Investigating the extremes of the continuum of paracrine functions in CD34⁻/CD31⁺ CACs across diverse populations

S100A8 and S100A9 proteins in concentrations secreted by CD34^–/CD31⁺ circulating angiogenic cells (CACs) with impaired function reduce endothelial cell capillary-like network formation. These effects appear to be mediated by Toll-like receptor 4 and are absent with S100A8 and S100A9 in concentrations secreted by healthy CD34^–/CD31⁺ CACs.

Abstract

Paracrine function of circulating angiogenic cells (CACs) is thought to contribute to vascular maintenance. We previously identified S100A8 and S100A9 secreted from physically inactive individuals’ CD34^–/CD31⁺ CACs as negative regulators of capillary-like network formation. The purpose of this study was to investigate further the extremes of the continuum of CAC paracrine actions using two distinctly different groups representing “healthy” and “impaired” CAC function. We aimed to determine how capillary-like network formation in human umbilical vein endothelial cells (HUVECs) is affected by S100A8 and S100A9 in concentrations secreted by CACs from different ends of the health spectrum. CD34^–/CD31⁺ CACs were isolated and cultured from 10 impaired function individuals defined as older (50–89 yr), non-ST-elevation myocardial infarction patients and 10 healthy individuals defined as younger (18–35 yr), healthy individuals, and conditioned media (CM) was generated. CM from the impaired function group's CACs significantly diminished network formation compared with CM from the healthy group (P < 0.05). We identified elevations in S100A8, S100A9, and S100A8/A9 in the CM from the impaired function group (P < 0.05). Pretreatment of HUVECs with inhibitors to a known S100A8 and S100A9 receptor, Toll-like receptor 4 (TLR4), but not receptor for advanced glycation end products, improved HUVEC network formation (P < 0.05) compared with CM alone in the impaired function conditions. Exposure of HUVECs to the TLR4 signaling inhibitor also blocked recombinant S100A8- and S100A9-mediated reductions in network formation. Collectively, the results suggest that the mechanisms behind impaired CAC CD34^–/CD31⁺ CM-mediated reductions in capillary-like network formation involve secretion of S100A8 and S100A9 and binding of these proteins to TLR4 receptors on HUVECs.

NEW & NOTEWORTHY S100A8 and S100A9 proteins in concentrations secreted by CD34^–/CD31⁺ circulating angiogenic cells (CACs) with impaired function reduce endothelial cell capillary-like network formation. These effects appear to be mediated by Toll-like receptor 4 and are absent with S100A8 and S100A9 in concentrations secreted by healthy CD34^–/CD31⁺ CACs.

cardiovascular disease (CVD) is the leading cause of death in developed countries (17). Numerous reports have found an inverse relationship between CVD and the number and function of circulating angiogenic cells (CACs), establishing CACs as novel CVD risk factors (4, 13, 14, 25). CAC is a broad term used to describe various subpopulations of circulating cells with angiogenic potential that are involved in the repair and maintenance of the vascular endothelium. CACs are identified and categorized based on the expression of known cell surface markers. These markers have included stem/progenitor cell markers, such as CD34 and the more immature CD133, as well as surface markers found on circulating endothelial cells, such as VEGF receptor 2 and CD31 (also known as platelet endothelial cell adhesion molecule-1). CD31⁺ CACs, in particular, have been found to contribute to angiogenesis and vasculogenesis to a comparable, if not greater, extent than the more commonly studied progenitor cell lines (23, 24). Indeed, treatment with CD31⁺ CACs lacking the CD34 marker (CD34^–/CD31⁺) improved mouse hindlimb ischemia to the same degree as CD34⁺ cells (24), making these more abundant CAC subtype targets for further investigation.

Initial studies in this field have focused on determining the frequency of CAC subtypes in peripheral blood, but more recent work has concentrated on understanding the function of specific CAC subtypes. Indeed, in addition to being lower in number, CACs from CVD patients have diminished functional capacity compared with those from their healthy counterparts (13, 22). Specifically, Kim et al. (22) demonstrated that CD31⁺ CACs from coronary artery disease (CAD) patients exhibited lower cell migration rates, lower adhesion, and greater rates of apoptosis compared with the CD31⁺ CACs from healthy controls. Additionally, Kushner et al. (26) reported a progressive, age-related decline in CD31⁺ T cell number and migration rates and greater rates of apoptosis in CACs from middle-aged and older adults compared with young adults. Finally, physical inactivity has been associated with impaired paracrine function (27) and altered redox balance (20) in CACs from younger adults compared with their physically fit counterparts. Collectively, these findings highlight the importance of considering the functional capacity of CACs, especially in circumstances associated with greater cardiovascular (CV) risk (i.e., advanced age, physical inactivity, and overt CVD).

Whereas some CACs may incorporate directly into the endothelial wall (2), it is now generally accepted that CACs primarily function through paracrine mechanisms, whereby they secrete factors that act on the pre-existing endothelium (23, 36). Despite this, paracrine functions of CACs remain poorly understood. As paracrine factors are believed to play a pivotal role in the overall mechanisms through which CACs contribute to vascular homeostasis, the characterization of secreted factors from CACs of populations exhibiting various health disparities and their role in vascular maintenance have become critical gaps in the literature.

