Skip to content

Omics and emerging science in thoracic aortic aneurysm

TL;DR — Between 2019 and 2026 the genetic architecture of thoracic aortic disease was rewritten from "a handful of Mendelian genes" to "a common, polygenic trait with ~100 loci". Deep learning applied to 4.6 million UK Biobank cardiac MRI frames turned aortic diameter into a scalable quantitative phenotype, yielding 82 loci for ascending and 47 for descending diameter (Pirruccello 2022, PMID 34837083); a case-control GWAS of 8,626 TAAD cases in the Million Veteran Program found 21 risk loci, 17 novel, and provided human genetic evidence that TAAD is a non-atherosclerotic disorder distinct from other vascular disease (Klarin 2023, PMID 37308786). Single-cell and single-nucleus atlases of human ascending aorta converge on one mechanism — smooth muscle cell phenotype modulation rather than the appearance of novel cell types (Chou 2022, PMID 36172868; Li 2020, PMID 33017217; Pedroza 2020, PMID 32698686). Polygenic scores now add 4.0–10.7 absolute percentage points of explained diameter variance over clinical factors across four cohorts (PMID 37662232), and automated imaging pipelines measure aortas better than they did two years ago (95.5% of diameters correct after retraining vs 42.1% concordance before; PMID 37716024). The unresolved question is translational: none of this has yet produced a drug target that has entered a human aortic trial, and the single genotype-directed trial that exists (nicotinamide riboside in ACTA2 R179 disease, NCT06280482) enrols 15 patients.

1. GWAS: from Mendelian genes to a polygenic trait

1.1 Deep-learning phenotyping made the sample size possible

The bottleneck for aortic genetics was never genotyping; it was phenotyping. Manual aortic measurement in tens of thousands of participants is infeasible, so the field was stuck with case-control designs on small ascertained cohorts. The unlock was training convolutional models to segment the aorta on cardiac MRI at biobank scale.

Study Phenotyping scale GWAS n Loci Key downstream result
Pirruccello 2022, Nat Genet (PMID 34837083) 4.6 million cardiac MR images, UK Biobank 39,688 82 ascending, 47 descending, 14 overlapping Transcriptome-wide analysis, rare-variant burden and human aortic single-nucleus RNA-seq prioritised genes including SVIL (descending). Ascending-diameter PRS: incident TAA HR 1.43 per SD (1.32–1.54), P=3.3×10⁻¹² in 385,621 participants
Nekoui 2022, JACC (PMID 35902171) 2.3 million cardiac MR images, 43,317 participants — 79 loci across LVOT, aortic root, and 6 ascending positions; 35 novel Most loci associated with only 1–2 of the 8 measured diameters — the genetics are spatially distinct. PRS for diameter ~13 mm above the sinotubular junction best predicted TAA (mean HR 1.42/SD, 1.34–1.50, P=6.67×10⁻⁵⁴, n=427,016); PRS for a smaller aortic root predicted aortic stenosis (HR 1.08/SD)
Klarin 2023, Nat Genet (PMID 37308786) Clinical TAAD diagnosis, Million Veteran Program 8,626 cases / 453,043 controls; replication 4,459 / 512,463 across 6 cohorts 21 loci, 17 novel Downstream analyses identified causal genes and cell types; established TAAD as non-atherosclerotic and "not solely inherited through protein-altering variants of large effect size"
Malhotra 2019, Nat Genet (PMID 31659325) CT-quantified aortic calcification 9,417 abdominal / 8,422 descending thoracic HDAC9, RAP1GAP (abdominal only) No SNP reached genome-wide significance for thoracic aortic calcification — an early signal that thoracic and abdominal aortic biology diverge

Two structural findings deserve emphasis over any individual locus:

  • Spatial specificity of aortic genetics. Nekoui found that most loci acted on one or two of eight measured diameters (PMID 35902171), and Pirruccello found only 14 of 82+47 loci shared between ascending and descending aorta (PMID 34837083). The embryological boundary between second-heart-field and neural-crest-derived aorta is visible in the GWAS signal. This reframes "thoracic aortic aneurysm" as several genetically distinct traits sharing a name — see anatomy and classification.
  • Non-atherosclerotic architecture. The MVP GWAS explicitly separated TAAD from coronary and peripheral arterial disease genetically (PMID 37308786). The absence of a thoracic calcification signal in Malhotra 2019 (PMID 31659325) points the same way. Practically: TAA should not be expected to respond to atherosclerosis therapeutics, and observational associations imported from AAA (a partly atherosclerotic disease) should be treated with suspicion.

