Introduction

In cases of acute myocardial infarction with ST-segment elevation (STEMI), morbidity and mortality have substantially decreased following the establishment of regional and national reperfusion networks, and the use of newer evidence-based drugs1,2. However, ventricular fibrillation (VF) during the acute phase of myocardial infarction—also known as primary ventricular fibrillation (PVF)—is still the leading cause of sudden prehospital cardiac death, and is a factor that predicts poor short-term prognosis3,4. In this context, numerous studies have attempted to identify predictors of PVF, without significant progresses. It is likely that susceptibility to VF during acute ischemia might be modulated by several factors, including hemodynamic dysfunction, electrolyte alterations, autonomic dysregulation, genetic factors, and certain environmental influences.

Particularly, an association between viral infections and acute myocardial infarction (AMI) has been proposed5,6. Furthermore other authors have found seasonal variations in sudden cardiac death (SCD)7, typically with a peak in winter, suggesting that viral exposure is a trigger of VF in patients suffering from acute ischemia8. However, only influenza virus and some enteroviruses have been investigated for roles in SCD, with contradictory results9,10. To date, evidence is also scarce and unclear regarding an association between inflammatory biomarkers and SCD in an asymptomatic population. For instance, interleukin 6 (IL-6) is reported as a predictor of sudden death in healthy men11. Additionally, growth differentiation factor 15 (GDF-15) has been described as a risk factor for SCD during the acute phase of myocardial infarction12, and as a predictor of short-term mortality in patients with PVF13. Nevertheless, the association between systemic inflammation at the time of STEMI and PVF remains unknown.

In the present study, we aimed to conduct a pilot study with two objectives: (1) to examine the circulating virome sequence, and (2) to explore the systemic inflammatory proteome in patients with STEMI, with and without PVF.

Methods

Patient population

The RUTI-STEMI-PVF cohort is a prospective single-centre registry of consecutive STEMI patients treated with primary percutaneous coronary intervention (PCI) and within the Codi IAM reperfusion network14,15. STEMI was defined according to the Third Universal Definition of Myocardial Infarction16. Patient management was decided by the physicians, following recommended guidelines17,18. Upon admission, blood samples were obtained by venipuncture and centrifuged, and then heparin-plasma was stored at − 80 °C until assay.

Patients were divided into two groups: those who had suffered PVF, and those who had not (non-PVF). PVF was defined as ventricular fibrillation occurring ≤ 24 h after diagnosis of myocardial infarction, and not preceded by heart failure or shock.

Blood samples for virome analyses were obtained from non-PVF (n = 9) and PVF (n = 11) STEMI patients at the time of primary PCI. The patients had a mean age of 60 ± 10 years and were 85% men. Patients with a first STEMI were selected, matched by sex, age, diabetes, and anterior myocardial infarction. Table 1 shows clinical and demographic characteristics of the studied groups. The inflammatory proteome was analysed among the same patients with available samples, PVF and non-PVF, as well as in non-STEMI healthy controls (without history of cardiovascular disease or cancer; mean age, 58.6 ± 1.2 years, 60% men). Written informed consent was obtained from all patients. The study was approved by the local ethics committee (The Ethics Committee of the Clinical Investigation of Germans Trias i Pujol Hospital) and was conducted in accordance with the Declaration of Helsinki.

Table 1 Clinical characteristics of PVF and non-PVF STEMI patients.
Full size table

Nucleic acid extraction

DNA was purified from total blood collected in BD Vacutainer EDTA tubes, using the FlexiGene DNA Kit (QIAGEN GmbH, Germany). DNA concentration was evaluated using the Qubit BR Assay Kit (Thermo Scientific, Wilmington, DE, USA). Integrity was checked by gel electrophoresis.

Molecular assays

We analysed a total of 20 human blood samples, from 9 patients with STEMI and non-PVF, and from 11 patients with STEMI and PVF, using VirCapSeq-VERT (Virome Capture Sequencing Platform for Vertebrate Viruses). VirCapSeq-VERT is a virome capture sequencing platform targeting viral taxa that infect vertebrates, using a database of 342,438 representative sequences spanning all virus sequence records19. When compared with other enrichment procedures, the utilized procedure allows for reduction of background human DNA, as well as a 100- to 10,000-fold enrichment in viral reads. This system enables the identification and genetic characterization of all known vertebrate viruses and their genetic variants (the genomes of 207 viral taxa known to infect vertebrates, including humans). Samples were processed using Illumina HiSeq/NovaSeq.

