Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Mar 3.
Published in final edited form as: J Vasc Res. 2025 Mar 3;62(4):204–210. doi: 10.1159/000544884

Unstable coronary artery plaque features in humans are associated with higher frequency of circulating CD56bright Natural Killer Cells

Micah Hunter-Chang a,*, Christine L Flora a,*, Hema Kothari a,*, Chantel C McSkimming a, Sydney D Blimbaum a,b, Corey M Williams a, Angela M Taylor c,d, Stefan Bekiranov a,e, Coleen A McNamara a,c,d
PMCID: PMC12324973  NIHMSID: NIHMS2067983  PMID: 40031876

Abstract

Introduction:

Unstable human artery plaques can suddenly rupture, leading to MI or stroke. Identification of blood markers associated with unstable plaque features is clearly needed. Humans with symptomatic carotid atherosclerotic plaques have increased infiltration of CD56bright Natural Killer (NK) cells into the plaque, yet whether subjects with unstable coronary artery plaque features have increased frequencies of circulating CD56bright NK cells is unknown.

Methods:

Coronary artery intravascular ultrasound (IVUS) was performed on subjects presenting for medically-indicated coronary angiography. 18 subjects stratified into high and low percent (%) necrotic core and matched for age, body mass index (BMI), and lipids underwent Mass Cytometry by Time of Flight (CyTOF) analysis on their peripheral blood mononuclear cells (PBMCs) collected prior to imaging. Clustering of major immune cell populations was performed on live singlets and CD56 bright and dim NK subsets were quantitated.

Results:

Subjects with high necrotic core had a significantly greater frequency of circulating CD56bright NK cells compared to subjects with low necrotic core (p=0.02). Additionally, the frequency of circulating CD56bright NK cells positively associated with IVUS-VH metrics of total atheroma volume (TAV) (p=0.0013), percent (%) atheroma burden (p=0.0048), % maximum stenosis (p=0.0021), % necrotic (p=0.0013), % calcium (p=0.0016)), % fatty (p = 0.0097) and negatively associated with % fibrous (p<0.0001), an IVUS-VH metric of plaque stability.

Conclusion:

These findings suggest that the frequency of CD56bright NK cells may be a safe, non-invasive marker of plaque volume and instability.

Keywords: Atherosclerosis, Natural Killer Cells, Intravascular Ultrasound, Mass Cytometry, Peripheral Blood Mononuclear Cells

Introduction

Natural Killer (NK) cells are innate immune lymphocytes broadly divided into CD56bright cells that secrete cytokines and CD56dim cells that secrete cytolytic factors [1, 2]. NK cells have been shown to reside in human carotid plaques [3, 4], with greater amounts of NK cells in the plaque of patients with more severe atherosclerosis compared with less severe disease [5, 6]. Moreover, subjects with symptomatic carotid atherosclerosis had increased CD56bright NK cells in their plaques [7]. Yet, whether CD56bright NK cells are implicated in unstable coronary artery plaques in humans is unknown.

Through novel pairing of intravascular ultrasound with virtual histology (IVUS-VH) – a plaque imaging modality – with immune profiling of peripheral blood mononuclear cells (PBMC) by mass cytometry (CyTOF) – a high-dimensional flow cytometry variant [8, 9] – quantitative in vivo investigation of relationships between plaque stability and circulating NK cell phenotypes can be conducted. IVUS-VH is a plaque imaging technique that provides the percentage of necrotic, calcification, fatty, and fibrous tissue present in the plaque. These measurements help characterize lesion stability as plaques with large necrotic cores are more prone to rupture [10, 11]. IVUS-VH also provides measurements of percent (%) maximum arterial stenosis, total atheroma volume (TAV), and % atheroma burden, which measure the amount and distribution of plaque in the vessel.

In this study, we test whether there is an association between the frequency of circulating CD56bright NK cells and IVUS-VH metrics including TAV, % atheroma burden, % maximum stenosis, % necrotic, % calcium, % fatty, and % fibrous in human subjects.

