The impact of COVID-19 on pulmonary, neurological, and cardiac outcomes: evidence from a Mendelian randomization study

Background Long COVID is a clinical entity characterized by persistent health problems or development of new diseases, without an alternative diagnosis, following SARS-CoV-2 infection that affects a significant proportion of individuals globally. It can manifest with a wide range of symptoms due to dysfunction of multiple organ systems including but not limited to cardiovascular, hematologic, neurological, gastrointestinal, and renal organs, revealed by observational studies. However, a causal association between the genetic predisposition to COVID-19 and many post-infective abnormalities in long COVID remain unclear. Methods Here we employed Mendelian randomization (MR), a robust genetic epidemiological approach, to investigate the potential causal associations between genetic predisposition to COVID-19 and long COVID symptoms, namely pulmonary (pneumonia and airway infections including bronchitis, emphysema, asthma, and rhinitis), neurological (headache, depression, and Parkinson’s disease), cardiac (heart failure and chest pain) diseases, and chronic fatigue. Using two-sample MR, we leveraged genetic data from a large COVID-19 genome-wide association study and various disorder-specific datasets. Results This analysis revealed that a genetic predisposition to COVID-19 was significantly causally linked to an increased risk of developing pneumonia, airway infections, headache, and heart failure. It also showed a strong positive correlation with chronic fatigue, a frequently observed symptom in long COVID patients. However, our findings on Parkinson’s disease, depression, and chest pain were inconclusive. Conclusion Overall, these findings provide valuable insights into the genetic underpinnings of long COVID and its diverse range of symptoms. Understanding these causal associations may aid in better management and treatment of long COVID patients, thereby alleviating the substantial burden it poses on global health and socioeconomic systems.

In the present study we used two sample MR to investigate the potential causal association of genetic predisposition to COVID-19 (exposure) with the onset of long COVID symptoms (outcomes), namely pulmonary disease (airway infections and pneumonia), neurological deficits (headache, depression, and parkinsonism), cardiac anomalies (chest pain and heart failure) and chronic fatigue.

Study design
Mendelian randomization (MR) relies on genetic variations as tools to investigate the lifelong and causal impacts of an exposure on an outcome (39).It closely resembles randomized controlled trials, as alleles are randomly assigned at conception, reducing susceptibility to reverse causation and unaccounted-for variables compared to traditional cohort studies.Consequently, it offers more robust evidence for establishing causal relationships.In the current study, we employed a two-sample MR approach by extracting exposure and outcome summary data from separate publicly available datasets.Notably, the effectiveness and reliability of these findings hinge on three key assumptions: (a) genetic instruments are linked to the exposure; (b) genetic instruments are not associated with any confounding factors that influence the exposure-outcome connection; (c) genetic instruments solely impact the outcome through the exposure, without involving other pathways (39).To statistically evaluate these assumptions, we performed pleiotropy and heterogeneity tests.The study design and underlying assumptions for MR is shown in Figure 1.

Data source: COVID-19
Genome Wide Association Study (GWAS) summary statistics were retrieved from one of the largest GWAS data set for COVID-19 (hospitalized vs. non-hospitalized) described by the COVID-19 Host Genetics Initiative Release 6: B1_ALL_leave_23andme, released on June 15, 2021.B1_ALL_leave_23andme consists of trans-ancestry GWAS summary statistics of individuals who were hospitalized (n = 14,480; cases) vs. those who were not hospitalized (n = 73,191; controls) due to COVID-19 (Table 1).1,35,110 highly significant SNPs were identified (p ≤ 0.01).To ensure the independence of these SNPs, they were first extracted from the Genome Asia dataset (40) and subsequently pruned for linkage disequilibrium (LD) using PLINK v1.9 (41).The LD proxies were limited to a minimum r 2 ≧ 0.6.1,02,354 SNPs were pruned out and the remaining 32,756 SNPs were used for downstream analyses.

Data source: disorders
Information about genetic association of pulmonary diseases (airway infections and pneumonia), neurological deficits (headache, depression and Parkinson's Disease), cardiac anomalies (chest pain and heart failure) and chronic fatigue were obtained from the ieu open gwas

Pleiotropy test
The intercept from MR-Egger regression (variants uncorrelated, random-effect model) implemented in the R package Study design and assumptions of Mendelian randomization (MR) analysis along with the assumptions.MendelianRandomization v0.9 (42), was utilized to test the pleiotropy of the SNPs associated with COVID-19 in pulmonary diseases (airway infections and pneumonia), neurological deficits (headache, depression and Parkinsonism), cardiac anomalies (chest pain and heart failure) and chronic fatigue (Table 2).In particular, MR assumes no pleiotropy, so a p > 0.05 implies no significant pleiotropy of the COVID-19 associated SNPs in above mentioned outcomes.Sensitivity analysis utilizing the "leave-one-out" method was performed using the mr_loo function in MendelianRandomization to evaluate whether the analysis could be influenced by a solitary SNP with significant and broad horizontal pleiotropic impact.In addition, funnel plots were generated using the mr_funnel function, were used to evaluate diversity among different SNPs.

