University of Groningen
Heart failure biomarkers: The importance of cardiac specificity Piek, Arnold
DOI:
10.33612/diss.146698618
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2020
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Piek, A. (2020). Heart failure biomarkers: The importance of cardiac specificity. University of Groningen. https://doi.org/10.33612/diss.146698618
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Chapter 6
A combined bioinformatics, experimental and clinical
approach to identify novel cardiac specific heart
failure biomarkers: Is Dickkopf-3 (DKK3) a possible
candidate?
A. Piek1, N. Suthahar1, A.A. Voors1, R.A. de Boer1, H.H.W. Silljé1
1. Department of Cardiology, University Medical Center Groningen, University of Groningen, The Netherlands
Adapted from: Eur J Heart Fail. 2020 Aug 18. doi 10.1002/ejhf.1988.
Chapter 6
122
ABSTRACT
Aims: Cardiac specificity provides an advantage in correlating heart failure (HF) biomarker
plasma levels with indices of cardiac function and remodeling, as shown for natriuretic peptides. Using bioinformatics, we explored the cardiac specificity of secreted proteins and investigated in more detail the relation of dickkopf-3 (DKK3) gene expression and DKK3 plasma concentrations with cardiac function and remodeling in (pre-)clinical studies.
Methods and results: The cardiac specificity of secreted proteins was determined using
RNAseq data for a large panel of organs and tissues. This showed that natriuretic peptides (NPPA and NPPB) are highly cardiac specific (> 99%) whilst other HF biomarkers, including galectin-3 (Gal-3, LGALS3) and growth differentiation factor-15 (GDF-15), lack cardiac specificity ( 4%). DKK3 was cardiac enriched (44%), warranting further investigation. In three different HF mouse models, cardiac Dkk3 expression was altered, but DKK3 plasma concentrations were not. In humans, DKK3 plasma concentrations were higher in HF patients (n=2090) in comparison with age- and sex-matched controls without HF (n=240) (46.4 vs. 36.3 ng/ml, P<0.001). Multivariate regression analysis revealed that DKK3 was strongly associated with HF risk factors and comorbidities, including age, kidney function and atrial fibrillation. After correction for existing prediction models, DKK3 did not independently predict HF outcome (all-cause mortality/HF hospitalization, hazard ratio 1.13 [0.79-1.61] per DKK3 doubling, P=0.503).
Conclusion: Of actively secreted HF biomarkers, only natriuretic peptides showed high
cardiac specificity. Despite a cardiac specificity of 44%, secreted DKK3 had limited additional diagnostic and prognostic value.
Identification of cardiac specific HF biomarkers
123
INTRODUCTION
Nowadays, natriuretic peptides as plasma biomarkers are indispensable in the diagnosis and treatment of heart failure (HF), and are included in European and American HF treatment guidelines1-3. The success of natriuretic peptides, which are actively released from the heart upon myocardial stretch, is predominantly explained by their cardiac specificity3-7. This also applies to cardiac specific troponins, which are released from the heart by passive leakage upon myocardial damage1,2,8. However, these biomarkers provide information only on myocardial stretch and myocardial damage. New biomarkers could potentially provide additional insight into ongoing cardiac remodeling processes and may improve diagnosis and patient stratification.
Recently, we argued that, in the context of relating circulating concentrations to cardiac remodeling and function, cardiac specificity is one of the most important characteristics of a cardiac biomarker4,9. Using several HF animal models, we showed that a lack of cardiac specificity is likely to be an important reason why many emerging HF biomarkers, including galectin-3 (Gal-3) and growth-differentiation factor-15 (GDF-15), have relatively limited clinical value4,9. Thus, for the identification of possible novel HF biomarkers that provide information on cardiac remodeling and/or function, cardiac specificity seems indispensable. This idea is further illustrated by established plasma biomarkers used in other fields, which often show high organ specificity, such as prostate specific antigen (PSA) for prostate cancer, alanine aminotransferase (ALAT) for liver disease, and pancreatic specific amylase and lipase for pancreas disease10-12.
Thus, in the search for novel HF biomarkers selecting for cardiac specificity seems, amongst other factors, a relevant step. In line with our recently proposed approach on HF biomarker studies4, we embarked on a search for novel cardiac specific biomarkers by performing an extensive bioinformatics analysis based on RNAseq data for a large panel of organs and tissues13. Two possible candidates were identified and molecular assays were available to perform further analyses on one of them, namely Dickkopf-3 (DKK3), for which previous studies suggested an attenuating effect on cardiac remodeling14-18. Expression levels of DKK3 were determined in three different HF mouse models and in plasma samples from HF patients.
METHODS
Ethics statement
The animal experimental protocols were approved by the Animal Ethical Committee of the University of Groningen (AVD105002016487, DEC6827A, DEC6920A and IVD16487-03-001) and conducted according to the existing guidelines for the care and use of laboratory
Chapter 6
124
animals. This study was performed according to the declaration of Helsinki. Informed consent was obtained from all human participants.
Bioinformatics
To determine the cardiac specificity of genes, several databases were combined. For gene expression of organs and tissues in healthy humans, RNAseq data as published by Fagerberg et. al.13 were used. Included organs and tissues are listed in the Supplemental Methods. Cardiac specificity of genes was calculated as follows: cardiac specificity (%) = (cardiac expression counts / expression counts in all organs) * 100. Expression counts are the RNAseq counts of each gene. To determine whether genes encode for secreted proteins, databases from proteinatlas.org were consulted. Analyses were performed for males and females separately. For a selection of genes, raw data and calculations are shown in Supplemental Table 1.
Animal studies
Three mouse models were included: A myocardial infarction (MI) model, a transverse aortic constriction (TAC) model and an obesity/hypertension model. The animal experimental protocols have been described in detail previously9,19, and are summarized in the Supplemental Methods.
RT-qPCR
Gene expression was determined by real-time quantitative polymerase chain reaction (RT-qPCR) in total RNA extracted from mouse tissues. The full procedure is described in the Supplemental methods. Oligonucleotide sequences are listed in Supplemental Table 2.
ELISA
Mouse DKK3 plasma concentrations were measured using mouse DKK3 enzyme-linked immunosorbent assay (ELISA) kits (OKCD02498-96, Aviva Systems Biology, USA). Mouse NT-proANP plasma concentrations as measured by NT-NT-proANP ELISA kits (BI-20892, BIOMEDICA, Austria) have been reported previously9 and are presented here for convenience. Human DKK3 plasma concentrations were measured using human DKK3 ELISA kits (DY1118, R&D systems, USA). For the DKK3 ELISAs, quality control and materials are described in the Supplemental Methods.
Human HF cohort
DKK3 plasma concentrations were measured in 2090 patients with HF. This cohort was described in detail elsewhere20. All patients were aged > 18 years and had evidence of HF (LVEF < 40%, or plasma concentrations of BNP > 400 pg/ml or NT-proBNP > 2000 pg/ml). Plasma samples were collected at baseline and stored at -80 °C until analysis. Patients were followed for a median of 21 months and endpoints were registered, including the primary
Identification of cardiac specific HF biomarkers
125
endpoint (HF hospitalization and all-cause mortality), all-cause mortality and cardiovascular mortality.
Age and sex matched control group
To compare plasma DKK3 concentrations between HF patients and controls, an age- and sex-matched control group of subjects without HF was composed from a large population cohort. This cohort has been described in detail elsewhere21 and is reported in brief, along with the protocol used to select it, in the Supplemental Methods. Plasma samples were collected at baseline and stored at -80 °C until analysis.