Methods

Human Cohort

Subjects referred for diagnostic coronary angiography by their physician for any reason were consented to participate in study IRB-HSR #15328. Clinical indications to perform angiography varied from chest pain syndromes to preoperative planning for heart surgery. Subjects with known inflammatory diseases (immune disorders, connective tissue diseases, inflammatory bowel diseases, arteritis or vasculitis, or ongoing serious infection), cancers, and/or pregnancy were excluded. Those in whom IVUS was deemed safe, such as those with without anemia (hematocrit greater than 30mg/dl and no blood donation within 56 days), were consented to have IVUS-VH.

For this study, 18 subjects were selected from IRB-HSR #15328 participants based on % necrotic core region with n=9 having high (>20%) and n=9 having low (<6%) necrotic tissue as a percentage of total plaque. Subjects were matched for age, body mass index (BMI), and lipid levels, which was verified to not be statistically significant using Wilcoxon tests (p>0.05 for all matched metrics). Wilcoxon tests were also conducted for additional patient data, including lipids (total cholesterol, triglycerides, HDL cholesterol, LDL cholesterol) and additional IVUS metrics used to determine CAD severity including TAV, % atheroma burden, and % maximum stenosis. A Fisher test was conducted to evaluate male-female significance (supplementary table 1).

IVUS-VH Imaging

At the time of cardiac catheterization, intravascular ultrasound was performed on subjects in accordance with the American College of Cardiology standards for acquisition of non-infarct related artery. IVUS-VH was performed on the least angulated vessel providing the longest length for evaluation. 150 mcg of intracoronary nitroglycerin was given prior to advancement of a 2.6F, 30 MHz IVUS-VH catheter (Volcano Corporation) into the target vessel. The transducer was positioned distally and a motorized pullback was performed at 0.5 mm/s with images being obtained at 30 frames/second for a minimum of 30mm. The R-100 pullback device (Volcano Corporation) was calibrated and inspected to meet the required accuracy of ± 0.03mm for pullback speeds of 0.5mm/sec. Digital images were stored for analysis off-line. Every 60th image was analyzed generating cross sections spaced 1.0mm apart. Histologic regions assigned by the Volcano IVUS-VH software were visually confirmed. Images were analyzed using ImageJ as provided by the National Institutes of Health. Measurements including % maximum lumen area stenosis, total atheroma volume, % atheroma burden, % necrotic, % fibrous, % fatty, and % calcium were obtained by a Cardiologist expert in IVUS-VH reading. This study (IRB-HSR #15328) was approved by the Human Institutional Review Board. All participants provided written informed consent prior to enrollment.

PBMC Isolation

Peripheral blood samples were drawn into BD K2 EDTA vacutainer tubes (CAT#: 367855, FisherScientific). Blood in vacutainers was centrifuged at 400g for 10 min to separate platelet rich plasma. PBMCs were isolated using Ficoll-Paquedensity-gradient centrifugation (CAT#: 17144003, Cytiva) and SepMate-50 IVD tubes (CAT#: 85460, StemCell). PBMCs were aliquoted and cryopreserved in freezing media (90% FBS and 10% DMSO; CAT#: A5256801 ThermoFisherScientific and CAT#: J66650.AK ThermoFisherScientific) and stored in liquid nitrogen until use to eliminate batch effects.