Bioinformatics data analysis

Bioinformatic data analysis involved the following workflow: identify and remove host background reads, quality check and trimming, de novo assembly, homology search for putative viral genomes, mapping of filtered reads and generation of counts, and analysis of viral communities. Kraken tools were used to remove sequenced human and bacterial reads from among the total sequencing reads generated for each sample20. A quality check and adapter trimming were performed using the quality control tool FASTQC21. Assembly of the host depleted trimmed reads was performed using SPAdes software22 version 3.15.2. Generation of the index and the mapping was done using BWA software23 version 0.7.17. Amplification duplicates that might confound the count were remove using SAMtools software24 version 1.12. Finally, mapping statistics were generated using the MultiQC tool25.

Epstein–Barr real time PCR and immunoassay

Real time PCR was performed in non-PVF (n = 6) and PVF (n = 9) DNA samples by EBV Amplification Reagent Kit (Abbott Molecular, 08N54-085).

IgG-class antibodies to Epstein–Barr virus nuclear antigen (EBV-EBNA-1) were determined in plasma. Non-PVF (n = 94) and PVF (n = 82) plasma samples were by Epstein–Barr Virus (EBNA-1) IgG ELISA (Demeditec, DE4246, lot 109G/K041).

Inflammation proteomic analysis

The inflammatory proteomic profiles of non-PVF and PVF patients were analysed using the Olink Inflammation panel, based on Proximity Extension Assay technology. This multiplex immunoassay enables analysis of 92 inflammation-related proteins26,27. Non-PVF patients (n = 7), PVF patients (n = 7), and healthy subjects (n = 5) were analysed using the Olink Inflammation panel.

Statistical analysis

Summary data were represented by mean and standard error of the mean (SEM), or by median and interquartile range (IQR) depending on the data normality. The D’Agostino and Pearson test was used to evaluate the normality of data. Two-groups comparisons were performed using the unpaired t-test or Mann Whitney test, and three-groups comparisons were performed using Kruskal–Wallis test or ANOVA, depending on the data normality. Fisher's exact test was used when required. Statistical significance was assumed when P was < 0.05. Statistical analyses were performed using Prism 9 for macOS version 9.0.2 (134) and 9.3.1 (350).

Results

Circulating virome

The virome capture sequencing platform VirCapSeq-VERT was used to target viral taxa in human blood samples from non-PVF patients (n = 9) and PVF patients (n = 11). The capture results were sequenced using Illumina HiSeq/NovaSeq. Human and bacterial reads were removed from the sequencing files, and the remaining reads ranged from 148–543 k pairs of reads per sample (Supplementary Table 1). Along all reads and samples, we found good quality per base position. However, we detected a high amount of PCR duplicates, due to the amplification and enrichment protocol (Supplementary Table 2). We also identified and removed common sequencing adapters. Details in the statistics regarding the trimming process for each sample are shown in Supplementary Table 3. The host depleted trimmed reads were assembled using SPAdes software to generate longer sequences, and for an additional and improved homology search (Supplementary Table 4). All viral reference genomes available in GenBank NCBI database were used to create a BLAST database, which was used for the homology search, with the generated assembled host depleted trimmed reads as input. We observed a total of 51 different genome entries (Table 2).

Table 2 Taxonomic information of the sequence identified.
Full size table

The identified genome entries were then used to create an alignment index, to map the reads corresponding to their exact position in the reference genomes. Nearly half of the reads did not map to retrieved viral sequences, likely because the host depleted reads may have contained archaea, yeast, or unclassified taxon reads. These results correspond with the low number of reads identified as viral (Supplementary Table 1). Supplementary Tables 5 and 6 summarize the length of each reference sequence, and the percent of base pairs covered (%), in non-PVF and PVF patients. Six genome entries found in the homology search did not generate reads mapping, since the better sensitivity of the mapping enabled more confident placing of a read compared to with BLAST.

Alpha diversity

We further assessed alpha diversity to determine the diversity and to enable comparisons of the type and quantity of virus species between non-PVF and PVF patients. Alpha diversity is a statistic used in this kind of sample, in which reads reflect the abundance of each of the identified operational taxonomical units (OTUs). Richness and diversity are alpha diversity metrics.

