Abstract
Aims This study assessed sex-specific proteomic profiles by
cardiovascular disease (CVD) phenotype (coronary artery disease [CAD]
vs coronary microvascular dysfunction [CMD]) and describe their role in
sex-specific pathways. Methods: In a secondary biobank analysis of the
Yale-CMD registry, adults with ischemic symptoms who underwent cardiac
positron emission test/computed tomography were categorized as a)
controls (normal coronary flow reserve (CFR) > 2 without perfusion
defect or coronary calcification), b) having CMD (CFR < 2 without
defect or calcification), or c) having CAD (known CAD or new perfusion
defect). Using proximity extension assays (Olink® Explore 3072), we
examined 2944 proteins. Differential protein expression was assessed
using linear regression models, adjusting for age, race, body mass
index, diabetes, dyslipidemia, hypertension, or smoking. Results: Of
190 patients, 91 provided blood samples (mean age, 56 years; 66 %,
females; 48 %, controls; 24 %, CAD; 27 %, CMD). Among controls, 15
proteins showed sex differences (5 proteins upregulated in females, 10
in males; false discovery rate [FDR < 0.05]). Upregulated in CAD
patients were FSHB in females and INSL3 and EDDM3B in males (FDR <
0.05). Among CMD patients, SCGB3A1 and HGFAC were higher in females;
INSL3, SPINT3, EDDM3B, and KLK3 were higher in males (FDR < 0.05). Per
pathway analysis, females showed upregulation of immune pathways in CAD
and lipid and glucose metabolism pathways in CMD. Males showed
upregulated endothelial regulation of blood flow in CAD and increased
angiogenesis in CMD. Conclusions: Sex differences exist in the
proteomic profiles of CAD and CMD patients, highlighting a need for
precision medicine.
Keywords: Proteomics, Sex differences, Coronary artery disease,
Coronary microvascular dysfunction, Olink, Precision medicine
1. Introduction
Cardiovascular disease (CVD) remains the leading cause of mortality in
the United States.[31][1] Sexual dimorphism exists in the onset,
progression, and outcomes of CVD, highlighting a need for improved
understanding of sex-specific underlying mechanisms. Males show a
higher prevalence of ischemic heart disease at younger ages, primarily
due to ischemia from obstructive coronary artery disease (CAD)
involving one of the epicardial arteries. In contrast, females present
at later ages, and show a higher prevalence of heart failure and
ischemia due to the coronary microvascular disease (CMD) involving the
microvasculature.[32][2] Even attribution of conventional cardiac risk
factors varies between sexes. Females with diabetes, obesity, or
smoking are at higher risk for CVD, whereas males with hypertension or
dyslipidemia demonstrate greater risk. Genetic factors (such as sex
chromosomes) and sex hormones explain some of these differences,
independently and through a complex interplay; for instance, estrogen
imparts a protective effect against ischemic heart disease in
premenopausal females.[33][3] However, genetics does not fully explain
CVD differences between postmenopausal females and similarly aged
males, implicating the role of environmental factors such as lifestyle
choices, social factors, and medications.[34][4] Novel tools are need
that allow a unified platform to study the influence of both factors on
CVD expression in males and females.
Proteomics investigates dynamic biological states of proteins,
influenced by both genetics and environmental factors, to explain the
sex differences in CVD phenotypes.[35][5] Blood proteins behave
differently in males and females.[36][6] Proteomics in population-based
cohorts indicate sex differences in incident development of CAD,[37][7]
and heart failure,[38][8] with sex-specific variations in biological
pathways.[39][9] However, sex comparisons in acute ischemia remain
scarce, or report conflicting results. In a female cohort, Prescott et
al.[40][10] found inflammation to be independently associated with CMD,
whereas Chandramouli et al.[41][11] found the opposite, with
inflammation playing a greater role in males with CMD in heart failure
patients. Thus, a discovery-based assessment of sex differences in
proteomics is needed to better elucidate the mechanism by phenotype.
Here, we used large-scale blood proteomics in emergency department (ED)
patients with symptoms of acute ischemia to understand sex-specific
pathways pertaining to CAD and CMD.
2. Methods
2.1. Study population
This study was a secondary analysis of patients enrolled in the
Yale-CMD registry, a prospective observational cohort study. Patients
were eligible if they met the following inclusion criteria: Adults
presenting to the Yale ED with chest pain or an angina equivalent and
admitted to the Yale observation unit between May 2014 and February
2020 for a cardiac positron emission test (PET) with computed
tomography (CT). We excluded patients who had acute myocardial
infarction, hemodynamic instability, or acute heart failure; were
currently receiving dialysis; had a febrile illness; or were unable to
consent. This study was approved by the Yale Institutional Review Board
(HIC 2000022866) and conducted in accordance with the ethical
principles outlined in the Declaration of Helsinki.
2.2. Data sources
A trained research associate (RA) interviewed consented patients for
sociodemographic characteristics, family history, symptoms,
reproductive history (in females), height, weight, medications, and
reviewed medical records for test results.
2.3. Blood sample collection, processing, and storage
Between 9am-12 pm, 10 ml of peripheral blood was collected after
overnight fasting in non-EDTA tubes, centrifuged at 3000 RPM (∼1200 g)
for 15 min into serum, aliquoted in 500-ul labeled cryovials, and
stored in a − 80 °C freezer within 2 h of collection.
2.4. Protein biomarker measurements
Deidentified blood samples (100 ul) were shipped to O-link by
standardized shipping procedures (overnight on dry ice) for proteomics
using Olink’s high-throughput Proximity Extension Assay (PEA)
technology.[42][12] Olink uses a proximity-dependent DNA polymerization
event, which generates a unique PCR target sequence. Subsequently, the
resulting DNA sequence is detected and quantified using a microfluidic
real-time PCR instrument (Biomark HD, Fluidigm Corporation, CA, USA).