PBMC Thawing and CyTOF Staining

Cryopreserved PBMCs were thawed with Deoxyribonuclease I from bovine pancreas (CAT#: D4513, Sigma-Aldrich), resuspended in warmed culture media (RPMI-1640 CAT#: 21870100 ThermoFisherScientific with 10% FBS CAT#: A5256801 ThermoFisherScientific and 1% penicillin/streptomycin CAT#: 15070063 ThermoFisherScientific), centrifuged (400g × 10 minutes), and resuspended. A cell count was then performed and the cells were washed with PBS, and incubated for 5 minutes using live-dead marker Cell-ID™ Cisplatin (CAT#: 201064, Standard Bio Tools). The cells were then washed with warmed media and incubated at 37C with 5% CO2 for one hour. Rested cells were then stained for 20 minutes with surface marker CyTOF antibodies (supplementary table 2) with PFA-sensitive epitopes after which the cells were fixed in 1.6% PFA. Following fixation, all centrifugation was performed at 800g × 10 minutes. Cells were then washed with cell staining buffer (CSB) (Maxpar® Cell Staining Buffer, CAT#: 201068, Standard Bio Tools), and stored at 4C overnight. The cells were then barcoded with the Maxpar Cell-ID™ 20-Plex Pd Barcoding Kit (CAT#: 201060, Standard Bio Tools) and pooled in CSB. Next, they were incubated for 5 minutes with Human TruStain FcX™ (CAT#: 422302, Bio Legend), and then stained for 30 minutes in the non-PFA sensitive surface stain antibodies (supplementary table 2) per manufacturer instructions. The cells were washed and permeabilized using a 10-minute methanol (MeOH) (CAT#: 67-56-1 FisherScientific) incubation with vortexing. The MeOH was quenched with CSB, and the cells were washed twice. Cells were then stained with a 45-minute intracellular antibody cocktail (supplementary table 2). Finally, the cells were incubated in Cell-ID™ Intercalator-Ir (CAT#: 201192A, Standard Bio Tools) for 30 minutes and resuspended in Maxpar® Cell Acquisition Solution (CAT#: 201240, Standard Bio Tools) to be run through the Helios Mass Cytometer.

CyTOF Data Preprocessing

CyTOF data was exported as FCS files. Samples were normalized using the Nolan lab MATLAB normalizer (http://github.com/nolanlab/bead-normalization/releases) and debarcoded using Zunder’s lab debarcoder (90) (https://github.com/zunderlab/single-cell-debarcoder). The files were uploaded to OMIQ, a cytometry analysis platform, and cleanup gating was performed to remove dead cells, doublets, debris, and low-quality barcoding.

Data Analysis & Clustering

Major cell type dimensionality reduction and clustering was performed using the 31 following markers: CD16, CD200R, CD26, CD11c, CD183, CD45RO, CD197, CD8, CD3, CD103, CD27, CD127, PD-1, CD19, CD123, CD38, CD4, CD11b, CD14, CD196, CD36, CD200, IL1R1, CD185, CD161, CD25, CD45, IgM, CD192, HLA-DR, and CD56. UMAP was performed using the uwot package in R with n_neighbors = 100, min_dist = 0.01, n_epochs = 200, and otherwise default parameters. The UMAP fuzzy graph was saved and prepared for use in clustering using the igraph package in R [12]. Leiden clustering [13] was performed in R using the leiden package with partition_type = “ModularityVertexPartition”, n_iterations = −1, resolution_parameter = 0.3, and otherwise default parameters. The CATALYST package was used for data input and plotting [14]. Clusters 3,4,7, and 11 were merged to define CD4 T cells, clusters 5 and 9 were merged to define CD8 T cells, cluster 2 defined B cells, cluster 8 defined NK cells, and clusters 1 and 6 were merged to define monocytes. Clusters 10, 12, and 13 were removed from analysis.

Following clustering, the NK cluster was re-uploaded to OMIQ and manually gated into CD56bright and CD56dim NK cells by population breaks on a combination of contour and density dot plots. The frequency of CD56bright NK cells was then calculated as a percentage of CD56bright cells divided by the NK cluster total.

Statistical Analysis

Wilcoxon tests were used for the following: comparisons between matched cohort characteristics (age, BMI, lipids; p>.05 for all matched characteristics), lipids (total cholesterol, triglycerides, HDL cholesterol, LDL cholesterol), and additional IVUS metrics used to determine CAD severity including TAV, % atheroma burden, and % maximum stenosis (supplementary table 1); comparisons of major circulating immune cell populations between low and high necrotic core groups; comparisons of CD56bright and CD56dim population frequencies between low and high necrotic core groups. A Fisher test was conducted to evaluate male-female significance (supplementary table 1). Spearman correlations were used to evaluate relationships between CD56bright NK frequency and IVUS-VH parameters across all subjects.