Statistical analysis
The homogeneity of variance and normality of data were analyzed using Levene’s test for homogeneity of variance, and Kolmogorov-Smirnov tests combined with QQ-plots, respectively. Animal data is presented as the meana ± standard error of the mean (SEM). For human data, normally distributed variables are presented as the mean ± standard deviation (SD), non-normally distributed variables as the median (interquartile range, IQR) and categorical variables as n (%). Differences between two groups were analyzed using independent-samples t-tests for normally distributed variables, Mann-Whitney-U tests for non-normally distributed variables and Fisher’s exact tests for categorical variables. To analyze differences among multiple groups, one-way analysis of variance (ANOVA) with Bonferroni post-hoc correction was used for normally distributed variables that showed homogeneity of variances and Kruskall-Wallis tests followed by separate Mann-Whitney-U tests were used otherwise. To analyze correlations between cardiac Dkk3 gene expression levels and DKK3 plasma concentrations with indices of cardiac remodeling in mice, Pearson’s correlation tests were used. For further analysis of the HF cohort, non-normally distributed variables were log2-transformed. Trends over DKK3 quintiles were tested using linear regression analysis for normally distributed variables, Jonckheere-Terpstra tests for non-normally distributed variables and Cochran-Armitage trend tests for categorical variables. Both crude and age-adjusted linear regression analyses were performed. Variables with a P-value < 0.1 in age-adjusted analysis were included in multivariable linear regression analysis, using forward selection and pairwise exclusion. Differences in survival between DKK3 quintiles were analyzed using Kaplan-Meier curves combined with log-rank tests. To analyze the association of DKK3 with clinical outcomes, Cox-regression analysis was performed. Using forward selection with entry based on significance and removal based on likelihood ratio the critically nullifying factors for DKK3 prediction capacity were determined. Harrell’s C-statistic, the net reclassification index (NRI) and the integrated discrimination index (IDI) were determined to investigate the predictive performance of DKK3.
A P-value < 0.05 was considered to indicate statistical significance. Statistical analysis was performed using SPSS software (IBM SPSS statistics, version 23, IBM, USA) and STATA (Version 14, Stata Corporation, USA).
Chapter 6
126
RESULTS
Several suggested HF biomarkers lack cardiac specificity
To determine the cardiac specificity of established heart failure (HF) biomarkers, we first determined the cardiac specificity of the total human genome as described in Figure 1A, using expression data of healthy tissues as published by Fagerberg et al.13. Of the 18.938 genes included in this analysis, only 24 genes (0.13%) showed cardiac-specific expression (> 90-100% cardiac derived (Figure 1B)). Based on the proteinatlas.org database, we subsequently selected for genes encoding for known and predicted secreted proteins, which might constitute potential HF biomarkers. Of these, only three genes, including natriuretic peptide precursor type-A (NPPA) and type-B (NPPB), were cardiac-specific, and five (0.24%) and 108 genes (5.1%) showed a cardiac specificity of respectively 50-90% and 10-50% (Figure 1B). The cardiac specificity of four HF biomarkers is visualized in Figure 1C, which shows that NPPB is highly cardiac-specific (99.9%), whereas other markers, including galectin-3 (Gal-3, LGALS3), growth differentiation factor-15 (GDF-15) and tissue inhibitor of metalloproteinase-1 (TIMP-metalloproteinase-1), have low cardiac specificity (3.2%, 0.06% and 3.6%, respectively). Similar results were acquired for females. (Supplemental Figure 1).
Next, cardiac specificity of secreted protein encoding genes was plotted against their total cardiac expression (Figure 2). Again, this confirmed that natriuretic peptides, including NPPA and NPPB, had high cardiac specificity (> 99%) and high absolute cardiac expression (respectively 6250.9 and 3030.7 cardiac expression counts). In contrast, several other HF biomarkers, including Gal-3 (LGALS3), interleukin-6 (IL-6), GDF-15, TIMP-1, insulin-like growth factor-binding protein-7 (IGFBP-7), myeloperoxidase (MPO) and human epididymis protein-4 (HE4) showed low absolute and low relative cardiac gene expression (Supplemental Figure 2). Similar results were acquired for females (Supplemental Figure 3-4).
Based on the bioinformatics analysis DKK3 showed biomarker potential
The bioinformatics approach was used to identify possible novel cardiac-specific HF biomarker candidates. Based on NPPA and NPPB, both cardiac specificity and high absolute cardiac expression were considered important characteristics. In line with these criteria, Dickkopf-3 (DKK3) and bone morphogenic protein 10 (BMP-10) were selected as biomarker candidates (Figure 2). DKK3 was cardiac-enriched with 44.1% of the expression counts coming from the heart. Moreover, total cardiac expression was relatively, high with an absolute 942.3 cardiac expression counts. In addition to the heart, DKK3 is also expressed in several other tissues, including the brain, gastro-intestinal tract, lung and kidney, amongst others (Supplemental Figure 5). BMP-10 showed greater cardiac specificity than DKK3 (97.9% cardiac-specific) but had lower absolute cardiac expression (300.5 cardiac expression counts). Due to analytical difficulties, we were not able to include BMP-10 in this study. Together, these results warranted further investigation of DKK3 as an HF biomarker.
Identification of cardiac specific HF biomarkers
127
Figure 1. Cardiac specificity of the human genome and HF biomarkers. (A) Schematic depiction of the
bioinformatics approach. Gene expression data as measured by RNAseq in all organs and tissues of healthy humans were used13. Included are adipose tissue, adrenal gland, appendix, bone marrow, brain, colon, duodenum, oesophagus, gall bladder, heart, kidney, liver, lung, lymph node, pancreas, salivary gland, skin, small intestine, spleen, stomach and thyroid gland. Sex specific organs/tissues were prostate and testis for male, and endometrium and ovary for female. The cardiac specificity was calculated using the formula as indicated. Finally, genes assumed to encode for secreted proteins were selected. (B) The cardiac specificity of the human genome (top) and of secreted protein encoding genes (bottom) showing the numbers of included genes. Coloured areas represent the amount of genes with the corresponding degree of specificity. Data include those for male tissues. (C) Cardiac specificity of several heart failure (HF) biomarkers. Coloured areas represent organ specificity. Data include those for male tissues. Gal-3 (LGALS3)=Galectin-3. GDF-15=Growth differentiation factor 15. GI-tract=Gastro-intestinal tract, including appendix, colon, duodenum, oesophagus, small intestine and stomach. NPPB=Natriuretic peptide precursor type B. TIMP-1=Tissue inhibitor of metalloproteinase 1. Sex-organs=Prostate and testis. The bioinformatics analysis was based on publicly available RNAseq data as previously published by Fagerberg et al.13.
A
B C
RNAseq: Expression in organs/tissues of
healthy humans
Calculation cardiac specificity: 𝑪𝒂𝒓𝒅𝒊𝒂𝒄 𝒔𝒑𝒆𝒄𝒊𝒇𝒊𝒄𝒊 % = 𝑬 𝒑𝒓𝒆𝒔𝒔𝒊𝒐𝒏 𝒄𝒐 𝒏 𝒔 𝒉𝒆𝒂𝒓 𝑬 𝒑𝒓𝒆𝒔𝒔𝒊𝒐𝒏 𝒄𝒐 𝒏 𝒔 𝒐 𝒂𝒍 𝒃𝒐𝒅 × Selection for secreted proteins Cardiac specificity of the human transcriptome
NPPB
Gal-3
GDF-15
TIMP-1
0-10% 10-50% 50-90% 90-100% All genes (n=18938) Cardiac specificity Heart Lung Kidney Liver GI-tract Adipose tissue Brain Adrenal gland Gall bladder Urinary bladder Sex organs OtherChapter 6
128
Figure 2. The cardiac specificity of secreted protein encoding genes. The total cardiac expression of secreted
protein encoding genes was plotted against their calculated cardiac specificity. All genes with cardiac specificity > 10% are shown. Gene name abbreviations are used for readability. Organs and tissues include adipose tissue, adrenal gland, appendix, bone marrow, brain, colon, duodenum, gall bladder, heart, kidney, liver, lung, lymph node, oesophagus, pancreas, salivary gland, skin, small intestine, spleen, stomach and thyroid gland, and sex‐ specific organs or tissues for males (prostate and testis). Data including those for male tissues. The bioinformatics analysis was based on publicly available RNAseq data as previously published by Fagerberg et al.13. 10 20 30 40 50 60 70 80 90 100 0 100 200 300 400 500 600 700 800 900 1000 1000 2000 3000 4000 5000 6000 7000 NPPA NPPB PTGDS DKK3 FHL1 GSN MFGE8 NFE2L1 BMP10 MXRA7 LAMB2 LPL GOT2 BGN SDHB CALU THBS4 DPT PEBP4 CDH13 SFRP1 PCOLCE2 AGT MASP1 LTBP1 EFEMP2 ASPN
OGNVWFFBLN2 ANK1 PLA2G5
C1QTNF1LOXL1 SRPXSPOCK1HAMPPRADC1COL8A1B3GALNT1
COL21A1 HTRA3
FSTL3LTBP2CILP
NTN1 VWC2
OMDHMSD COLQ VSTM2LFAM180AADAMTSL5WNT9A
APLNLCN6ENAM EMILIN3NGFADAMTS7ANGPTL7 FGF18 DAND5 CRLF1C1QTNF2VWDEGDF6COL6A6MMP21ANGPTL5NRTNTLL2GHRHTAC4C1QTNF9 LCN1 PRSS42UCMA CSHL1GH2FSHB Cardiac specificity (%) To ta l c ar d ia c ex p re ss io n 10-50% 50-90% 90-100% Cardiac specificity
Identification of cardiac specific HF biomarkers
129
Figure 3. Cardiac gene expression and plasma concentrations of ANP and DKK3 in three different heart failure mouse models. (A-B) Included heart failure (HF) models are transverse aortic constriction (TAC, top
row), myocardial infarction (MI, middle row) and an obese/hypertension mouse model with HF with preserved ejection fraction characteristics (HFD+AngII, bottom row). (A) Natriuretic peptide precursor type A (Nppa) left ventricular (LV) gene expression (graphs on the left) and N-terminal pro-Atrial natriuretic peptide (NT-proANP) plasma concentrations (graphs on the right). Data on Nppa and NT-proANP have been published previously9, but are here presented for convenience. (B) Dickkopf-3 (Dkk3) gene expression (graphs on the left) and DKK3 plasma levels (graphs on the right). Gene expression values are corrected for 36b4 gene expression and presented as fold change. Bars represent means. Error bars represent standard error of the mean. *P 0.05 vs. sham or LFD. #P 0.05 vs. HFD. AngII=Angiotensin-II. HFD=High fat diet. LFD=Low fat diet.
