toxins and CV pro-
teomics could illuminate the pathophysiology of uremic toxins with respect to
the CV system. Given that the gut microbiota produces many uremic toxins, a
study to link the gut microbiome, uremic toxins, CV protein biomarkers, and
52
clinical CV phenotype would be reasonable. A CV risk prediction model
based on different omics data could be constructed to improve risk stratifica-
tion in patients with kidney disease. Based on the principles of precision med-
icine, we expect an early diagnosis and early treatment to overcome CVD in
patients with CKD.
53
Summary in Swedish (sammanfattning på svenska)
Denna avhandling bidrar till att identifiera biomarkörer som är relaterade till
allmän dödlighet, kardiovaskulär (CV) dödlighet, samt kardiovaskulära hän-
delser hos patienter med eller utan njursvikt. Ett särskilt fokus har varit på
biomarkörer involverade i ben-kärl-axeln.
Bland äldre män i Sverige (Sweden MrOs kohort) var två potentiellt skadliga
SNPs (rs9536314 och rs9527025) i Klotho-genen inte associerade till allmän
eller specifik CV-dödlighet (studie I). En sammansatt bedömning av förhöjda
FGF23-nivåer, D-vitaminbrist och njurfunktionsnedsättning var dock associ-
erad med dödlighet i allmänhet och även specifikt CV-dödlighet. Bland ben-
relaterade faktorer var förhöjda nivåer av FGF23 den viktigaste markören för
att förutsäga mortalitet, men denna faktor var tydligt relaterad till njurfunkt-
ionsstatus (studie II). Hos hemodialys (HD) patienter i en Dansk kohort var
osteoprotegrin det benrelaterade protein som var starkast relaterat till CV-hän-
delser oberoende av annan cytokinaktivitet (studie III). Med hjälp av PEA-
baserad metod för proteomik kunde vi slå fast att högre nivåer av IL-8, TIM-
1 och CCL20, samt lägre nivåer av SCF och GAL, var förknippade med ökad
mortalitetsrisk och även risk för CV händelser (studie IV).
Fokus i studierna har varit på proteinmarkörer, speciellt inom ben-kärl-axeln,
och kardiovaskulär sjuklighet. Vi har utvärderat sambanden både på gennivå
och vad gäller cirkulerande proteinnivåer. Studierna utfördes både inom den
allmänna befolkningen och i patienter med njursjukdom. Baserat på dessa
forskningsresultat föreligger det ett tydligt samband mellan benomsättning
och vaskulära komplikationer. Forskningsprojektet har även identifierat ett
flertal proteiner som kan betraktas som potentiella biomarkörer för hjärtkärl-
sjukdom hos HD-patienter.
54
Acknowledgments
There are many I would like to thank. This thesis is the result of many collab-
orations and the support and effort of many people. I am most grateful to
Östen Ljunggren, my main supervisor, has introduced me to the concept of
mineral bone disease. I appreciate your positive attitude to help me with the
research work and allow me to take lots of different courses.
Bengt Fellström, my co-supervisor and the senior mentor, your engagement,
determination, and enthusiasm for science have been invaluable for my devel-
opment as a researcher. You have always been supportive when I struggled
and encountered difficulties. Thank you for sharing your experience with car-
diovascular disease research in nephrology. You allow me to join this fantastic
field.
Torbjörn Linde (my co-supervisor) and My Hanna Sofia Svensson (my co-
author for study III and study IV), your wise advice in scientific writing and
commentaries on our research works were inspiring.
Rie Io Glerup and Per-Anton Westerberg, your data sharing and detailed in-
troduction for the prospective cohort helped me a lot.
I want to thank all colleagues in Kaohsiung Medical University Hospital, Tai-
wan. Mei-Chuan Kuo, my main supervisor in Taiwan, I must express my grat-
itude to you for guiding me in the thinking process, clinical experience, and
knowledge of nephrology. Yi-Wen Chiu, you are brilliant, knowledgeable,
and the best leader to support my research work in Taiwan. I appreciate your
patience and guidance.
A special thanks to my family for supporting me to study abroad and helping
me to take care of our baby. It’s such a blessing that I have you in my life.
55
References
1. Gansevoort RT, et al. Chronic kidney disease and cardiovascular risk: epidemiology, mechanisms, and prevention. Lancet. Jul 27 2013;382(9889):339-52.
