Transcriptomic profiling of endothelial progenitor cells in post-COVID-19 patients: Insights at 3 and 6-month post-infection

Summary

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused significant global morbidity since 2019. Long COVID, characterized by persistent symptoms after acute infection, may involve endothelial injury. We analyzed endothelial colony-forming cells (ECFCs) from post-COVID-19 patients at 3- and 6-month post-infection, comparing them with healthy controls and stratifying by prior pulmonary embolism (PE). Transcriptomic profiling identified differentially expressed genes (DEGs) associated with endothelial homeostasis, inflammation, oxidative stress, and thrombosis. Post-COVID ECFCs showed downregulation of NOS3, KLF2, ANGPT1, PIK3R3, GBX2, GDF6, SMAD6, SRC, and TGFB1, and upregulation of CASP1, CXCL5, IL12A, SOD2, TIMP3, and TLR2. Minimal differences were observed between 3 and 6-month samples. PE patients showed downregulation of thrombosis-related genes such as PTGS2 and ACKR3. These findings indicate sustained endothelial dysfunction and inflammation up to 6 months post-infection, highlighting the importance of long-term monitoring and potential therapeutic strategies to support vascular health in post-COVID-19 patients.

Graphical abstract

Highlights

• •Transcriptomic profiling reveals long-term endothelial dysfunction in post-COVID-19
• •Enrichment analyses highlight disrupted pathways of endothelial homeostasis
• •Minimal changes between 3 and 6 months suggest sustained endothelial damage
• •Post-COVID patients with PE display unique profiles with reduced thrombosis markers

Immunology; Transcriptomics

Introduction

Coronavirus disease 2019 (COVID-19), caused by SARS-CoV-2 infection, has represented a significant public health challenge, leading to widespread morbidity and mortality worldwide.¹ Multiple investigations have demonstrated that COVID-19 has implications not exclusively for the respiratory system but also impacts various organs, particularly the cardiovascular system.^2,3 Indeed, thromboembolic events, prominently influenced by endothelial dysfunction, have been markedly common in patients diagnosed with COVID-19.⁴ Additionally, long COVID syndrome, characterized by persistent or newly emerging symptoms that persist for more than 12 weeks following the infection, was also related to persistent endothelial cell activation.⁵ Common symptoms include fatigue, pain, dyspnea, sleep disturbances, physical limitations, psychological distress, and cognitive impairments.^6,7 These manifestations closely resemble those observed in patients with myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS), highlighting possible similarities in disease mechanisms.^6,7

The vascular endothelium, lining the innermost layer of blood vessels, is vital for maintaining tissue homeostasis. Its functions encompass regulating vascular tone, preserving barrier integrity, and managing inflammation and coagulation.¹ Notably, a range of viral species, including SARS-CoV-2, can infect endothelial cells (ECs), triggering endothelial dysfunction and shifting its homeostasis to a pro-inflammatory and pro-coagulant state.^1,8 It is known that SARS-CoV-2 can impact the endothelium through direct infection, disrupting its anti-thrombogenic and barrier properties.⁸ Alternatively, it can indirectly contribute to endothelial injury by triggering a local cytokine storm and a systemic inflammatory response.⁸

Endothelial colony-forming cells (ECFCs) are a rare cell population in peripheral blood circulation with strong proliferative potential capable of forming human blood vessels in vivo, and promoting re-endothelialization.⁹ ECFC number and function act as key biomarkers for evaluating vascular damage and predicting cardiovascular risk.¹⁰ In pathological conditions, ECFCs undergo mobilization from the bone marrow or from their niche within the vessel wall to areas of vascular damage to facilitate vascular regeneration.¹¹ Previously, our group reported an abnormal increase in ECFC production, which was maintained up to 12-month after discharge in post-COVID-19 patients with severe pneumonia, compared to healthy individuals.^12,13,14

High-throughput RNA sequencing (RNA-seq) stands out as a crucial method for comprehensive transcriptome profiling, employing advanced deep-sequencing technologies.¹⁵ It enables the quantification of genes and their processed or partially degraded products, facilitating the identification of both coding and non-coding RNA.¹⁵ Furthermore, it can contribute to the discovery of relevant biomarkers and potential targets for drug development, revealing the molecular mechanisms driving persistent vascular issues in individuals recovering from COVID-19.

To date, gathered transcriptomic analysis in samples obtained from patients during COVID-19 infection has provided significant insights. Motta et al.¹⁶ performed a transcriptomic analysis of human brain microvascular endothelial cells (HBMECs) exposed to SARS-CoV-2. They identified endothelial activation pathways, particularly through the NF-κB pathway, and observed alterations in the secretion of angiogenic factors. Their findings revealed prominent up-regulation of genes associated with endothelial activation, including members of the chemokine family such as CXCL1, CXCL2, CXCL3, CCL20, PTX3, as well as ICAM1 and TNF. Margaroli et al.¹⁷ conducted transcriptomic analysis on coronary ECs from severe COVID-19 patients, revealing distinct expression patterns. Additionally, Daamen et al.¹⁸ performed a comprehensive transcriptomic analysis on blood, lung, and airway samples from COVID-19 patients, highlighting inflammation, complement activation, and alterations in coagulation factors and fibrinogen genes. Moreover, another study investigated molecular signaling networks through blood transcriptomics in COVID-19 patients, identifying evidence of endothelial injury.¹⁹ Collectively, these studies consistently indicate significant endothelial injury and activation during SARS-CoV-2 infection, involving key pathways, such as chemokine activation, inflammation, coagulation, angiogenesis, and immune response.

Despite the presence of several studies assessing gene expression in COVID-19 patients, there has been limited exploration into the gene expression profiles of post-COVID-19 patients. To our knowledge, this is the first investigation providing RNA-seq data from isolated patient-derived endothelial progenitor cells obtained from 3 and 6-month post-COVID-19 patients. Accordingly, this study aims to apply a comprehensive transcriptomic profiling approach to ECFCs isolated from post-COVID-19 patients, with assessments conducted at 3 months post-infection and repeated at 6 months as part of longitudinal follow-up.