As a result, species richness did not significantly differ between non-PVF patients and PVF patients (26.44 ± 2.69 vs. 20.45 ± 2.39; P = 0.112) (Fig. 1A). PVF and non-PVF patients also did not significantly differ in other richness indexes, such as the Chao1 Richness Estimate (28.23 ± 2.63 vs. 21.93 ± 2.54; P = 0.105) (Fig. 1B) and Abundance Coverage Estimator (ACE) index (29.02 ± 2.69 vs. 22.00 ± 2.44; P = 0.069) (Fig. 1C). We also used a simple linear regression model to explore whether the species richness correlated with the number of raw read pairs sequenced. We identified a slight correlation between the number of raw read pairs sequenced and the observed richness (R2 = 0.14; P = 0.099) (Fig. 1D), Chao1 (R2 = 0.13; P = 0.113) (Fig. 1E), and ACE index (R2 = 0.20; P = 0.049) (Fig. 1F). Concerning the species diversity, we found no significant differences between non-PVF and PVF patients using Shannon’s Diversity Index (0.614 ± 0.044 vs. 0.552 ± 0.018; P = 0.252) (Fig. 2A), the Simpson Index (0.239 ± 0.019 vs. 0.212 ± 0.006; P = 0.456) (Fig. 2B), or the Inverse Simpson Index (1.322 ± 0.036 vs. 1.271 ± 0.011; P = 0.423) (Fig. 2C).

Figure 1
figure 1

Species richness alpha diversity. (A–C) Species richness represented by the following metrics: (A) Observed richness values, (B) Chao1 Richness Estimate (Chao1), and (C) Abundance Coverage Estimator (ACE). (D–F) Simple linear regression model between the number of sequenced raw read pairs and (D) observed richness, (E) Chao 1, and (F) ACE.

Full size image
Figure 2
figure 2

Species diversity alpha diversity. Species diversity represented by (A) Shannon’s Diversity Index, (B) the Simpson Index, and (C) the Inverse Simpson Index.

Full size image

Frequent sequences

In addition, due to species richness of viruses did not significantly differ between non-PVF and PVF patients, we explored whether specific virus families were differentially expressed. Supplementary Table 7 summarizes the top 10 OTUs. The predominant OTU was Human endogenous retrovirus K113 (NC_022518.1) (Fig. 3A), and its frequency did not significantly differ between non-PVF and PVF patients (0.869 ± 0.012 vs. 0.885 ± 0.004; P = 0.381).

Figure 3
figure 3

Heatmap representations of (A) the most frequent OTUs at genus level normalized by Human endogenous retrovirus K113 (NC_022518.1) frequency and (B) the frequencies of the 27 Torque teno virus identified. Prism 9 for macOS version 9.3.1 (350).

Full size image

Among the most frequent OTUs, the genus most commonly found was Lymphocryptovirus (Fig. 3A), belonging to the Herpesviridae family. In particular, we detected the complete genomes of three viruses of this family—NC_007605.1, NC_009334.1, and NC006146.1—corresponding to Human gammaherpesvirus 4 (Epstein–Barr virus), Human herpesvirus 4 type 2 (Epstein–Barr virus type 2), and Macacine gammaherpesvirus 4 (Rhesus lymphocryptovirus), respectively (Table 2).

Non-PVF and PVF patients only significantly differed in the frequencies of Human herpesvirus 4 type 2 and Macacine gammaherpesvirus 4. As shown in Fig. 4A, Human herpesvirus 4 type 2 (NC_009334.1) was significantly more common in non-PVF patients than in PVF patients (0.0011 ± 0.000 vs. 0.0004 ± 0.000; P = 0.042), while Macacine gammaherpesvirus 4 (NC_006146.1) was significantly less frequent in non-PVF patients than in PVF patients (0.009 ± 0.000 vs. 0.011 ± 0.000; P = 0.024) (Fig. 4B). Non-PVF and PVF patients did not significantly differ in the frequency of Human gammaherpesvirus 4 (NC_007605.1) (0.0045 ± 0.000 vs. 0.0042 ± 0.000; P = 0.939) (Fig. 4C). Human gammaherpesvirus 4 (NC_007605.1) results were also validated by RT-PCR and ELISA.

Figure 4
figure 4

Relative frequency of (A) Human herpesvirus 4 type 2 (NC_009334.1), (B) Macacine gammaherpesvirus 4 (NC006146.1), (C) Human gammaherpesvirus 4 (NC_007605.1), (D) Torque teno virus 18 and (NC_043414.1), and (E) Torque Teno virus 8 (NC_014084.1).