2. Go AS, et al. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med. Sep 23 2004;351(13):1296-305.
3. Gargiulo R, et al. Cardiovascular disease and chronic kidney disease. Dis Mon. Sep 2015;61(9):403-13.
4. Jha V, et al. Chronic kidney disease: global dimension and perspectives. Lancet. Jul 20 2013;382(9888):260-72.
5. Chronic Kidney Disease Prognosis C, et al. Association of estimated glomerular filtration rate and albuminuria with all-cause and cardiovascular mortality in general population cohorts: a collaborative meta-analysis. Lancet. Jun 12 2010;375(9731):2073-81.
6. Hemmelgarn BR, et al. Relation between kidney function, proteinuria, and adverse outcomes. JAMA. Feb 3 2010;303(5):423-9.
7. Briasoulis A, et al. Chronic kidney disease as a coronary artery disease risk equivalent. Curr Cardiol Rep. Mar 2013;15(3):340.
8. Irie F, et al. The relationships of proteinuria, serum creatinine, glomerular filtration rate with cardiovascular disease mortality in Japanese general population. Kidney Int. Apr 2006;69(7):1264-71.
9. Ninomiya T, et al. Chronic kidney disease and cardiovascular disease in a general Japanese population: the Hisayama Study. Kidney Int. Jul 2005;68(1):228-36.
10. Liao LN, et al. Three-year incidence of elevated albuminuria and associated factors in a population-based cohort: The Taichung Community Health Study. Eur J Prev Cardiol. Jun 2015;22(6):788-97.
11. Husain-Syed F, et al. Cardio-Pulmonary-Renal Interactions: A Multidisciplinary Approach. J Am Coll Cardiol. Jun 9 2015;65(22):2433-48.
12. Barreto DV, et al. Coronary calcification in hemodialysis patients: the contribution of traditional and uremia-related risk factors. Kidney Int. Apr 2005;67(4):1576-82.
13. London GM, et al. Association of bone activity, calcium load, aortic stiffness, and calcifications in ESRD. J Am Soc Nephrol. Sep 2008;19(9):1827-35.
14. Adijiang A, et al. Indoxyl sulphate promotes aortic calcification with expression of osteoblast-specific proteins in hypertensive rats. Nephrol Dial Transplant. Jun 2008;23(6):1892-901.
15. Sun CY, et al. Suppression of Klotho expression by protein-bound uremic toxins is associated with increased DNA methyltransferase expression and DNA hypermethylation. Kidney Int. Apr 2012;81(7):640-50.
56
16. Vickery S, et al. Prognostic value of cardiac biomarkers for death in a non-dialysis chronic kidney disease population. Nephrol Dial Transplant. Nov 2008;23(11):3546-53.
17. Matsushita K, et al. Subclinical atherosclerosis measures for cardiovascular prediction in CKD. J Am Soc Nephrol. Feb 2015;26(2):439-47.
18. Snaedal S, et al. Comorbidity and acute clinical events as determinants of C-reactive protein variation in hemodialysis patients: implications for patient survival. Am J Kidney Dis. Jun 2009;53(6):1024-33.
19. Chonchol M, et al. Low Vitamin D and High Fibroblast Growth Factor 23 Serum Levels Associate with Infectious and Cardiac Deaths in the HEMO Study. J Am Soc Nephrol. Jan 2016;27(1):227-37.
20. Moe SM, et al. Cinacalcet, Fibroblast Growth Factor-23, and Cardiovascular Disease in Hemodialysis: The Evaluation of Cinacalcet HCl Therapy to Lower Cardiovascular Events (EVOLVE) Trial. Circulation. Jul 07 2015;132(1):27-39.
21. Wang AY, et al. N-terminal pro-brain natriuretic peptide: an independent risk predictor of cardiovascular congestion, mortality, and adverse cardiovascular outcomes in chronic peritoneal dialysis patients. J Am Soc Nephrol. Jan 2007;18(1):321-30.
22. D'Marco L, et al. Cardiovascular biomarkers in chronic kidney disease: state of current research and clinical applicability. Dis Markers. 2015;2015:586569.
23. Bos D, et al. Calcification in major vessel beds relates to vascular brain disease. Arterioscler Thromb Vasc Biol. Oct 2011;31(10):2331-7.
24. de Oliveira RB, et al. Vascular calcification in chronic kidney disease: a review. J Bras Nefrol. Apr-Jun 2013;35(2):147-61.
25. Durham AL, et al. Role of smooth muscle cells in vascular calcification: implications in atherosclerosis and arterial stiffness. Cardiovasc Res. Mar 15 2018;114(4):590-600.