Through this investigation, we seek to elucidate the molecular mechanisms underlying post-COVID-19 endothelial impairment, thereby advancing our understanding of long-term vascular sequelae. Ultimately, these insights may inform the development of preventive strategies and targeted therapeutic interventions if required.

Results

Population characteristics

Table S2 displays general clinical characteristics of both post-COVID-19 patients and healthy controls. Subjects were meticulously selected and displayed no significant differences in any of the examined characteristics, including age, gender, BMI, and the presence of comorbidities such as AHT, DM, or DLP.

Identification of differentially expressed genes

To understand the vascular abnormalities following SARS-CoV-2 infection, we analyzed transcriptomic data from ECFCs obtained from individuals at 3 and 6 months post-COVID-19 infection, as well as from healthy controls. This analysis enabled us to characterize the post-infection sequelae in the vascular system. Principal-component analysis (PCA) revealed a clear separation between ECFCs isolated from healthy controls and those from post-COVID-19 patients at 3 months post-infection. This difference was significantly maintained at 6-month after the infection (Figures 1A and 1B). No significant segregation was observed between 3 and 6-month post-COVID-19 patients.

Figure 1.

Principal-component analysis (PCA)

(A) PCA of post-COVID-19 patients and healthy controls.

(B) PCA stratified by COVID-19 status and time since infection.

(C) PCA of 3- and 6-month PE or non-PE post-COVID-19 patients and healthy controls.

Component 1 explains 66.7% of the variance, and component 2 explains 6.7%.

Additionally, PCA of post-COVID-19 patients was stratified based on the occurrence of acute PE during initial hospitalization. The analysis revealed a clear separation between patients with and without PE at both 3- and 6-month’ post-infection, indicating distinct gene expression profiles between the two groups (Figure 1C).

Differentially expressed genes (DEGs) were identified for each contrast using the following thresholds: FDR ≤0.05, | Log2FC | ≥ 1, and p value ≤0.05.

Contrast 1: 3-Month post-COVID-19 patients vs. healthy controls

991 DEGs were identified between 3-month post-COVID-19 patients and healthy controls, with 454 upregulated and 537 downregulated genes (Figure 2A; Table 1). From this contrast, several enriched pathways were detected. Significant pathways were related to interleukin-17 (IL-17) signaling pathway, phosphoinositide 3-kinase (PI3K)-Akt signaling pathway, ribosome subunits structure, leukocyte transendothelial migration, vascular endothelial growth factor (VEGF) signaling pathway, regulation of apoptosis, tight junction, calcium signaling pathway, and cyclic guanosine monophosphate-protein kinase G (cGMP-PKG) signaling pathway, among others (Figure 2A; Tables S3 and S8).

Figure 2.

Volcano plots representing the differential expressed genes (DEGs) and pathway enrichment analysis using the KEGG database

(A–E) Volcano plots and pathway enrichment analysis representing the comparison of 3-month post-COVID-19 vs. healthy control (A), 6-month post-COVID-19 vs. healthy control (B), 3-month post-COVID-19 vs. 6-month post-COVID-19 (C), 3-month PE post-COVID-19 vs. 3-month non-PE post-COVID-19 (D), 6-month PE post-COVID-19 vs. 6-month non-PE post-COVID-19 (E).

P adjusted value <0.05 and |Log2FC| > 1 defined as cutoff. Blue dots represent downregulated genes and red dots upregulated genes.

Table 1.

DEGs summary of the different contrasts. DEGs cutoff defined as FDR≤0.05, |log2FC|≥1, adjusted p value≤0.05

Contrast	Upregulated	Downregulated
3-month post-COVID-19 vs. CTRL	453	537
6-month post-COVID-19 vs. CTRL	1590	1926
3-month post-COVID-19 vs. 6-month post-COVID-19	1	2
3-month post-COVID-19 PE vs. 3-month post-COVID-19 non-PE	464	449
6-month post-COVID-19 PE vs. 6-month post-COVID-19 non-PE	8	7

Contrast 2: 6-Month post-COVID-19 patients vs. healthy controls

3516 DEGs (1590 upregulated and 1926 downregulated) were found when 6-month post-COVID-19 samples were compared to healthy non-COVID-19 controls (Figure 2B; Table 1). Similar to 3 months contrast, pathway enrichment analysis revealed pathways mainly related to leukocyte transendothelial migration, PI3K-Akt signaling pathway, VEGF signaling pathway, and IL-17 signaling pathway (Figure 2B; Tables S4 and S9).

Contrast 3: 3-Month post-COVID-19 vs. 6-month post-COVID-19 patients

As shown in Figures 3A and 3B, minimal differences in gene expression were observed between 3 and 6 months post-COVID-19 within the same cohort of patients. Only three genes met the criteria for differential expression when comparing these two time points in followed individuals. Two displayed downregulation, PCDHA9 and AGBL4, while one, GRASP1, exhibited upregulation (Figure 2C; Table 1). Pathway enrichment analysis showed that the enriched pathways were related to cytokine-cytokine receptor interaction, antigen processing and presentation, shear stress, PI3K-Akt, and VEGF signaling pathway (Figure 2C; Tables S5 and S10).

Figure 3.

Venn diagram showing the overlapping DEGs between contrasts by up- or downregulation, labeling the number and percentage of overlapped or non-overlapped genes

(A) Upregulated genes of 3- or 6-month post-COVID-19 vs. healthy controls, and 3- vs. 6-month post-COVID-19.

(B) Downregulated genes of 3 or 6-month post-COVID-19 vs. healthy control and 3 vs. 6-month post-COVID-19.

(C) Upregulated genes of 3- or 6-month post-COVID-19 with PE vs. non-PE.