Results

Our cohort was composed of 18 individuals referred for cardiac catheterization at the University of Virginia Hospital who consented to have coronary artery IVUS-VH. This cohort was broken into two groups, 9 with high necrotic core and 9 with low necrotic core. Wilcoxon tests were conducted on cohort characteristics to verify matching for age, BMI, and lipids and no significant differences were found for these characteristics (p>0.05; Supplementary Table 1). Wilcoxon tests were also conducted comparing additional cohort characteristics including lipids (total cholesterol, triglycerides, HDL, and LDL; p>0.05) and additional IVUS metrics used in conjunction with % necrotic to determine CAD severity. This included total atheroma volume (p<0.001***), % atheroma burden (p<0.001***), and percent maximum stenosis (p<0.001***). A Fisher test was conducted to evaluate male-female differences between low and high necrotic core groups, providing borderline significance (p=0.0498; Supplementary Table 1).

No major broad immune cell population abundance (CD4 T cells, CD8 T cells, B cells, Monocytes, NK cells) was significantly different between high and low necrotic core groups (p>0.05 for all comparisons by Wilcoxon tests). As subtypes of immune cells can have distinct functions and CD56bright NK cells are in greater abundance in symptomatic carotid plaques [7], we looked at the NK subtypes.

Subjects in the high (>20%) necrotic core group had a significantly greater frequency of circulating CD56bright NK cells compared to subjects in the low (<6%) necrotic core group (4.62% ± 1.87% vs 2.48% ± 1.11%, respectively; p=0.02*; Fig. 1. a). CD56dim NK cell frequency was not significantly different between high and low necrotic core subjects (p=0.34). Positive correlative associations were found between IVUS-VH metrics and CD56bright NK cell frequency for % necrotic (R = 0.6966, p = 0.0013**), % calcium (R = 0.6877, p = 0.0016**), % fatty (R = 0.5913, p = 0.0097***), % maximum stenosis (R = 0.6760, p = 0.0021***), total atheroma volume (TAV) mm2 (R = 0.6966, p = 0.0013**), and % atheroma burden (R = 0.6326, p = 0.0048**). Negative correlative association was found between CD56bright NK cell frequency and IVUS-VH metric % fibrous (R = −0.8039, p = 0.00006***; Fig 1. b). Note that *p < 0.05, **p < 0.01, and ***p < 0.001.

Fig. 1.

Fig. 1.

Open in a new tab

Circulating CD56bright natural killer cell frequency is associated with the size of the necrotic core in human coronary plaques and other unstable plaque features. (Fig.1. a) CD56bright NK cell frequency (%) was significantly higher in subjects with high necrotic core (n = 9) in comparison to those with low necrotic core (n=9; p = 0.02*). (Fig. 1. b) Spearman correlation graphs of IVUS-VH metrics with CD56bright NK cell frequency. Positive significant correlations between CD56bright frequency and percent (%) necrotic (R = 0.6966, p = 0.0013**), % calcium (R = 0.6877, p = 0.0016**), % fatty (R = 0.5913, p = 0.0097***), % maximum stenosis (R = 0.6760, p = 0.0021***), total atheroma volume (TAV) mm2 (R = 0.6966, p = 0.0013**), % atheroma burden (R = 0.6326, p = 0.0048**). Negative significant correlation with % fibrous (R = −0.8039, p = 0.0001***). *p < 0.05, **p < 0.01, ***p < 0.001. IVUS-VH = Intravascular Ultrasound; TAV= Total Atheroma Volume.

Discussion

Identification of a safe, non-invasive marker of individuals at risk for unstable coronary artery plaque features that may lead to acute coronary syndromes (ACS) is of clear significance. Studies in human carotid atherosclerotic plaques, which are more easily obtained due to the ability to excise plaques with carotid endarterectomy, have shown increased CD56bright NK cells in symptomatic compared to asymptomatic subjects. While plaque excision is not routinely performed in coronary arteries, IVUS-VH has emerged as a powerful tool for identifying plaques with features that predict plaque rupture and MI [11]. While this technology does not allow for identification of individual cell types in the lesions, it does allow for characterization of the plaque features.

Through pairing of IVUS-VH with mass cytometry on circulating PBMCs, we provide the first evidence that the frequency of circulating CD56bright NK cells is associated with indicators of coronary artery plaque instability – including plaque volume, maximum stenosis (%), atheroma burden (%), and the percentage of the plaque that is necrotic, fatty, or calcified – and inversely associated with the percentage that is fibrous (a marker of plaque stability) in a small cohort of 18 subjects matched for age, BMI, and lipids. These findings suggest that the frequency of CD56bright NK cells may be a safe, non-invasive marker of plaque volume and instability.