26. Thompson B, et al. Arterial calcification and bone physiology: role of the bone-vascular axis. Nat Rev Endocrinol. Sep 2012;8(9):529-43.
27. Hu MC, et al. Klotho deficiency causes vascular calcification in chronic kidney disease. J Am Soc Nephrol. Jan 2011;22(1):124-36.
28. Lu X, et al. Klotho/FGF23 Axis in Chronic Kidney Disease and Cardiovascular Disease. Kidney Dis (Basel). Jul 2017;3(1):15-23.
29. Moe SM, et al. The pathophysiology of early-stage chronic kidney disease-mineral bone disorder (CKD-MBD) and response to phosphate binders in the rat. J Bone Miner Res. Nov 2011;26(11):2672-81.
30. Samelson EJ, et al. RANKL inhibition with denosumab does not influence 3-year progression of aortic calcification or incidence of adverse cardiovascular events in postmenopausal women with osteoporosis and high cardiovascular risk. J Bone Miner Res. Feb 2014;29(2):450-7.
31. Eastell R, et al. Biomarkers of bone health and osteoporosis risk. Review. The Proceedings of the Nutrition Society. May 2008;67(2):157-62.
32. Civitelli R, et al. Bone turnover markers: understanding their value in clinical trials and clinical practice. Review. Osteoporosis international. Jun 2009;20(6):843-51.
33. Vasikaran S, et al. Markers of bone turnover for the prediction of fracture risk and monitoring of osteoporosis treatment: a need for international reference standards. Osteoporosis international. Feb 2011;22(2):391-420.
34. Malluche HH, et al. Bone mineral density and serum biochemical predictors of bone loss in patients with CKD on dialysis. CJASN. Jul 2014;9(7):1254-62.
57
35. Gill RM, et al. Cardiac diastolic dysfunction in conscious dogs with heart failure induced by chronic coronary microembolization. Am J Physiol Heart Circ Physiol. Dec 2006;291(6):H3154-8.
36. Querejeta R, et al. Serum carboxy-terminal propeptide of procollagen type I is a marker of myocardial fibrosis in hypertensive heart disease. Circulation. Apr 11 2000;101(14):1729-35.
37. Ho CY, et al. Myocardial fibrosis as an early manifestation of hypertrophic cardiomyopathy. N Engl J Med. Aug 5 2010;363(6):552-63.
38. Ihm SH, et al. Serum carboxy-terminal propeptide of type I procollagen (PIP) is a marker of diastolic dysfunction in patients with early type 2 diabetes mellitus. Int J Cardiol. Nov 30 2007;122(3):e36-8.
39. Sharma UC, et al. Galectin-3 marks activated macrophages in failure-prone hypertrophied hearts and contributes to cardiac dysfunction. Circulation. Nov 9 2004;110(19):3121-8.
40. Calvier L, et al. The impact of galectin-3 inhibition on aldosterone-induced cardiac and renal injuries. JACC Heart Fail. Jan 2015;3(1):59-67.
41. Drechsler C, et al. Galectin-3, Renal Function, and Clinical Outcomes: Results from the LURIC and 4D Studies. J Am Soc Nephrol. Sep 2015;26(9):2213-21.
42. Aslam B, et al. Proteomics: Technologies and Their Applications. J Chromatogr Sci. Feb 2017;55(2):182-196.
43. Ponomarenko EA, et al. The Size of the Human Proteome: The Width and Depth. Int J Anal Chem. 2016;2016:7436849.
44. Smith JG, et al. Emerging Affinity-Based Proteomic Technologies for Large-Scale Plasma Profiling in Cardiovascular Disease. Circulation. Apr 25 2017;135(17):1651-1664.
45. Lundberg M, et al. Homogeneous antibody-based proximity extension assays provide sensitive and specific detection of low-abundant proteins in human blood. Nucleic Acids Res. Aug 2011;39(15):e102.
46. Assarsson E, et al. Homogenous 96-plex PEA immunoassay exhibiting high sensitivity, specificity, and excellent scalability. PLoS One. 2014;9(4):e95192.
47. Wu PH, et al. The association between Single Nucleotide Polymorphisms of Klotho Gene and Mortality in Elderly Men: The MrOS Sweden Study. Sci Rep. Jun 24 2020;10(1):10243.