DKK3 in HF mouse models
To investigate the relation between cardiac DKK3 expression and DKK3 plasma concentrations, left ventricular (LV) Dkk3 gene expression levels and DKK3 plasma concentrations were determined in three different HF mouse models, including a myocardial infarction (MI) model, a transverse aortic constriction (TAC) model and an obesity/hypertension model. Nppa gene expression and NT-proANP plasma concentrations were determined as controls. These models have been described in detail previously9,19 and data on cardiac remodeling, including hypertrophy, fibrosis, cardiac function and cardiac dimensions are listed in Supplemental Table 4. As shown previously9, LV Nppa gene expression was increased in all models and was associated with increased NT-proANP plasma concentrations (Figure 3A). In addition, LV Dkk3 gene expression levels were increased in MI and TAC mice, but were decreased in obese/hypertensive animals (Figure 3B). However, the changes in LV Dkk3 gene expression did not translate into altered DKK3 plasma concentrations in any of these models (Figure 3B). In liver, kidney and lung tissue Dkk3 expression did not change in association to the disease perturbations (Supplemental Figure 6). Similarly findings in humans, mice Dkk3 gene expression was relatively high in the
A ANP B
LV gene expression Blood plasma level
LFD HFD HFD+AngII 0 1 2 3 4 N P P A / 36 B 4 * # * LFD HFD HFD+AngII 0 1 2 3 N T-p ro A N P ( n M ) # * Sham TAC 0 5 10 15 20 N P P A / 36 B 4 * Sham TAC 0 2 4 6 8 N T-p ro A N P ( n M ) * Sham MI 0 5 10 15 20 N P P A / 36 B 4 * Sham MI 0 2 4 6 N T-p ro A N P ( n M ) * DKK3
LV gene expression Blood plasma level
LFD HFD HFD+AngII 0 100 200 300 D K K 3 (n g /m l) LFD HFD HFD+AngII 0.0 0.5 1.0 1.5 D K K 3 / 3 6B 4 * Sham TAC 0 50 100 150 200 D K K 3 (n g /m l) Sham TAC 0 2 4 6 D K K 3 / 3 6B 4 * Sham MI 0 1 2 3 D K K 3 / 3 6B 4 * Sham MI 0 50 100 150 D K K 3 (n g /m l) H F D + A ng II TA C MI LFD HFD HFD+AngII 0.0 0.5 1.0 1.5 D K K 3 / 3 6B 4 *
Tab le 1. B asel in e c h ar ac te ri sti cs o f h e ar t fai lu re p atien ts ac co rd in g t o b lo o d p lasm a DK K 3 q u in til e s. D KK3 (n g/ ml ) P f o r tr e n d Q 1: < 38.8 Q 2: 38.8 -45.8 Q 3: 45.8 -53.3 Q 4: 53.3 -65.8 Q 5: > 65.8 (n =418) (n =418) (n =418) (n =418) (n =418) C lin ic al c h ar ac te ri sti cs A ge ( ye ars ) 61 ±13 66 ±12 70 ±11 72 ±11 73 ±10 < 0.001* Mal e s ex , n ( % ) 300 (71.8) 323 (77.3) 306 (73.2) 304 (72.7) 301 (72.0) 0.552 Cau cas ian race , n ( % ) 415 (99.3) 411 (98.3) 414 (99.0) 415 (99.3) 412 (98.6) 0.767 BMI (kg /m 2 ) 28 (25 -33) 28 (24 -31) 27 (24 -31) 26 (24 -30) 26 (23 -29) < 0.001* SBP ( mmH g) 120 (110 -138) 125 (110 -140) 124 (110 -140) 120 (110 -140) 120 (108 -130) 0.002* eG FR ( ml /mi n /1.73m 2 ) 77.8 (59.3 -92.2) 63.1 (47.2 -79.3) 59.6 (45.2 -75.5) 53.5 (42.1 -69.6) 45.8 (31.9 -62.2) < 0.001* Cu rr en t A F, n ( % ) 80 (20.9) 123 (33.4) 146 (41.6) 163 (48.1) 180 (53.3) < 0.001* N YH A 3 -4, n ( % ) 232 (57.1) 231 (57.0) 258 (62.9) 257 (64.4) 270 (66.0) 0.001* Per ip h era l e d ema 168 (50.8) 188 (53.9) 207 (57.3) 218 (64.1) 241 (68.9) < 0.001* Cu rr en t smo ki n g, n ( % ) 91 (21.8) 78 (18.7) 55 (13.2) 44 (10.6) 40 (9.6) < 0.001* Ec h o car d io gr ap h y LV EF ( % ) 30 (25 -36 ) 30 (25 -35 ) 30 (25 -38 ) 30 (25 -35 ) 30 (25 -38 ) 0.069 LV EDD ( mm) 62 ±10 62 ±9 61 ±9 60 ±10 60 ±10 0.003* LV ES D (m m) 51 ±12 50 ±11 50 ±11 50 ±11 49 ±12 0.134 In te rve n tri cu lar WT ( mm) 10 (9 -12 ) 10 (9 -12 ) 10 (9 -12 ) 10 (9 -12 ) 10 (9 -12 ) 0.207 Po ste ri o r WT ( mm ) 10 (9 -11 ) 10 (9 -11 ) 10 (9 -12 ) 10 (9 -11 ) 10 (9 -12 ) 0.328 Le ft atri al d iamet e r (mm ) 46 ±7 47 ±7 48 ±7 47 ±9 48 ±9 0.009* Mi tr al val ve re gu rg itati o n , n ( % ) 149 (38.3 ) 190 (48.1 ) 191 (47.4 ) 194 (48.4 ) 201 (50.8 ) 0.002 * M e d ic al h ist o ry , n ( % ) H F h o sp ital iz ati o n p as t ye ar 113 (27.0) 132 (31.6) 122 (29.2) 136 (32.5) 150 (35.9) 0.009* Myo card ial In farc ti o n 140 (33.5) 155 (37.1) 153 (36.6) 165 (39.5) 164 (39.2) 0.063 Str o ke 31 (7.4) 33 (7.9) 44 (10.5) 36 (8.6) 58 (13.9) 0.003* Per ip h era l arte ri al d is eas e 45 (10.8) 45 (10.8) 48 (11.5) 43 (10.3) 48 (11.5) 0.843 Di ab ete s 133 (31.8) 138 (33.0) 134 (32.1) 141 (33.7) 121 (28.9) 0.486 CO PD 59 (14.1) 73 (17.5) 85 (20.3) 75 (17.9) 64 (15.3) 0.621 M e d ic ati o n , n ( % ) Bet ab lo cke r 358 (85.6) 352 (84.2) 344 (82.3) 335 (80.1) 348 (83.3) 0.127 A CE -i /A R B 325 (77.8) 313 (74.9) 286 (68.4) 287 (68.7) 288 (68.9) < 0.001* A ld o ste ro n a n tag o n is t 248 (59.3) 213 (51.0) 229 (54.8) 220 (52.6) 187 (44.7) < 0.001* Di u re ti cs 418 (100.0) 418 (100.0) 416 (99.5) 418 (100.0) 418 (100.0) 1.000 Bl o o d lab o rato ry v al u e s H emo gl o b in ( g/d L) 13.4 ±1.8 13.3 ±1.9 13.2 ±1.9 13.1 ±1.9 12.8 ±1.9 0.541 H emato cri t (% ) 40.5 ±4.9 40.3 ±5.4 40.0 ±5.3 39.8 ±5.4 39.1 ±5.4 0.880 BUN ( mmo l/L ) 8.3 (6.0 -13.7) 10.4 (7.5 -17.1) 11.1 (7.4 -17.9) 11.1 (8.0 -18.2) 13.9 (9.3 -22.5) < 0.001*