1.2 Polygenic risk and prediction

Model What it adds Performance Caveat
Clinical score, 11 covariates that would not independently prompt imaging (Pirruccello 2022 JAMA, PMID 36378208) Baseline Explained 28.2% (UK Biobank validation), 30.8% (Framingham), 32.6% (Mass General Brigham) of diameter variance; AUROC for ≥4.0 cm 0.770 / 0.813 / 0.766 At a fixed threshold of 3.537, sensitivity was only 8.9% / 11.3% / 18.8%; the model significantly over- or under-estimated diameter in external validation
AORTA Gene: clinical + 1.1M-variant PRS (Pirruccello 2023, PMID 37662232) Genetics Variance explained 39.9% vs 29.2% (UK Biobank), 36.5% vs 32.5% (MGB), 41.8% vs 33.0% (FHS), 34.9% vs 28.9% (All of Us); AUROC ≥4 cm 0.834 vs 0.765 (P=7.3×10⁻¹⁰) Preprint at time of curation; authors note "larger and more diverse cohorts will be needed to develop more powerful and equitable scores" — the PRS was derived in a predominantly European cohort
PRS-stratified screening economics (Kelemen 2024, PMID 39277617) Cost-effectiveness modelling Leveraging pleiotropy improved PRS R² by 22.7% vs using AAA alone; high vs low PRS tertile HR 3.70 (2.86–4.80) adjusted for clinical risk factors Abdominal aortic aneurysm, not thoracic — included here as the methodological template, not as TAA evidence

The Kelemen framework — invite high-PRS male current smokers for screening earlier than 65, screen high/intermediate-PRS female smokers (currently not invited at all) at 65 and 70 (PMID 39277617) — is the clearest published statement of what a PRS is actually for: reallocating imaging, not replacing it. The equivalent analysis has not been done for the thoracic aorta.

1.3 Mendelian randomization

MR in this space splits cleanly by risk factor: blood pressure is causal for the thoracic aorta, lipids are not.

  • Blood pressure is causal for aortic size, and the relationship is bidirectional. Two-sample MR on GWAS summary statistics found genetically predicted diastolic BP (β 0.0272 mm per mmHg, 95% CI 0.0224–0.0320, P<0.001), mean arterial pressure (0.0168, 0.0130–0.0206, P<0.001) and systolic BP (0.0041, 0.0008–0.0074, P=0.02) all associated with a larger ascending diameter, while pulse pressure was inversely associated (−0.0155, −0.0213 to −0.0096, P<0.001). Reverse-direction MR showed genetically predicted ascending diameter lowers pulse pressure (−2.07 mmHg/mm) and systolic BP (−1.29 mmHg/mm) while raising diastolic BP (+0.82 mmHg/mm) — the haemodynamic consequence of a wider conduit rather than a disease effect (DePaolo 2023, PMID 36601961). Note the magnitude: roughly 0.03 mm of ascending diameter per mmHg of diastolic pressure.
  • Hypertension is causal for events. Genetic liability to hypertension was associated with aortic dissection (OR 1.81, 1.27–2.58) and aortic aneurysm (OR 1.43, 1.22–1.66); per-SD genetically determined diastolic BP with dissection (OR 1.14, 1.09–1.19) and aneurysm (OR 1.07, 1.05–1.09), while systolic BP was null for both (Yang 2023, PMID 37443259). The aneurysm phenotype in that analysis is not thoracic-specific.
  • Lipids are causal for abdominal, not thoracic, aneurysm. Among 368,139 UK Biobank participants followed a median 13.65 years (1,634 AAA, 698 TAA, 184 dissection events), remnant cholesterol predicted AAA (adjusted HR 1.65, 1.36–1.99; MR OR 2.08, 1.70–2.56, surviving adjustment for LDL-C) but showed no association with TAA or dissection in either the observational or the MR analysis (Zhou 2025, PMID 39962497). This is the sharpest direct test of the atherosclerotic hypothesis in the thoracic aorta, and it is negative — concordant with the non-atherosclerotic genetic architecture of TAAD (PMID 37308786).
  • Klarin 2023 used downstream analytic methods including Mendelian randomization infrastructure (S. Burgess, V. Zuber among authors) to argue TAAD is aetiologically distinct from atherosclerotic vascular disease (PMID 37308786).
  • Genetically predicted IL-6 receptor signalling was associated with a composite cardiovascular endpoint that included aortic aneurysm, at HR 1.11 per 1 mg/dL increment in absolute hsCRP (1.06–1.17) in 397,060 UK Biobank participants (Georgakis 2022, PMID 35948913). This is a composite endpoint, not aorta-specific, and should not be cited as MR evidence for inflammation causing TAA.
  • Clonal haematopoiesis of indeterminate potential (CHIP) was associated with incident atherosclerotic disease "across multiple beds" in UK Biobank and Mass General Brigham, driven by DNA-damage-repair gene mutations (Zekavat 2023, PMID 36949957) — relevant as a candidate age-related mechanism, but again not TAA-specific.