48. Wu PH, et al. Osteoprotegerin predicts cardiovascular events in patients treated with hemodialysis. Nephrol Dial Transplant. Jun 4 2021;
49. Barrett JC, et al. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics. Jan 15 2005;21(2):263-5.
50. Yamazaki Y, et al. Increased circulatory level of biologically active full-length FGF-23 in patients with hypophosphatemic rickets/osteomalacia. J Clin Endocrinol Metab. Nov 2002;87(11):4957-60.
51. Flodin M, et al. Evaluation of Gentian cystatin C reagent on Abbott Ci8200 and calculation of glomerular filtration rate expressed in mL/min/1.73 m(2) from the cystatin C values in mg/L. Scand J Clin Lab Invest. 2007;67(5):560-7.
52. Siilin H, et al. Prevalence of primary hyperparathyroidism and impact on bone mineral density in elderly men: MrOs Sweden. World J Surg. Jun 2011;35(6):1266-72.
53. Staley JR, et al. A comparison of Cox and logistic regression for use in genome-wide association studies of cohort and case-cohort design. Eur J Hum Genet. Jun 2017;25(7):854-862.
58
54. Arnlov J, et al. Higher fibroblast growth factor-23 increases the risk of all-cause and cardiovascular mortality in the community. Kidney Int. Jan 2013;83(1):160-6.
55. Levey AS, et al. Definition and classification of chronic kidney disease: a position statement from Kidney Disease: Improving Global Outcomes (KDIGO). Kidney Int. Jun 2005;67(6):2089-100.
56. Eckardt KU, et al. Definition and classification of CKD: the debate should be about patient prognosis--a position statement from KDOQI and KDIGO. Am J Kidney Dis. Jun 2009;53(6):915-20.
57. Ross AC, et al. The 2011 report on dietary reference intakes for calcium and vitamin D from the Institute of Medicine: what clinicians need to know. J Clin Endocrinol Metab. Jan 2011;96(1):53-8.
58. Pilz S, et al. Vitamin D, cardiovascular disease and mortality. Clin Endocrinol (Oxf). Nov 2011;75(5):575-84.
59. Schierbeck LL, et al. Parathyroid hormone and vitamin D--markers for cardiovascular and all cause mortality in heart failure. Eur J Heart Fail. Jun 2011;13(6):626-32.
60. Hagstrom E, et al. Plasma-parathyroid hormone is associated with subclinical and clinical atherosclerotic disease in 2 community-based cohorts. Arterioscler Thromb Vasc Biol. Jul 2014;34(7):1567-73.
61. Hagstrom E, et al. Plasma parathyroid hormone and the risk of cardiovascular mortality in the community. Circulation. Jun 2 2009;119(21):2765-71.
62. Breiman L. Random Forests. Machine Learning. 2001;45(1):5-32. 63. Menze BH, et al. A comparison of random forest and its Gini importance
with standard chemometric methods for the feature selection and classification of spectral data. BMC Bioinformatics. Jul 10 2009;10:213.
64. Pencina MJ, et al. Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers. Stat Med. Jan 15 2011;30(1):11-21.
65. Pencina MJ, et al. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med. Jan 30 2008;27(2):157-72; discussion 207-12.
66. Holme I, et al. Prognostic model for total mortality in patients with haemodialysis from the Assessments of Survival and Cardiovascular Events (AURORA) study. J Intern Med. May 2012;271(5):463-71.
67. Arking DE, et al. KLOTHO allele status and the risk of early-onset occult coronary artery disease. Am J Hum Genet. May 2003;72(5):1154-61.
68. Majumdar V, et al. Association of the functional KL-VS variant of Klotho gene with early-onset ischemic stroke. Biochem Biophys Res Commun. Dec 17 2010;403(3-4):412-6.
69. Haiming Zhang YS, Fukang Ma, Lijun Wang, Yahong Hou, Zhijun Zhu. Association of Klotho single nucleotide polymorphisms with cardiovascular diseases: a systematic review and meta-analysis. Int J Clin Exp Med. 2017;10(3):5721-5741.
70. Arking DE, et al. Association of human aging with a functional variant of klotho. Proc Natl Acad Sci U S A. Jan 22 2002;99(2):856-61.
71. Dubal DB, et al. Life extension factor klotho enhances cognition. Cell Rep. May 22 2014;7(4):1065-76.
72. Tucker Zhou TB, et al. Biochemical and functional characterization of the klotho-VS polymorphism implicated in aging and disease risk. J Biol Chem. Dec 20 2013;288(51):36302-11.