Dat a ar e p re se n ted as m ean s ± sta n d ar d d ev iat ion f o r n o rm al ly d is tri b u ted v ar iab le s, as m ed ian s (i nt erq u ar ti le ra n ge ) fo r n o n -n o rm al ly d is tri b u ted v ar iab le s an d as p erce n ta ge (n ) fo r cat e gori ca l var iab le s. AC E-i= An gi ot en si n con ve rti n g en zy m e in h ib ito r. AF= Atr ial f ib ri llat ion . ARB =A n gi ot en si n r ece p to r b lock er. BMI= Bod y m as s in d ex. BU N =B loo d u re a n itro ge n . COPD= Chro n ic o b stru cti ve p u lm o n ar y d is eas e. DKK 3 =Di ckk o p f-3. e G FR =E sti m at ed gl om eru lar fi ltra ti o n ra te. H DL=Hi gh d en si ty l ipo p ro tei n. H F= H ear t fai lu re . LVE DD= Le ft ve n tri cu lar en d -d ias to lic d iame ter. LVE SD =Le ft ve n tri cu lar e n d -s ys to lic d ia m eter . LV EF= Lef t ve n tri cu lar ej ecti o n f ra cti o n . N T -p ro BNP =N -term in al p ro h o rm o n e o f B -ty p e n at ri u re ti c p ep ti d e. N YH A = N ew Y o rk h ear t as so ci ati o n cl ass . S BP= Sy sto lic b loo d p re ss u re . WT =Wa ll t h ickn e ss . *P < 0. 05 for tren d a n al ys is . So d iu m ( mmo l/L ) 140 (137 -141) 139 (137 -141) 140 (137 -142) 140 (138 -141) 139 (136 -142) 0.964 Po tas si u m ( mmo l/L ) 4.2 (3.9 -4.5) 4.3 (3.9 -4.7) 4.3 (3.9 -4.6) 4.2 (3.9 -4.6) 4.2 (3.9 -4.6) 0.945 H DL ( mm o l/L ) 1.01 (0.80 -1.32) 1.06 (0.86 -1.27) 1.02 (0.83 -1.25) 1.09 (0.88 -1.40) 1.02 (0.86 -1.32) 0.689 NT -p ro BN P (p g/mL ) 1444 (636 -3177) 2345 (1062 -4884) 2689 (1158 -5337) 3533 (1819 -6299) 4503 (2048 -9322) < 0.001* Se ru m c re ati n in e ( µ g/d l) 1.00 (0.81 -1.19) 1.14 (0.95 -1.40) 1.13 (0.97 -1.44) 1.22 (1.00 -1.56) 1.40 (1.10 -1.83) < 0.001* Tro p o n in -T ( p g/mL ) 21.0 (13.1 -37.4) 28.4 (17.9 -49.5) 30.3 (20.2 -50.2) 34.8 (23.2 -56.2) 43.8 (27.6 -72.2) < 0.001*
Chapter 6
132
heart and low in liver and kidney, however, mouse lung tissue showed relatively high Dkk3 gene expression (Supplemental Figure 7). LV Dkk3 gene expression was associated with cardiac hypertrophy, cardiac dilatation and cardiac dysfunction (Supplemental Figure 8-10). However, LV Dkk3 gene expression levels showed no association with plasma DKK3 levels. Further, no direct associations between plasma DKK3 levels and indices of cardiac remodeling were observed (Supplemental Figure 8-10). Thus, in these models there was no simple relation between circulating DKK3 with either cardiac Dkk3 expression or with cardiac remodeling.
Plasma DKK3 concentrations were higher in HF patients
Plasma concentrations of DKK3 were determined in HF patients and in age- and sex-matched controls without HF. These results showed that median DKK3 plasma concentrations were higher in HF patients than in controls (46.4 ng/ml [IQR 38.6-57.4] vs. 36.3 ng/ml [IQR 31.5-41.3], P<0.001) (Supplemental Figure 11A and Supplemental Table 5). Moreover, in HF patients, DKK3 concentrations increased with increasing NT-proBNP quartiles, indicating a relationship with HF severity (Supplemental Figure 11B).
Associations between DKK3 and patient characteristics
Next, the association of DKK3 with patient characteristic was further investigated in the total HF cohort. Mean ± SD age of the cohort was 68.3 ± 12.1 years. Median estimated glomerular filtration rate (eGFR) and LVEF were 59.5 mL/min [IQR 44.2-77.6 mL/min] and 30% [IQR 25-36%], respectively. Median DKK3 plasma concentration was 49.4 ng/ml [IQR 40.5-61.8]. Baseline characteristics according to DKK3 quintiles are shown in Table 1. Patients in higher DKK3 quintiles were generally older, and had a lower body mass index (BMI) and a lower eGFR. Moreover, these patients showed more severe HF, as indicated by a higher New York Heart Association (NYHA) class, higher prevalence of peripheral edema, and higher circulating levels of NT-proBNP and blood urea nitrogen (BUN). As shown by echocardiography, higher plasma DKK3 levels were associated with smaller left ventricular end-diastolic diameters (LVEDD), larger left atrial diameters and higher prevalence of mitral valve regurgitation. Finally, this group’s medical record showed a more vulnerable patient,
Table 2. Multivariable model of variables associated with DKK3 plasma levels
Variable β ± SE sβ P-value eGFR† - 0.137 ± 0.027 - 0.201 < 0.001* NT-proBNP † 0.041 ± 0.010 0.158 < 0.001* Age 0.004 ± 0.001 0.155 < 0.001* Current AF 0.098 ± 0.024 0.149 < 0.001* BMI† - 0.142 ± 0.061 - 0.084 0.020*
Multivariable model including DKK3-associated factors. R2adj=0.212. Constructed with forward selection and pairwise exclusion. Variables with P < 0.10 in univariate regression analysis were included in constructing this model. DKK3 plasma levels were log2-transformed. β=Beta coefficient. Sβ=Standardized beta coefficient. SE=Standard error. Other abbreviations as in Table 1. †=log2-transformed. *P < 0.05.
Identification of cardiac specific HF biomarkers
133
with higher prevalence of HF hospitalizations in the past year, and higher prevalence of atrial fibrillation and chronic obstructive pulmonary disease (COPD).
Supplemental Table 6 shows the associations between plasma DKK3 and clinical variables. Next, a multivariable model was composed (Table 2). The strongest correlates with higher DKK3 concentrations were lower eGFR, higher age, higher NT-proBNP level, presence of atrial fibrillation and lower BMI. The adjusted R2 for the model was 0.212.