With the use of a discovery-based approach, we recently reported that S100A8 and S100A9 proteins were secreted from cultured CD34^–/CD31⁺ CACs. We further found that elevations in S100A8 and S100A9, due to physical inactivity, were, at least partially, responsible for reductions in capillary-like network formation (27). S100A8 and S100A9 are members of the S100 family of calcium-binding proteins and are best known for their roles in inflammatory diseases (3, 7, 8, 33). Although circulating plasma concentrations of S100A8 and S100A9 are elevated in chronic conditions, the molecular mechanisms associated with the direct effects of S100A8 and S100A9 exposure on vascular function and angiogenesis are not completely understood, with most reports restricted to statistical associations (7, 8). Additionally, to the best of our knowledge, we are the only laboratory to report anti-angiogenic actions of S100A8 and S100A9 secreted specifically by CD34^–/CD31⁺ CACs, a cell type considered for autologous cell therapies to treat ischemic diseases due to their reputed proangiogenic properties (24). Thus further insight into the mechanisms through which CD34^–/CD31⁺ paracrine actions on endothelial cells take place—specifically in populations that would be considered candidates for potential therapies—is necessary to explain these seemingly contradictory findings. The purpose of this study was to examine the mechanisms behind paracrine-mediated differences in human umbilical vein endothelial cell (HUVEC) capillary-like network formation caused by secreted factors derived from CD34^–/CD31⁺ CACs, specifically S100A8 and S100A9, using two distinctly different populations representing the far ends of the spectrum of CAC functions: “healthy” (young, healthy, active) and “impaired” (older, inactive, CVD).

METHODS

Ethical approval.

The University of Maryland College Park and University of Maryland Baltimore Institutional Review Boards approved all study procedures, and subjects provided written, informed consent. The study procedures conformed to the standards set by the Declaration of Helsinki.

Screening.

Non-ST segment elevation myocardial infarction (NSTEMI) patients were selected as our older, inactive CVD group, representing impaired function. This group was selected over STEMI patients, due to the timing requirement needed to obtain consent before catheterization. Potential NSTEMI patients were recruited from those admitted to the Baltimore Veterans Affairs Medical Center. The patients participating in this study were men, 50–89 yr of age, with a body mass index (BMI) between 18 and 40 kg/m², a history of or current symptomatic CAD, and a recent, uncomplicated myocardial infarction without ST segment elevation. Participants were not excluded based on race, ethnicity, or medications, due to their necessity in this population. Medications that were currently being taken by the NSTEMI patients included blood pressure-lowering medications (100%), statins (54%), diabetes medication (54%), acid-reflux medication (31%), noninflammatory doses of aspirin (85%), steroids (31%), nitrates (15%), diuretics (23%), bronchodilators (15%), blood thinners (15%), anti-depressants (23%), and anti-seizure medications (15%).

Physically fit, nonsmoking men, 18–39 yr of age, with no history of CVD or metabolic disease, were selected as our group representing healthy function. The purpose of selecting this distinctly different group was to establish a cohort of subjects with large, measurable differences in function, which we hypothesized would allow us to determine mechanistically if functional differences exist in CACs from our impaired group. Potential subjects were initially screened by telephone or e-mail and reported to the laboratory following an overnight fast for a screening visit to verify eligibility. Fitness was determined based on their reported physical activity over the last 5 yr and confirmed by maximal oxygen consumption testing. Specifically, the healthy group (n = 10) reported performing >4 h/wk of moderate- to high-intensity endurance exercise. Exclusion criteria for the healthy group were systolic blood pressure ≥ 130 mmHg, diastolic blood pressure ≥ 90 mmHg, serum total cholesterol ≥ 200 mg/dl, LDL–cholesterol ≥ 130 mg/dl, HDL–cholesterol ≤ 35 mg/dl, or fasting glucose ≥ 100 mg/dl.

Blood sampling.

For the healthy group, a screening blood sample was obtained for assessment of fasting serum triglyceride, lipoprotein lipids, and glucose levels (Quest Diagnostics, Baltimore, MD). Height, weight, seated blood pressure, and BMI were measured, and body composition was assessed using the 7-site skinfold procedure (19). On the day of blood sampling for CACs, the healthy subjects reported to the laboratory in the morning after an overnight (~12 h) fast. A sample of 50 ml blood was drawn using EDTA tubes (Becton Dickinson, Franklin Lakes, NJ) for isolation of CD34^–/CD31⁺ CACs and plasma. Blood sampling for the impaired function group was performed at the Baltimore Veterans Affairs Medical Center. Briefly, 50 ml blood samples were obtained in EDTA tubes between 24 and 72 h after presentation to the Emergency Room before cardiac catheterization procedures. Patients fasted overnight before blood sampling. Samples were then immediately transported to the University of Maryland College Park for processing. Samples from both groups were handled in the same fashion before and during CAC isolation.

Immunomagnetic cell separation.