(D) Downregulated genes of 3- or 6-month post-COVID-19 with PE vs. non-PE comparisons.

Contrast 4: PE post-COVID-19 patients vs. non-PE post-COVID-19 patients at 3 and 6 months

The number of DEGs between COVID-19 patients who experienced PE during hospitalization and those who did not was evaluated at 3- and 6-month post-infection. A total of 913 DEGs were identified at 3-month post-COVID-19 between patients who suffered from PE and those who did not. Of these, 464 genes upregulated and 449 downregulated in the PE group compared to non-PE group (Figure 2D; Table 1). Enriched pathways in PE, as compared to non-PE, were related to Notch signaling (including Notch4), MAPK1/MAPK3 signaling, protein digestion and absorption, drug metabolism—cytochrome P450, and signaling pathways regulating pluripotency of stem cells (Figure 2D, Tables S6 and S11).

However, as shown in Figures 3C and 3D, most of the DEGs identified at 3 months that differentiated post-COVID-19 patients with PE from those without PE were not sustained at 6 months. By 6 months, only 15 DEGs were identified—8 upregulated and 7 downregulated—differentiating the two groups (Figure 2E; Table 1). Pathways such as VEGFA, TCR signaling, and interferon signaling, which were altered in 3-month post-COVID patients with PE compared to non-PE patients, had returned to similar levels between the two subgroups by 6 months post-infection, indicating a normalization of PE-associated pathways (Figure 4, Tables S7 and S12). Notably, only one gene, RGN, remained consistently upregulated at both 3 and 6 months. The top 50 most significantly enriched pathways from the preceding contrasts are summarized in Figure 4 and detailed in Tables S3–S12.

Figure 4.

Heatmap of Top 50 Upregulated and Downregulated Reactome Pathways Across Study Comparisons

(A and B) Heatmap of the Top 50 (A) upregulated and (B) downregulated Reactome pathways, sorted by normalized enrichment scores (NES) absolute value. The analysis includes five comparisons: 3-month post-COVID-19 vs. healthy controls, 6-month post-COVID-19 vs. healthy controls, 6-month vs. 3-month post-COVID-19, 3-month post-COVID-19 with PE vs. without PE, and 6-month post-COVID-19 with PE vs. without PE. Significance: p adjusted value <0.05 (∗∗) or p value <0.05 (∗).

Color gradient indicates degree of up- or downregulation. Pathways are clustered into different functionals groups.

Contrast 5: Long COVID vs. non-long COVID

No significant differences in gene expression were observed between patients who suffered long COVID syndrome at the time of sample acquisition and those who did not. Despite extensive analysis of various gene markers and pathways, the comparison did not reveal any substantial variations in the expression profiles of the two groups.

Top DEGs analysis

Furthermore, an in-depth analysis of the top DEGs in each contrast was conducted based on Log2Fold-change values.

GBX2 (Log2FC: −8.89, adj. p value: 0.012) was the most significantly downregulated gene at 3-month post-COVID-19 samples against healthy controls, while CXCL5 (Log2FC: 7.74, adj. p value: 0.022) exhibited the highest upregulation within the same contrast (Tables S13 and S14. DEGs statistical results, related to Figure 2A; Table 1). When comparing 6-month post-COVID-19 and healthy controls, MPZL2 (Log2FC: −9.58, adj. p value: 0.0001) displayed the lowest Log2 FC value, and TIMP3 (Log2FC: 9.27, adj. p value: 0.0001) stood out as the most upregulated gene (Tables S15 and S16. DEGs statistical results, related to Figure 2A; Table 1).

Minor differences were observed between 3 and 6-month post-COVID-19 patients, however, the genes PCDHA9 and GPRASP1 were uniquely upregulated and downregulated, respectively, in 3-month post-COVID-19 patients compared to those at 6 months post-COVID-19 (Tables S17 and S18. DEGs statistical results, related to Figure 2A; Table 1).

Additionally, in the PE vs. non-PE contrast at 3-month, PE samples showed a downregulation of SLITRK4 and an upregulation of IRX2 when compared to non-PE samples (Tables S19 and S20. DEGs statistical results, related to Figure 2A; Table 1). In addition, POTED was noted as the most downregulated gene and GSTM1 was identified as the most upregulated gene at 6-month PE vs. non-PE (Tables S21 and S22. DEGs statistical results, related to Figure 2A; Table 1).

Protein-protein interaction network analysis

A protein-protein interaction network (PPI) was created based on the overlapping genes between the 3 or 6-month post-COVID-19 group versus healthy controls, thus identifying genes that remained altered after SARS-CoV-2 infection (Figure 5A). From the PPI network, enrichment analysis was performed, revealing pathways involved in immunologic response, homeostasis and cellular response, cellular structures and complexes, physiological and metabolic processes, cellular assembly and organization, signaling transduction, and other biological processes (e.g., hematopoietic cell lineage development, identical protein binding) (Table S23. Related to Figure 5 and Cluster identification and Hub genes from STAR Methods). Then, from the PPI, smaller protein clusters were identified (2 clusters) based on the defined thresholds and Cluster 2 was used for studying candidate genes (Figure 5B) due to the higher number of genes identified (12), and its high statically significance at 3 and 6 months (Figure S1; Tables S24 and S25. Related to Figure 5 and Cluster identification and Hub genes from STAR Methods).

Figure 5.

Protein-protein interactions (PPI) networks

(A) PPI network of overlapping genes from 3- and 6-month post-COVID-19. Network connection width is based on STRING database interaction scores between proteins, log2FC is represented by a blue (downregulation) to red (upregulation) color palette, with node size corresponding to Log2FC from the 3-month post-COVID-19 vs. healthy control comparison.

(B) Cluster 2 identified from the defined PPI network.