Data were quality controlled and normalized using an internal extension
control and an inter-plate control to adjust for intra- and inter-run
variation. The extension control is composed of an antibody coupled to
a unique pair of DNA tags that serve as a synthetic control added to
every sample well. This approach adjusts for technical variation
introduced in the extension step and hence reduces intra-assay
variability. Protein biomarker levels are recorded as normalized
protein expression values on a log2 scale. Olink® Explore 3072
incorporating 2944 plasma proteins for eight panels (Cardiometabolic,
Cardiometabolic II, Inflammation, Inflammation II, Neurology, Neurology
II, Oncology, and Oncology II) were analyzed. All assay validation data
for the proteins for these panels (detection limits, intra- and
inter-assay precision data, accuracy, reproducibility, and validity)
are available on O-link website.[43][12], [44][13], [45][14], [46][15]
Quality control process were applied both per manufacturer
recommendation and with coefficient of variation analysis for both
inter- and intra-assays. We excluded biomarkers with > 15 % of values
below the detection limit. Patients with samples that failed quality
control were excluded. A complete list of 2944 plasma proteins is
provided in [47]Supplemental Table 1.
2.5. Disease phenotype classification
Disease phenotyping was based on quantitative analysis of dynamic
3-dimensional rubidium-82 cardiac PET/CT. We defined disease phenotypes
in accordance with classification by the Coronary Vasomotor Disorders
International Study (COVADIS) Group:[48][16] 1) controls (normal
coronary flow reserve [CFR ≥ 2]) without perfusion defect or coronary
calcification, 2) having coronary microvascular dysfunction (CMD,
CFR < 2 without defect, calcification or cardiomyopathy, or if
confirmed by vasoreactivity testing on follow up angiogram), or 3)
having CAD (prior CAD or revascularization, new perfusion
defect).[49][17] We reported this methodology in detail
previously.[50][17].
2.6. Statistical analysis
Baseline characteristics of patients were compared between males and
females within each group using Wilcoxon rank sum tests for continuous
variables and chi-squared tests for categorical variables. Baseline
characteristics were compared among the three groups (control, CAD, and
CMD) using Kruskal-Wallis tests for continuous variables and
chi-squared tests for categorical variables.
To assess differential protein expression between males and females, we
used linear regression models, adjusting for age, race, body mass index
(BMI), comorbidities (history of diabetes, dyslipidemia, or
hypertension), and smoking. Assumptions of linear regression models,
including normality, linearity, and homoscedasticity, were assessed and
no significant violations were found. For each protein biomarker, the
regression coefficient of sex scaled by the standard deviation of
biomarker expression was reported. This scaling allows for more
interpretable results by reporting changes in units of standard
deviation rather than raw values, which varied between proteins.
Multiple hypothesis testing was corrected using the false discovery
rate (FDR). Proteins with an FDR < 0.05 were considered significant.
Differential protein expression analyses were performed separately for
the CAD, CMD, and control patients. All statistical analyses were
conducted using R (version 4.4.0).
2.7. Pathway and network analyses
To understand the biological functions of proteins that are
differentially expressed in males versus females, functional pathway
analysis was performed using ingenuity pathway analysis (IPA).[51][18]
Differentially expressed proteins with P-value < 0.05 and |β| >1 were
included in the analysis. Upregulated proteins in males and females
were analyzed separately to identify enriched biological pathways. For
the CAD and CMD groups, proteins were excluded from pathway analysis if
they were also differentially expressed in the control group. All
biomarkers measured on the protein assay in this study were used as the
enrichment background. Fisher’s exact test was used to determine the
overrepresentation of a specific functional pathway among the selected
differentially expressed proteins. Pathways with P-value < 0.05 were
considered significant. The STRING database[52][19] was used to
identify protein–protein interactions among biomarkers that were
upregulated in males and females separately. The Markov cluster (MCL)
algorithm[53][20] was used to identify clusters in the interaction
networks with the parameter values: granularity = 4.0 and array
source = combined_score. Proteins in each cluster were used to
determine their associated biological processes.[54][21], [55][22].
2.8. Weighted correlation network analysis of proteomics
We performed weighted correlation network analysis (WGCNA) to construct
protein correlation networks and to identify modules of highly
correlated proteins in the CAD, CMD, and control patients
separately.[56][23] WGCNA is an unsupervised method for identifying
clusters of highly correlated biomarkers. Network was constructed using
a minimum module size of 30 biomarkers. The protein expression profiles
of modules were summarized by eigenproteins, the first principal
component of the biomarker expression. Spearman correlation was used to
assess the correlation between each module’s eigenprotein and sex.
Proteins in the top modules with P-value < 0.05 were selected for
pathway analysis. Pathways with P-value < 0.05 were considered
significant.
3. Results
3.1. Baseline characteristics
Of the 189 eligible patients in the Yale-CMD registry, 91 contributed
blood samples for proteomic analysis. These included 22 patients with
CAD, 25 patients with CMD, and 44 patients with normal flow. This
sample selection was made to maintain an approximate 1:2 ratio of CAD
or CMD patients (cases) to normal flow patients (controls). Patients
who provided blood samples were similar to those who participated in
the Yale-CMD registry (median age 56 vs 57 years; 66 % versus 67 %
female; with 48 % versus 50 % controls, respectively). Baseline
characteristics indicate no significant sex difference in
sociodemographic or clinical characteristics ([57]Table 1), similar to
the full cohort ([58]Supplemental Table 2). Patients with CMD were
predominantly females compared with those with CAD (88 % versus 50 %, P
= 0.012), younger (52 versus 62 years, P = 0.002), and more obese (mean
BMI 41 versus 32, P < 0.001). Female patients across all three groups,
CAD, CMD, and control, did not show significant differences in
reproductive factors associated with CVD risk including status of
menopause, polycystic ovary syndrome (PCOS), an history of pregnancy
loss. The test results and p-values for all factors comparing across
groups are provided in [59]Supplemental Table 3.