Peripheral blood mononuclear cells were isolated from the venous blood samples using density gradient centrifugation (Ficoll; GE Healthcare, Pittsburgh, PA). The CD34⁺ fraction was purified using multiple rounds of immunomagnetic cell separation, according to the manufacturer’s instructions (EasySep immunomagnetic cell separation kits; Stemcell Technologies, Vancouver, BC, Canada), using an antibody specific for CD34 (Stemcell Technologies). CD31⁺ cells were selected from the CD34⁻ fraction of cells and purified (hereby referred to as CD34^–/CD31⁺), as described above, using an antibody specific for CD31 (BD Biosciences, San Jose, CA). With the use of flow cytometry analysis, our laboratory reports a purity of ~60% for our CD34^–/CD31⁺ CACs after immunomagnetic separation (Fig. 1, A and B). We have previously shown that our cell purity after immunomagnetic isolation using similar methods is equivalent to and in some cases, up to 30% greater than other published results also using nonmobilized peripheral blood (27).

Fig. 1.

Flow cytometry profiles representing FITC-conjugated CD34/phycoerythrin (PE)-conjugated CD31 cells within total peripheral blood mononuclear cells (PBMCs; A), freshly isolated, selected CD34⁻/CD31⁺ PBMCs using immunomagnetic cell separation procedures (B), and selected CD34⁻/CD31⁺ PBMCs after 48 h of culture (C). Multiple trials have resulted in an isolation purity of ~60%. FITC- and PE-A, FITC- and PE-area.

CAC culture and conditioned media.

CD34^–/CD31⁺ CACs were resuspended in unsupplemented endothelial growth medium free of growth factors or serum [endothelial basal medium 2 (EBM-2); Lonza, Basel, Switzerland] with 1% antibiotic–antimycotic (Thermo Fisher Scientific, Waltham, MA), each at a density of 300,000 cells/well in a 96-well plate, as determined by an average of multiple hemocytometer counts. Cultures were maintained for 48 h in a humidified incubator at 37°C and 5% CO₂. Flow cytometry analyses indicate that the cells generally retain their CD34^–/CD31⁺ markers after 48 h of culture (Fig. 1C). After incubation, the conditioned medium (CM) from all wells was withdrawn, combined, and clarified by spinning at 2,500 g for 20 min to remove cells and debris from the medium. CM, not immediately used for experiments, was flash frozen and stored at −80°C until further analyses.

Capillary-like network formation assay.

The capillary-like network formation assay (1, 9, 15, 30) was performed, as previously described (27), to assess the pro- or anti-angiogenic effects of CD34^–/CD31⁺ CM from healthy and impaired function individuals. Briefly, culture plates were coated with BD Matrigel Growth Factor Reduced (BD Biosciences), and the Matrigel was left to solidify for 30 min at 37°C and 5% CO₂. Each condition was performed in duplicate. In a 96-well plate, each well contained 20,000 HUVECs and either EBM-2 free of growth factors and serum or CM from CD34^–/CD31⁺ CACs. For some conditions, HUVECs were pretreated with a Toll-like receptor 4 (TLR4) signaling inhibitor (500 nM; TAK-242) and/or a receptor for advanced glycation end products (RAGE) antagonist (1 μM; FPS-ZM1) for 10 min before exposure to CM. Optimal concentrations of each inhibitor were determined through pilot experiments in our laboratory using ligands for each receptor and a range of concentrations of each drug treatment based on manufacturer recommendations, as well as previous studies using these inhibitors for treatment of HUVECs. Additional wells were prepared with a similar amount of fresh basal media (EBM-2) as a positive control and DMSO vehicle controls. The average HUVEC passage used in the angiogenesis assay for healthy and impaired groups was 4.1 ± 0.4 and 4.2 ± 0.3, respectively (not significant). HUVEC cultures were then visualized 16 h after seeding under a light microscope at 5× magnification, and five images were photographed from standardized locations within each well. These images were then coded and blindly assessed for HUVEC network length and number of nodes (defined as a connecting point between network segments) using the ImageJ Angiogenesis Analyzer software program (National Institutes of Health, Bethesda, MD) (6). Results are presented as each condition normalized to the basal condition to control for daily variability in HUVEC growth and passage number within each assay.

Mass spectrometry, Western blot analysis, and S100A8 and S100A9 expression.