Full size image

Human gammaherpesvirus 4 (NC_007605.1) viral load of 15 patients was evaluated by RT-PCR. Viral load was detected in 13 patients and no significant differences were observed between non-PVF and PVF patients (2147 ± 887.2 vs. 1113 ± 683.6 UI/mL; P = 0.731) (Fig. 5A).

Figure 5
figure 5

(A) Epstein–Barr viral load (UI/mL) and (B) IgG-class antibodies to Epstein–Barr nuclear antigen levels (units) in non-PVF patients and PVF patients.

Full size image

IgG-class antibodies to Epstein–Barr nuclear antigen were detected in 94.8% of the analyzed population. Similar to the non-significant differences found in Human gammaherpesvirus 4 (NC_007605.1) frequencies between non-PVF and PVF patients by sequencing, IgG-class antibody detection levels did not significantly differ between non-PVF and PVF patients (31.42 ± 1.121 vs. 33.40 ± 1.125 Units; P = 0.196) (Fig. 5B).

Supplementary Table 8 details the relative frequencies of the rest of the identified OTUs. Among the less common OTUs, the Alphatorquevirus genus (Fig. 3B), belonging to the Anelloviridae family, was most prominently represented, as we identified the complete genome or specific genes of up to 27 Torque teno virus, also referred to as transfusion transmitted viruses or TTVs. PVF and non-PVF patients showed significantly different frequencies of Torque teno virus 18 (NC_043414.1) and Torque teno virus 8 (NC_014084.1). The relative frequency of Torque teno virus 18 was significantly higher in non-PVF patients than in PVF patients (0.00058 ± 0.0002 vs. 0.00013 ± 0.00006; P = 0.0297) (Fig. 4D). Similarly, the relative frequency of Torque teno virus 8 was significantly increased in non-PVF patients compared to PVF patients (0.00038 ± 0.00011 vs. 0.00004 ± 0.00003; P = 0.0097) (Fig. 4E).

Systemic inflammation

The inflammatory proteomic profile was analysed using the Olink Inflammation panel (Supplementary Table 9). Our results did not show that PVF and non-PVF patients significantly differed in any of the analysed inflammatory-related proteins (Table 3). Indeed, 53 of the 92 analysed proteins showed no differences between any of the studied groups (Supplementary Figs. 1, 2, 3 and 4). On the other hand, 16 of the 92 proteins significantly differed between healthy subjects versus both non-PVF and PVF patients (Supplementary Fig. 5). No differences were found between non-PVF and PVF patients; however, some proteins significantly differed between the healthy control group and one of the AMI groups. Compared to healthy subjects, PVF patients showed significantly higher circulating levels of CUB domain containing protein 1 (CDCP1) (P = 0.0364) (Fig. 6A) and Interleukin-18 receptor 1 (IL18-R1) (P = 0.0488) (Fig. 6B).

Table 3 Inflammatory-related protein values in healthy subjects and non-PVF and PVF patients.
Full size table
Figure 6
figure 6

Protein levels (pg/mL) of (A) CUB domain containing protein 1 (CDCP1), (B) Interleukin-18 receptor 1 (IL18-R1), (C) Monocyte chemotactic protein 1 (MCP-1), (D) C–C motif chemokine 4 (CCL4), (E) Interleukin 10 (IL-10), (F) Tumor necrosis factor receptor superfamily member 9 (TNFRSF-9), (G) Neurotrophin-3 (NT-3), and (H) C–C motif chemokine 19 (CCL19) in healthy subjects, non-PVF patients, and PVF patients. *P < 0.05; **P < 0.01.

Full size image

Monocyte chemotactic protein 1 (MCP-1), C–C motif chemokine 4 (CCL4), Tumor necrosis factor receptor superfamily member 9 (TNFRSF-9), Interleukin-10 (IL-10), Chemokine (C–C motif) ligand 19 (CCL19), and Neurotrophin-3 (NT-3) each showed a different expression profile between healthy subjects and non-PVF patients, but did not exhibit significant differences when compared with PVF patients. MCP1 (P = 0.0221) (Fig. 6C), CCL4 (P = 0.0281) (Fig. 6D), IL-10 (P = 0.0279) (Fig. 6E), and TNFRSF-9 (P = 0.0499) (Fig. 6F) levels were significantly promoted in non-PVF patients in comparison with the control groups. Conversely, NT-3 (P = 0.0092) (Fig. 6G) and CCL19 (P = 0.0093) (Fig. 6H) levels were significantly lower in non-PVF patients compared with in healthy subjects.