59
73. Kim HK, et al. Lack of functional KL-VS polymorphism of the KLOTHO gene in the Korean population. Genet Mol Biol. Jul-Sep 2016;39(3):370-3.
74. Rhee EJ, et al. Relationship between polymorphisms G395A in promoter and C1818T in exon 4 of the KLOTHO gene with glucose metabolism and cardiovascular risk factors in Korean women. J Endocrinol Invest. Jul-Aug 2006;29(7):613-8.
75. Shimoyama Y, et al. KLOTHO gene polymorphisms G-395A and C1818T are associated with lipid and glucose metabolism, bone mineral density and systolic blood pressure in Japanese healthy subjects. Clin Chim Acta. Aug 2009;406(1-2):134-8.
76. Rhee EJ, et al. The differential effects of age on the association of KLOTHO gene polymorphisms with coronary artery disease. Metabolism. Oct 2006;55(10):1344-51.
77. Friedman DJ, et al. Klotho variants and chronic hemodialysis mortality. J Bone Miner Res. Nov 2009;24(11):1847-55.
78. Zittermann A, et al. Vitamin D deficiency and mortality risk in the general population: a meta-analysis of prospective cohort studies. Am J Clin Nutr. Jan 2012;95(1):91-100.
79. Garland CF, et al. Meta-analysis of all-cause mortality according to serum 25-hydroxyvitamin D. Am J Public Health. Aug 2014;104(8):e43-50.
80. Gaksch M, et al. Vitamin D and mortality: Individual participant data meta-analysis of standardized 25-hydroxyvitamin D in 26916 individuals from a European consortium. PLoS One. 2017;12(2):e0170791.
81. Al Mheid I, et al. Vitamin D and cardiovascular disease: is the evidence solid? Eur Heart J. Dec 2013;34(48):3691-8.
82. Tomson J, et al. Vitamin D and risk of death from vascular and non-vascular causes in the Whitehall study and meta-analyses of 12,000 deaths. Eur Heart J. May 2013;34(18):1365-74.
83. Zhang Y, et al. Association between vitamin D supplementation and mortality: systematic review and meta-analysis. BMJ. Aug 12 2019;366:l4673.
84. Charoenngam N, et al. Oral vitamin D3 supplementation increases serum fibroblast growth factor 23 concentration in vitamin D-deficient patients: a systematic review and meta-analysis. Osteoporosis international. Nov 2019;30(11):2183-2193.
85. Razzaque MS. The FGF23-Klotho axis: endocrine regulation of phosphate homeostasis. Nat Rev Endocrinol. Nov 2009;5(11):611-9.
86. Erben RG. Physiological Actions of Fibroblast Growth Factor-23. Front Endocrinol (Lausanne). 2018;9:267.
87. Wolf M. Update on fibroblast growth factor 23 in chronic kidney disease. Kidney Int. Oct 2012;82(7):737-47.
88. Drew DA, et al. Association between Soluble Klotho and Change in Kidney Function: The Health Aging and Body Composition Study. J Am Soc Nephrol. Jun 2017;28(6):1859-1866.
89. Zou D, et al. The role of klotho in chronic kidney disease. BMC Nephrol. Oct 22 2018;19(1):285.
90. Hasegawa H, et al. Direct evidence for a causative role of FGF23 in the abnormal renal phosphate handling and vitamin D metabolism in rats with early-stage chronic kidney disease. Kidney Int. Nov 2010;78(10):975-80.
91. Isakova T, et al. A blueprint for randomized trials targeting phosphorus metabolism in chronic kidney disease. Kidney Int. Oct 2009;76(7):705-16.
60
92. Isakova T, et al. Fibroblast growth factor 23 is elevated before parathyroid hormone and phosphate in chronic kidney disease. Kidney Int. Jun 2011;79(12):1370-8.
93. Marthi A, et al. Fibroblast Growth Factor-23 and Risks of Cardiovascular and Noncardiovascular Diseases: A Meta-Analysis. J Am Soc Nephrol. Jul 2018;29(7):2015-2027.
94. Hu MC, et al. Renal Production, Uptake, and Handling of Circulating alphaKlotho. J Am Soc Nephrol. Jan 2016;27(1):79-90.
95. Marcais C, et al. Circulating Klotho Associates With Cardiovascular Morbidity and Mortality During Hemodialysis. J Clin Endocrinol Metab. Sep 1 2017;102(9):3154-3161.