Mendelian randomization analysis
To evaluate the causal association between COVID-19 and various pulmonary, neurological, and cardiac disorders and chronic fatigue, we performed MR analysis using inverse variance weighted (IVW), Penalized IVW, Robust-IVW, Penalized robust IVW, Penalized MR-Egger, Robust MR-Egger, Mr-Egger, Penalized robust MR-Egger, weighted median, simple median and Penalized weighted median estimators implemented in the R package MendelianRandomization v0.9.p value < 0.05 represents a causal association of COVID-19 with various disorders.

Single SNP effect analysis
The mr_plot function in MendelianRandomization v0.9 was used to visualize the individual potential causal effects of COVID-19 associated SNPs on various pulmonary, neurological, and cardiac disorders and chronic fatigue.Furthermore, the mr_forest function was used to determine single SNP effect size for COVID-19 on various disorders.

Pleiotropy test
In terms of pleiotropy and sensitivity, the MR-Egger regression analysis demonstrated no indications of biased pleiotropy for any of the disorders under study (Table 2).The leave-one-out analysis and funnel plots also indicated no breaches of the instrumental variable assumptions (Supplementary Figure S1, respectively).The leaveone-out sensitivity analysis depicted that removing a specific SNP among COVID-19 SNPs did not change the results for any of the disease under study (Supplementary Figure S1).
The overall Mendelian Randomization analysis results are summarized in Figure 2.Only penalized methods (penalized weighted median, penalized IVW and penalized MR-Egger) are shown.The plot was generated using the R package "forestplot."
We tested the effect of COVID-19 on risk of development of Parkinson's disease.While the simple median (β: 0.785, 95% CI 0.633 to 0.938, p < 0.0001) and IVW based methods, i.e., weighted median (β: 0.801, 95% CI 0.697 to 0.905, p < 0.0001), penalized weighted median (β: 1.073, 95% CI 0.965 to 1.181, p < 0.0001) suggested that the occurrence of COVID-19 is positively associated with an increased risk of Parkinson's disease.This was also supported by the penalized IVW (β: 1.122, 95% CI 1.066 to 1.177, p < 0.0001) and penalized robust Forrest plot of MR analysis evaluating causal effects of genetic liability to COVID-19 on various pulmonary, cardiovascular and nueropsychiatric disorders.The plot was generated using the R package forestplot.The x-axis shows MR effect size for COVID-19 on various disorders.The y-axis shows the analysis for all SNPs together in a single instrument with the penalized IVW, penalized MR-Egger and penalized weighted median methods.4C, Figure 4C).