Label-free proteomics analysis was used to compare proteins present in the CD34^–/CD31⁺ CM from the healthy and impaired function groups. Proteins with substantially different expression levels were identified using spectral counting for n = 3/group (28). Subsequently, these findings were quantitatively confirmed using immunoblotting analyses on n = 10 samples/group. Briefly, media were concentrated using Amicon Ultra-0.5 3 kD centrifugal filter devices (EMD Millipore, Billerica, MA). The soluble proteins in the supernatant were subjected to digestion with Trypsin Lys-C (Promega, Madison, WI). Samples were analyzed using a Fusion Lumos Tribrid mass spectrometer (Thermo Fisher Scientific) interfaced to a U3000 nano HPLC system. The resulting data were searched against the UniProtKB human protein database (2014.12.19) using Sequest HT and Mascot search engines through Proteome Discoverer (c1.4; Thermo Fisher Scientific). Search results were further compiled with Scaffold software (Proteome Software, Portland, OR) to apply strict parsimony and for spectral counting analysis. Western blot analyses were used to assess differences in two identified S100 proteins with specific antibodies for S100A8 (R&D Systems, Minneapolis, MN) and S100A9 (Santa Cruz Biotechnology, Dallas, TX) using our complete cohort of samples (n = 10/group) as done previously (27). Membranes were washed and then incubated with horseradish peroxidase-conjugated anti-mouse IgG (1:1,000; Cell Signaling Technology, Danvers, MA) or horseradish peroxidase-conjugated anti-goat IgG (1:5,000; Novus Biologicals, Littleton, CO) secondary antibodies. Blots were developed using SuperSignal enhanced chemiluminescent reagents (Thermo Fisher Scientific), and bands were visualized and quantified using ChemiDoc Imaging System and Software (Bio-Rad Laboratories, Hercules, CA). Values were normalized to the 300,000 cells/well used to generate the CM. CM calprotectin levels (S100A8/S100A9 heterodimer complex) were measured using the Legend Max ELISA Kit (BioLegend, San Diego, CA). This kit is designed to detect the heterodimer complex using a capture antibody that recognizes an epitope present on the heterodimer complex but that is not present on either of the monomers. The average intra-assay coefficient of variation was 7.4%. To avoid inter-assay variability, all samples were assayed on the same plate. This assay has a sensitivity of 0.62 ± 0.34 ng/ml.

Assessment of gene expression by RT-PCR.

Total RNA was extracted from freshly isolated CACs using the TRIzol Reagent, quantified using a spectrophotometer (Synergy H1 Hybrid Reader; BioTek, Winooski, VT) and reverse transcribed to cDNA (Thermo Fisher Scientific). Quantitative RT-PCR was performed using the Applied Biosystems 7300 Real Time PCR System (Thermo Fisher Scientific). Primer assays were purchased from Integrated DNA Technologies (Coralville, IA), and optimal concentration for efficacy of >90% was determined. Primer sequences are as follows: S100A8 (forward: 5′-CCCATCTTTATCACCAGAATGAG-3′, reverse: 5′-CCGAGTGTCCTCAGTATATCAG-3′); S100A9 (forward: 5′-CCTCCATGATGTGTTCTATGACC-3′, reverse: 5′-CAACACCTTCCACCAATACTCT-5′); GAPDH (forward: 5′-TGTAGTTGAGGTCAATGAAGGG-3′, reverse: 5′-ACATCGCTCAGACACCATG-3′). Each reaction was performed in duplicate on a 96-well plate and contained iTaq Universal Probes Supermix (Bio-Rad Laboratories), respective primer probe, and the cDNA template. The PCR conditions used were as follows: 95°C for 3 min, followed by 50 cycles of 95°C for 15 s and 60°C for 45 s. mRNA expression values are presented as 2^−ΔCT, where ΔC_T is the cycle threshold of the target gene minus GAPDH control for each condition. GAPDH primers were used as a control gene, and GAPDH C_T were not different across time in the present study.

Recombinant S100A8 and S100A9 HUVEC treatment.

To confirm the direct effects of S100A8 and S100A9 on HUVEC network formation, the concentrations and proportions estimated in the CM of the healthy and impaired function groups’ CD34^–/CD31⁺ CACs were used in a capillary-like network formation assay. With the use of recombinant human (rh)S100A8/A9 (Thermo Fisher Scientific), a standard curve was established using immunoblotting to allow us to estimate the concentration of each protein to which the HUVECs were exposed in a subset of CM samples from each group (n = 3/group). Based on these estimations, 2.9 μg/ml rhS100A8 (ProSpec, East Brunswick, NJ) and 0.8 μg/ml rhS100A9 (Thermo Fisher Scientific) were used to simulate the healthy conditions, and 6.7 μg/ml rhS100A8 and 1.4 μg/ml rhS100A9 were used to match the impaired function conditions. These proteins were added to a HUVEC-based, capillary-like network formation assay and compared with a positive control prepared with EBM-2 and a vehicle control.

To confirm the role of TLR4 in the inhibitory actions of these proteins on HUVECs, a condition, in which the HUVECs were pretreated with 500 nM TAK-242 before exposure to concentrations of S100A8 and S100A9 estimated to be present in the impaired function group's CM, was also included. In these experiments, each condition was assessed in samples collected from n = 6 independent cell culture wells from multiple culture plates collected on different days to confirm replication of our findings. All experiments were conducted on HUVECs from the same passage number (P4) for each condition described above.

Plasma S100A8/A9 ELISA.

Plasma calprotectin levels were measured using an ELISA kit (BMA Biomedicals, Augst, Switzerland). This kit is designed to detect the heterodimer complex using a capture antibody that recognizes an epitope present on the heterodimer complex but that is not present on either of the monomers. The average intra-assay coefficient of variation was 6.9%, and all samples were assayed on the same plate to avoid inter-assay variability. This assay has a sensitivity of ~1 ng/ml.

Statistics.

Statistical analyses were completed using IBM SPSS Statistics 21 (IBM, Armonk, NY). Assumptions of homoscedasticity and normality were verified for all outcome measures. Capillary-like network formation data were analyzed using ANOVA with pairwise comparisons where appropriate. Independent sample t-tests were used for Western blot, ELISA, and quantitative PCR analyses. Statistical significance was accepted at P ≤ 0.05. Values are expressed as means ± SE.