Candidate genes

Given that ECs play a pivotal role in preserving vascular equilibrium, which encompasses vital functions like inflammation, coagulation, and angiogenesis, this study focused on analyzing a specific subset of DEGs, considered as differentially expressed with the defined thresholds, associated with cellular proliferation, apoptosis, inflammation, and angiogenesis. Accordingly, ANGPT1, CASP1, CXCL5, GBX2, GDF6, IL12A, KFL2, NOS3, PIK3R3, SMAD6, SOD2, SRC, TGFB1, TIMP3, TLR2, VEGFA, PTGS2, and ACKR3 were selected for validation through qPCR (Figure 6).

Figure 6.

Relative mRNA expression levels of selected genes normalized to β-actin, as measured by qPCR

(A) Comparison between 3- and 6-month post-COVID-19 patients and healthy controls.

(B) Comparison between post-COVID-19 patients with and without pulmonary embolism (PE) at 3-month post-infection.

Values expressed as mean ± SD (n = 4–10/group). p < 0.05∗, p < 0.1∗∗, p < 0.001∗∗∗, Unpaired t test for parametric tests, Mann-Whitney test for non-parametric tests.

Post-COVID-19 patients vs. healthy controls

Among all the genes studied, GBX2, GDF6, KFL2, NOS3, SMAD6, SRC, and TGFB1 were downregulated at both 3- and 6-month post-infection compared to healthy subjects. In addition, ANGPT1 was downregulated in patient samples 3 months post-COVID-19 compared to healthy controls, and PIK3R3 was downregulated at 6 months post-COVID-19. Conversely, CASP1, CXCL5, IL12A, SOD2, TIMP3, TLR2, and VEGFA were upregulated in post-COVID-19 patients compared to healthy controls. Specifically, IL12A was upregulated at 3-month, VEGFA at 6-month, and CASP1, CXCL5, SOD2, TIMP3, and TLR2 at both 3 and 6-month post-infection (Figure 6).

PE post-COVID-19 patients vs. non-PE

Interestingly, PTGS2, and ACKR3 were downregulated in 3-month post-COVID-19 patients with PE compared to non-PE post-COVID-19 patients (Figure 6).

Long COVID vs. non-long COVID

No differences were found between patients who suffered long COVID syndrome and those who did not (Figure 6).

Discussion

Our recent previous findings showed that 3-month post-COVID-19 patients exhibited elevated circulating ECFCs compared to healthy subjects, which persisted up to 6 and 12 months, indicating abnormal ECFC mobilization in response to persistent vascular damage months after SARS-CoV-2 infection.^12,13 In this study, we investigated for the first time the transcriptomic profile of patient-derived ECFCs obtained from 3 and 6-month post-COVID-19 patients, and we aimed to identify key gene expression patterns linking endothelial dysfunction following COVID-19 to persistent vascular complications.

SARS-CoV-2 can infect ECs, directly or indirectly, leading to endothelial dysfunction, a key factor contributing to various cardiovascular disorders commonly associated with COVID-19. Endothelial dysfunction, is characterized by the upregulation of chemokines, adhesion molecules, and other proteins governing cell-cell interactions.²⁰ This process contributes to the prothrombotic and proinflammatory state within blood vessels.²⁰

Our findings revealed that the DEGs in post-COVID-19 patients at 3 and 6 months compared to healthy subjects were associated with pathways related to endothelial homeostasis. These pathways include VEGF, PI3K-Akt, leukocyte transendothelial migration, IL-17, regulation of apoptosis, calcium, and cGMP-PKG signaling pathways.

Comparable gene expression profiles were observed between patients at 3 and 6 months following COVID-19 infection, indicating that endothelial injury and associated transcriptional alterations identified at 3 months persist through to 6 months post-infection. This is consistent with our previous findings, showing a significant increase in ECFCs and clinical biomarkers, such as troponin, NT-proBNP, ferritin, and angiogenesis-related proteins at 3 months, which persisted up to 12 months. These results suggest that endothelial damage is sustained up to one-year post-infection.¹³

Our study revealed key alterations in genes associated with angiogenesis, apoptosis, and inflammation, pathways crucial for endothelial homeostasis.

NOS3, KFL2, ANGPT1, PIK3R3, GBX2, GDF6, SMAD6, SRC, and TGFB1 were downregulated at 3-month post-COVID-19 compared to healthy subjects and remained reduced up to 6-month post-infection. The functional roles of these genes and their relevance to the mechanisms investigated in this study are discussed in the following paragraphs.

Nitric oxide (NO) is an essential molecule for endothelial function and vascular homeostasis, regulating vessel tone. In ECs, NO has athero-protective effects, acting as an antioxidant, anti-apoptotic, anti-thrombotic, and anti-inflammatory agent. NO production in ECs is controlled by endothelial nitric oxide synthase (eNOS) expression.²¹ Under pathological conditions, such as inflammation, the expression of eNOS is reduced in response.²¹ Various regulators of eNOS have been identified in this analysis, including kruppel like factor 2 (KLF2), a primary activator of eNOS expression and activity.²¹ Our data reported a downregulation in NOS3 and KFL2 expression in post-COVID-19 patients compared to controls, possibly leading to a reduction in vasodilation and the subsequent endothelial dysfunction observed in these patients. In the same line, Lei et al.²² reported that the spike (S) protein of SARS-CoV-2 damages ECs, leading to a reduced eNOS expression and NO bioavailability. Additionally, Xue et al.²³ demonstrated that human ECs treated with serum from COVID-19 patients showed increased monocyte adhesion, decrease expression of KLF2 and eNOS, and increased expression of ICAM1, and VCAM1. Other studies have also documented decreased endothelial KLF2 expression in lung autopsies of COVID-19 patients compared to control subjects.²⁴ As KLF2 is crucial for endothelial functions, including cell migration, angiogenesis and barrier integrity, its downregulation, along with the reduced eNOS expression, may lead to pro-thrombotic, oxidant, and pro-inflammatory conditions.