The fact that there were more females in the low compared to high necrotic core group could be a confounder. However, prior studies have shown that females have a higher proportion of CD56bright NK cells than males [15, 16, 17]. As such, one would expect the group with more female subjects (low necrotic core group) to have a greater proportion of CD56bright NK cells, and this is not the case. In fact, it is the opposite, potentially strengthening the unique finding that the frequency of CD56bright NK cells is associated with high necrotic core. Our small sample size is a limitation of our study. However, these novel findings provide rationale for performing a larger study with IVUS-VH or other advanced coronary artery imaging techniques to confirm the result and potentially understand covariates that may affect this finding.

Supplementary Material

Suppl_01

Acknowledgements

We thank M. Solga and the UVA Flow Cytometry Core for excellent technical assistance and acquisition of CyTOF data. We also thank M. Brown for contributing his expertise in natural killer cell biology.

Funding Sources

This study was funded by the American Heart Association Innovative Project Award (HK, CAM), R01HL148109 (CAM), grant T32AI00749 (SDB), and UVA iPRIME (AMT, SB, CAM).

The funders had no role in the design, data collection, data analysis, or reporting of this study.

Footnotes

Statement of Ethics

Study approval statement: These studies (IRB-HSR #15328; IRB-HSR #16017) were approved by the Human Institutional Review Board. All participants provided written informed consent prior to enrollment.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Data Availability Statement

Human clinical and imaging data are not made publicly available due to patient confidentiality but can be made available to interested parties by contacting the corresponding author.

Cohort characteristics and antibody panel are available in the supplemental materials.

Clustered CyTOF data is publicly available: https://data.mendeley.com/datasets/6n8cv6st4s/1.

Any other associated data can be made available upon request by contacting the corresponding author.