Together these give the field its clearest causal asymmetry: pressure control has genetic support in the thoracic aorta; lipid lowering does not.

2. Single-cell atlases of the human aorta

2.1 The datasets

Study Tissue and n Cells / nuclei Resolution Central finding
Li 2020, Circulation (PMID 33017217) 8 ascending TAA (4F/4M) + 3 controls scRNA-seq 11 major cell types; reclustered to 40 subtypes Multiple SMC, macrophage and T-lymphocyte subtypes. TAA tissue had fewer non-immune and more immune cells, especially T lymphocytes. Extensive mitochondrial dysfunction signature. Integration with GWAS + promoter capture Hi-C implicated ERG
Pedroza 2020, ATVB (PMID 32698686) Fbn1^C1039G/+ mouse root/ascending + human Marfan root aneurysm scRNA-seq All aortic cell types Identified a distinct modulated SMC (modSMC) cluster present only in adult mutant aortic aneurysm. MFS-specific signature vs atherosclerosis included Serpine1 (PAI-1) and Klf4; 481 genes differentially expressed modSMC vs SMC. modSMCs absent from young mutant aortas despite small aneurysm, and absent from non-dilated descending aorta
Dawson 2022, Genes (PMID 35052435) 3 Marfan ascending aneurysms + 4 controls scRNA-seq SMC, fibroblast, EC populations Increased de-differentiated proliferative SMCs. TGF-β ligand up in two fibroblast clusters, but TGF-β receptor genes and downstream genes down in SMC, fibroblast and EC — i.e. impaired signalling despite ligand excess
Chou 2022, ATVB (PMID 36172868) 13 human donors: 6 TAA, 7 non-aneurysmal 71,689 nuclei, snRNA-seq 14 clusters / 11 cell types; 7 VSMC and 6 fibroblast subclusters "Sporadic aortic aneurysm is characterized by differential gene expression within known cellular classes rather than by the appearance of novel cellular forms." Quiescent fibroblasts almost disappear in aneurysm. Used DEGs to prioritise genes at aortic diameter and distensibility GWAS loci: LTBP1, IL34 (fibroblasts); PDLIM5, ACTN4 (VSMC); plus macrophage genes
Zhen 2026, Clin Transl Med (PMID 42007493) 8 ATAA + 9 controls (6 from GEO) 187,163 high-quality immune cells; scRNA-seq + Visium HD spatial 8 major immune cell types, near-single-cell spatial resolution by deconvolution First high-resolution immune atlas of human ATAA; immune cells enriched vs controls; CellChat implicated interferon, AP-1 and NF-κB pathways
Bramel 2024, preprint (PMID 38883722) Tgfbr1-mutant Loeys-Dietz mouse aorta scRNA-seq + spatial transcriptomics VSMC heterogeneity by aortic segment A less-differentiated, proinflammatory VSMC subset localises to the aortic root regardless of genotype; postnatal VSMC-specific deletion of the identified factor reduced root dilation. A similar population identified in a human aortic scRNA-seq dataset. Preprint — treat as provisional