Furthermore, ANGPT1, PIK3R3, SRC, TGFβ1, and SMAD6 were found to be downregulated in post-COVID-19 patients compared to healthy individuals in our study. The downregulation of SMAD6 may exacerbate the effects of reduced TGFβ1 signaling, further compromising endothelial barrier integrity and promoting inflammatory responses. Several investigations highlight the critical role of PI3K/AKT signaling in angiopoietin 1 (Ang-1)-mediated processes such as cell migration, survival, and angiogenesis.²⁵ Ang-1 triggers PI3K activation, subsequently activating AKT and eNOS, which in turn promotes EC survival, migration, and sprouting.²⁵ SRC kinase activation serves as a key hub in the signal transduction of various growth factor receptors, acting as a regulator of EC signaling, with downstream pathways including PI3K.²⁶ The downregulation of these proteins in post-COVID-19 patients could negatively impact EC function and survival.

TGFβ1 plays a crucial role in vascular ECs, inducing apoptosis, inhibiting proliferation and migration and promoting anigiogenesis.²⁷ Montalvo-Villalba et al. recently reported that SARS-CoV-2 infected patients showed lower TGFβ1 expression compared to controls.²⁸ TGF-β1 induces eNOS expression by promoting Smad2 translocation to the nucleus, where it interacts with the eNOS promoter.²⁹ Smad transcription factors are key mediators in transmitting signals from the plasma membrane to the nucleus. In post-COVID-19 patients, both TGFβ1 and SMAD6 were downregulated compared to healthy individuals. Smad6, crucial for blood vessel integrity, represses TGF-β signaling.³⁰ This downregulation of TGFβ1 likely contributes to the decreased eNOS expression reported in our series. Additionally, GDF6 (also known as BMP13), a TGF-β superfamily member, play a role in vascular integrity. Krispin et al.³¹ proposed that GDF6 enhances vascular stabilization by modulating VEGF signaling. We found a significant decrease in GDF6 expression, indicating a disturbance in vascular stabilization in patients after SARS-CoV-2 infection.

Additionally, plasma proteomic analyses have suggested that long COVID is associated with vascular alterations driven by hypoxia, potentially impairing both cardiac and neurological functions. For example, Iosef et al.³² reported an upregulation of ANGPT1, a key regulator of vascular homeostasis, with its dysregulated expression contributing to endothelial dysfunction. In contrast, our analyses did not reveal significant differences between long COVID and non-long COVID cohorts.

All the aforementioned findings indicate a clear disturbance in the TGF-β/Smad and PI3K/eNOS pathways in post-COVID-19 patients, leading to decreased NO bioavailability and disrupted EC functions that persist up to 6 months after the infection. These results are consistent with our prior work, which demonstrated downregulation of angiogenesis-related proteins in serum samples at 3, 6, and 12 months following infection, thereby indicating sustained impairment of angiogenic capacity attributable to unresolved endothelial injury.¹³ Collectively, these molecular and functional alterations may represent a mechanistic basis through which long-COVID promotes endothelial dysfunction, underscoring the urgent need for further research into targeted therapeutic strategies.

In our study, we also reported an upregulation in CXCL5, IL12A, TLR2, CASP1, SOD2, TIMP3, and VEGFA in COVID-19 recovered patients compared to controls.

It is known that SARS-CoV-2 induces an inflammatory response leading to the release of various inflammatory molecules including cytokines and chemokines such as CXCL5 or IL12A, which are important for neutrophil recruitment and accumulation.³³ According to our findings, several studies have demonstrated that COVID-19 patients showed elevated levels of these molecules, leading to EC pathological activation.³³ Under inflammatory conditions, EC caspase-1 activation, a key biomarker mediating inflammation in COVID-19, leads to pyroptosis and endothelial dysfunction. Tissue inhibitor of metalloproteinase-3 (TIMP3) exhibits anti-angiogenic effects by interacting with VEGF receptor-2 (KDR), inhibiting EC functions, and promoting apoptosis. In our data, both TIMP3 and CASP1 were upregulated in post-COVID-19 patients, indicating increased EC apoptosis and decreased angiogenesis. Caspase-1 is a key biomarker mediating inflammation in COVID-19, with studies showing increased inflammasome and caspase-1 activation in COVID-19 patients.³⁴ Increased VEGF levels in SARS-CoV-2-infected individuals contribute to vascular hyperpermeability and systemic inflammation.^35,36 We also found upregulated VEGF expression in post-COVID-19 patients, likely due to the inflammatory environment of COVID-19.³⁶ In addition, stimulation of toll-like receptors (TLRs) pathways, a component of innate immunity, also triggers the release of pro-inflammatory cytokines and has a role in the pathogenesis of SARS-CoV-2 infection.³⁷ Different studies have reported that TLR2 mediated SARS-CoV-2 inflammation.^38,39 In line with these data, we reported an upregulation of TLR2 in post-COVID-19 patients, indicating ongoing inflammation and immunity response in patients up to 6-month after recovery.

SARS-CoV-2 can also disturb redox homeostasis, increasing oxidative stress. Immunohistochemical studies have shown that COVID-19 elevates levels of the primary antioxidant enzyme manganese superoxide dismutase (SOD-2). In our study, we observed an upregulation of SOD2 in post-COVID-19 patients compared to healthy controls, indicating a need to counteract the elevated oxidative stress following infection.

SOD2 and TLR2 have been also linked to muscle-related problems such as myalgic encephalomyelitis. Mitochondria, as central regulators of cellular energy metabolism, have been increasingly implicated in the pathophysiology of post-viral fatigue syndromes. A reduction in SOD2, a mitochondrial-specific enzyme, has been linked to skeletal muscle dysfunction and is associated with significant impairments in exercise capacity, muscle strength, and mitochondrial activity.^40,41 Notably, decreased SOD2 levels have also been observed in some cases of ME/CFS.^40,41 Zheng et al.⁴² also identified TLR-2 as a key target in their analysis of genes associated with ME/CFS.