Discussion
The COVID-19 pandemic remains an ongoing challenge globally with ~77,04,37,327 confirmed cases worldwide that have not only led  to ~69,56,900 deaths (43) (last accessed 11th September, 2023), but has been followed by the emergence of long COVID, which is a complex, multi-systemic entity affecting at least 10% of SARS-CoV-2 infections (2).The heterogeneity of symptoms (≥200) in long COVID patients makes it challenging to dissect which symptoms arise as a cause of SARS-CoV-2 infection vs. those resulting from exacerbation of pre-existing or coincidental conditions.These factors present significant challenges for understanding the pathomechanisms at play and for developing treatment strategies for long COVID.Here we tested the causal association of a genetic predisposition to COVID-19 with several health problems observed among subjects with long COVID, including respiratory, neurological, and cardiovascular dysfunction.Two-sample MR studies in this study suggest that a genetic predisposition for COVID-19 is causally associated with increased risk of fatigue, development of airflow blockage and respiratory problems including bronchitis, emphysema, asthma and rhinitis, pneumonia, headache, and heart failure.However, the association of COVID-19 with Parkinson's disease, depression and chest pain was inconclusive.Persistent fatigue is one of the most reported long COVID symptoms worldwide (1,16).Studies showed that a subset of subjects, ~20% who recover from SARS-COV-2 infections, particularly mild infections, manifest with dyspnea and persistent fatigue despite normal cardiac and pulmonary function (44,45).Especially in cohorts of long COVID patients with a history of SARS-CoV-2 induced hospitalization chronic fatigue was observed in a higher proportion ~40-60% of affected individuals (46,47).Myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) that is triggered by many pathogens, e.g., Epstein-Barr (EBV) and Giardia lamblia in a subset of infections (48,49) was also noted in a subgroup, ~50% of long COVID subjects (44) including in young individuals with mild or moderate SARS-CoV-2 infections (50).Scatter plot of SNPs associated with COVID-19 and chronic fatigue (tiredness/lethargy).The slopes of each line represent the causal association for each method.The plot was generated using the R package MendelianRandomization v0.9.Congruently, MR studies here showed a strong positive association of a genetic predisposition for COVID-19 and persistent fatigue (tiredness/lethargy).While fatigue is likely multisystemic, the pathomechanisms underlying chronic fatigue in long COVID involve neural dysregulation, such as underactivity in particular cortical circuits, abnormal autonomic function, and myopathic changes in skeletal muscle (51).This has also been correlated with structural changes in the thalamus and basal ganglia in long COVID patients with sustained fatigue (52).
A spectrum of neurological symptoms has been reported at varying frequencies in the post-infection phase among long COVID patients.This is consistent with the neurotropic, neuroinvasive and neurovirulent nature of SARS-CoV-2 (53).In some subjects long COVID neuropathogenesis is marked by structural brain anomalies, such as pronounced decrease in its overall size, reduction in gray matter thickness and tissue damage in primary olfactory cortex (54), encephalopathy (55), hemorrhagic posterior reversible encephalopathy (56), and demyelinating lesions in the central nervous system (CNS) (57) in addition to cognitive decline (25).Other factors influencing the neuropathology of long COVID may include neuroinflammation, anti-neural auto-immune dysfunction, hypometabolism of the brain and brain stem, and abnormal cerebrospinal fluid (58)(59)(60)(61).Congruently neuroinflammation injury and apoptosis, brain hypoxia and microhaemorrhages have been observed in non-human models of SARS-CoV-2 infection (62).Further neurological and cognitive impairment marked by structural brain abnormalities have also been noted among long COVID subjects following mild or moderate initial SARS-CoV-2 infections (54,61).
Frequency of chronic headache range widely from ~8-40% among long COVID cases (46,(63)(64)(65).Individuals mildly affected by SARS-CoV-2 infections seemed more prone to post-COVID headaches (64,66), which were exacerbated with a prior history of migraines (63, 64).Our MR analysis showed a strong positive causal association between COVID-19 and chronic headache in long COVID patients.Recent evidence suggests that long COVID headaches appear to be triggered by hyperinflammation and are sustained by chronic inflammatory activation, and dysregulation of neurotransmitters and metabolic inflammation (66).
Based on observational studies, among individuals affected with long COVID, ~20% may develop mood disorders, e.g., anxiety and depression (47).In one study the risk of mood disorders returned to baseline within 2 months following initial COVID-19, but some conditions such as cognitive impairment, seizures, psychosis, and dementia lingered up to 2 years (25).Another study noted that limbic atrophy and significantly abnormal cerebral functional connectivity underlie anxiety and depression among long COVID patients with mild SARS-CoV-2 infections (67).In contrast several studies inferred that anxiety and depression were not strongly linked to COVID-19 (68) with neuropsychiatric symptoms elevated disproportionately in long COVID subjects with acute initial SARS-CoV-2 infection (68, 69) or other post-infection health complications and psychiatric history (70).The onset and progression of mood disorders may also be modulated by dysfunctional regulatory cells of the innate and adaptive immune system that may contribute to chronic systemic and neuroinflammation (71).Therefore, the observations of increased psychiatric manifestations in long COVID subjects could be a result of compromised immunoregulatory mechanisms.Using MR in the present study the causal association of genetic predisposition to COVID-19 with anxiety and depression remained obscure, warranting further research to clarify this.
Parkinson's disease is a progressive motor disorder that is highly prevalent in older adults (72).It may be triggered following infections by viruses, e.g., Influenza A, EBV and Herpes simplex virus 1 (73).Some cases of post-COVID parkinsonism have also been recorded (26,27).Nevertheless, we detected no causal association of COVID-19 with Parkinson's disease using MR studies.
A range of cardiovascular manifestations have been noted in long COVID patients, including those without any evidence of pre-existing cardiovascular disease or risk factors or COVID-19 related hospitalization (29).While the pathophysiology of long COVID linked cardiovascular disease is far from certain, it may involve viral invasion of cardiomyocytes and cell death, downregulation of Angiotensin-converting enzyme 2 (ACE2), endothelial cell infection, complement activation, deregulation of renin-angiotensin-aldosterone system, autonomic abnormalities, myocarditis and cardiac tissue fibrosis (74-79).The proportion of long COVID subjects with chest pain varied from ~3-20% up to 6 months post initial SARS-CoV-2 infection in different cohorts (80,81).While the overall association of genetic predisposition to COVID-19 with chest pain in this study was inconclusive, five out of twelve MR strategies showed a strong positive causal link that was statistically significant, two techniques showed a positive association that was non-significant, and the rest showed negative association.This strongly warrants further studies to explore a putative causal link of COVID-19 with chest pain.
Heart failures as part of cardiovascular sequalae in long COVID have been observed in patients with pre-existing cardiac complications (82,83), as well as those without (84)(85)(86).This is also supported by anecdotal accounts of increased cardiovascular abnormalities particularly among mildly symptomatic or asymptomatic subjects in the post SARS-CoV-2 infection period.Studies note that autonomic imbalance and increased sympathetic activity in the post-acute phase in ~30 days after mild SARS-CoV-2 infection may explain the cardiovascular complications during this time (87,88).Although, the involvement of autonomic dysfunction in heart complications in long COVID is uncertain (87).All 11 methods employed for MR analysis in this study suggested a positive causal association between genetic predisposition to COVID-19 and heart failures.Ten out of eleven methods concur on this with high significance (p < 0.0001).Moreover, the effect size (β) range is highly consistent across all methods with narrow confidence intervals.Taken together MR analysis suggested a strong and positive causal association between genetic predisposition to COVID-19 and heart failure.
Finally, respiratory complications, e.g., cough are one of the commonest features of long COVID subjects (4) and are exacerbated in patients with preexisting respiratory comorbidities (89).Subjects with moderate-to-severe SARS-CoV-2 related pneumonia sustain lung abnormalities, such as parenchymal lung disease, fibrosis, and bronchiectasis even a year following initial infection (90).New onset of airway abnormalities, e.g., asthma is not common but has been strongly associated with COVID-19 (91).Increased lung emphysema was noted at 30 days following initial SARS-CoV-2 infection (92).In this study MR results support a strong positive causal association of a history of COVID-19 with increased risk of disorders with chronic lung and airway inflammation, as well as pneumonia.
This study has some limitations.First since it utilizes publicly available genetic datasets, the vaccination and BTI status of the included individuals is unknown.Accordingly, it cannot account for how these factors influence long COVID outcomes.It is noteworthy that BTI of SARS-CoV-2 following vaccination may be influenced by several factors, e.g., age and previous infection (93).Moreover, the serological response to vaccines was also reported to be insufficient in individuals with two or more pre-existing chronic conditions (94).These factors may influence long COVID phenotypes and cannot be evaluated in the present study.Second, the dataset for chest pain used here is non-specific, and may include subjects where it is caused due to cardiac, lung or abdominal organ abnormalities.This may also explain why the association of a history of COVID-19 with chest pain in this study was inconclusive despite it being a commonly noted long COVID complexity (2).
In conclusion, this study has employed MR to shed light on the genetic underpinnings of long COVID, uncovering significant causal associations between genetic predisposition to COVID-19 and several prevailing health complications.MR findings in this study corroborate the multiple adverse outcomes of long COVID, linking it to an increased risk of developing pneumonia, airway infections, headache, chronic fatigue, and heart failure.Our findings on Parkinson's disease, depression, and chest pain were inconclusive.These insights are critical in enhancing our understanding of the lasting health implications of COVID-19.As we continue to grapple with the longterm challenges posed by the COVID-19 pandemic, a deeper comprehension of the genetic factors contributing to long COVID will be a precursor to developing targeted strategies to support those symptomatic with these long COVID symptoms.Further research is warranted to explore the intricate web of genetic determinants underlying the diverse manifestations of long COVID, which will spearhead global efforts in combating this healthcare crisis effectively.