Table 1.
Baseline characteristics of patients for proteomic analysis.
Control (n = 44) CAD (n = 22) CMD (n = 25)
Male Female P-value Male Female P-value Male Female P-value
N 17 27 11 11 3 22
Social demographics
Age, years 52.0 (42.0–––59.0) 54.0 (51.0–––57.5) 0.188 61.0
(59.0–––67.0) 62.0 (60.0–––74.5) 0.621 49.0 (45.5–––50.0) 55.5
(49.5–––62.8) 0.143
Ethnicity, Hispanic (%) 1 (5.9 %) 4 (14.8 %) 0.312 2 (18.2 %) 2
(18.2 %) 1 0 (0 %) 2 (9.1 %) 1
Race, White (%) 11 (64.7 %) 18 (66.7 %) 1 6 (54.5 %) 6 (54.5 %) 1 2
(66.7 %) 12 (54.5 %) 1
Insurance, Private Insured (%) 11 (64.7 %) 19 (70.4 %) 0.952 7 (63.6 %)
6 (54.5 %) 1 3 (100 %) 9 (40.9 %) 0.192
Educational level, Less than college (%) 8 (47.1 %) 19 (70.4 %) 0.713 8
(72.7 %) 10 (90.9 %) 0.58 2 (66.7 %) 17 (77.3 %) 1
Clinical risk factors
Hypertension 8 (47.1 %) 17 (63.0 %) 0.469 10 (90.9 %) 10 (90.9 %) 1 3
(100 %) 17 (77.3 %) 0.878
Hyperlipidemia 9 (52.9 %) 15 (55.6 %) 1 8 (72.7 %) 10 (90.9 %) 0.58 1
(33.3 %) 14 (63.6 %) 0.706
Hyperglycemia 10 (58.8 %) 13 (48.1 %) 0.704 5 (45.5 %) 6 (54.5 %) 1 2
(66.7 %) 12 (54.5 %) 1
Known CAD 0 (0 %) 0 (0 %) NA 8 (72.7 %) 7 (63.6 %) 1 0 (0 %) 3 (13.6 %)
1
Cerebral Vascular Accident (CVA) 0 (0 %) 1 (3.7 %) 1 1 (9.1 %) 1
(9.1 %) 1 0 (0 %) 4 (18.2 %) 1
Dementia 3 (17.6 %) 3 (11.1 %) 0.87 1 (9.1 %) 1 (9.1 %) 1 1 (33.3 %) 8
(36.4 %) 1
Pulmonary Embolism (PE) & Deep Vein Thrombosis (DVT) 0 (0 %) 1 (3.7 %)
1 1 (9.1 %) 2 (18.2 %) 1 0 (0 %) 1 (4.5 %) 1
Thyroid 2 (11.8 %) 11 (40.7 %) 0.087 3 (27.3 %) 0 (0 %) 0.214 0 (0 %) 7
(31.8 %) 0.641
Autoimmune 0 (0 %) 4 (14.8 %) 0.26 1 (9.1 %) 0 (0 %) 1 0 (0 %) 3
(13.6 %) 1
Thrombolysis in myocardial infarction (TIMI) score 0.631 0.497 0.878
0 9 (52.9 %) 11 (40.7 %) 2 (18.2 %) 1 (9.1 %) 0 (0 %) 5 (22.7 %)
1–3 8 (47.1 %) 16 (59.3 %) 8 (72.7 %) 7 (63.6 %) 3 (100 %) 17 (77.3 %)
4 0 (0 %) 0 (0 %) 1 (9.1 %) 3 (27.3 %) 0 (0 %) 0 (0 %)
Smoker (%) 8 (47.1 %) 13 (48.1 %) 1 8 (72.7 %) 6 (54.5 %) 0.658 2
(66.7 %) 15 (68.2 %) 1
BMI, kg/m2 40.5 (37.3–––44.5) 38.6 (33.9–––43.3) 0.555 31.1
(28.5–––35.9) 32.6 (28.8–––36.9) 0.748 36.0 (34.4–––47.5) 40.9
(36.3–––49.6) 0.783
Family History of Myocardial Infarction (MI) 5 (29.4 %) 10 (37.0 %)
0.847 2 (18.2 %) 5 (45.5 %) 0.36 2 (66.7 %) 4 (18.2 %) 0.261
Chest pain type 0.126 0.801 0.575
Atypical angina 11 (64.7 %) 10 (37.0 %) 5 (45.5 %) 4 (36.4 %) 2
(66.7 %) 13 (59.1 %)
Typical angina 2 (11.8 %) 10 (37.0 %) 5 (45.5 %) 5 (45.5 %) 0 (0 %) 4
(18.2 %)
Non-anginal 4 (23.5 %) 7 (25.9 %) 1 (9.1 %) 2 (18.2 %) 1 (33.3 %) 3
(13.6 %)
Female-specific risk factors
Menopause / 21 (77.8 %) / 11 (100 %) / 15 (68.2 %)
Polycystic ovary syndrome (PCOS) / 2 (7.4 %) / 2 (18.2 %) / 0 (0 %)
Pregnancy loss / 12 (44.4 %) / 7 (63.6 %) / 9 (40.9 %)
Physiological data
Coronary Flow Reserve (CFR) 3.62 (2.68–––3.88) 2.79 (2.51–––3.25) 0.085
2.51 (2.37–––2.91) 2.12 (1.63–––2.57) 0.173 1.84 (1.83–––1.93) 1.81
(1.40–––1.96) 0.359
Average heart rate, beats/min 93.3 (85.1–––100) 96.5 (84.8–––109) 0.448
88.5 (79.3–––95.8) 89.0 (86.0–––98.0) 0.39 113 (111–––115) 96.0
(90.0–––112) 0.19
Average systolic blood pressure, mmHg 128 (116–––139) 126 (113–––150)
0.95 128 (105–––137) 131 (128–––140) 0.39 175 (147–––178) 122
(109–––143) 0.138
Medications
Angiotensin-converting enzyme (ACE) inhibitor 9 (52.9 %) 13 (48.1 %) 1
4 (36.4 %) 5 (45.5 %) 1 3 (100 %) 8 (36.4 %) 0.143
Aspirin 6 (35.3 %) 7 (25.9 %) 0.746 8 (72.7 %) 9 (81.8 %) 1 2 (66.7 %)
9 (40.9 %) 0.823
Beta blocker 2 (11.8 %) 6 (22.2 %) 0.635 8 (72.7 %) 8 (72.7 %) 1 2
(66.7 %) 12 (54.5 %) 1
Calcium channel blocker (CCB) 2 (11.8 %) 7 (25.9 %) 0.453 5 (45.5 %) 3
(27.3 %) 0.658 0 (0 %) 10 (45.5 %) 0.379
Antiplatelet 0 (0 %) 0 (0 %) NA 4 (36.4 %) 5 (45.5 %) 1 0 (0 %) 0 (0 %)
NA