Furthermore, we subdivided post-COVID-19 patients in those who have suffered PE during admission and those who did not, in order to assess the differences related to PE. In our series, PE subjects showed a downregulation in PTGS2, and ACKR3 expression compared to non-PE subjects. Prostaglandin-endoperoxide synthase (PTGS)-2 role in arterial thrombosis has been well established.⁴³ Amadio et al.⁴³ demonstrated that depletion of PTGS2 induces a hypercoagulable state, predisposing to venous thrombosis. ACKR3 also plays a role in platelet activation and thrombosis. Rohlfing et al.⁴⁴ reported that ACKR3 is a critical regulator of platelet activation inhibiting thrombus formation. In our cohort, PE post-COVID-19 showed lower PTGS2, and ACKR3 expression after 3 months SARS-CoV-2 infection when compared to non-PE patients, indicating higher risk of thrombosis and coagulation in patients who have undergone PE. Also, we observed that Notch signaling pathway was downregulated at both 3 and 6-month post-COVID-19 with PE, suggesting that COVID-19 may hyperactivate IL-6 and pro-inflammatory mediators, with persistent Notch downregulation even 6 months after the infection on those who suffered PE during hospitalization.⁴⁵

Finally, we examined gene expression differences among post-COVID-19 patients categorized based on the presence or absence of long COVID symptoms. Long COVID syndrome, which affects multiple organ systems and shares clinical features with conditions such as myalgic encephalomyelitis and chronic fatigue syndrome, has been associated with persistent endothelial dysfunction and related pathophysiological processes.⁷ Additionally, emerging evidence suggests that mitochondrial dysfunction may contribute to systemic inflammation, a common feature in conditions such as ME/CFS and long COVID.⁷ This association between impaired mitochondrial function and chronic inflammation may underlie the persistent fatigue and wide-ranging symptoms observed in these syndromes, potentially through mechanisms involving disrupted energy homeostasis and sustained inflammatory signaling.⁷ However, despite this well-documented association between long-COVID and endothelial dysfunction, our analysis revealed no significant differences in gene expression between the two groups. The lack of significant differences in gene expression between long-COVID and non-long COVID patients suggests that the persistent symptoms observed in some individuals may not be directly attributable to major alterations in the endothelial transcriptomic profile. These findings underscore the need for multi-omics approaches, including proteomics and metabolomics, to better understand the biological basis of long-COVID and its clinical manifestations.

Additionally, emerging evidence supports a pivotal role for mitochondrial dysfunction in cardiovascular disease.⁴⁶ In the context of long COVID, persistent endothelial impairment may be amplified by altered mitochondrial dynamics and bioenergetic deficits in cardiac tissue.⁴⁶ Specifically, studies have reported disrupted mitochondrial function after SARS-CoV-2 infection, such as diminished ATP synthesis and increased reactive oxygen species, which may contribute to cardiovascular pathologies like myocarditis and microvascular injury.⁴⁶

Future research should focus on elucidating the role of mitochondrial signaling pathways in post-COVID syndrome. This includes investigating alterations in mitochondrial function, bioenergetics, and signaling mechanisms such as reactive oxygen species production and mitophagy. Understanding how these pathways contribute to persistent inflammation may reveal relevant biomarkers and therapeutic targets.

Overall, our findings identify differential expression patterns associated with disrupted EC functions in ECFCs derived from post-COVID-19 patients. Our findings corroborate and expand upon existing evidence of endothelial dysfunction in COVID-19 patients, demonstrating decreased angiogenesis capacity, increased apoptosis, and oxidative stress. This confirms persistent endothelial damage after recovery and may help explain certain pathological events observed in these individuals months or even years after infection.

Limitations of the study

A key limitation of this study is the relatively small sample size, which may constrain statistical power. Although standardized protocols for ECFC isolation, culture, and RNA extraction were applied to minimize variability and enhance reproducibility, the inherent biological heterogeneity of patient-derived ECFCs remains a persistent challenge in this field. Increasing the cohort size in future studies will be essential to validate these findings and strengthen the robustness of the conclusions. Another limitation pertains to the restricted number of paired samples available at both the 3- and 6-month post-infection time points. Despite initial enrollment of post-COVID-19 patients for longitudinal follow-up, attrition occurred due to voluntary withdrawal or mortality. Additionally, in certain cases, ECFCs isolated from patient-derived samples exhibited insufficient proliferative capacity, thereby limiting their utility in downstream analyses.

Resource availability

Lead contact

Further information and requests should be directed to and will be fulfilled by the lead contact, Olga Tura-Ceide (olgaturac@gmail.com).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •RNA-seq data reported in this paper have been deposited in the Genome-Phenome Archive (EGA) database (accession ID: EGAD50000001452) and are available under restricted access to protect individual information. The data is publicly available as of the date of publication.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This research was supported by the funding from two Miguel Servet type grants from the Institute of Health Carlos III (CP17/00114, CPII22/00006) (Co-funded by European Social Fund “Investing in your future”), Spanish Society of Respiratory Medicine (SEPAR), Menarini laboratories, Convocatoria de Becas Gilead a la Investigación Biomédica,GLD24/00100, Catalan Society of Pneumology (SOCAP), Catalan Pneumology Foundation (FUCAP) and from the Institute of Health Carlos III (PI21/01212). P. Poyatos was a recipient of a Banco Santander-University of Girona grant (IFUdG2021). Cofunding was provided by the Fondo Europeo de Desarrollo Regional (FEDER); “Una manera de hacer Europa”. The authors acknowledge the Clinical laboratory from Parc Hospitalari Martí i Julià of Salt for their support, healthy volunteers for providing the samples and Andreu Cardona for his assistance.