DKK3 and HF outcome
During a median follow-up of 21 months 869 patients (41.6%) reached the primary endpoint (HF hospitalization or all-cause mortality) and 555 patients (26.6%) died, 319 (57.5%) of whom from a cardiovascular cause. Kaplan-Meier curves for crude DKK3 quintiles showed that higher quintiles were associated with lower survival rates for all endpoints (Supplemental Figure 12, Log-rank for all endpoints P<0.001). These results were corroborated by crude Cox-regression analysis, showing that baseline DKK3 plasma concentrations were predictive for the primary endpoint (HR 2.42 [1.74-3.38] per DKK3 doubling, P 0.001), all-cause mortality (HR 2.78 [2.11-3.67] per DKK3 doubling, P 0.001) and cardiovascular mortality (HR 2.71 [1.86-3.93] per DKK3 doubling, P 0.001) (Table 3, model A). These remained significant after correction for age for all endpoints (Table 3, model B), including the primary endpoint (HR 1.83 [1.29-2.60] per DKK3 doubling, P 0.001), all-cause mortality (HR 2.01 [1.05-2.72] per DKK3 doubling, P 0.001) and cardiovascular mortality (HR 2.03 [1.36-3.04] per DKK3 doubling, P=0.001). However, after adjusting for established HF prediction models22 (for predicting primary endpoint and all-cause mortality) or the multivariable model (Table 2, for predicting cardiovascular mortality), DKK3 did not independently predict HF outcome (Table 3, model E-F). Critically nullifying factors for DKK3 prediction capacity were age + NT-proBNP for the primary endpoint and cardiovascular mortality, and age + NT-proBNP + BUN for all-cause mortality (Table 3, model C-D). Harrell’s Table 3. Cox proportional hazards analysis for prediction of HF outcome by doubling of DKK3
Model Primary endpoint All-cause mortality Cardiovascular mortality
HR (95% CI) P-value HR (95% CI) P-value HR (95% CI) P-value
A 2.42 (1.74-3.38) < 0.001* 2.78 (2.11-3.67) < 0.001* 2.76 (1.90-4.00) < 0.001*
B 1.83 (1.29-2.60) 0.001* 2.02 (1.50-2.72) < 0.001* 2.07 (1.39-3.10) < 0.001*
C 1.34 (0.93-1.92) 0.116 1.43 (1.05-1.94) 0.024* 1.44 (0.95-1.92) 0.088
D N.A. N.A. 1.24 (0.91-1.70) 0.172 N.A. N.A.
E 1.13 (0.79-1.61) 0.503 1.27 (0.94-1.73) 0.122 N.A. N.A.
F N.A. N.A. N.A. N.A. 1.15 (0.74-1.80) 0.531
Model A=Univariable DKK3. Model B=Model A corrected for age. Model C=Model B corrected for NT-proBNP. Model D=Model C corrected for BUN. Model E=Model A corrected for previously published HF risk prediction
models22: For primary endpoint this includes age, previous HF hospitalization in the past year, peripheral edema, SBP, NT-proBNP, hemoglobin, HDL, sodium and betablocker usage. For all-cause mortality this includes age, NT-proBNP, BUN, hemoglobin and betablocker usage. Model F=Model A corrected for multivariable model as presented in Table 2. CI=Confidence interval. HR=Hazard ratio. Other abbreviations as in Table 1. *P < 0.05.
Chapter 6
134
C-statistic, the net reclassification index (NRI) and the integrated discrimination index (IDI) confirmed that DKK3 did not improve prediction models of HF outcome (Supplemental Table 7).
DISCUSSION
By performing a bioinformatics analysis, we determined the cardiac specificity of known HF biomarkers, and aimed to identify novel HF biomarkers of cardiac remodeling by selecting for cardiac specificity. We confirmed that several suggested HF biomarkers, including Gal-3 and GDF-15, lack cardiac specificity, whereas natriuretic peptides are characterized by both high relative and absolute cardiac expression. A search for novel HF biomarkers with higher cardiac specificity and absolute cardiac expression showed that DKK3 is cardiac-enriched and hence could be a potential HF biomarker. However, in three different HF mouse models, no relationship between cardiac Dkk3 expression and circulating DKK3 concentrations was observed. Moreover, although DKK3 plasma concentrations were increased in HF patients compared with age- and sex-matched controls, plasma concentrations were predominantly associated with HF risk factors and comorbidities and showed no independent association with clinical outcomes. Overall, DKK3 has limited value as an HF biomarker in addition to natriuretic peptides, and it appears that proteins must have a high degree of cardiac specificity to have potential as diagnostic HF biomarkers.
In recent decades, HF biomarkers have been studied extensively, but only natriuretic peptides (released by active secretion) and troponins (released by passive leakage) are now used in the clinic, whereas many other investigated markers fail to provide additional clinical value1-5,7,8. It is known that comorbidities influence the plasma levels of several biomarkers, which hampers the correlating of plasma concentrations with processes of cardiac remodeling9. Lack of cardiac specificity was suggested to be one of the main reasons why novel biomarkers fail to add clinical value4,9. Our analysis confirms the lack of cardiac specificity of several suggested HF biomarkers, including Gal-3, GDF-15, HE4, IGFBP-7, IL-6, MPO and TIMP-1. Further, this study suggests that, although DKK3 can be marked as cardiac-enriched and shows relatively high absolute cardiac expression, DKK3 appears not to be suitable biomarker of cardiac remodeling in HF. Our study suggests that only BMP-10 shows a similar profile as natriuretic peptides (high cardiac absolute expression, high cardiac specificity, encoding a secreted protein). However, whether BMP-10 expression and secretion are altered upon pathological stress conditions is yet unknown and there is currently no proper method of detecting plasma BMP-10. Thus, to date natriuretic peptides are the only known secreted proteins with high absolute cardiac expression, high cardiac specificity and increased expression in HF conditions.
Identification of cardiac specific HF biomarkers
135
Based on previous research and our bioinformatics analysis, DKK3 showed HF biomarker potential. Our mouse studies revealed elevated cardiac Dkk3 gene expression in mouse models of heart failure with reduced ejection fraction (HFrEF), but these changes were not reflected in the plasma pool. This indicates that cardiac contribution to the total plasma pool is limited and/or that cardiac-derived DKK3 does not enter the circulation. In this regard, it should be noted that DKK3 is a relatively large glycoprotein of approximately 50 kDa, and is much larger than the small natriuretic peptides (< 4 kDa), which possibly impedes its passage over the tight cardiac endothelial barrier23. Protein size may therefore also be an important consideration in HF biomarker research. In addition, other protein characteristics, such as half-life, degradation, metabolization, internalization and protein interactions, may also influence plasma levels and hence biomarker function. The fact that plasma DKK3 concentrations were elevated in HF patients compared with control subjects may be a result of comorbidities present in HF patients. In these patients, increased DKK3 expression and/or secretion by other affected tissues may cause elevated plasma levels. This may also be partly dependent on the degree of HF, as severe HF will more strongly affect other organs and tissues. In this respect, HF may indirectly affect plasma DKK3 concentrations. The HF mouse models used are in general “clean” models, which lack the complexity of many comorbidities and/or other diseases present in the human population, which may explain the absence of elevated plasma DKK3 concentrations in HF mouse models. Only the obesity/hypertension mouse model included obesity as a comorbidity and interestingly showed lower cardiac Dkk3 gene expression and a tendency towards lower DKK3 plasma levels. This may reflect the inverse relation of DKK3 plasma concentrations and BMI, as observed in the investigated patient cohort, and deserves further investigation. These results indicate that cardiac-specific gene expression is an important determinant, but not the only factor that determines plasma concentrations of biomarkers.
Our results suggest that cardiac specificity appears to be an important factor in a diagnostic HF biomarker. In the animal models investigated, no direct relation between cardiac Dkk3 expression and plasma DKK3 concentrations was observed. Hence, the diagnostic value of DKK3, at least in the types of HF investigated, seems limited. Possibly, DKK3 might serve as a diagnostic atrial biomarker as DKK3 was associated with left atrial diameter and atrial fibrillation. However, as HF is a systemic syndrome, high cardiac specificity is probably less important for a prognostic biomarker. This is illustrated by the fact that existing HF prediction models contain a broad variety of predictors, also including some not specifically associated with the heart. The prediction models used in this study are already strong prediction models. DKK3 did not improve these models and we believe it will be difficult to identify additional and independent predictors to add to these established predictors of HF outcome.
Previously, the use of HF mouse models confirmed an association between DKK3 and cardiac remodeling14-16. Dkk3 overexpression attenuated cardiac hypertrophy and dilatation,
Chapter 6
136
whereas Dkk3 deletion aggravated these processes and resulted in increased mortality14-16. Thus, Dkk3 has cardioprotective qualities. Although our analysis was not designed to investigate the role of Dkk3 in cardiac remodeling, the observation that LV Dkk3 gene expression was increased in MI and TAC mice and was associated with indices of cardiac remodeling at least suggests a role for Dkk3 in HF pathophysiology.
DKK3 is an emerging biomarker and has been investigated as a biomarker for non-cardiac diseases in several studies. Firstly, a recent study in patients with prevalent renal dysfunction showed that urinary levels of DKK3 independently predict future loss of renal function24. Further, in patients about to undergo major cardiac surgery, preoperative urinary DKK3 levels are associated with postoperative acute kidney injury and future loss of renal function25. Animal experiments showed that renal tubular epithelial cells appear to be the main source of urinary DKK3 and DKK3 in urine has been proposed as a marker for renal tubular damage and renal tubules at risk24-27. In this study, which investigated DKK3 levels in plasma rather than in urine, a relationship was identified between kidney function and DKK3 plasma levels. Secondly, a previous study showed that DKK3 plasma levels are inversely related to atherosclerosis28. The current study was not designed to specifically investigate the association of DKK3 with atherosclerosis, but it did not identify any independent relationship with any variable associated with atherosclerosis. Finally, DKK3 has been found previously to be associated with age29. The current study also found a strong correlation between DKK3 and age, thereby, corroborating previous observations.