FIGURE 3
FIGURE 3Scatter plot of SNPs associated with COVID-19 and pulmonary diseases.(A) pneumonia and (B) a group of diseases associated with airflow blockage and breathing-related problems including bronchitis, emphysema, asthma and rhinitis.The slopes of each line represent the causal association for each method.The plot was generated using the R package MendelianRandomization v0.9.

FIGURE 4
FIGURE 4Scatter plot of SNPs associated with COVID-19 and neurological abnormalities.(A) anxiety/depression, (B) Parkinson's Disease, and (C) headache.The slopes of each line represent the causal association for each method.The plot was generated using the R package MendelianRandomization v0.9.

FIGURE 5
FIGURE 5Scatter plot of SNPs associated with COVID-19 and cardiovascular diseases.(A) heart failure and (B) chest pain.The slopes of each line represent the causal association for each method.The plot was generated using the R package MendelianRandomization v0.9.

TABLE 1 Various
disorders and COVID-19 genetic summary data sources.

TABLE 3B The
MR estimates of the causal effect of COVID-19 on group of diseases associated with airflow blockage and breathing-related problems.

TABLE 4A The
MR estimates of the causal effect of COVID-19 on anxiety/depression.

TABLE 5A The
MR estimates of the causal effect of COVID-19 on heart failure.

TABLE 4B The
MR estimates of the causal effect of COVID-19 on Parkinson's disease.

TABLE 5B The
MR estimates of the causal effect of COVID-19 on chest pain.

TABLE 6
The MR estimates of the causal effect of COVID-19 on fatigue (tiredness/lethargy).