2.2 What the atlases agree on

  1. Phenotype modulation, not new cells. Chou's formulation (PMID 36172868) is the strongest version: the aneurysmal aorta contains the same cell classes doing different things. This is a substantive claim because it means therapeutic targets are transcriptional programmes and chromatin states, not cell populations to be depleted.
  2. The contractile-to-synthetic axis is the dominant signal, recurring in Marfan mouse and human (PMID 32698686, 35052435), in sporadic disease (PMID 36172868, 33017217), and mechanistically in the epigenetic work below. See pathophysiology.
  3. Fibroblasts matter more than the SMC-centric literature implied. Quiescent fibroblast populations near-vanish in aneurysm (PMID 36172868); LTBP1 and IL34 were prioritised in fibroblasts, not SMCs.
  4. Regional intrinsic vulnerability. The aortic root harbours a less-differentiated VSMC population independent of genotype (PMID 38883722), and second-heart-field-derived cells map onto exactly the regions where aortopathy occurs (PMID 35143327). This is the cell-biological correlate of the spatially distinct GWAS signal in §1.1. See anatomy and classification and animal models.
  5. TGF-β cannot be read as a single scalar. Dawson's finding of up-regulated ligand with down-regulated receptors and downstream genes in the same tissue (PMID 35052435) is why plasma TGF-β never worked as a biomarker — see biomarkers.

2.3 Single-cell data as a GWAS-interpretation engine

The most productive use of these atlases has not been descriptive but interpretive: intersecting cell-type-specific expression with GWAS loci to name causal genes. Chou 2022 (PMID 36172868) and Pirruccello 2022 (PMID 34837083, which used human aortic single-nucleus RNA-seq to prioritise SVIL) both do this, and Li 2020 combined scRNA-seq with GWAS and promoter capture Hi-C to implicate ERG (PMID 33017217). This closes a loop that neither method could close alone: GWAS localises to a region, single-cell data says which gene in which cell.

3. Spatial transcriptomics

Spatial methods are early but address the field's central blind spot — the aortic wall is layered and regionally heterogeneous, and dissociation destroys exactly that information.

  • Human ATAA immune microenvironment. Visium HD spatial transcriptomics deconvolved to near-single-cell resolution, integrated with scRNA-seq of 187,163 immune cells from 8 ATAA and 9 control aortas (Zhen 2026, PMID 42007493). Interferon, AP-1 and NF-κB signalling were nominated as targets.
  • Regional VSMC identity in Loeys-Dietz. Spatial transcriptomics localised a proinflammatory, less-differentiated VSMC subset to the aortic root (Bramel 2024, PMID 38883722, preprint).
  • The transmural gradient. Human ascending aneurysm pathology is predominantly in the outer media and adventitia, mirrored in angiotensin II-infused mice and coincident with second-heart-field-derived cell distribution; PAI-1 immunostaining showed a transmural gradient matching the pathology gradient in both mouse and human aneurysmal aorta (Sawada 2022, PMID 35143327).

4. Proteomics

Tissue proteomics has been more informative than plasma proteomics so far, and the best-executed study used it to catch the disease before it was visible:

  • Mass-spectrometry-assisted proteomics of angiotensin II-infused mouse ascending aortas without overt pathology showed down-regulation of smooth muscle cell proteins and differential expression of extracellular matrix proteins including several LRP1 ligands. Integrating proteomics with single-cell transcriptomics identified PAI-1 as the most increased protein in second-heart-field-derived SMCs and fibroblasts. SHF-specific Lrp1 deletion augmented aneurysm and rupture; SHF-specific Tgfbr2 deletion caused embryonic lethality at E12.5 with outflow-tract dilatation (Sawada 2022, PMID 35143327).

The pre-pathology design is the transferable idea: sampling at the point where the molecular programme has changed but the geometry has not is how a growth-predictive marker would be discovered.

The human plasma equivalent does not exist. No plasma proteomic study of human TAA with a prospective growth or event endpoint could be retrieved (searched 2026-08-28); every human dataset is cross-sectional and diameter-anchored:

  • Targeted proteomics in 52 ATAA patients stratified by diameter (4.0–4.5 cm n=23; 4.6–5.0 cm n=20; >5.0 cm n=9) plus 30 matched controls found CCL5, HBD1, ICAM1, IL-8, TNFα and TGF-β1 elevated versus controls (p<0.0001), with AUCs of 0.83–0.84 for the best three (Daskalopoulou 2023, PMID 37238945).
  • Untargeted plus targeted proteomics and extracellular-vesicle analysis in 80–85 matched patients identified 1,037 and 1,077 proteins respectively, of which 9 correlated with aortic diameter (ACTN1, CRP, TGM3, KRT84, IGHG3, DPYSL2, TSPAN8, IGKV3D-11, VDAC1) and 3 with the grade of medial degeneration; the authors call explicitly for external validation (Arndt 2025, PMID 41061941).