RESULTS

Subject characteristics.

Subject characteristics can be found in Table 1. The CVD group was relatively lean with well-controlled blood pressure and plasma lipoprotein-lipid profiles.

Table 1.

Subject characteristics

	Healthy, n = 10	CVD, n = 10
Age, yr	27 ± 2	72 ± 3*
BMI, kg/m2	23 ± 0.6	27 ± 2*
Body fat, %	9.5 ± 1
SBP, mmHg	127 ± 1.5	128 ± 5
DBP, mmHg	70 ± 3	70 ± 3
MAP, mmHg	89 ± 2	89 ± 3
Glucose, mg/dl	79 ± 3	114 ± 11*
Cholesterol, mg/dl	170 ± 11	177 ± 23
HDL-C, mg/dl	66 ± 5	42 ± 3*
LDL-C, mg/dl	88 ± 7	101 ± 15
VLDL-C, mg/dl	16 ± 5
TC/HDL	2.6 ± 0.2	4.2 ± 0.5*
LDL/HDL	1.4 ± 0.1	2.6 ± 0.4*
Triglycerides, mg/dl	53 ± 7	135 ± 29*

SBP, systolic blood pressure; DBP, diastolic blood pressure; MAP, mean arterial pressure; LDL- and HDL-C, LDL- and HDL-cholesterol; TC, total cholesterol.

^*Significantly different from healthy group.

Capillary-like network formation with CM.

With the confirmation of the impaired function of CACs from the CVD group, CM, derived from these CACs, resulted in 9% lower total network length (P < 0.05; Fig. 2A) and a tendency for 18% fewer nodes (P = 0.08; Fig. 2B) compared with CM from healthy individuals’ CACs. Representative images of HUVEC networks after CM exposure are presented in Fig. 2C.

Fig. 2.

HUVEC capillary-like network formation is reduced after culture with CM from CD34⁻/CD31⁺ CACs of the impaired function group compared with healthy controls. CM from CACs was collected from each subject and then used in a HUVEC-based, capillary-like network formation assay. Average total network length (A) and number of nodes (B) were quantified with image analysis software. Representative images from each condition are provided (C). Results are presented as each condition normalized to the basal (EBM-2) condition to control for daily variability in HUVEC growth and passage number with each assay. Images were taken under 5× original magnification. *P < 0.05, statistically significant difference from the healthy group.

Detection of secreted proteins in CM.

To compare secreted paracrine factors from CACs of the healthy and impaired function groups, CM from CD34^–/CD31⁺ cells from each group was analyzed by mass spectrometry. A full list of identified proteins can be found in the Supplemental data. Previous work in our laboratory indicates that S100A8 and S100A9 proteins secreted by CD34^–/CD31⁺ CACs can affect capillary-like network formation in younger adults as a function of habitual exercise training (27). As such, we chose to focus on S100A8 and S100A9 to gain more information about these secreted proteins in the present study. Spectrum counts from S100A8 and S100A9 were 78% and 76% greater, respectively, in the impaired function group compared with the healthy group (not significant; Fig. 3, A and B). We quantitatively confirmed these findings in our complete cohort of samples (n = 10/group) using immunoblotting and determined that S100A8 content was 165% higher in the CM from the impaired function group compared with the healthy group (P < 0.05; Fig. 3C) and that there was 153% more S100A9 content in CD34^–/CD31⁺ CM of impaired function individuals compared with that of the healthy subjects (P < 0.05; Fig. 3D). Concentrations of the S100A8/S100A9 heterodimer complex were also assessed, and we found that S100A8/S100A9 levels were ~3× greater in the CD34^–/CD31⁺ CM of the impaired function group compared with the healthy group (P < 0.05; Fig. 3E).

Fig. 3.

S100A8 and S100A9 content is greater in CM derived from the impaired function group's CACs compared with the CM from healthy individuals’ CACs. S100A8 (A) and S100A9 (B) were identified in CM of cultured CD34⁻/CD31⁺ cells using mass spectrometry spectrum counts (n = 3/group). Immunoblotting was used to confirm mass spectrometry spectrum counts for S100A8 (C) and S100A9 (D) in CM from CD34⁻/CD31⁺ cells from the impaired function group compared with the healthy group (n = 10/group). Representative images for S100A8 and S100A9 immunoblots can be found above (C and D, respectively). Values were normalized to cell number used to generate the CM (300,000 cells/well). E: S100A8/S100A9 heterodimer levels were measured using an ELISA. *P ≤ 0.05, statistically different from the healthy group.

S100A8 and S100A9 mRNA expression.

Quantitative PCR analyses of CD34^–/CD31⁺ CACs indicated that there was significantly greater mRNA expression of both S100A8 and S100A9 in the CD34^–/CD31⁺ CACs of the impaired function group compared with those of the healthy group (P < 0.05 for both; Fig. 4, A and B).

Fig. 4.

mRNA expression of S100A8 and S100A9 is higher in impaired function individuals’ CACs. S100A8 (A) and S100A9 (B) mRNA expression in CD34⁻/CD31⁺ CACs (n = 10). All data were normalized to GAPDH mRNA. *P < 0.05, statistically significant.