Discussion

Primary ventricular fibrillation (PVF) is among the leading causes of prehospital sudden cardiac death. It is presently unknown what factors increase the probability of PVF development during acute ischemia, complicating the identification of PVF predictors. We thus aimed to evaluate possible PVF predictors or triggers, including the complete DNA virome and the inflammatory proteome in PPCI-treated STEMI patients.

A growing number of viruses have been determined to be associated with inflammatory cardiomyopathy. Previous data suggest that viral exposure could increase PVF susceptibility, although this has not been conclusively proven. In this context, Andréoletti et al. identified coxsackievirus B infection in post-mortem endomyocardial tissue of patients who died suddenly due to AMI10. Additionally, the AGNES (Arrhythmia Genetics in the NEtherlandS) study showed that PVF during first STEMI was most significantly associated with SNP rs2824292 at chromosome 21q21, where the CXADR gene is found. CXADR encodes the coxsackie and adenovirus receptor protein, which has been implicated in myocarditis28, dilated cardiomyopathy28, and ventricular conduction and arrhythmia vulnerability29. However, this association was not replicated in at least two additional studies30,31. Extreme influenza epidemics are also reportedly associated with out-of-hospital cardiac arrest8. However, no other relationships have been found between PVF occurrence and enterovirus or influenza exposure9.

The present pilot study is the first to include a circulating virome analysis of all DNA viruses that infect vertebrates. Our findings indicate that non-PVF and PVF patients significantly differed only in the levels of Macacine gammaherpesvirus 4 (Rhesus lymphocryptovirus), Human herpesvirus 4 type 2 (Epstein–Barr virus type 2), and Torque teno viruses 8 and 18 (transfusion transmitted viruses).

Gamma-herpesvirinae family viruses are lymphotropic viruses that infect lymphoid cells. Epstein–Barr virus (EBV) is a highly ubiquitous herpesvirus which asymptomatically infect over 90% of the population32. Once infected, EBV persists in B-cells for life and could be reactivated in immunosuppression cases33. In terms of the heart, EBV reportedly induces severe infection of T-cells in the myocardium of patients with ongoing myopericarditis34,35, as well as in abdominal or coronary aneurysms36,37. EBV infection may also influence the development of atherosclerosis38. Here we identified EBV (Human gammaherpesvirus 4) and EBV type 2 (Human herpesvirus 4 type 2). The relative frequency of EBV did not significantly differ between non-PVF and PVF patients. Along this line, we did not find significant differences in the viral load or in the IgG-class antibodies to EBV between non-PVF and PVF patients measured by RT-PCR and ELISA, respectively. On the other hand, the EBV type 2 frequency was significantly higher in non-PVF patients than in PVF patients, and is thus not a risk factor for second-hit ischaemia-driven cardiac arrest. Any of the patients analysed took immunosuppressive treatment or had any malignancy.

Furthermore, Torque teno viruses (TTVs) are small DNA viruses that have been detected in many mammalian hosts, and whose prevalence in humans is > 90%39. It is not clear that TTVs act as primary pathogens, and it appears that TTVs usually establish chronic infections without causing pathology. It has been suggested that TTVs could be used as markers of viral environmental contamination, since TTVs are potential contaminants in water sources40 and hospitals41, including in the blood supply42. This may explain why we detected 27 species of TTVs in the presented study. Remarkably, among 20 human samples, only 1 tested negative for all detected TTV species. Although they are not among the 10 most frequent relative entries, TTV-8 and TTV-18 were the most frequently detected TTVs, and their frequencies significantly differed between non-PVF and PVF patients. However, the relative frequencies of TTV-8 and TTV-18 were significantly higher in non-PVF patients than in PVF patients, and thus do not provide information to predict sudden cardiac arrest. Takeuchi et al. detected one TTV sequence read in a patient with acute myocarditis, but could not establish it as a potential pathogen of myocarditis43. Both our results and Takeuchi’s findings support the widespread idea that TTVs are unlikely to act as primary pathogens.