Diabetes medications 6 (35.3 %) 5 (18.5 %) 0.371 3 (27.3 %) 3 (27.3 %)
1 2 (66.7 %) 9 (40.9 %) 0.823
Lipid lowering drugs 6 (35.3 %) 11 (40.7 %) 0.965 9 (81.8 %) 11 (100 %)
0.458 1 (33.3 %) 12 (54.5 %) 0.941
[60]Open in a new tab
Continuous variables are shown as median (interquartile range), and
categorical variables are shown as n (%). P-values were derived from
Wilcoxon rank sum test and chi-squared test for continuous and
categorical variables respectively; CVA: Cerebral Vascular Accident;
PE: Pulmonary Embolism; DVT: Deep Vein Thrombosis; TIMI: Thrombolysis
in myocardial infarction; PCOS: Polycystic ovary syndrome; CFR:
Coronary Flow Reserve; ACE: Angiotensin-converting enzyme; CCB: Calcium
channel blocker;
Missing values are quantified as a percentage of the cohort totals. For
variables applicable to all participants: Educational Level (6.59%),
Chest pain type (2.20%), Coronary Flow Reserve (3.30%), Average Heart
Rate (8.79%), and Average Systolic Blood Pressure (5.49%). For
female-specific variables, percentages reflect missing data within the
female cohort subset: Menopause (1.67%) and Pregnancy Loss (8.33%).
3.2. Sex differences in protein biomarker expression
We assessed sex differences in plasma protein biomarker expression.
[61]Table 2 lists the 20 proteins with the smallest FDR in each group.
Whereas expression of reproductive proteins was the highest across all
three groups, CAD patients had higher upregulation of inflammatory
proteins (CXCL9, ICAM2, SELL, CD80, SLAMF7), while proteins from
multiple mechanisms were higher in CMD patients ranging from
inflammation (SCGB3A1), rho kinase activity (OPHN1), extracellular
matrix remodeling (MFAP5, XG, CDCP1, IGDCC4), to cell metabolism
(ANGPTL3). Results for all proteins are shown in [62]Supplemental Table
4.
Table 2.
Top 20 plasma protein biomarkers that are differentially expressed
between males and females in three groups.
Control
Biomarker Protein Name Scaled
[MATH: β :MATH]
SE FDR
INSL3 Insulin-like 3 −2.039 0.171 1.81E-25
SPINT3 Kunitz-type protease inhibitor 3 −1.985 0.319 3.04E-17
EDDM3B Epididymal secretory protein E3-beta −1.850 0.182 3.30E-12
KLK3 Prostate-specific antigen −1.853 0.308 3.02E-09
PZP Pregnancy zone protein 1.685 0.357 3.48E-06
FSHB Follitropin subunit beta 1.378 0.193 3.57E-04
XG Glycoprotein Xg 1.476 0.128 4.57E-04
ACRV1 Acrosomal protein SP-10 −1.465 0.317 5.08E-04
LEP Leptin 1.163 0.261 6.19E-04
TEX101 Testis-expressed protein 101 −1.390 0.259 9.61E-04
PROK1 Prokineticin-1 −1.146 0.231 1.19E-03
CALCA Calcitonin −1.176 0.238 1.59E-02
MYLPF Myosin regulatory light chain 2, skeletal muscle isoform −1.146
0.615 4.41E-02
GFAP Glial fibrillary acidic protein 1.000 0.191 4.41E-02
OBP2B Odorant-binding protein 2b −1.152 0.226 4.41E-02
SPESP1 Sperm equatorial segment protein 1 −1.231 0.247 5.08E-02
PAFAH2 Platelet-activating factor acetylhydrolase 2, cytoplasmic 1.094
0.268 5.40E-02
PSPN Persephin −1.099 0.393 7.93E-02
PTPRR Receptor-type tyrosine-protein phosphatase R 1.111 0.158 9.99E-02
CLMP CXADR-like membrane protein 0.916 0.090 1.76E-01
CAD
Biomarker Protein Name Scaled
[MATH: β :MATH]
SE FDR
INSL3 Insulin-like 3 −1.766 0.667 5.98E-03
FSHB Follitropin subunit beta 1.764 0.217 7.60E-03
EDDM3B Epididymal secretory protein E3-beta −1.435 0.339 4.26E-02
SPINT3 Kunitz-type protease inhibitor 3 −1.563 0.884 8.82E-02
KLK3 Prostate-specific antigen −1.382 0.800 9.10E-02
PTPRC Receptor-type tyrosine-protein phosphatase C 1.456 0.144 9.10E-02
CTSZ Cathepsin Z 1.148 0.168 9.10E-02
CXCL9 C-X-C motif chemokine 9 1.087 0.275 1.57E-01
BPIFB2 BPI fold-containing family B member 2 1.176 0.275 1.71E-01
ICAM2 Intercellular adhesion molecule 2 1.157 0.137 1.92E-01