Capillary-like network formation with targeted inhibition of TLR4 and RAGE.

HUVECs were pretreated with inhibitors for known S100A8 and S100A9 receptors, TLR4 (TAK-242) and RAGE (FPS-ZM1). TAK-242 pretreatment of HUVECs exposed to CM from impaired function individuals significantly improved total network length (P < 0.05) and number of nodes (P < 0.05) compared with HUVECs exposed to CM from the impaired function group alone (Fig. 5, A–C). In contrast, neither FPS-ZM1 treatment nor combined TAK-242 + FPS-ZM1 treatments significantly altered network formation in the impaired function group (P > 0.05). Inhibitor treatments had no effect on HUVEC network formation in the healthy subjects’ CM conditions (P > 0.05).

Fig. 5.

Inhibition of TLR4, but not RAGE, normalizes HUVEC network formation after exposure to CM from CD34⁻/CD31⁺ CACs of the impaired function group. HUVEC network length (A) and number of nodes (B) after exposure to CM from CD34⁻/CD31⁺ CACs of the impaired function group compared with the healthy group are rescued in the presence of a TLR4 inhibitor but not a RAGE antagonist. The data in the first column for A and B, displaying CD34⁻/CD31⁺ CM from each group, are presented for comparison purposes and have already been presented in Fig. 2, A and B. CM from CACs was collected from each subject and then used in a HUVEC capillary-like network formation assay, with or without pretreatment of HUVECs with TAK-242 (TAK; TLR4 signaling inhibitor) or FPS-ZM1 (FPS; RAGE antagonist). C: representative images from each condition are provided. Results are presented as each condition normalized to the basal (EBM-2) condition to control for daily variability in HUVEC growth and passage number with each assay. Images were taken under 5× original magnification. *P < 0.05, statistically significant difference from the healthy group.

rhS100A8 and rhS100A9 treatment on HUVECs.

To determine if the effects of the CM could be explained by S100A8 and S100A9 content, we repeated the HUVEC experiments with rhS100A8 and rhS100A9, which were added to a HUVEC-based, capillary-like network formation assay in the concentrations and proportions that were estimated to be present in the CD34^–/CD31⁺ CM of the healthy and impaired function groups. There were no significant differences in network formation parameters between the basal condition and the condition using the concentrations of rhS100A8 and rhS100A9, estimated to be present in the healthy subjects’ CD34^–/CD31⁺ CM. However, when using concentrations of rhS100A8 and rhS100A9, estimated to be present in the impaired function group's CM, there was a 23% reduction in network length (P < 0.05; Fig. 6A) and a 41% reduction in number of nodes (P < 0.05; Fig. 6B) compared with the basal condition. This effect was rescued when the HUVECs were pretreated with TAK-242 before exposure to the recombinant proteins, such that network length and number of nodes were significantly higher than basal conditions and the healthy CM conditions (P < 0.05; Fig. 6, A and B). TAK-242 treatment alone also significantly improved network length and number of nodes compared with basal and healthy conditions (P < 0.05; data not shown).

Fig. 6.

Experiments using rhS100A8 and rhS100A9 mimic the effects of the CM from impaired function and healthy CACs. Exposure of HUVECs to rhS100A8 and rhS100A9 levels equivalent to that found in CM from the impaired function group reduces HUVEC network length (A) and number of nodes (B). However, pretreatment of the HUVECs with TAK-242 prevents the reduced network length and number of nodes induced by S100A8 and S100A9. Concentrations of S100A8 and S100A9, estimated to be present in the CM of CD34⁻/CD31⁺ CACs from the healthy group, were 2.9 and 0.8 μg/ml, respectively, and from the impaired function group, were 6.7 and 1.4 μg/ml, respectively. HUVEC network formation was assessed for total length (A) and number of nodes (B). C: representative images from each condition are provided. In these experiments, each condition was assessed in samples collected from 6 independent cell culture wells from multiple culture plates collected on different days. All experiments were conducted on cells from the same passage number (P4). *P ≤ 0.05, statistically different basal condition; #P ≤ 0.05, statistically different than rhS100 healthy condition; $P ≤ 0.05, statistically different than vehicle condition; %P ≤ 0.05, statistically different than rhS100 impaired function condition.

Plasma levels of S100A8/A9.

The plasma S100A8/A9 heterodimer complex tended to be elevated in the impaired function group compared with the healthy group (5,249 ± 831 vs. 4,381 ± 622 ng/ml), however the difference was not statistically significant (P > 0.05; data not shown).

DISCUSSION

In this study, we show alterations in the CAC secretome in individuals representing impaired CAC function and support the notion that paracrine actions of CD34^–/CD31⁺ CACs alter capillary-like network length. The data presented here provide novel extensions of our previous work (27) by demonstrating that inhibitory activities of S100A8 and S100A9 on capillary-like network length are also present in the CM of CD34^–/CD31⁺ CACs further along the continuum of health disparities in an older population exhibiting increased CV risk. With the extension on our previous findings, our data further suggest that elevated S100A8 and S100A9 inhibit HUVEC network formation by activating TLR4 signaling. Collectively, our results indicate potential molecular mechanisms associated with the paracrine actions of impaired CD34^–/CD31⁺ CACs.

