Article Figures & Data
Figures
- Figure S1. Study design.
This illustration describes the steps of the present investigation.
- Figure S2. Protein profiles of HPA045005 in every sample set.
The solid lines illustrate the trends estimated by linear regression, and the shades around them show 95% confidence intervals of fitted values. At the bottom right corner, all trend lines were overlapped to present the increasing trends in all nine sample sets.
- Figure 1. Meta-analysis from nine different sample sets.
In the forest plot, the numbers in parenthesis indicate the age range of the included subjects. For each sample set, the estimated effect of age on HPA045005-derived profiles from the linear regression model, 95% confidence interval of it, and study weight in the meta-analysis are shown in the middle as a tick, a line, and a gray box, respectively. The numeric value of the effect is clarified at the right side.
- Figure 2. Genome-wide association study results and histidine-rich glycoprotein (HRG) domains.
(A) Manhattan plot. The significance of association between genotypes and HPA045005 profiles is presented vertically. The dashed guide line marks the stringent threshold of P-value for genome-wide association study, which is P = 0.01 after Bonferroni correction. One peak in chromosome 3 indicates strong association of a locus with the molecular phenotype. (B) LocusZoom (Pruim et al, 2010) on associated locus. The illustration shows the elements of chromosome 3 associated with HPA045005 profiles. (A) Zooming in on the peak of the Manhattan plot in (A), the genes around the locus are presented together with the associated single-nucleotide polymorphisms. (C) Box plots to show the association between genotypes of rs9898 and two antibody profiles, HPA045005 and BSI0137. The trends were opposite. (D) Representation of HRG to illustrate how two separate domains of the HRG protein affected the profiles of the antibodies. Protein domains, glycosylation sites and disulfide bonds of HRG were schematically illustrated using black round, purple dots and yellow lines, respectively. The figure was prepared based on the schematic representation of Poon et al (2011).
- Figure S3. Additional validation of molecular target.
(A) Comparative sandwich immunoassays for histidine-rich glycoprotein (HRG). To confirm the molecular binding of HPA045005 to HRG, dilution of spiked-in HRG protein levels were studied. Different colors represent additional capture antibodies, such as the anti-HRG binder HPA054598. All were used for a multiplexed assessment of HRG binding. For the detection, a labeled version of HPA054598 was used. (B) High-density protein array analysis of HPA045005. The signal intensities from the interaction of the antibody with any of the 16,728 immobilized unique protein fragments on the glass slide are illustrated. For the antibody HPA045005, one peak was exclusively observed for the protein fragment that had been used for the production of the antibody (red bar), as well as for the positive controls detecting the presence of the rabbit IgG antibody (orange and green bars).
- Figure S4. Manhattan plot and LocusZoom plot for BSI0137 antibody profile.
Because the P-values of the top single-nucleotide polymorphisms in the genome-wide association study for BSI0137 was too low, the computation tool for this analysis, PLINK, gave us 0 s. (A, B) The values were manually set to the lowest values the visualization software allowed, 1 × 10−320 in (A) or 1 × 10−350 in (B).
- Figure S5. Probabilistic identification of causal SNPs (PICS) analysis for the associated single-nucleotide polymorphisms with HPA045005 and BSI0137.
The top two figures display the distribution of -log of P-values obtained from the linear regression model for the effects of the genotype of rs1042464 in 100,000 permutations, assuming rs9898 was causal for individual antibody profiles. The bottom figures show the results of rs9898 assuming rs1042464 was causal genetic variant. The green vertical lines indicate the observed significances.
- Figure S6. Detection of histidine-rich glycoprotein peptides in human serum by MS/MS analysis.
Peptides from histidine-rich glycoprotein detected in serum samples using a bottom-up proteomic experimental workflow and liquid chromatography tandem mass spectrometry for read-out. Top panel: five different enzymes (y-axis; GluC, LysN, LysC, chymotrypsin, and trypsin) and the peptide amino acid sequence coverage (x-axis) of their experimental peptides. Four single-nucleotide polymorphism variants (rs10770, rs9898, rs2228243, and rs1042464) have been indicated by vertical black lines. Bottom panel: experimental peptides reported by the PeptideAtlas (accessed 16 May, 2017). The peptide amino acid sequence coverage (x-axis) is visualized in relation to experimental peptides for each experiment (y-axis). Highlighted in this example, rs9898 has only a few theoretical peptide sequences that can be observed with the bottom up method (because of sequence length) using this set of enzymes. Either one tryptic peptide with one missed cleavage NCPRHHFPR (identified once highlighted in the figure) or the chymotryptic peptide SCRNCPRH (based on low enzymatic specificity). For both panels, blue indicates the most abundant protein form and red is for any variation from the form.