SELL L-selectin 1.454 0.099 2.11E-01
SOX9 Transcription factor SOX-9 1.381 1.090 2.11E-01
CD80 T-lymphocyte activation antigen CD80 1.311 0.190 2.11E-01
TJP3 Tight junction protein ZO-3 1.242 0.264 2.11E-01
INHBB Inhibin beta B chain 1.293 0.348 2.17E-01
GGH Gamma-glutamyl hydrolase 1.194 0.125 2.17E-01
MMP7 Matrilysin 1.299 0.068 2.24E-01
SLAMF7 SLAM family member 7 1.447 0.319 2.24E-01
OSCAR Osteoclast-associated immunoglobulin-like receptor 1.121 0.150
2.29E-01
ENPP2 Ectonucleotide pyrophosphatase/phosphodiesterase family member 2
1.118 0.182 2.63E-01
CMD
Biomarker Protein Name Scaled
[MATH: β :MATH]
SE FDR
INSL3 Insulin-like 3 −2.797 0.311 1.72E-08
SPINT3 Kunitz-type protease inhibitor 3 −2.603 0.515 9.62E-06
EDDM3B Epididymal secretory protein E3-beta −2.854 0.301 2.43E-04
KLK3 Prostate-specific antigen −2.565 0.422 2.98E-03
SCGB3A1 Secretoglobin family 3A member 1 2.698 0.161 9.21E-03
HGFAC Hepatocyte growth factor activator 2.670 0.140 9.21E-03
SNCG Gamma-synuclein 1.302 0.278 6.16E-02
PZP Pregnancy zone protein 2.467 0.877 2.12E-01
FOXJ3 Forkhead box protein J3 −1.685 0.411 2.12E-01
OPHN1 Oligophrenin-1 −2.231 0.607 2.56E-01
ANGPTL3 Angiopoietin-related protein 3 2.094 0.274 2.95E-01
CDCP1 CUB domain-containing protein 1 −1.904 0.355 3.57E-01
IGDCC4 Immunoglobulin superfamily DCC subclass member 4 1.958 0.191
4.82E-01
CA9 Carbonic anhydrase 9 −2.085 0.365 4.92E-01
COL9A2 Collagen alpha-2(IX) chain 1.947 0.464 5.12E-01
MFAP5 Microfibrillar-associated protein 5 1.822 0.327 5.49E-01
PSAPL1 Proactivator polypeptide-like 1 1.629 0.387 6.58E-01
XG Glycoprotein Xg 1.593 0.294 6.58E-01
ATP5PO ATP synthase subunit O, mitochondrial −2.020 0.489 6.58E-01
SERPINF2 Alpha-2-antiplasmin 2.035 0.344 6.58E-01
[63]Open in a new tab
In the control group, 15 proteins were differentially expressed between
males and females (FDR < 0.05; [64]Fig. 1A). Among them, five proteins
had higher expression in females, with the largest difference observed
in PZP (β = 1.69), followed by XG, FSHB, LEP, and GFAP (β > 0.99).
These proteins are involved in the following biological functions:
pregnancy and fertility (PZP − Pregnancy zone protein, FSHB −
Follitropin subunit beta), blood homeostasis and immune processes (XG −
Xg glycoprotein), hormonal regulation (LEP − Leptin), and neurological
function (GFAP − Glial fibrillary acidic protein). The other 10
proteins had higher expression in males, with the largest difference
observed in INSL3 (β = -2.04), followed by SPINT3, KLK3, EDDM3B, ACRV1,
TEX101, CALCA, OBP2B, MYLPF, and PROK1 (β < -1.15). These proteins are
related to hormone response (INSL3 − Insulin-like 3 A chain, EDDM3B −
Epididymal secretory protein E3-beta, KLK3 − Prostate-specific antigen,
ACRV1 − Acrosomal vesicle protein 1, TEX101 − Testis-expressed protein
101, PROK1 − Prokineticin-1, CALCA − Calcitonin), enzymatic inhibition
(SPINT3 − Serine peptidase inhibitor, Kunitz type 3), muscle
contraction (MYLPF − Myosin light chain, phosphorylatable, fast
skeletal muscle), and olfactory processes (OBP2B − Odorant-binding
protein 2b).
Fig. 1.
[65]Fig. 1
[66]Open in a new tab
Differential protein expression between males and females in three
groups. (A-C) Volcano plots of relative protein expressions in the
control (A), CAD (B), and CMD (C) patients. Positive x-values represent
proteins that are higher in females (red points), and negative x-values
represent proteins that are higher in males (blue points). The dashed
line represents FDR < 0.05. (D) Venn diagram showing the intersection
of differentially expressed proteins in the control, CAD, and CMD
groups. Overlapping regions indicate proteins that show sex
differences, either upregulated in males or females, across groups.