96. Hu MC, et al. Klotho and phosphate are modulators of pathologic uremic cardiac remodeling. J Am Soc Nephrol. Jun 2015;26(6):1290-302.
97. Hu MC, et al. Fibroblast growth factor 23 and Klotho: physiology and pathophysiology of an endocrine network of mineral metabolism. Annu Rev Physiol. 2013;75:503-33.
98. Liu CL, et al. Cysteine protease cathepsins in cardiovascular disease: from basic research to clinical trials. Nat Rev Cardiol. Jun 2018;15(6):351-370.
99. Carlsson AC, et al. Endostatin, Cathepsin S, and Cathepsin L, and Their Association with Inflammatory Markers and Mortality in Patients Undergoing Hemodialysis. Blood Purif. 2015;39(4):259-65.
100. Moallem SA, et al. Correlation between cathepsin D serum concentration and carotid intima-media thickness in hemodialysis patients. Int Urol Nephrol. Sep 2011;43(3):841-8.
101. Ozkayar N, et al. Relation between serum cathepsin D levels and endothelial dysfunction in patients with chronic kidney disease. Nefrologia. 2015;35(1):72-9.
102. Skau E, et al. GDF-15 and TRAIL-R2 are powerful predictors of long-term mortality in patients with acute myocardial infarction. Eur J Prev Cardiol. Oct 2017;24(15):1576-1583.
103. Hao Y, et al. Cardiac hypertrophy is exacerbated in aged mice lacking the osteoprotegerin gene. Cardiovasc Res. May 1 2016;110(1):62-72.
104. Mori K, et al. Tumor necrosis factor-related apoptosis-inducing ligand as an independent predictor of mortality in hemodialysis patents. Cytokine. Mar 2013;61(3):912-6.
105. Bernardi S, et al. Roles and Clinical Applications of OPG and TRAIL as Biomarkers in Cardiovascular Disease. Biomed Res Int. 2016;2016:1752854.
106. Schoppet M, et al. Localization of osteoprotegerin, tumor necrosis factor-related apoptosis-inducing ligand, and receptor activator of nuclear factor-kappaB ligand in Monckeberg's sclerosis and atherosclerosis. J Clin Endocrinol Metab. Aug 2004;89(8):4104-12.
107. Liabeuf S, et al. The circulating soluble TRAIL is a negative marker for inflammation inversely associated with the mortality risk in chronic kidney disease patients. Nephrol Dial Transplant. Aug 2010;25(8):2596-602.
108. Mesquita M, et al. Plasma osteoprotegerin is an independent risk factor for mortality and an early biomarker of coronary vascular calcification in chronic kidney disease. Clin Chem Lab Med. 2009;47(3):339-46.
109. Sigrist MK, et al. Elevated osteoprotegerin is associated with all-cause mortality in CKD stage 4 and 5 patients in addition to vascular calcification. Nephrol Dial Transplant. Oct 2009;24(10):3157-62.
61
110. Svensson M, et al. Osteoprotegerin as a predictor of renal and cardiovascular outcomes in renal transplant recipients: follow-up data from the ALERT study. Nephrol Dial Transplant. Jun 2012;27(6):2571-5.
111. Forne C, et al. Machine learning analysis of serum biomarkers for cardiovascular risk assessment in chronic kidney disease. Clin Kidney J. Aug 2020;13(4):631-639.
112. Morena M, et al. Plasma osteoprotegerin is associated with mortality in hemodialysis patients. J Am Soc Nephrol. Jan 2006;17(1):262-70.
113. Speer G, et al. Serum osteoprotegerin level, carotid-femoral pulse wave velocity and cardiovascular survival in haemodialysis patients. Nephrol Dial Transplant. Oct 2008;23(10):3256-62.
114. Winther S, et al. Osteoprotegerin and mortality in hemodialysis patients with cardiovascular disease. Clin Nephrol. Sep 2013;80(3):161-7.
115. Ford ML, et al. FGF-23 and osteoprotegerin are independently associated with myocardial damage in chronic kidney disease stages 3 and 4. Another link between chronic kidney disease-mineral bone disorder and the heart. Nephrol Dial Transplant. Feb 2012;27(2):727-33.
116. Huang QX, et al. Elevated Osteoprotegerin Concentration Predicts Increased Risk of Cardiovascular Mortality in Patients with Chronic Kidney Disease: A Systematic Review and Meta-Analysis. Kidney Blood Press Res. 2020;45(4):565-575.