Author contributions

Author contributions are as follows: P.P.: isolation of ECFC from all patients and healthy controls, experiments performance, data analysis, drafting of manuscript, editing and critical revision. M.G.: RNA-sequencing analysis, data analysis, drafting and editing of manuscript. D.A.: RNA-sequencing analysis. N.L.: isolation of ECFC from all patients and healthy controls and critical revision of the manuscript. M.B.-V. and S.E.: patient recruitment of all subjects included in the study. M.C. and R.O.: critical revision of the manuscript. O.T.-C.: report conception and design, and critical revision of the manuscript.

Declaration of interests

All the authors have read the journal’s policy on conflicts of interest, declaring no conflicts of interest. All the authors have read the journal’s authorship agreement.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Peripheral blood samples from post-COVID-19 patients and healthy controls	Pneumology Group - Hospital Universitari de Girona Dr. Josep Trueta	N/A
Chemicals, peptides, and recombinant proteins
ECM-2 Medium	ScienceCell Research Laboratories	Cat# 1001-4
Collagen 6-well tissure culture plates	Corning	Cat# 354400
Fetal Bovine Serum (FBS) Hyclone	Cytiva	Cat# SH30071.02
Penicilin-Streptomycin (P/S)	Lonza	Cat# 15070063
Ficoll-Paque PLUS	Lonza	Cat# GE17144002
Dimethyl sulfoxide (DMSO)	Pan Biotech	Cat# P6036720100
Critical commercial assays
RNeasy Mini Kit	Qiagen	Cat# 74104
High-Capacity cDNA Reverse Transcription Kit	Qpplied Biosystems	Cat# 4368814
Power SYBR Green PCR Kit	Applied Biosystems	Cat# 4309155
Deposited data
RNA-seq raw data	This Paper	EGAD50000001452
Protein-related interaction data	STRING v12.0 database	https://string-db.org/
RNA-related pathway data 1	Kyoto Encyclopedia of Genes and Genomes (KEGG)	https://www.genome.jp/kegg/
RNA-related pathway data 2	Reactome Knowledgebase	https://reactome.org
Experimental models: Cell lines
Endothelial Colony-Forming Cells (ECFCs): derived from peripheral blood of COVID-19 patients and healthy controls	This paper	N/A
Oligonucleotides
qRT-PCR primers	This paper	See Table S1
Software and algorithms
STAR aligner v2.7.9a	Dobin et al. 2013	https://github.com/alexdobin/STAR
RStudio	R core team	https://www.R-project.org
RSEM	Li & Dewey 2011	https://github.com/deweylab/RSEM
Trim Galore v0.6.7	Martin Marcel 2011	https://cutadapt.readthedocs.io/en/stable/
BBDuk	Bushnell et al. 2017	https://github.com/BioInfoTools/BBMap
Limma	Ritchie et al. 2015	https://bioconductor.org/packages/release/bioc/html/limma.html
ComBat-seq	Zhang et al. 2020	https://github.com/zhangyuqing/ComBat-seq
FGSEA	Korotkevich et al. 2019	https://github.com/alserglab/fgsea
Cytoscape v3.10.2	Shannon et al. 2003	https://cytoscape.org/
Cytoscape’s Molecular Complex Detection (MCODE)	Bader Lab	https://apps.cytoscape.org/apps/mcode
RNA-seq (NextSeq2000) sequencing service	GraphPad Software	https://www.graphpad-prism.cn/
Other
RNA-seq (NextSeq2000) sequencing service	Centre of Genomic Regulation (CRG), Barcelona	https://www.crg.eu

Experimental model and study participant details

Study population and human samples

For this genomic study, post-COVID-19 patients were recruited at two distinct time points after recovery: 3 months (n=6) and 6 months (n=6). Three individuals participated at both time points, while the remaining patients were different at each time point. Since not all samples corresponded to the same individuals, the 3- and 6-month post-COVID-19 groups were treated as independent cohorts in the statistical analysis. These groups were compared with a control group of 5 healthy individuals (CL). Patients were further classified as long-COVID or non-long-COVID based on the WHO definition.⁴⁷ Three patients in the 3-month group and two in the 6-month group met the criteria for long-COVID. All participants were matched based on age, gender, body mass index (BMI), and clinical comorbidities like arterial hypertension (AHT), diabetes mellitus (DM), or dyslipidemia (DLP). COVID-19 patients were hospitalized and admitted during the months of March and April 2020, when the pandemic began. None of the subjects had received prior vaccination and all of them were infected with the same SARS-CoV-2 Wuhan-variant. All patients were discharged from the Pulmonary Medicine Service with severe pneumonia and a diagnosis of COVID-19 by positive PCR. Furthermore, COVID-19 patients were categorized depending on whether they received a pulmonary embolism diagnosis detected by CT examination. Healthy control subjects, who were non-hospitalized and non-staff volunteers from the same health region as the COVID-19 patients, tested negative for SARS-CoV-2 infection at the time of ECFC isolation through PCR. They also had no history of prior COVID-19 infection. Detailed clinical information of patients is provided in Table S2.

The study was approved by the Clinical Research Ethic Committee from Hospital Universitari de Girona Dr. Josep Trueta (CEIm_COVID-Pneumo 2020.0099) in accordance with the Declaration of Helsinki. All subjects gave written informed consent before inclusion to the study.

Cell lines

Cells were seeded onto type-1 rat-tail collagen-coated six-well tissue culture plates (BD Biosciences) and incubated at 37°C, 5% CO₂, and 95% relative humidity for 3–4 weeks.⁴⁸ The medium was changed every 2 days until the colonies appeared. Cells were expanded in ECM-2 culture medium supplemented with 10% FBS and cryopreserved in 90% FBS with 10% DMSO.

Method details

Isolation of ECFC

ECFC isolation from all participants was conducted following established protocols.^48,49 Peripheral blood mononuclear cells (PBMCs) were separated through buoyant density centrifugation using Ficoll-Paque Plus (GE Healthcare), resuspended in endothelial cell medium (ECM-2 medium, ScienceCell, Research Laboratories) supplemented with 20% fetal bovine serum (FBS hyclone, Cytiva), and 1% penicillin-streptomycin (P/S, Lonza).