Some limitations should be mentioned. The bioinformatics approach is based on RNAseq data derived from many organs and tissues, but does not correct for organ size. Small organs may have been over-represented and large organs under-represented in this analysis. Nevertheless, the values observed for natriuretic peptides and several other markers (including Gal-3, GDF-15, TIMP-1 and DKK3) were in line with previous findings9,30. Based on the bioinformatics approach, BMP-10 also appeared to be an HF biomarker candidate. Unfortunately, we were not able to measure BMP-10 with acceptable precision and reproducibility, an issue that has been reported previously and which is thought to be caused by interference of BMP-9 with BMP-1031. We note that BMP-10 assays may become available in the near future as shown by a recent patent by Roche (WO/2020/035605, Roche Diagnostics, Germany).
CONCLUSION
Herein, we have shown that several suggested HF biomarkers are not cardiac-specific. Although DKK3 is cardiac-enriched and the secreted protein shows relatively high absolute cardiac expression, including in comparison with other suggested markers, such as Gal-3 and GDF-15, in HF mouse models a simple relation between the cardiac and circulating DKK3
Identification of cardiac specific HF biomarkers
137
pools was absent. Moreover, DKK3 was not independently associated with HF outcome. Based on this study, the clinical potential of DKK3 as a biomarker for HF seems limited. Finally, we believe that high cardiac specificity is an important characteristic in a protein if it is to have potential as a diagnostic HF biomarker.
Acknowledgements
The authors would like to thank Martin Dokter, Weijie Du, Frouke Houtsma, Noa Keizer and Marloes Schouten for excellent technical assistance, and Haye van der Wal for reviewing the statistical methods.
Funding
The authors acknowledge the support of the Netherlands Heart Foundation (CVON Predict2, grant no. 2018-30; CVON DOSIS, grant no. 2014–40), the Foundation Leducq (CurePLaN) and the European Research Council (ERC CoG 818715, SECRETE‐HF).
Conflicts of interest
The University Medical Centre Groningen, which employs A.P., N.S., A.A.V., R.A.dB., and H.H.W.S. has received research grants and/or fees from AstraZeneca, Abbott, Bristol‐Myers Squibb, Novartis, Novo Nordisk and Roche. R.A.dB. has received personal fees from Abbott, AstraZeneca, Novartis and Roche. A.A.V. received consultancy fees and educational grants from Roche Diagnostics. The other authors have nothing to disclose.
Chapter 6
138
REFERENCES
1. Yancy CW, Jessup M, Bozkurt B, et al. 2017 ACC/AHA/HFSA focused update of the 2013 ACCF/AHA
guideline for the management of heart failure: A report of the american college of cardiology/american heart association task force on clinical practice guidelines and the heart failure society of america. J Card
Fail. 2017;23(8):628-651.
2. Ponikowski P, Voors AA, Anker SD, et al. 2016 ESC guidelines for the diagnosis and treatment of acute and
chronic heart failure: The task force for the diagnosis and treatment of acute and chronic heart failure of the european society of cardiology (ESC)developed with the special contribution of the heart failure association (HFA) of the ESC. Eur Heart J. 2016;37(27):2129-2200.
3. Levin ER, Gardner DG, Samson WK. Natriuretic peptides. N Engl J Med. 1998;339(5):321-328.
4. Piek A, Du W, de Boer RA, Sillje HHW. Novel heart failure biomarkers: Why do we fail to exploit their
potential? Crit Rev Clin Lab Sci. 2018;55(4):246-263.
5. Kinnunen P, Vuolteenaho O, Ruskoaho H. Mechanisms of atrial and brain natriuretic peptide release from
rat ventricular myocardium: Effect of stretching. Endocrinology. 1993;132(5):1961-1970.
6. Liang F, Gardner DG. Mechanical strain activates BNP gene transcription through a
p38/NF-kappaB-dependent mechanism. J Clin Invest. 1999;104(11):1603-1612.
7. Liang F, Wu J, Garami M, Gardner DG. Mechanical strain increases expression of the brain natriuretic
peptide gene in rat cardiac myocytes. J Biol Chem. 1997;272(44):28050-28056.
8. Omland T, Rosjo H, Giannitsis E, Agewall S. Troponins in heart failure. Clin Chim Acta. 2015;443:78-84.
9. Du W, Piek A, Schouten EM, et al. Plasma levels of heart failure biomarkers are primarily a reflection of
extracardiac production. Theranostics. 2018;8(15):4155-4169.
10. De Angelis G, Rittenhouse HG, Mikolajczyk SD, Blair Shamel L, Semjonow A. Twenty years of PSA: From prostate antigen to tumor marker. Rev Urol. 2007;9(3):113-123.
11. Dufour DR, Lott JA, Nolte FS, Gretch DR, Koff RS, Seeff LB. Diagnosis and monitoring of hepatic injury. I. performance characteristics of laboratory tests. Clin Chem. 2000;46(12):2027-2049.
12. Ismail OZ, Bhayana V. Lipase or amylase for the diagnosis of acute pancreatitis? Clin Biochem. 2017;50(18):1275-1280.
13. Fagerberg L, Hallstrom BM, Oksvold P, et al. Analysis of the human tissue-specific expression by genome-wide integration of transcriptomics and antibody-based proteomics. Mol Cell Proteomics. 2014;13(2):397-406.
14. Zhang Y, Liu Y, Zhu XH, et al. Dickkopf-3 attenuates pressure overload-induced cardiac remodelling.
Cardiovasc Res. 2014;102(1):35-45.
15. Zhai CG, Xu YY, Tie YY, et al. DKK3 overexpression attenuates cardiac hypertrophy and fibrosis in an angiotensin-perfused animal model by regulating the ADAM17/ACE2 and GSK-3beta/beta-catenin pathways. J Mol Cell Cardiol. 2018;114:243-252.
16. Lu D, Bao D, Dong W, et al. Dkk3 prevents familial dilated cardiomyopathy development through wnt pathway. Lab Invest. 2016;96(2):239-248.
17. Bao MW, Cai Z, Zhang XJ, et al. Dickkopf-3 protects against cardiac dysfunction and ventricular remodelling following myocardial infarction. Basic Res Cardiol. 2015;110(3):25-015-0481-x. Epub 2015 Apr 4.
18. Cox EJ, Marsh SA. A systematic review of fetal genes as biomarkers of cardiac hypertrophy in rodent models of diabetes. PLoS One. 2014;9(3):e92903.
19. Piek A, Koonen DPY, Schouten EM, et al. Pharmacological myeloperoxidase (MPO) inhibition in an obese/hypertensive mouse model attenuates obesity and liver damage, but not cardiac remodeling. Sci
Rep. 2019;9(1):18765-019-55263-y.
20. Voors AA, Anker SD, Cleland JG, et al. A systems BIOlogy study to TAilored treatment in chronic heart failure: Rationale, design, and baseline characteristics of BIOSTAT-CHF. Eur J Heart Fail. 2016;18(6):716-726.
Identification of cardiac specific HF biomarkers
139
21. Pinto-Sietsma SJ, Janssen WM, Hillege HL, Navis G, De Zeeuw D, De Jong PE. Urinary albumin excretion is associated with renal functional abnormalities in a nondiabetic population. J Am Soc Nephrol. 2000;11(10):1882-1888.
22. Voors AA, Ouwerkerk W, Zannad F, et al. Development and validation of multivariable models to predict mortality and hospitalization in patients with heart failure. Eur J Heart Fail. 2017;19(5):627-634.
23. Aird WC. Phenotypic heterogeneity of the endothelium: I. structure, function, and mechanisms. Circ Res. 2007;100(2):158-173.
24. Zewinger S, Rauen T, Rudnicki M, et al. Dickkopf-3 (DKK3) in urine identifies patients with short-term risk of eGFR loss. J Am Soc Nephrol. 2018;29(11):2722-2733.
25. Schunk SJ, Zarbock A, Meersch M, et al. Association between urinary dickkopf-3, acute kidney injury, and subsequent loss of kidney function in patients undergoing cardiac surgery: An observational cohort study.