117. Rochette L, et al. The Role of Osteoprotegerin in Vascular Calcification and Bone Metabolism: The Basis for Developing New Therapeutics. Calcif Tissue Int. Sep 2019;105(3):239-251.
118. Demir P, et al. Serum Osteoprotegerin Levels Related With Cardiovascular Risk Factors in Chronic Kidney Disease. J Clin Lab Anal. Nov 2016;30(6):811-817.
119. Lin WC, et al. Serum osteoprotegerin level is positively associated with peripheral artery disease in patients with peritoneal dialysis. Ren Fail. Nov 2020;42(1):131-136.
120. Feldreich T, et al. Circulating proteins as predictors of cardiovascular mortality in end-stage renal disease. J Nephrol. Nov 29 2018;
121. Feldreich T, et al. The association between plasma proteomics and incident cardiovascular disease identifies MMP-12 as a promising cardiovascular risk marker in patients with chronic kidney disease. Atherosclerosis. Aug 2020;307:11-15.
122. Honda H, et al. Serum albumin, C-reactive protein, interleukin 6, and fetuin a as predictors of malnutrition, cardiovascular disease, and mortality in patients with ESRD. Am J Kidney Dis. Jan 2006;47(1):139-48.
123. Rao M, et al. Plasma interleukin-6 predicts cardiovascular mortality in hemodialysis patients. Am J Kidney Dis. Feb 2005;45(2):324-33.
124. Panichi V, et al. Chronic inflammation and mortality in haemodialysis: effect of different renal replacement therapies. Results from the RISCAVID study. Nephrol Dial Transplant. Jul 2008;23(7):2337-43.
125. Panichi V, et al. Interleukin-8 is a powerful prognostic predictor of all-cause and cardiovascular mortality in dialytic patients. Nephron Clin Pract. 2006;102(2):c51-8.
126. Shiotsu Y, et al. Plasma S100A12 level is associated with cardiovascular disease in hemodialysis patients. Clin J Am Soc Nephrol. Apr 2011;6(4):718-23.
62
127. Libetta C, et al. Hepatocyte growth factor (HGF) and hemodialysis: physiopathology and clinical implications. Clin Exp Nephrol. Jun 2016;20(3):371-8.
128. Lonn J, et al. Hepatocyte growth factor in patients with coronary artery disease and its relation to periodontal condition. Results Immunol. 2011;2:7-12.
129. Rajpathak SN, et al. Hepatocyte growth factor and the risk of ischemic stroke developing among postmenopausal women: results from the Women's Health Initiative. Stroke. May 2010;41(5):857-62.
130. Malatino LS, et al. Hepatocyte growth factor and left ventricular geometry in end-stage renal disease. Hypertension. Jan 2003;41(1):88-92.
131. Anan F, et al. Predictors for silent cerebral infarction in patients with chronic renal failure undergoing hemodialysis. Metabolism. May 2007;56(5):593-8.
132. Rampino T, et al. Hemodialysis stimulates hepatocyte growth factor release. Kidney Int. May 1998;53(5):1382-8.
133. Lebherz-Eichinger D, et al. Increased chemokine excretion in patients suffering from chronic kidney disease. Transl Res. Dec 2014;164(6):433-43 e1-2.
134. Safa A, et al. Higher circulating levels of chemokines CXCL10, CCL20 and CCL22 in patients with ischemic heart disease. Cytokine. Jul 2016;83:147-157.
135. Calvayrac O, et al. CCL20 is increased in hypercholesterolemic subjects and is upregulated by LDL in vascular smooth muscle cells: role of NF-kappaB. Arterioscler Thromb Vasc Biol. Nov 2011;31(11):2733-41.
136. Elnabawi YA, et al. CCL20 in psoriasis: A potential biomarker of disease severity, inflammation, and impaired vascular health. J Am Acad Dermatol. Apr 2021;84(4):913-920.
137. Tesmer LA, et al. Th17 cells in human disease. Immunol Rev. Jun 2008;223:87-113.
138. Taleb S, et al. IL-17 and Th17 cells in atherosclerosis: subtle and contextual roles. Arterioscler Thromb Vasc Biol. Feb 2015;35(2):258-64.
139. Wan W, et al. Genetic deletion of chemokine receptor Ccr6 decreases atherogenesis in ApoE-deficient mice. Circ Res. Aug 5 2011;109(4):374-81.