Sex differences were also observed in the CVD phenotypes. In the CAD
group, three proteins were differentially expressed between males and
females (FDR < 0.05; [67]Fig. 1B). Among them, FSHB had higher
expression in females (β = 1.76), whereas INSL3 (β = -1.77) and EDDM3B
(β = -1.44) were higher in males. All three proteins are hormone
related and were also differentially expressed in the control group
([68]Fig. 1D). In the CMD group, six proteins were differentially
expressed between males and females (FDR < 0.05; [69]Fig. 1C), all with
at least a 2.5-SD difference. Among them, secretoglobin family 3A
member 1 (SCGB3A1; β = 2.70) and hepatocyte growth factor activator
short chain (HGFAC; β = 2.67) had higher expression in females. SCGB3A1
is involved in secretion processes and HGFAC is involved in
liver-associated growth modulation. The other four proteins had higher
expression in males, with the largest difference observed in EDDM3B
(β = -2.85), followed by INSL3, SPINT3, and KLK3 (β < -2.56). All four
proteins upregulated in males are hormone related and are also
differentially expressed in the control group ([70]Fig. 1D).
3.3. Pathway enrichment analyses
In the control group, 25 proteins had a P-value < 0.05 and |β| >1.
Among them, 10 were upregulated in females and 15 were upregulated in
males. No pathway was found to be enriched in these proteins.
The pathway enrichment analysis for CAD samples showed sex differences.
After excluding proteins that were also upregulated in females in the
control group ([71]Supplemental Fig. 1A), the 112 proteins upregulated
in females in the CAD group were enriched for pathways associated with
immune system processes, such as cell adhesion, immune cell
communication, and inflammation ([72]Fig. 2A). Multiple chemokines
(CCL14, CCL15, CCL16, CCL19, CCL20, CCL25, CCL7, CXCL9, XCL1), adhesion
molecules (ICAM1, ICAM2), and T-lymphocyte activation antigens (CD80,
CD86) were upregulated and enriched in the top pathways. These proteins
also highly interact with each other in the protein–protein
interactions (PPI) network. Moreover, four clusters were identified in
the PPI network ([73]Fig. 2C). Pathways related to immune response were
enriched in protein cluster 1. In cluster 2, proteins were enriched for
pathways related to immune response in addition to cytokine- and
chemokine-mediated signaling. In contrast, after excluding proteins
that were also upregulated in males in the control group
([74]Supplemental Fig. 1B), the 11 proteins upregulated in males were
enriched for pathways involved in the endothelial regulation of blood
flow ([75]Fig. 2B). However, due to the limited number of proteins
upregulated in males, there were not enough interactions to construct a
PPI network for this group.
Fig. 2.
[76]Fig. 2
[77]Open in a new tab
Pathway overrepresentation in sex-specific protein upregulation in the
CAD group. (A) Overrepresented pathways upregulated in females among
CAD patients. The red line represents p < 0.05. (B) Overrepresented
pathways upregulated in males among CAD patients. The red line
represents p < 0.05. (C) Protein-protein interaction network for
biomarkers that are higher in females with CAD. Red represents protein
cluster 1, yellow represents protein cluster 2, blue represents protein
cluster 3, and green represents protein cluster 4.
For the CMD group, in the 57 proteins upregulated in females, after
excluding proteins that were also upregulated in females in the control
group ([78]Supplemental Fig. 1A), there was a substantial emphasis on
lipid and glucose metabolism, as seen in the enriched pathways related
to liver X receptor/ farnesoid X receptor/ retinoid X receptor
(LXR/FXR/RXR) activation, Clathrin-mediated endocytosis signaling,
3β-hydroxysterol Δ24-reductase (DHCR24) signaling, and atherosclerosis
signaling ([79]Fig. 3A). The upregulated proteins in these pathways,
especially apolipoproteins (APOA1 and APOD), suggest an active
regulatory mechanism related to cholesterol transport and synthesis. In
contrast, after excluding proteins that were also upregulated in males
in the control group ([80]Supplemental Fig. 1B), the 59 proteins
upregulated in males were enriched for pathways involved in
angiogenesis and some lipid pathways ([81]Fig. 3B).
Fig. 3.
[82]Fig. 3
[83]Open in a new tab
Pathway overrepresentation in sex-specific protein upregulation in the
CMD group. (A) Overrepresented pathways upregulated in females among
CMD patients. The red line represents p < 0.05. (B) Overrepresented
pathways upregulated in males among CMD patients. The red line
represents p < 0.05.
3.4. Weighted correlation network analysis
Protein correlation network was constructed to identify modules of
highly correlated proteins using WGCNA. The identified protein modules
and their correlation with sex were assessed in both CAD and CMD
groups. Eleven protein modules were identified in the CAD group, among
which the green module ([84]Supplemental Fig. 2A) was positively
correlated with sex (r = 0.49, P = 0.02). The proteins in the green
module were significantly enriched for pathways related to immune
response and inflammation, including granulocyte and agranulocyte
adhesion and diapedesis, the complement system, phagosome formation
([85]Supplemental Fig. 2B). In contrast, no module was significantly
correlated with sex in the CMD group.
4. Discussion
In a novel proof-of-concept study, we used an unbiased, discovery-based
plasma proteomics approach to compare sex differences by CVD phenotype
for acute ischemia. We noted important differential protein expression
in reproductive hormones between male and female patients across all
groups, as well as multiple biological pathways that shed light onto
sexual divergence in CVD development. Among CAD patients, females
showed upregulation of inflammatory and immune pathways while males
showed upregulation of proteins related to vascular flow. Among CMD
patients, females had upregulation of glucose and lipid pathways while
males had upregulation of angiogenesis and endothelial flow. To our
knowledge, this is the first study to identify sex differences in
proteomics of acutely symptomatic patients being assessed for cardiac
ischemia in the ED.