We used rhS100A8 and rhS100A9 in concentrations estimated to be present in each group's CM to confirm their isolated effects on capillary-like network formation. Concentrations of rhS100A8 and rhS100A9, estimated to be present in the CM from the impaired function individuals’ CACs, significantly reduced HUVEC network formation compared with the basal and healthy group's conditions. However, equivalent concentrations of rhS100A8 and rhS100A9, as found in healthy subjects’ CM, did not significantly alter network formation. These results support our hypothesis that higher secretion of S100A8 and S100A9 is sufficient to contribute to the reduced angiogenic actions observed in the CD34^–/CD31⁺ CACs from the impaired function group. Although elevated S100A8/A9 plasma levels have been identified as risk factors for CVD (7, 16), this is the first study, to our knowledge, to identify elevated secretion of S100A8 and S100A9 from CD34^–/CD31⁺ CACs from a population considered to have impaired CV function acting as a negative regulator of aspects of the angiogenic process.

It is noteworthy that the use of rhS100A8 and rhS100A in isolation appeared to have a greater inhibitory effect on capillary-like network formation than use of the CM from the impaired function group. This is likely the result of other factors found in the CM from both groups. Specifically, the proteomic analyses identified over 250 proteins in the CM from the CD34^–/CD31⁺ CACs. As this specific CAC subtype has known roles in angiogenesis, it is likely that other secreted proteins either directly affect the function of S100A8 and/or S100A9 to blunt their inhibitory actions or promote capillary-like network formation through separate mechanisms, leading to an overall mitigated effect compared with the isolated effects of S100A8 and S100A9. Further exploration of other secreted proteins will be necessary to assess fully the paracrine actions of CD34^–/CD31⁺ CACs.

To date, two major receptors for S100A8 and S100A9—RAGE and TLR4—have been identified on endothelial cells (5, 11, 33–35). In this study, we demonstrate that inhibition of TLR4, but not RAGE, attenuated the negative actions of CM from the impaired function group on capillary-like network formation. These results suggest that the major factor(s) present in the CM of impaired function subjects’ CACs are acting primarily through TLR4. Furthermore, our results show that the pretreatment of HUVECs with the TLR4 inhibitor before exposure to rhS100A8 and rhS100A9 in concentrations present in impaired function subjects’ CD34^–/CD31⁺ CM eliminates the attenuation in capillary-like network formation caused by these proteins. Although it is well known that TLR4 is a major receptor for S100A8 and S100A9, our data demonstrate, at least in an in vitro environment, that it is TLR4 and not the other major receptor for S100A8 and S100A9—RAGE—that is mediating our observed differences in capillary-like network formation. Upon binding to TLR4, S100A8 and S100A9 activate downstream MAPK signaling, leading to increased expression and secretion of inflammatory factors, such as CXCL1 and IL-8, and decreased expression of the intercellular junction proteins (32). This, in turn, can cause damage to endothelial barrier function and cytoskeletal structure, ultimately impairing endothelial integrity (31, 32, 34, 35). Some studies have found that S100A8 and S100A9 preferentially bind to TLR4 or RAGE, respectively, on HUVECs (33, 35). As such, we can speculate that the higher concentrations of S100A8 in the CM may be driving our observed differences in network formation, although further studies investigating the independent effects of S100A8 and S100A9 and their affinity for specific receptors are necessary.

Of note, we also found that TAK-242 treatments alone significantly improved network formation. TAK-242 acts by blocking TLR4 signaling of the intracellular domain (18), and previous results have shown that TLR4-deficient mice have impaired expression of proinflammatory cytokines (29). Thus it is not unreasonable to suggest that TLR4 signaling may be a negative regulator of capillary-like network formation, and use of inhibitors, such as TAK-242, may be a possible method for improving angiogenesis. Collectively, our results indicate that S100A8 and S100A9, secreted by the CD34^–/CD31⁺ CACs from individuals exhibiting impaired CAC function, reduce HUVEC capillary-like network length and that these actions appear to be mediated by TLR4.

Limitations.

As indicated, the intent of this study was to compare young, healthy individuals with older CVD patients as models of healthy and impaired function (respectively) to determine mechanistically factors contributing to impairments of CAC paracrine function and compare the paracrine function of these CACs across a broad continuum of CV risk. It should be noted that with our study design, we are unable to make statements regarding the individual effects of age, CVD, or physical activity level. As the current study provides evidence that differences in CD34^–/CD31⁺ paracrine function exist between these two distinctly different groups, future studies, investigating a broad range of CAC functions to delineate whether our observed differences are a result of physical activity habits, age, or CVD, are necessary. Previous studies in our laboratory have found that acute exercise and exercise training provide a beneficial stimulus that aids in the reduction of oxidative stress (20, 21) and improves paracrine function (27) in younger, healthy adults. Future studies should determine if exercise could be used as a means of altering CAC paracrine actions in populations with impaired function.