140. de Bono D. Significance of raised plasma concentrations of tissue-type plasminogen activator and plasminogen activator inhibitor in patients at risk from ischaemic heart disease. Br Heart J. Jun 1994;71(6):504-7.
141. Lowe GD, et al. Tissue plasminogen activator antigen and coronary heart disease. Prospective study and meta-analysis. Eur Heart J. Feb 2004;25(3):252-9.
142. Segarra A, et al. Circulating levels of plasminogen activator inhibitor type-1, tissue plasminogen activator, and thrombomodulin in hemodialysis patients: biochemical correlations and role as independent predictors of coronary artery stenosis. J Am Soc Nephrol. Jun 2001;12(6):1255-63.
143. Diaz-Cabiale Z, et al. Neurochemical modulation of central cardiovascular control: the integrative role of galanin. Exp Suppl. 2010;102:113-31.
144. Palkeeva M, et al. Galanin/GalR1-3 system: A promising therapeutic target for myocardial ischemia/reperfusion injury. Biomed Pharmacother. Jan 2019;109:1556-1562.
145. Martinelli I, et al. Galanin promotes autophagy and alleviates apoptosis in the hypertrophied heart through FoxO1 pathway. Redox Biol. Apr 2021;40:101866.
63
146. Bjorkbacka H, et al. Plasma stem cell factor levels are associated with risk of cardiovascular disease and death. J Intern Med. Dec 2017;282(6):508-521.
147. Wallentin L, et al. Plasma proteins associated with cardiovascular death in patients with chronic coronary heart disease: A retrospective study. PLoS Med. Jan 2021;18(1):e1003513.
148. Lennartsson J, et al. Stem cell factor receptor/c-Kit: from basic science to clinical implications. Physiol Rev. Oct 2012;92(4):1619-49.
149. Lutz M, et al. Local injection of stem cell factor (SCF) improves myocardial homing of systemically delivered c-kit + bone marrow-derived stem cells. Cardiovasc Res. Jan 2008;77(1):143-50.
150. Frankenfield DL, et al. Differences in intermediate outcomes for Asian and non-Asian adult hemodialysis patients in the United States. Kidney Int. Aug 2003;64(2):623-31.
151. Yoshino M, et al. International differences in dialysis mortality reflect background general population atherosclerotic cardiovascular mortality. J Am Soc Nephrol. Dec 2006;17(12):3510-9.
152. Goodkin DA, et al. Association of comorbid conditions and mortality in hemodialysis patients in Europe, Japan, and the United States: the Dialysis Outcomes and Practice Patterns Study (DOPPS). J Am Soc Nephrol. Dec 2003;14(12):3270-7.
153. Yan G, et al. Racial and Ethnic Variations in Mortality Rates for Patients Undergoing Maintenance Dialysis Treated in US Territories Compared with the US 50 States. CJASN. Jan 7 2020;15(1):101-108.
154. Held PJ, et al. Five-year survival for end-stage renal disease patients in the United States, Europe, and Japan, 1982 to 1987. Am J Kidney Dis. May 1990;15(5):451-7.
155. Wong JS, et al. Survival advantage in Asian American end-stage renal disease patients. Kidney Int. Jun 1999;55(6):2515-23.
156. Noori N, et al. Racial and ethnic differences in mortality of hemodialysis patients: role of dietary and nutritional status and inflammation. Am J Nephrol. 2011;33(2):157-67.
157. Eisenga MF, et al. Iron deficiency, elevated erythropoietin, fibroblast growth factor 23, and mortality in the general population of the Netherlands: A cohort study. PLoS Med. Jun 2019;16(6):e1002818.
158. Vervloet M. Renal and extrarenal effects of fibroblast growth factor 23. Nat Rev Nephrol. Feb 2019;15(2):109-120.
159. Francis C, et al. Inflammation regulates fibroblast growth factor 23 production. Curr Opin Nephrol Hypertens. Jul 2016;25(4):325-32.
Acta Universitatis Upsaliensis Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine 1785
Editor: The Dean of the Faculty of Medicine
A doctoral dissertation from the Faculty of Medicine, Uppsala University, is usually a summary of a number of papers. A few copies of the complete dissertation are kept at major Swedish research libraries, while the summary alone is distributed internationally through the series Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine. (Prior to January, 2005, the series was published under the title “Comprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine”.)
Distribution: publications.uu.se urn:nbn:se:uu:diva-456907
ACTA UNIVERSITATIS
UPSALIENSIS UPPSALA
2021