References

  • 1.Caligiuri MA (2008). Human natural killer cells. Blood, 112(3), 461–469. 10.1182/blood-2007-09-077438 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Vivier E, Tomasello E, Baratin M, Walzer T, & Ugolini S (2008). Functions of natural killer cells. Nature Immunology, 9(5), 503–510. 10.1038/ni1582 [DOI] [PubMed] [Google Scholar]
  • 3.Bobryshev YV, & Lord RSA (2005). Identification of natural killer cells in human atherosclerotic plaque. Atherosclerosis, 180(2), 423–427. 10.1016/j.atherosclerosis.2005.01.046 [DOI] [PubMed] [Google Scholar]
  • 4.Fernandez DM, Rahman A, Fernandez NF, Chudnovskiy A, Amir ED, Amadori L, Khan NS, Wong CK, Shamailova R, Hill CA, Wang Z, Remark R, Li J, Pina C, Faries C, Awad AJ, Moss N, Björkegren J, Kim-Schulze S, Gnjatic S, Ma’ayan A, Mocco J, Faries P, Merad M, Giannarelli C (2019). Single-cell immune landscape of human atherosclerotic plaques. Nature Medicine, 25(10), 1576–1588. 10.1038/s41591-019-0590-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.van Dijk RA, Rijs K, Wezel A, Hamming JF, Kolodgie FD, Virmani R, Schaapherder AF, & Lindeman JHN (2016). Systematic Evaluation of the Cellular Innate Immune Response During the Process of Human Atherosclerosis. Journal of the American Heart Association, 5(6). 10.1161/jaha.115.002860 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Clerc G, & Rouz PM (1997). Lymphocyte Subsets in Severe Atherosclerosis before Revascularization. Annals of Internal Medicine, 126(12), 1004–1005. 10.7326/0003-4819-126-12-199706150-00028 [DOI] [PubMed] [Google Scholar]
  • 7.Bonaccorsi I, Spinelli D, Cantoni C, Barillà C, Pipitò N, De Pasquale C, Oliveri D, Cavaliere R, Carrega P, Benedetto F, & Ferlazzo G (2019). Symptomatic Carotid Atherosclerotic Plaques Are Associated With Increased Infiltration of Natural Killer (NK) Cells and Higher Serum Levels of NK Activating Receptor Ligands. Frontiers in Immunology, 10, 1503–1503. 10.3389/fimmu.2019.01503 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Bandura DR, Baranov V, Ornatsky O, Antonov A, Kinach R, Lou X, Pavlov S, Vorobiev S, Dick JE, & Tanner SD (2009). Mass cytometry: Technique for real time single cell multitarget immunoassay based on inductively coupled plasma time-of-flight mass spectrometry. Analytical Chemistry, 81(16), 6813–6822. 10.1021/ac901049w [DOI] [PubMed] [Google Scholar]
  • 9.Bendall SC, Simonds EF, Qiu P, Amir ED, Krutzik PO, Finck R, Bruggner RV, Melamed RD, Trejo A, Ornatsky O, Balderas RS, Plevritis SK, Sachs K, Pe’er D, Tanner SD, & Nolan GP (2011). Single-Cell Mass Cytometry of Differential Immune and Drug Responses Across a Human Hematopoietic Continuum. Science, 332(6030), 687–696. 10.1126/science.1198704 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Finn AV, Nakano M, Nakano Masataka, Nakano Masataka, Nakano M, Narula J, Kolodgie FD, & Virmani R (2010). Concept of Vulnerable/Unstable Plaque. Arteriosclerosis, Thrombosis, and Vascular Biology, 30(7), 1282–1292. 10.1161/atvbaha.108.179739 [DOI] [PubMed] [Google Scholar]
  • 11.Nissen SE, & Yock P (2001). Intravascular Ultrasound. Circulation, 103(4), 604–616. 10.1161/01.CIR.103.4.604 [DOI] [PubMed] [Google Scholar]
  • 12.Csárdi G, Nepusz T (2006). The igraph software package for complex network research. InterJournal, 1695–1704. https://igraph.org/ [Google Scholar]
  • 13.Traag VA, Waltman L, van Eck NJ, & van Eck NJ (2018). From Louvain to Leiden: Guaranteeing well-connected communities. Scientific Reports. 10.1038/s41598-019-41695-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Crowell HL, Chevrier Stéphane, Chevrier S, Jacobs Andrea, Jacobs A, Sivapatham Sujana, Sivapatham S, Bodenmiller Bernd, Robinson Mark D., Bodenmiller B, & Robinson MD (2020). An R-based reproducible and user-friendly preprocessing pipeline for CyTOF data. F1000Research, 9, 1263. 10.12688/f1000research.26073.1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Phan M-T, Chun S, Kim S-H, Ali AK, Lee S-H, Kim S, Kim S-H, & Cho D (2017). Natural killer cell subsets and receptor expression in peripheral blood mononuclear cells of a healthy Korean population: Reference range, influence of age and sex, and correlation between NK cell receptors and cytotoxicity. Human Immunology, 78(2), 103–112. 10.1016/j.humimm.2016.11.006 [DOI] [PubMed] [Google Scholar]
  • 16.Timmons BW, Tarnopolsky MA, & Bar-Or O (2006). Sex-based effects on the distribution of NK cell subsets in response to exercise and carbohydrate intake in adolescents. Journal of Applied Physiology, 100(5), 1513–1519. 10.1152/japplphysiol.01125.2005 [DOI] [PubMed] [Google Scholar]
  • 17.Al-Attar A, Presnell SR, Peterson CA, Thomas DT, & Lutz CT (2016). The effect of sex on immune cells in healthy aging: Elderly women have more robust natural killer lymphocytes than do elderly men. Mechanisms of Ageing and Development, 156, 25–33. 10.1016/j.mad.2016.04.001 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Suppl_01
NIHMS2067983-supplement-Suppl_01.docx (18.8KB, docx)

Data Availability Statement

Human clinical and imaging data are not made publicly available due to patient confidentiality but can be made available to interested parties by contacting the corresponding author.

Cohort characteristics and antibody panel are available in the supplemental materials.

Clustered CyTOF data is publicly available: https://data.mendeley.com/datasets/6n8cv6st4s/1.

Any other associated data can be made available upon request by contacting the corresponding author.

RESOURCES