In the current study, we used a gain-of-function approach for assessing the role of S100A8 and S100A9. However, future studies using approaches to ablate S100A8 and S100A9 function, directly using neutralizing antibodies, will provide further support for their roles. Unfortunately, genetic manipulation of cultured CD34^–/CD31⁺ CACs has proven to be challenging, as the cells are not amenable to most standard forms of plasmid DNA delivery. Furthermore, our studies are largely dependent on the specificity of TAK-242 toward TLR4; thus future studies using other small molecule inhibitors of TLR4 or other approaches that exclusively block ligand-dependent signaling of TLR4, such as the small interfering RNA-mediated knockdown approach, are necessary to understand fully the role of this receptor in angiogenesis. In the current study, we used the capillary-like network formation assay as our assessment of angiogenic function, as it is a comprehensive assay that measures several components of the angiogenic process (1, 10). However, our results should be interpreted carefully, as additional assays measuring specific aspects of the angiogenic process, such as migration and proliferation, are necessary to confirm how elevated CAC-derived S100A8 and S100A9 affect angiogenesis or vasculogenesis as a whole and the role of TLR4 in this process. Although our findings present valuable insight into the altered paracrine actions of CD34^–/CD31⁺ CACs from populations exhibiting impaired function, further research is needed to validate whether CAC paracrine factors exhibit similar effects on other endothelial cells and ultimately, to determine the in vivo effects of CAC paracrine actions.

It is possible that the paracrine activity of CD34^–/CD31⁺ CACs is altered by mechanisms associated with other CAC functions or characteristics. Differences in the composition of CACs may vary across groups, despite selecting for CD34^–/CD31⁺. For example, Kim et al. (22) found that fewer immunomagnetically selected CD31⁺ CACs from CAD patients coexpressed CD3, and more coexpressed CD14 and CD11b compared with healthy controls. At this time, we cannot discount the possibility that differences in distribution profile between groups may result in differences in paracrine activity. Similarly, differences in rates of apoptosis between CD34^–/CD31⁺ CACs of different health statuses may also affect the secretion profile of these cells (22). However, further research is necessary to test these hypotheses empirically to identify individual subpopulations within these CACs and examine rates of apoptosis that may be driving the alterations in paracrine action and subsequently, capillary-like network formation.

Another potential limitation of this study is that the individuals from the impaired function group were taking a number of medications. Several medications have been found to affect CAC number and function (12), and it is possible that some of the medications taken by this group affected our outcomes. We did not control the medications in the impaired function group, because it would be unethical to have the patients stop taking these medications. Additionally, as one of our long-term goals is to understand the function of CACs from CVD patient populations that might benefit from autologous cell therapies, it is more physiologically relevant to study CACs from this group in the conditions at which they would be used if implementing these treatments.

Conclusions.

In summary, we found that CD34^–/CD31⁺ CACs from our impaired function group exhibit reduced paracrine-mediated, capillary-like network length of HUVECs compared with our healthy group. We found that S100A8 and S100A9 protein content was substantially higher in the CM from the impaired function group compared with the healthy group. Experiments using rhS100A8 and rhS100A9 in the concentrations estimated to be present in the CD34⁻CD31⁺ CM of the impaired function group, but not the healthy group, further confirmed that these proteins reduced network formation by activation of TLR4. Collectively, our results identify molecular mechanisms associated with CD34^–/CD31⁺ CACs from individuals exhibiting high CV risk that involves secreted S100A8, S100A9, and TLR4 and suggest that regulation of S100A8 and S100A9 may serve as a potential target for treating inflammatory diseases, such as CVD.

GRANTS

Support for R. Q. Landers-Ramos was provided by the National Institute on Aging Predoctoral Institutional Training Grant T32AG000268 (to J. M. Hagberg) and by the Ann G. Wylie Dissertation Fellowship and the University of Maryland Summer Research Fellowship (to R. Q. Landers-Ramos). Support for S. J. Prior was provided by a Paul B. Beeson Career Development Award in Aging (K23-AG040775) and the American Federation for Aging Research. Support for this study was provided by the National Heart, Lung, and Blood Institute (Grant R21-HL98810; to J. M. Hagberg), the American College of Sports Medicine Doctoral Research Grant (to R. Q. Landers-Ramos), and the Baltimore VA Geriatric Research, Education, and Clinical Center.

DISCLOSURES

E. R. Chin is the founder and chief scientific officer of MyoTherapeutics, a University of Maryland-based start-up company. No other conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

R.Q.L-R., E.R.C., E.E.S., S.J.P., and J.M.H. conceived and designed research; R.Q.L-R., R.M.S., E.V., J.M., and Y.W. performed experiments; R.Q.L-R., R.M.S., Y.W., E.R.C., E.E.S., S.J.P., and J.M.H. analyzed data; R.Q.L-R., R.M.S., Y.W., E.R.C., E.E.S., S.J.P., and J.M.H. interpreted results of experiments; R.Q.L-R. prepared figures; R.Q.L-R. drafted manuscript; R.M.S., S.R., Y.W., E.R.C., E.E.S., S.J.P., and J.M.H. edited and revised manuscript; R.Q.L-R., R.M.S., E.V., J.M., S.R., Y.W., E.R.C., E.E.S., S.J.P., and J.M.H. approved final version of the manuscript.