Both are the design diagnosed in biomarkers §3.3 — discriminating a large aorta from a normal one, which imaging already does. The gap is conspicuous given that plasma proteomic platforms are now routine in biobanks with the imaging and follow-up needed to do this properly.

5. Machine learning

5.1 Automated measurement — the near-term deliverable

Automated aortic measurement is the ML application closest to clinical use, and its recent history is a useful corrective to hype:

  • Retraining transformed performance. In 995 non-ECG-gated chest CTs (392 contrast-enhanced, 603 non-contrast) previously classified as dilated, a retrained deep-learning tool produced correct diameters in 8,539/8,948 (95.5%) measurements. Only 3,765/8,948 (42.1%) were correct in both the original and retrained versions; 4,456 (49.8%) were correct only after retraining. The aortic root was the hardest site (aortic sinus correct in only 221/995, 22%, before retraining; 564/995, 57%, after). Residual failures: 228 (2.5%) tilted planes, 181 (2.0%) over/under-segmentation, concentrated at the sinus (Lo Piccolo 2023, PMID 37716024).
  • Volumetric and zonal segmentation. An open-source model segmenting Society for Vascular Surgery / STS aortic zones in 59 acute uncomplicated type B dissections achieved a mean Dice coefficient of 0.73 (best: zone 9 at 0.91, zone 5 at 0.84). Comparing 33 rapid growers (≥5 mm/yr) with 26 slow/no growers over a median 1.07 yr, there were no baseline zone-volume differences between groups (Krebs 2024, PMID 38851467).

That last negative result matters: better measurement of baseline anatomy did not, in this cohort, predict who would grow. Automation solves reproducibility, not prognosis.

5.2 Phenotyping at biobank scale

Deep learning's demonstrated value has been as a phenotype factory feeding genetics: 4.6 million images for aortic diameter (PMID 34837083), 2.3 million for spatially resolved LVOT/root/ascending diameters (PMID 35902171), and 47,223 participants for velocity-encoded flow phenotypes including ascending aortic diameter, with reference ranges derived in up to 31,909 healthy individuals (Kany 2023, PMID 37205587, preprint).

5.3 Growth and event prediction — where the claims outrun the evidence

A 2025 comprehensive review of AI in acute aortic syndrome and aneurysm found that ML/DL models — particularly ensemble algorithms and biomarker-integrated approaches — frequently outperformed EuroSCORE II and GERAADA for mortality prediction, and that AI has been extended to growth-rate prediction, post-repair remodelling, rupture risk and location, and morbidity endpoints including acute kidney injury, stroke and paraplegia. The same review states plainly that most studies are limited by small sample sizes, single-centre design, and absence of external validation, and calls for TRIPOD-AI-compliant reporting and multicentre validation (Ayhan 2025, PMID 41375721).

Read together with §5.1, the honest summary is: AI reads aortas reliably; AI does not yet predict aortic futures reliably.

6. Epigenetics and gene-targeted therapy horizons

The strongest mechanism-to-therapy chain in this field runs through chromatin, not through TGF-β.