- Figure 3. Survival analysis comparing upper and lower quarters of histidine-rich glycoprotein levels.
The individuals of sample set 3 were divided into four subsets by the quartiles of histidine-rich glycoprotein levels. Differential mortality across follow-up time is illustrated by the survival curves, where age was used as the time scale (Thiébaut & Bénichou, 2004). The number at risk at 10 yr intervals were displayed below the survival curve. Some detailed statistics related to this survival analysis are presented in Table S4.
- Figure S7. Survival curves for women and men, comparing two extreme quartiles of set 3 by histidine-rich glycoprotein profiles.
Detailed statistics are provided in Table S5.
- Figure S8. Location of the 204th residue of histidine-rich glycoprotein.
The amino acid variant that affected HPA045005 profiles is highlighted in a 3D model predicted by SWISS-MODEL for the peptide of 19–250th amino acids of the protein using the homology with the template 6ht9.1B “Fetuin-B” (Waterhouse et al, 2018).
- Figure S9. Difference between serum and plasma in sample sets 1 and 2.
Each plot displays the first two principal components of the MFIs of one assay, where the profiles of a set of 384 antibodies were obtained. The points of the scatter plot were colored by sample preparation types, such as plasma and serum.
Tables
Study set Age (yr) Gender (F:M) Indicationa Cohort name Sample Type References Set 1 50–92 78:78 Population TwinGene Serum Lichtenstein et al (2002) and Magnusson et al (2013) Set 2 9–63 102:102 Population LifeGene Plasma Almqvist et al (2011) Set 3 48–93 1,613:1,386 Population TwinGene Serum Lichtenstein et al (2002) and Magnusson et al (2013) Set 4 51–86 50:0 Breast cancer Set 5 56–75 0:50 Prostate cancer Set 6 55–78 16:27 Cardiovascular disease IMPROVE Plasma Baldassarre et al (2010) Set 7 41–60 12:31 Myocardial infarction SCARF Plasma Samnegård et al (2005) Set 8 48–73 20:23 Acute coronary heart syndrome CHAPS Plasma Odeberg et al (2014) Set 9 40–73 600:0 Mammography KARMA Plasma Gabrielson et al (2017) ↵a Subjects included in the presented study did not include individuals diagnosed with the disease of the indication area, but the subjects assigned as controls for the different disease cohorts.
Trait HPA045005 Bsi0137 Estimatea P-value Estimatea P-value Aging Age 0.010 1.53 × 10−6 0.0017 0.44 Mortality risk 1.25 6.45 × 10−5 1.03 0.59 Genetic/protein variants rs9898 (Pro204Ser) 0.15 2.35 × 10−97 −0.23 1.85 × 10−177 rs1042464 (Asn493Ile) 0.10 1.9 × 10−44 −0.33 <1 × 10−300 Clinical trait APOA1 −0.26 9.47 × 10−6 −0.01 0.76 APOB −0.23 8.46 × 10−4 0.30 6.81 × 10−10 C-reactive protein 0.012 7.33 × 10−5 0.0003 0.88 Glucose −0.028 0.061 0.013 0.21 Hb −0.008 6.45 × 10−9 0.0004 0.68 HbA1C −0.047 0.070 0.030 0.092 High density lipoprotein −0.035 0.41 −0.120 5.65 × 10−5 Low density lipoprotein −0.044 1.53 × 10−2 0.071 1.90 × 10−8 TC −0.060 1.37 × 10−4 0.060 4.92 × 10−8 TG −0.086 2.73 × 10−5 0.074 2.36 × 10−7 ↵a The estimates from selected models for individual associations. For clinical traits, the values are the estimated slope from linear regression models with adjustment for age and the top single-nucleotide polymorphism (rs9898 for HPA045005 and rs1042464 for BSI0137). Linear models were also used for age and genetic/protein variants, whereas Cox models for mortality risk (more details in the Materials and Methods section). For the trait, hazard ratios are presented in the column.
The two distinct histidin-rich glycoprotein profiles were compared with respect to the association with various traits in set 3, which are 2,999 samples from the TwinGene cohort (Lichtenstein et al, 2002; Magnusson et al, 2013).
Supplemental Data 1.
Materials and methods.[LSA-2020-00817_Supplemental_Data_1.docx]
Table S6 Summary of survival analyses.