The second objective of this study was to examine the systemic inflammatory response, which is known to play important roles in the pathophysiology of acute coronary syndrome and atherosclerosis. Notably, in recent years, its involvement in SCD has also been studied, although attempts to find predictive biomarkers have yielded inconclusive results44. The Physicians’ Health Study showed that C-reactive protein (CRP) levels are an independent risk factor for SCD (OR, 2.78; 95% CI, 1.35–5.72)45. In contrast, the Nurses’ Health Study did not confirm any significant correlation between SCD and highly sensitive CRP46. Among healthy European middle-aged men who participated in the PRIME Study, higher IL-6 was a strong predictor of sudden death, with an OR of 3.06 (95% CI, 1.20–7.81)11, but CRP was not shown to predict SCD, as in the Nurses’ Health Study. Furthermore, our group identified growth differentiation factor 15 (GDF-15) as a predictor of mortality and CV morbidity47, and Andersson et al. detected GDF-15 as a risk factor for sudden cardiac death in the acute phase of MI, with an OR of 1.47 (95% CI, 1.11–1.95)12.

Our analyses revealed no significant differences between non-PVF and PVF patients for any of the analysed inflammatory-related proteins. We did identify differential protein expression between healthy subjects and STEMI patients (including both non-PVF and PVF patients) (Supplementary Fig. 5). Compared to healthy subjects, STEMI patients showed significantly higher levels of inflammatory proteins related to cell adhesion, chemotaxis, and cellular response to cytokine stimulus, and cell activation proteins involved in immune response, such as IL-6, IL-8 CXCL11, CCL11, MCP3, MCP4, and ENRAGE. The roles of IL-6 and IL-8 in AMI have been previously described48,49. CCL11 has potent eosinophil chemoattractant activity, and is expressed by cardiac macrophages50. Here we found that CCL11 levels were increased in STEMI patients compared to healthy subjects, thus confirming the previously observed association between CCL11 and myocardial infarction51,52. MCP-3 plays an important role in cell recruitment to inflammatory sites, specifically, it has been described that MCP-3 recruits mesenchymal stem cells and improved cardiac remodeling53. Mao et al., found that MCP-3 levels were decreased in patients with cardiac remodeling after AMI compared to MI and control groups; in addition, MCP-3 values were not differential between MI and healthy subjects54. These results do not agree with what was found in our pilot study, so delving into the role that MCP-3 plays in STEMI patients would be interesting.

Although no differences were found between non-PVF and PVF patients, some proteins significantly differed between the healthy control group and one of the STEMI groups. For example, CDCP1 and IL18-R1 were significantly higher in PVF patients than in healthy subjects. Shia et al. conducted genome-wide association analyses, and identified variations in the DNA sequence that affect the expression of 3p21.31 (CDCP1), which were associated with myocardial infarction55. Those authors did not specify whether the patients had PVF. In the other hand, Ponasenko et al. also found that a polymorphic variant of IL18R1 was associated with an increased risk of MI in CAD patients with coronary artery disease56. Based on our results, it would be interesting to further examine into the studies related to CDCP1 or IL18-R1 and PVF. In contrast, MCP-1, CCL4, TNFRSF-9, and NT-3 showed different expression profiles in healthy subjects compared to non-FVP patients, but not compared with FVP patients. The association of some of them with cardiovascular disorders has already been previously described by other authors. MCP-1, which recruit circulating monocytes, plays a major role in the immunologic profile of ischaemia/reperfusion injury in the heart57; CCL4 is directly involved in the atheroma plaque stabilization58; and elevated NT-3 plasma levels are associated with an increased risk atrial fibrillation recurrence59. However, there remains a need to elucidate potential key roles of these proteins in inflammatory process development in AMI; and they do not seem to be involved in PVF.

This study has several limitations. It was a pilot study with a limited sample size. Despite comprehensive examination of both the virome and the proteome, we did not identify any clear trend. The VirCapSeq-VERT panel can capture both DNA and RNA viruses; however, due to the storage conditions and available blood material, we cannot fully exclude the presence of undetected RNA viruses. In addition, we have not been able to make the correlation between the OTUs and the inflammatory protein levels because, although the population is the same, some samples were used to the virome screening study and others to the inflammation analyses. Lastly, to confirm the presence of a viral genome within the myocardium during the acute phase of STEMI, we would need to perform endomyocardial biopsies, which is ethically unacceptable.

In conclusion, our observations revealed no clear trend in associations between the circulating virome or inflammatory proteome and PVF in STEMI. Hence, there remains a critical need for new strategies to better elucidate the possible triggers of PVF, and to identify individuals at high risk of SCD.