  • A shared epigenetic node. Mutations affecting TGF-β signalling and mutations affecting the VSMC actomyosin cytoskeleton both converge on formation of a ternary complex of the histone deacetylase HDAC9, the chromatin remodeller BRG1, and the long non-coding RNA MALAT1. This complex binds chromatin and represses contractile protein genes with gain of H3K27me3. Disrupting Malat1 or Hdac9 restored contractile protein expression, improved aortic mural architecture, and inhibited experimental aneurysm growth (Lino Cardenas 2018, PMID 29520069). This unifies the two clinical groups described in genetics of TAA.
  • Druggable by oligonucleotide or small molecule. In the stenotic-vascular counterpart of the same biology, pharmacological targeting via MALAT1 antisense oligonucleotides or inhibition of the methyltransferase EZH2 (recruited by the HDAC9 complex) reduced neointimal formation (Lino Cardenas 2019, PMID 30674723). ASOs and EZH2 inhibitors are both clinically available modalities in other diseases — this is the most credible existing route to a targeted aortic drug.
  • HDAC9 arrived independently from human genetics. HDAC9 was the first genome-wide-significant locus for abdominal aortic calcification, with functional follow-up showing HDAC9 overexpression promoted calcification and reduced contractility in human aortic SMCs, and Hdac9 loss reduced aortic calcification by 40% in matrix Gla protein-deficient mice (Malhotra 2019, PMID 31659325). Two independent lines converging on one gene is the strongest target-validation signal in the field.
  • Redox and cofactor rescue. Anti-TGF-β treatment and, independently, oral folic acid attenuated aneurysm formation in Fbn1^C1039G/+ mice by recoupling eNOS through restored DHFR and tetrahydrobiopterin; circulating BH4 levels tracked tissue BH4 and correlated negatively with aortic root expansion, proposed as a biomarker (Huang 2020, PMID 33126053). Mouse only — see animal models.
  • The one genotype-directed human trial. Nicotinamide riboside in ACTA2 Arg179 smooth muscle dysfunction syndrome (NCT06280482, phase 1, n=15, UT Health Houston, recruiting, primary completion 2027-07; entry requires a confirmed ACTA2 arginine-179 variant). Its registered title targets moyamoya-like cerebrovascular disease, and the primary outcomes are correspondingly cerebrovascular and systemic — cerebral oxygenation and perfusion, cognition, autonomic symptoms, headache impact, spirometry, systolic blood pressure — alongside echocardiographic aortic diameter at 8 weeks; blood NAD⁺/NR are secondary, and reduced aortic glucose uptake appears among the stated aims rather than in the outcome list. It is small, single-centre, non-randomised and only partly aortic, but it is the only trial retrieved that targets a molecular mechanism in a genetically defined aortopathy.
  • Gene editing in this field builds models rather than treating them. CRISPR/Cas9 was used to knock a patient-derived Fbn1 p.G234D variant into mice, producing spontaneous dissection with 50% mortality by 5 weeks and endothelial mechanosensing failure from 1 week of age (Kimura 2025, PMID 40365676); mosaicism in an SMC-specific Acta2 R179C knock-in has been discussed explicitly for what it implies about editing efficiency thresholds in future therapy (Kaw 2022, PMID 35878552). No in-vivo editing rescue of an aortic phenotype in an Fbn1 model, and no registered gene-editing or antisense interventional trial for TAA, could be retrieved (PubMed and ClinicalTrials.gov, searched 2026-08-28). The closest antisense precedent is vascular but not thoracic: Nudt6 antisense oligonucleotides limited disease progression in three murine and one porcine model of AAA and carotid disease (Winter 2023, PMID 37147804). The gene-therapy "horizon" for TAA remains entirely preclinical.

7. What these approaches could unlock

Approach Plausible deliverable Blocking problem
Biobank GWAS + PRS Reallocating surveillance imaging by genetic risk, as modelled for AAA (PMID 39277617) No thoracic equivalent of that cost-effectiveness analysis; PRS derived in European-ancestry cohorts (PMID 37662232)
Segment-specific genetics Treating root, ascending, and descending disease as separate entities with separate thresholds (PMID 34837083, 35902171) Guidelines and trials still pool them; TITAN:SvS randomises on a single ascending diameter (NCT03536312)
Single-cell + spatial atlases Targets that are transcriptional programmes (PMID 36172868), plus GWAS gene prioritisation Human aortic tissue is available almost exclusively at operation, so every atlas is an end-stage snapshot
Pre-pathology proteomics A growth-predictive marker discovered before geometry changes (PMID 35143327) Done in mice; every human plasma dataset is cross-sectional and diameter-anchored (PMID 37238945, 41061941)
Automated imaging Reproducible diameters, volumes, and growth rates at population scale (PMID 37716024, 38851467) Aortic root remains the least reliable site — and it is the site that determines surgery in syndromic disease
Epigenetic targeting (HDAC9/MALAT1/EZH2) A first disease-modifying drug for non-syndromic TAA (PMID 29520069, 30674723, 31659325) No human aortic trial has started