While proteomics, particularly untargeted genome-wide approaches, is
well-established in chronic cardiovascular research, its application in
acute ischemic events remains limited. Previous work, such as Mazidi et
al,[86][24] demonstrated the integration of proteomics and genetic data
in ischemic heart disease, these studies focused primarily on chronic
stable state of ischemia and population-level analyses. We extended
this approach to an acute clinical setting, specifically examining
sex-based proteomic differences among symptomatic patients being
assessed for acute cardiac ischemia in the ED and builds upon existing
proteomic research by investigating a broader panel of
proteins.[87][7], [88][9] We identified a much larger number of
upregulated proteins among female patients than among males with CVD,
confirming the findings of a previous study.[89][25] Experimental rat
models of ischemia shed light into the underlying mechanism of this
observation.[90][26] Compared with males, female rats after ischemia
show significantly higher upregulation of proteins linked with
extracellular matrix remodeling and inflammation – an estrogen led
effect leading to faster replenishment of contractile structures and
more rapid healing.[91][26] This finding aligns with De Bakker et
al.[92][27], who reported extracellular matrix organization as dominant
in females and regulation of apoptotic processes as prevalent in males
with heart failure with reduced ejection fraction. These consistent
observations suggest shared biological mechanisms involving tissue
remodeling, immune regulation, and vascular maintenance across
different cardiovascular conditions.
We also identified upregulated immune and cytokine-mediated proteins in
CAD patients most pronounced among females. Our findings are consistent
with the functional pathways noted in the UK Biobank registry as well
as the Framingham Heart Registry.[93][7], [94][28] Although the
specific proteins in the UK study were different from those found in
our study, reflecting the dynamic nature of a changing proteome in
acute versus chronic ischemic states, we found similar biological
pathways involved in CAD among females. Immune activation has been
shown to accelerate plaque development and atherosclerosis.[95][29] In
females, immune pathways appear to affect plaque through dyslipidemia
pathways[96][30], effects that could be related to hormonal changes.
For instance, we observed elevated FSHB in females with CAD and among
controls, which could reflect the high proportion of menopausal females
in both groups. FSH has been implicated in the development of CAD
through its role in the development of dyslipidemia in perimenopausal
and menopausal females, even with similar estrogen levels.[97][25] In
fact, the blockade of FSH signaling helps to lower serum cholesterol
levels by 30 %, potentially offering a sex-specific therapeutic target
after menopause.[98][31].
We also observed upregulation of proteins in males with CAD, primarily
in relation to vascular flow. Arginine and endothelin (ET) pathways,
unregulated in our study, are related to vascular flow and blood
pressure regulation and might explain increased risk for CAD in younger
males.[99][32] Sex differences have been noted in the endothelin (ET)
pathway of the vasculature, heart, and kidneys of humans and
experimental animals with higher levels noted in males compared to
females.[100][33] In addition to the effects of sex chromosomes,
testosterone positively modulates ET-1 expression, whereas estrogen
downregulates it.[101][34] Our finding corroborates the multiple shared
biological pathways observed with CAD in both males and females in the
UK biobank, with male-specific upregulation of vascular flow and
hormonal response (insulin-like growth factor response)
pathways.[102][25].
Another key finding is that in CMD, proteins with a primarily
sex-linked heterogeneous profile of mechanisms were upregulated. In
females, we observed upregulation in glucose and LXR pathways, which
maintain cholesterol homeostasis for protein targets, such as APOA1 and
APOD. Postmenopausal females, due to declining estrogen production,
experience disruptions in their evolutionarily optimized lipid and
cholesterol metabolism, increasing their susceptibility to
obesity-related diseases, including CVD.[103][35], [104][36] Two thirds
of our female CMD group was postmenopausal, and majority morbidly obese
(median BMI of 40.9), hence upregulation of these pathways underscores
the well-known connection among menopausal status, obesity, and
CMD.[105][37] Schindler[106][38] enriches this argument by explaining
the role of different adipose tissue types in obesity and suggests that
increases in visceral adipose tissue, as opposed to subcutaneous fat,
lead to greater metabolic abnormalities and CMD. Abnormal lipid
metabolism may cause endothelial dysfunction in CMD patients.[107][39]
This effect is particularly relevant to postmenopausal females, who
tend to lose the protective benefits of subcutaneous fat accumulation
with age, thus potentially increasing CMD risks. The Mayo Clinic
Proteomic Markers of Arteriosclerosis Study reported sex differences
with the HDL component of ApoA-1 to be higher in females with
CMD.[108][40] Obesity status has also been associated with higher
levels of free fatty acids in circulating blood, likely due to
increased sympathetic drive, causing activation of lipolysis in adipose
tissue.[109][41] High levels of free fatty acid reduce cardiac glucose
utilization, leading to increased oxygen consumption and making the
heart more susceptible to ischemic events.
Our findings regarding the upregulation of the LXR/RXR pathway align
with prior evidence of sex-dimorphic regulation in lipid metabolism, as
described by Rando and Wahli.[110][42] Nuclear receptors such as PPARα,
LXR, and ERα modulate lipid and glucose metabolism through
hormone-sensitive regulatory networks, with females exhibiting more
robust compensatory mechanisms. This overlap suggests that the observed
upregulation of LXR/RXR pathways in CMD females could reflect both
intrinsic sex-based differences and disease-specific metabolic
disruption. This convergence emphasizes the complex interplay between
biological sex and CVD pathophysiology, supporting further
investigation into sex-specific regulatory networks.