Lancet. 2019;394(10197):488-496.
26. Schunk SJ, Speer T, Petrakis I, Fliser D. Dickkopf 3-a novel biomarker of the 'kidney injury continuum'.
Nephrol Dial Transplant. 2020:doi: 10.1093/ndt/gfaa003.
27. Federico G, Meister M, Mathow D, et al. Tubular dickkopf-3 promotes the development of renal atrophy and fibrosis. JCI Insight. 2016;1(1):e84916.
28. Yu B, Kiechl S, Qi D, et al. A cytokine-like protein dickkopf-related protein 3 is atheroprotective.
Circulation. 2017;136(11):1022-1036.
29. Zenzmaier C, Sklepos L, Berger P. Increase of dkk-3 blood plasma levels in the elderly. Exp Gerontol. 2008;43(9):867-870.
30. Krupnik VE, Sharp JD, Jiang C, et al. Functional and structural diversity of the human dickkopf gene family.
Gene. 1999;238(2):301-313.
31. Tillet E, Ouarne M, Desroches-Castan A, et al. A heterodimer formed by bone morphogenetic protein 9 (BMP9) and BMP10 provides most BMP biological activity in plasma. J Biol Chem. 2018;293(28):10963-10974.
32. Hagdorn QAJ, Bossers GPL, Koop AMC, et al. A novel method optimizing the normalization of cardiac parameters in small animal models: The importance of dimensional indexing. Am J Physiol Heart Circ
Physiol. 2019.
Chapter 6
140
SUPPLEMENTAL METHODS
Included organs and tissues in the bioinformatics analysis
In the bioinformatics analysis, gene expression data of organs and tissues in healthy humans as published by Fagerberg et al13 were used. The following organs and tissues were included: Adipose tissue, adrenal gland, appendix, bone marrow, brain, colon, duodenum, oesophagus, gall bladder, heart, kidney, liver, lung, lymph node, pancreas, salivary gland, skin, small intestine, spleen, stomach, thyroid gland and urinary bladder. Sex specific organs/tissues were prostate and testis for male, and endometrium and ovary for female.
Animal experimental protocol
Three mouse models were included as described before9,19. An 8 weeks myocardial infarction (MI) model was induced by ligation of the left anterior descending artery (LAD). An 8 weeks transverse aortic constriction (TAC) model, a model of cardiac pressure overload, was induced by tying a suture around a 27 gauge blunt needle and the aorta, resulting in a reproducible stenosis after removal of the needle. An obese/hypertensive mouse model with characteristics of heart failure with preserved ejection fraction (HFpEF) was induced by 16 weeks of high fat diet (HFD, 60% kcal% fat, D12492, Research Diets, USA) combined with 4 weeks of Angiotensin-II (AngII, 1 mg/kg/day; Bachem, Switzerland) infusion during the last 4 weeks of diet using osmotic minipumps (Alzet 1004, Durect corporation, USA). Control mice were sham operated or received saline infusion and a control low fat diet (LFD, 10% kcal% fat, D12450J, Research Diets, USA). Mice were housed in a 12/12 h day/night cycle with ad libitum access to water and chow. Mice were randomized based on body weight. Mice with abnormal bodyweight (> 2 SD) and/or abnormal behavior were excluded. For anesthesia, 2% isoflurane/oxygen was used. To alleviate post-operative wound pain, 5.0 mg/kg Carprofen was administered subcutaneously prior to surgery. Blood plasma and LV were snap frozen in liquid nitrogen and stored at -80 °C until further analysis. LV weight was corrected for tibia length to the power of 332. Phenotyping of these mouse models using cardiac magnetic resonance imaging (CMR) measurements, PV-loop measurements, and histological analyses were described before9,19, and performed in a blinded fashion.
RT-qPCR
Total mouse RNA was extracted from powdered LV, lung, liver and kidney tissue using trizol reagent (Invitrogen, Thermo Fisher Scientific, USA). RNA concentration was determined by spectrophotometry (NanoDrop 2000, Thermo Scientific, the Netherlands). cDNA was synthesized using QuantiTect Reverse Transcriptional kits (Qiagen, the Netherlands). A total of 7.5 ng cDNA was used to determine relative gene expression by real-time quantitative polymerase chain reaction (RT-qPCR) using Absolute SYBR Green mix (Thermo Scientific, the Netherlands) and the BIO-Rad CFx384 real time system (Bio-Rad, the Netherlands). Gene expression was corrected for ribosomal protein, large, P0 (36B4) reference gene expression and is presented as relative expression to controls.
Identification of cardiac specific HF biomarkers
141 Mouse DKK3 ELISA
Mouse DKK3 plasma concentrations were measured using mouse DKK3 enzyme-linked immunosorbent assay (ELISA) kits (mouse DKK3 ELISA, OKCD02498-96, Aviva Systems Biology, USA). Assays were performed according to the manufacturer’s manual. Plasma samples were diluted 1:50. The lower standard of this assay is 0.156 ng/ml and the manufacturer’s reported intra- and inter-assay coefficient of variances (CV) of the mouse DKK3 ELISA are 10 and 12%, respectively.
Human DKK3 ELISA
DKK3 plasma concentrations were measured using human DKK3 ELISA kits (DY1118, R&D systems, USA). Assays were performed in half area 96 wells plates (3690, Costar half area 96 well polystyrene assay plates, Corning, USA). Standard curves were included in duplicate on every ELISA plate. Plasma samples were diluted 1:100 in 1% bovine serum albumin (BSA, 11930.03, Serva, Germany) in phosphate buffered saline (PBS). As a wash buffer, 0.05% Tween (Tween 20, P1379, Sigma-Aldrich, USA) in PBS was used. Substrate solution was prepared with TMB (Tetramethylbenzidine-dihydrochloride, T3405, Sigma-Aldrich, USA), H2O2 (Hydrogen peroxide 30%, Emsure, 1.07209.0250, VWR, USA), ultra-purified H2O and NaAc (Sodiumacetate, S2889, Sigma-Aldrich, USA). As a stop solution, 2 N H2SO4 (Sulfuric acid, 7102, Avantor, USA) was used. The lower standard of this assay is 31 pg/ml. Quality control analysis as performed by the manufacturer, as mentioned in the assay manual, showed no cross-reactivity or interference of human DKK1, human DKK4, human LRP-6-Fc chimera, mouse DKK1, mouse DKK3 and mouse DKK4 in this assay. Additional quality control was performed by our lab. In our hands, the coefficient of variance (CV) and inter-assay reproducibility of this assay were 8.9% uncorrected for CV) and 97.2% (uncorrected for CV), respectively. Stability of DKK3 in plasma sample over four freeze-thaw cycles, storage at 4 °C for 48 hours and storage at room temperature for 48 hours was determined. Although, storage resulted in somewhat lower levels as compared to fresh material (albeit not significant), there was no difference between 48 hours storage at 4 ˚C or room temperature and upon multiple freeze thaw cycles, indicating that DKK3 is relatively stable (Supplemental Table 3).
Age and sex matched control group
An age and sex matched control group of subjects without HF was composed to be able to compare plasma DKK3 concentrations between HF patients and controls. Subjects were selected from a large population cohort also including many subjects without HF. This cohort study was described in detail elsehere21. The cohort was designed to investigate the natural course of microalbuminuria and the relation of microalbuminuria with the development of future renal and cardiovascular diseases. In short, a questionnaire consisting of questions regarding cardiovascular risk factors and morbidity and a vial to collect an early-morning urine sample were send to a total of 85421 subjects of the general population aged 28-75 years, of which 40856 subjects responded. Pregnant women and subjects with diabetes
Chapter 6
142
mellitus were excluded from participation. Subjects with a morning urinary albumin concentration (UAC) > 10 mg/L and a selection of subjects with an UAC < 10 mg/L were included in the cohort, which consisted of in total 8952 subjects. At baseline, a plasma sample was collected and stored at -80 °C until analysis. The control group for this study was composed as follows: Subjects with a history of myocardial infarction, cerebrovascular events or renal diseases and subjects with an estimated glomerular filtration rate (eGFR) < 60 ml/min were excluded. At baseline, the maximum age of subjects in the used population cohort was 75 years, whilst in the HF patient cohort also subjects with the age > 75 years were included. To allow an accurate HF-control comparison independent of age and sex, the control group was composed for only those patients from the HF cohort aged 75 years (n=1434). In total 240 subjects were randomly included in the control group.