Open questions

  • Do the 21 TAAD risk loci (PMID 37308786) and the ~100 aortic-diameter loci (PMID 34837083, 35902171) implicate the same biology, or is "genetics of aortic size" a different trait from "genetics of dissection"? The two GWAS families have not been formally reconciled.
  • Given that most diameter loci act on only one or two aortic segments (PMID 35902171), should segment-specific polygenic scores replace a single aortic PRS — and would a root-specific score outperform for syndromic disease?
  • Does a polygenic score add predictive value conditional on measured diameter, the comparison that determines clinical utility? AORTA Gene predicts diameter (PMID 37662232); it has not been shown to predict events in people whose diameter is already known.
  • Is the modulated SMC state (PMID 32698686, 36172868) causal for wall failure or a reactive response to it? Every human atlas is cross-sectional and taken at operation, so temporality is unresolved.
  • Why does the aortic root harbour an intrinsically less-differentiated VSMC population regardless of genotype (PMID 38883722, preprint), and does that explain the root's disproportionate vulnerability in Marfan and Loeys-Dietz disease?
  • Can HDAC9/MALAT1/EZH2 inhibition (PMID 29520069, 30674723) be delivered to the aortic wall at a tolerable systemic epigenetic cost? No human aortic trial exists.
  • MR now separates the two classical risk factors: blood pressure is causal for ascending diameter and for dissection (PMID 36601961, 37443259) while remnant cholesterol is causal for abdominal but not thoracic aneurysm (PMID 39962497). Does that asymmetry mean lipid lowering has no place in thoracic aortic care, or only that the phenotype and event counts are still too small to detect an effect (698 TAA events versus 1,634 AAA)?
  • Can an ML model predict aortic growth from baseline imaging at all? The best-controlled attempt found no baseline zonal volume differences between rapid and slow growers (PMID 38851467).