Two additional pathways in relation to CMD deserve attention. First, we
noted a differential upregulation of HGFAC in females with CMD as
compared to controls and CAD patients. Hepatocyte growth factor has
long been identified as a potent mitogen for hepatocytes promoting
angiogenic growth. Furthermore, high HCF levels counter response to
endothelial dysfunction, offering both a prognostic and therapeutic
target.[111][43], [112][44] Second, SCGB3A1, a protein in females with
CMD, is implicated in the immune response to ischemia. Interestingly,
nocarandil, a smooth muscle vasodilator, suppresses SCGB3A1 in acute
ischemia,[113][45] offering a potential biological pathway for improved
microvascular flow. Together, these findings support further research
to identify sex-specific molecular targets by CVD phenotype as an
important step in advancing personalized CVD care.
The rationale for using WGCNA, rather than individual protein-level
statistical tests, is its ability to uncover biologically meaningful
modules of co-expressed proteins that function collectively within the
same biological pathways. This network-based approach can reveal hidden
relationships among proteins that may be missed by conventional
univariate statistical methods, enhancing interpretation of
sex-specific molecular mechanisms. Importantly, what we found through
WGCNA aligns with the findings from individual protein-level analysis.
Specifically, proteins involved in immune and inflammatory pathways
were upregulated in females with CAD in both approaches. This
convergence supports the robustness of our findings, it suggests that
proteins within the same biological processes are not only individually
significant but also co-regulated, strengthening the case for these
pathways’ involvement in sex-specific CAD mechanisms.
5. Limitations
This was a single site study limited to a single time point assessment.
As such, the results show only an association without evidence of
causality. We also did not find diagnostic differentiation of proteomic
profile between controls and CVD phenotype. This could also be due to
the heterogeneity of mechanisms among controls who were symptomatic
patients with overlapping CVD comorbidities as opposed to healthy
individuals. However, consistent with prior investigations that
established sex differences in proteome profiles among healthy
individuals, we identified differential expression of proteins in males
and females in our controls as well as among diseased groups, making a
sex-specific comparison valid.[114][6].
The limited sample size in this single-site study restricted the
statistical power to detect differentially expressed proteins. For
example, under the threshold FDR < 0.05, we identified fewer
differentially expressed proteins in CVD phenotypes compared to
controls, likely due to the control group being about twice as large as
the CAD or CMD groups. However, in the pathway analysis using the
criteria P-value < 0.05 and |β| >1, we observed a greater number of
proteins in CAD and CMD than in the control group, suggesting that
sex-specific proteomic differences can persist or even become more
pronounced in the presence of disease. Additionally, we observed
differences in medication use, such as antiplatelets and aspirin,
across groups. Differences in medications could influence protein
expression and pathways. While we did not find a statistically
significant difference in medication use between males and females, we
acknowledge that the overall differences in medication use across
groups may influence the proteomic comparisons. Also, the inclusion or
exclusion of posttranslational modifications (PTMs) of proteins
presents a methodological challenge. PTMs can substantially alter
protein function and interactions, impacting the interpretability of
proteomic data in the context of CVD phenotypes. Our analysis did not
account for PTMs, which may affect the comprehensiveness and accuracy
of our findings regarding protein function and regulation.
We used unadjusted p-values for pathway enrichment analysis in
exploratory proteomics to increase sensitivity in detecting pathways of
interest, though this might increase the risk of false positives. While
differentially expressed proteins were identified using FDR control,
the same adjustment was not applied in pathway analysis to avoid
missing potentially important biological pathways. Future studies could
incorporate adjusted p-values or validate key findings in independent
samples. Finally, we acknowledge that tissue-based proteomics is more
sensitive in capturing local physiology of myocardial changes or
vascular function related to cardiac ischemia. While tissue samples
provide greater insights, they are invasive and difficult to obtain
especially in less sick patients limiting its generalizability. A study
of circulating protein biomarkers in patients with ischemia as we and
others have done,[115][10], [116][11] should be considered hypothesis
generating for insights into the biological pathways for ischemia. It
also provides a relative distribution of proteins and thus
interpretation of pathways should consider both upregulation as well as
downregulation of proteins relative to sexes. Further investigation of
proteomics in cardiac tissue is necessary to gain deeper mechanistic
insight into CVD phenotypes as well as absolute differences.
6. Conclusions
In this exploratory analysis, we demonstrated that sex differences
exist in the proteome of CVD phenotypes. We revealed sex differences in
proteomics and biological pathways for acutely symptomatic CMD and CAD,
illustrating they are distinct diseases with potential for different
therapeutic targets. Further validation studies that incorporate omics
into clinical pathways would help advance a precision-based approach
for personalized CVD care.
CRediT authorship contribution statement
Yihan Liu: Writing – review & editing, Writing – original draft, Formal
analysis, Data curation, Conceptualization. Zuoheng Wang: Writing –
review & editing, Formal analysis. Sean P. Collins: Writing – review &
editing, Conceptualization. Jeffery Testani: Writing – review &
editing, Conceptualization. Basmah Safdar: Writing – review & editing,
Writing – original draft, Visualization, Validation, Supervision,
Resources, Project administration, Methodology, Investigation, Funding
acquisition, Formal analysis, Data curation, Conceptualization.
Funding
Dr. Safdar is supported by NIH grants (1OTHL56812-01; U24NS129500,
1OT2HL162110-01), a CDC grant (75D30121F0002), as well as institutional
grants from Comprehensive Research Associates. Dr. Wang is supported by
NIH grant (R01LM014087). This publication was made possible by CTSA
Grant Number UL1 TR001863 or KL2 TR001862 or TL1 TR001864 (as
appropriate) from the National Center for Advancing Translational
Science (NCATS), a component of the National Institutes of Health
(NIH). Its contents are solely the responsibility of the authors and do
not necessarily represent the official view of NIH.
Declaration of competing interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to
influence the work reported in this paper.
Acknowledgement