Identification of cardiac specific HF biomarkers
143
SUPPLEMENTAL TABLES
Supplemental table 1. Raw data of a selection of genes as analyzed in the bioinformatics analysis.
Gene Expression level Cardiac specificity (%)
Heart Total body
Male Female Male Female
BMP-10 300.5 307.0 307.1 97.9 97.8 DKK3 942.3 2135.2 2077.7 44.1 45.4 GDF-15 0.2 334.4 282.0 0.1 0.1 HE4 1.4 2032.2 2206.5 0.1 0.1 IGFBP-7 215.2 3773.8 3717.6 5.7 5.8 IL-6 6.7 291.8 282.0 2.3 2.4 LGALS3 143.0 4515.6 4683.5 3.2 3.1 MPO 0.4 1654.7 1653.7 0.0 0.0 NPPA 6250.9 6297.4 6306.4 99.3 99.1 NPPB 3030.7 3034.7 3034.5 99.9 99.9 TIMP-1 160.1 4465.3 4556.9 3.6 3.5
Expression levels are represented as absolute read values in RNAseq as performed and published by Fagerberg et al13. Included organs and tissues in total body expression levels are adipose tissue, adrenal gland, appendix, bone marrow, brain colon, duodenum, esophagus, gall bladder, heart, kidney, liver, lung, lymph node, pancreas, salivary gland, skin, small intestine, spleen, stomach, thyroid gland, and prostate and testis for male, and endometrium and ovary for female. Cardiac specificity was calculated using the following formula: Cardiac
specificity = (expression heart / expression total body) * 100. BMP-10=Bone morphogenic protein 10.
DKK3=Dickkopf-3. GDF-15=Growth differentiation factor 15. HE4=Human epididymis protein 4. IGFBP-7=Insulin growth factor binding protein 7. IL-6=Interleukin-6. LGALS3=Gene encoding galectin-3. MPO=Myeloperoxidase. NPPA=Natriuretic peptide precursor type A. NPPB=Natriuretic peptide precursor type B. TIMP-1=Tissue inhibitor of metalloproteinase-1.
Supplemental table 2. Oligonucleotide pairs used for RT-qPCR.
Gene 5' - 3' forward 5' - 3' reverse
36B4 AAGCGCGTCCTGGCATTGTC GCAGCCGCAAATGCAGATGG
DKK3 ACCAGGTACTGCCAGTTCTC TCACTGTCTCGGGTGCATAG
NPPA ATGGGCTCCTTCTCCATCAC TCTACCGGCATCTTCTCCTC
Supplemental table 3. Stability of human DKK3 in plasma.
Test DKK3 (ng/ml)
Control 30.8 ± 2.4
Four freeze-thaw cycles 28.4 ± 5.1
Stored at 4°C 48h 28.6 ± 4.7
Stored at room temperature 48h 27.9 ± 2.8
Stability of human DKK3 in plasma. Included treatments are four freeze-thaw cycles, storage of plasma samples at 4°C for 48 hours and storage of a plasma sample at room temperature for 48 hours. DKK3 values are shown as mean ± standard deviation. N=3 per group. DKK3=Dickkopf-3.
Su p p le m e n tal tab le 4. Hem o d yn am ic s, c ar d iac d im e n si o n s an d left ve n tr ic u lar r e m o d el in g in T A C, M I an d o b e se/ h yp e rte n si ve m ic e . Gr o u p Car d iac r e m o d e lin g P -c ath e te r CM R LV we ig h t (µ g/ m m 3 ) LV fi b ro si s (% ) LVE SP (m m H g) d P/d Tma x (1/ s) d P/d Tmi n (1/ s) LVE SV (µ L) LVE D V (µ L) LVE F (% ) Sh am 20.4 ± 0.8 0.7 ± 0.2 100.3 ± 2. 6 81.7 ± 4.6 -78.3 ± 5.4 26.4 ± 1.9 62.5 ± 2.6 58.1 ± 1.6 MI 29.4 ± 1.2 * 24.1 ± 3.0 * 84.2 ± 3.8 * 55.9 ± 1.4 * -46.0 ± 1.8* 148.2 ± 2 2.9* 175.1 ± 2 1.7* 18.3 ± 3.3 * Sh am 21.0± 0. 4 1.2 ± 0.1 90.8 ± 3.0 82.5 ± 3.9 -85.4 ± 3.6 22.6 ± 1.2 59.0 ± 1.9 62.0 ± 1.1 TA C 40.6 ± 1.4 * 2.8 ± 0.4 * 119.1 ± 9. 7* 44.9 ± 1.6 * -39.9 ± 1.6* 77.4 ± 7.0 * 100.2 ± 6. 5* 23.6 ± 2.2 * LFD 22.5 ± 0.5 0.8 ± 0.1 97.3 ± 2.3 74.8 ± 4.7 -66.4 ± 6.4 22.7 ± 1.8 49.9 ± 2.1 55.1 ± 1.8 H FD 23.7 ± 1.2 0.8 ± 0.1 105.1 ± 3. 9 72.5 ± 4.1 -60.5 ± 3.6 17.8 ± 0.7 46.3 ± 1.4 61.7 ± 0.9 * H FD +A n gII 30.1 ± 1.5 * # 1.6 ± 0.3 * # 114.1 ± 9. 2* 59.0 ± 4.8 * -47.4 ± 4.1* 24.7 ± 3.1 51.5 ± 2.5 53.7 ± 3.7 Dat a ar e p re se n ted as m ean s ± SE M. An gII= An gi ot en si n -II (f in al 4w ks o f 16 w ks H FD) . CMR = Card iac m agn eti c re so n an ce i m agi ng. d P/d Tmax =M easure fo r le ft ve n tri cu lar m ax im al con tra cti o n cap aci ty , h ere corre cted fo r m ax im al v en tri cu lar p re ss u re . d P/d Tmi n =Me as u re f o r le ft ve n tri cu lar m ax im al r el axa ti o n cap ac ity , h ere correct ed f o r m ax im al v en tri cu lar p re ss u re . H FD= H igh fat d ie t (16 w ks) . LFD= Lo w f at d ie t (1 6w ks ). LV = Lef t ve n tri cl e. LVE DV =Le ft ve n tri cu lar en d -d ias to lic v o lu m e. LVE F= Le ft ve n tri cu lar ejec ti o n f ra cti o n . LVE SP=L ef t ve n tri cu lar e n d -s ys to lic p re ss u re . LVE SV =Le ft ve n tri cu lar e n d -s ys to lic v o lu m e. MI =My o car d ial in far cti o n ( 8w ks ). P -cat h eter= Pr es su re cat h eter. TAC =T ra n sv er se ao rti c con stri cti o n (8w ks ). LV w ei ght i s cor re cted fo r ti b ia le n gth to th e p o w er o f 3 32 . N = 6 -2 0. *P < 0. 05 ve rs u s re sp ecti ve s h am gr o u p f o r TAC an d MI, o r LFD for H FD a n d H FD+ An gII. # P < 0.05 v er su s H FD. Da ta a s p u b lis h ed b ef o re in T h era n o sti cs 9 an d Sci en ti fi c R ep o rts 19 , b u t p re se n ted h ere f o r con ve n ie n ce .
Identification of cardiac specific HF biomarkers
145
Supplemental table 5. Baseline characteristics of patients with heart failure, and age and sex matched controls.
Characteristic Control (N=240) Heart failure (N=1434) P-value
Age (years) 62.6 ±10.0 62.4 ±9.7 0.854 Male sex, % (n) 189 (78.8) 1133 (79.0) 0.927 Caucasian race, % (n) 234 (97.5) 1412 (98.5) 0.280 BMI (kg/m2) 27 (25-29) 28 (25-32) 0.001* eGFR (ml/min/1.73m2) 89.8 (80.1-96.1) 66.2 (49.1-82.5) < 0.001* DKK3 (ng/ml) 36.3 (31.5-41.3) 46.4 (38.6-57.4) < 0.001*
Data are presented as means ± SD for normally distributed variables, as medians ± IQR for non-normally distributed variables and as percentage (n) for categorical variables. BMI=Body mass index. DKK3=Dickkopf-3. eGFR=Estimated glomerular filtration rate. Maximum age used for these comparisons was 75 yrs. Control subjects were selected for having no history of myocardial infarction, cerebrovascular accidents or renal disease and no eGFR < 60ml/min/1.73m2. *P < 0.05.