RNA isolation and transcriptomic analysis

Total RNA was purified from 3 and 6-months post-COVID-19 patients and healthy controls using RNeasy Mini kit (Qiagen), according to the manufacturer’s recommendations. RNA concentration and purity were evaluated using a NanoDrop (ThermoFisher Scientific). The integrity of RNA was detected using a BioAnalyzer 2100 instrument (AgilentTechnologies). All samples had an RNA integrity number above 9.5. Total RNA was sequenced in the Genomics facility from Centre for Genomic Regulation (CRG) using Illumina’s NextSeq2000.

Libraries were prepared using NEBNext® Poly(A) mRNA Magnetic Isolation Module (ref. e7490) and NEBNext® Ultra II Directional RNA Library Prep Kit for Illumina (24reactions ref. e7760 or 96 reactions ref. e7765) according to the manufacturer’s protocol, to convert total RNA into a library of template molecules of known strand origin and suitable for subsequent cluster generation and DNA sequencing. Briefly, 90 to 100ng of total RNA were used for poly(A)-mRNA selection using poly-T oligo attached magnetic beads using two rounds of purification. During the second elution of the poly-A RNA, the RNA was fragmented under elevated temperature and random primed. Then, the cleaved RNA fragments were copied into first strand cDNA using reverse transcriptase. After that, second strand cDNA was synthesized, removing the RNA template and synthesizing a replacement strand, incorporating dUTP in place of dTTP to generate ds cDNA using DNA Polymerase I and RNase H. dsDNA was subjected to the addition of “A” bases to 3′ ends (A-tailing) and ligation of the NEB adapter. USER enzyme was used to excise the circular adapters. Finally, PCR selectively enriched those DNA fragments that had adapter molecules on both ends. The PCR was performed using NEBNext Multiplex Oligos for Illumina (96 Unique Dual Index Primer Pairs, ref. E6440, E6442, E6444, E6446) and the master mix provided with the kit. Final libraries were analyzed using Bioanalyzer DNA 1000 or Fragment Analyzer Standard Sensitivity (ref: 5067-1504 or ref: DNF-473, Agilent) to estimate the quantity and validate the size distribution, and were then quantified by qPCR using the KAPA Library Quantification Kit KK4835 (REF. 07960204001, Roche). Libraries were sequenced 1 ∗ 50+8+8 bp on Illumina’s NextSeq2000.

Mapping and expression quantification

Raw fastq files were processed as follows: (1) BBDuk was used to identify and remove possible rRNA contamination⁵⁰; (2) Trim-Galore 0.6.7 was used to remove adapters⁵¹; (3) Reads were mapped to the GRCh38 human genome with STAR 2.7.9a.⁵² Gene expression levels were quantified with RSEM.⁵³ Gene expression was transformed to logCPM and normalized using the quantile method with voom from limma package⁵⁴ in R. Genes with no variation across samples (sd = 0) were removed. Non-protein-coding genes were also removed. We applied ComBat⁵⁵ to adjust expression data for covariates (Sex, Smoke, Age, BMI).

Differential expression analysis and functional annotation

Differential expression analysis was carried out with lmFit from limma package. A gene was considered as differentially expressed with |log2FC| >1 and adjusted P < 0.05. Pathway enrichment analysis was performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database⁵⁶ and the Reactome Knowledgebase,⁵⁷ through FGSEA algorithm.⁵⁸

Protein-protein interaction network (PPI) and clustering identification

Overlapping genes between contrasts were graphically identified through Venn Diagram, based on that dataset, PPI Network from overlapping genes was created using the online database of Search Tool for the Retrieval of Interacting Genes (STRING), version 12.0 (string-db.org).⁵⁹ PPI networks of upregulated and downregulated DEGs were generated as full string networks, with edges indicate both physical and functional interactions with the highest confidence score of 0.900 and hidden the unconnected nodes in the data. Then, PPI network was imported to Cytoscape, a software for visualization bimolecular interaction networks (version 3.10.2).⁶⁰

Cluster identification and hub genes

Core groups from the PPI network were recognized. Relevant modules were extracted from the PPI network using Cytoscape’s Molecular Complex Detection (MCODE) plugin, a novel graph theoretic clustering algorithm was used for the identification of protein clusters, with a degree cut-off of 2, max depth of 100, node score cut-off of 0, and k-core of 2, MCODE score ≥ 4.5, and nodes ≥ 10 thresholds.

Quantitative Real Time PCR

Reverse transcription was carried out using the Applied Biosystems high-capacity cDNA reverse transcription kit. For qRT-PCR, a power SYBR® Green PCR Kit (Applied Biosystems), along with specific primers, was employed on the QuantStudio 7 Flex thermocycler (Applied Biosystem). Normalization of cDNA copy numbers was performed relative to the genomic DNA level of endogenous β-actin and analyzed using the 2^-ΔΔCt method. All primers were provided by IDT, and their sequences are detailed in Table S1.

Quantification and statistical analysis

Statistical analyses were performed using GraphPad Prism 7 software, version 7.0e. The statistical tests applied are indicated in each figure legend. Clinical data and differences in gene expression between groups were compared using Student’s t-test for normally distributed variables, or Mann Whitney U test for non-normally distributed variables, and Chi squared test in categorical variables. More than two groups were compared using One-way ANOVA with Tukey's post-hoc test or non-parametric analysis of variance Kruskal–Wallis test with a Dunn's post-hoc multiple comparison test. Significance thresholds were set at p < 0.05 (∗), p < 0.01 (∗∗), p < 0.001 (∗∗∗), and p < 0.0001 (∗∗∗∗). Sample sizes (n) are reported in the corresponding figure legends. Data is presented as mean ± SD.

Published: October 8, 2025