References

  1. Pirruccello JP, Chaffin MD, Chou EL, et al. Deep learning enables genetic analysis of the human thoracic aorta. Nat Genet. 2022;54:40-51. PMID 34837083
  2. Nekoui M, Pirruccello JP, Di Achille P, et al. Spatially Distinct Genetic Determinants of Aortic Dimensions Influence Risks of Aneurysm and Stenosis. J Am Coll Cardiol. 2022;80:486-497. PMID 35902171
  3. Klarin D, Devineni P, Sendamarai AK, et al. Genome-wide association study of thoracic aortic aneurysm and dissection in the Million Veteran Program. Nat Genet. 2023;55:1106-1115. PMID 37308786
  4. Malhotra R, Mauer AC, Lino Cardenas CL, et al. HDAC9 is implicated in atherosclerotic aortic calcification and affects vascular smooth muscle cell phenotype. Nat Genet. 2019;51:1580-1587. PMID 31659325
  5. Pirruccello JP, Lin H, Khurshid S, et al. Development of a Prediction Model for Ascending Aortic Diameter Among Asymptomatic Individuals. JAMA. 2022;328:1935-1944. PMID 36378208
  6. Pirruccello JP, Khurshid S, Lin H, et al. AORTA Gene: Polygenic prediction improves detection of thoracic aortic aneurysm. medRxiv. 2023. PMID 37662232
  7. Kelemen M, Danesh J, Di Angelantonio E, et al. Evaluating the cost-effectiveness of polygenic risk score-stratified screening for abdominal aortic aneurysm. Nat Commun. 2024;15:8063. PMID 39277617
  8. DePaolo J, Levin MG, Tcheandjieu C, et al. Relationship Between Ascending Thoracic Aortic Diameter and Blood Pressure: A Mendelian Randomization Study. Arterioscler Thromb Vasc Biol. 2023;43:359-366. PMID 36601961
  9. Yang T, Yuan X, Gao W, et al. Causal effect of hypertension and blood pressure on aortic diseases: evidence from Mendelian randomization. Hypertens Res. 2023;46:2203-2212. PMID 37443259
  10. Zhou T, Lin W, Yang B, et al. Remnant cholesterol and risk of aortic aneurysm and dissection: a prospective cohort study from the UK Biobank study and Mendelian randomization analysis. Lipids Health Dis. 2025;24:53. PMID 39962497
  11. Georgakis MK, Malik R, Richardson TG, et al. Associations of genetically predicted IL-6 signaling with cardiovascular disease risk across population subgroups. BMC Med. 2022;20:245. PMID 35948913
  12. Zekavat SM, Viana-Huete V, Matesanz N, et al. TP53-mediated clonal hematopoiesis confers increased risk for incident atherosclerotic disease. Nat Cardiovasc Res. 2023;2:144-158. PMID 36949957
  13. Li Y, Ren P, Dawson A, et al. Single-Cell Transcriptome Analysis Reveals Dynamic Cell Populations and Differential Gene Expression Patterns in Control and Aneurysmal Human Aortic Tissue. Circulation. 2020;142:1374-1388. PMID 33017217
  14. Pedroza AJ, Tashima Y, Shad R, et al. Single-Cell Transcriptomic Profiling of Vascular Smooth Muscle Cell Phenotype Modulation in Marfan Syndrome Aortic Aneurysm. Arterioscler Thromb Vasc Biol. 2020;40:2195-2211. PMID 32698686
  15. Dawson A, Li Y, Li Y, et al. Single-Cell Analysis of Aneurysmal Aortic Tissue in Patients with Marfan Syndrome Reveals Dysfunctional TGF-β Signaling. Genes (Basel). 2022;13:95. PMID 35052435
  16. Chou EL, Chaffin M, Simonson B, et al. Aortic Cellular Diversity and Quantitative Genome-Wide Association Study Trait Prioritization Through Single-Nuclear RNA Sequencing of the Aneurysmal Human Aorta. Arterioscler Thromb Vasc Biol. 2022;42:1355-1374. PMID 36172868
  17. Zhen K, Du Z, Wang P, et al. Spatial and single-cell RNA sequencing reveals the immune microenvironment of human ascending thoracic aortic aneurysms. Clin Transl Med. 2026;16:e70650. PMID 42007493
  18. Bramel EE, Espinoza Camejo WA, Creamer TJ, et al. Intrinsic expression sensitizes the aortic root to dilation in a Loeys-Dietz syndrome mouse model. Res Sq [preprint]. 2024. PMID 38883722
  19. Sawada H, Katsumata Y, Higashi H, et al. Second Heart Field-Derived Cells Contribute to Angiotensin II-Mediated Ascending Aortopathies. Circulation. 2022;145:987-1001. PMID 35143327
  20. Daskalopoulou A, Giotaki SG, Toli K, et al. Targeted Proteomic Analysis of Patients with Ascending Thoracic Aortic Aneurysm. Biomedicines. 2023;11(5). PMID 37238945
  21. Arndt N, Mair T, Riedner M, et al. Thoracic aortic diseases: Identification of diagnostic biomarkers using proteomic analysis. Cardiovasc Pathol. 2025;80:107785. PMID 41061941
  22. Lo Piccolo F, Hinck D, Segeroth M, et al. Impact of retraining a deep learning algorithm for improving guideline-compliant aortic diameter measurements on non-gated chest CT. Eur J Radiol. 2023;168:111093. PMID 37716024
  23. Krebs JR, Imran M, Fazzone B, et al. Volumetric analysis of acute uncomplicated type B aortic dissection using an automated deep learning aortic zone segmentation model. J Vasc Surg. 2024;80:1025-1034.e4. PMID 38851467
  24. Kany S, Rämö JT, Hou C, et al. Assessment of valvular function in over 47,000 people using deep learning-based flow measurements. medRxiv. 2023. PMID 37205587
  25. Ayhan C, Mekhaeil M, Channawi R, et al. Applications of Artificial Intelligence as a Prognostic Tool in the Management of Acute Aortic Syndrome and Aneurysm: A Comprehensive Review. J Clin Med. 2025;14:8420. PMID 41375721
  26. Lino Cardenas CL, Kessinger CW, Cheng Y, et al. An HDAC9-MALAT1-BRG1 complex mediates smooth muscle dysfunction in thoracic aortic aneurysm. Nat Commun. 2018;9:1009. PMID 29520069
  27. Lino Cardenas CL, Kessinger CW, Chou E, et al. HDAC9 complex inhibition improves smooth muscle-dependent stenotic vascular disease. JCI Insight. 2019;4:e124706. PMID 30674723
  28. Huang K, Wang Y, Siu KL, Zhang Y, Cai H. Targeting feed-forward signaling of TGFβ/NOX4/DHFR/eNOS uncoupling/TGFβ axis with anti-TGFβ and folic acid attenuates formation of aortic aneurysms. Redox Biol. 2020;38:101757. PMID 33126053
  29. Kimura K, Motoyama E, Kanki S, et al. Novel Aortic Dissection Model Links Endothelial Dysfunction and Immune Infiltration. Circ Res. 2025;137:26-42. PMID 40365676
  30. Kaw A, Pedroza AJ, Chattopadhyay A, et al. Mosaicism for the smooth muscle cell (SMC)-specific knock-in of the Acta2 R179C pathogenic variant: Implications for gene editing therapies. J Mol Cell Cardiol. 2022;171:102-104. PMID 35878552
  31. Winter H, Winski G, Busch A, et al. Targeting long non-coding RNA NUDT6 enhances smooth muscle cell survival and limits vascular disease progression. Mol Ther. 2023;31:1775-1790. PMID 37147804