SNVs and COVID-19
The first major finding of this study is that there are no direct effects of IL18 rs360717, IL18 rs187238, NLPR3 rs10157379, and NLRP3 rs10754558 SNVs on critical COVID-19 outcome although there are indirect effects of these genetic variants which are mediated by inflammatory biomarkers (for IL18 SNVs) and the path from SSC to SARS (for NLRP3 SNVs). In other words, if we had evaluated genotypes in relation to diagnostic WHO categories, this genotypic research would have shown negative findings. This is an example of "indirect-only mediation," in which the mediators, namely SSC, SARS, and inflammatory pathways, mediate the link between the gene variants and critical COVID-19 even when no direct or combined (direct + indirect) effects exist [33]. All in all, our results indicate that these SNVs participate in the response to SARS-CoV-2 through indirect effects on the inflammatory response, SSC, and SARS.
Indeed, we showed that homozygosity or heterozygosity for the A allele of the IL18 rs360717 (G > A) conferred approximately 44.0% protection against the development of moderate and severe COVID-19, whereas the GG homozygous genotype predicts an increased severity in COVID-19 [24]. The most likely explanation is that this SNV is located in the potentially regulatory 5' untranslated region (5'UTR) of the IL18 gene, which may impact gene expression and altered levels of IL-18 [21]. Additionally, we demonstrated a strong correlation between the NLPR3 rs10157379 T > C variation and SARS severity, with CT genotypes scoring higher than CC or CC + TT combined. Moreover, there was a substantial correlation between the NLPR3 rs10754558 C > G variant and SSC severity, with CG genotype scoring worse than GG and CC genotypes. Additionally, both the NLPR3 rs10157379 CT and NLPR3 rs10754558 GG genotypes predicted SARS symptoms, with NLPR3 rs10754558 being related with tiredness, myalgia, appetite loss, and headache [25].
COVID-19 severity is linked to higher levels of proinflammatory cytokines such as IL-1β, IL-6, IL-18, G-CSF, IFN-γ, and TNF-α [34, 35]. IL-18 is a cytokine that belongs to the IL-1 family and is involved in both innate and adaptive immune responses. The activation of inflammasomes, particularly the NOD-, LRR-, and NLRP3 inflammasomes, is required for their stimulation [34, 36]. Previous research found that patients with COVID-19 had higher IL-18 levels, which was associated with disease severity [14]. These authors found that IL-18 levels, but not caspase 1, were higher in patients who required mechanical ventilation compared with patients who did not. Moreover, this study reported that the levels of IL-18, but not caspase 1, were higher in lethal cases of COVID-19 compared with survivors [14]. As a result, the pathophysiology of severe COVID-19 may be influenced by genetic NLRP3 variations that impact cytokine release.
Other predictors of a critical outcome of COVID-19
It should be emphasized that critical COVID-19 was predicted in the present research not only by SNVs and inflammation, but also by decreased SpO2, increased CCTA, increasing age, male sex and BMI, SARS severity, lowered SSC, and medical comorbidities such as hypertension and T2DM, whereby multiplicative and multistep mediated effects of all those variables predict 49.5% of the variance in COVID-19 outcome.
It is worth noting that our PLS pathway analysis found that lung lesions and related decreases in SpO2 were associated with greater peripheral inflammation, suggesting that the combined impact of these variables may exacerbate COVID-19. The severity of CCTA in COVID-19 indicates continuous lung inflammation, as well as the more severe conditions such as bronchiolitis, pneumonia, and lung fibrosis. These lung infection locations may result in the recruitment of a variety of immune cells, exacerbating pro-inflammatory reactions [35]. CCTA may result in reduced oxygenation [37]. Additionally, in COVID-19, decreased SpO2 is related with higher levels of IL-6, IL-10, CRP, and soluble receptor for advanced glycation end products (AGE), as well as decreased levels of albumin, magnesium, and calcium [37]. It is well established that hypoxia causes inflammation [38]. There were no significant connections between the NLRP3 SNVs and these three associated covariates in the current investigation.
Our investigation found that NLR, CRP, and ferritin levels are much greater in critical disease than in non-critical disease, which is consistent with prior research [39, 40]. Patients with severe illness and fatal outcomes had a lower lymphocyte/WBC ratio at admission and during hospitalization, compared to survivors [39, 40]. CRP and ferritin levels were shown to be linked with the severity of SARS-CoV-2 infection, probably because of a hyperactivated Th1 response [40].
In line with prior research, we discovered that traditional risk factors for SARS-CoV-2 infection were related with moderate and severe COVID-19 infection [41]. Numerous risk factors for COVID-19 development to a severe and critical stage have been identified, including advanced age, male sex, and underlying comorbidities such as hypertension, T2DM, chronic lung illnesses, and heart, liver, and kidney problems [41, 42]. Additionally, our research discovered that a worse COVID-19 result was predicted by T2DM, hypertension, and a higher BMI. In this respect, it is interesting to note that T2DM is associated with an active NLRP3 inflammasome, as shown by higher baseline NLRP3 activity in monocytes and enhanced caspase-1 activation and IL-1β and IL-18 production in peripheral blood mononuclear cells (PBMCs) [43]. When combined with IL-12, IL-18 may induce the release of IFN-γ from Th1 cells, nonpolarized T cells, natural killer (NK) cells, B cells, dendritic cells, and macrophages. Additionally, IL-18 stimulates CD8+ T cells directly and enhances the cytotoxic activity of NK cells and cytotoxic T cells, which exert their cytocidal effect through the utilization of perforin and Fas ligand (FasL). FasL, in turn, promotes apoptosis in Fas-expressing target cells, and perforin is a highly pore-forming protein that acts directly on the target cell membrane, therefore facilitating viral clearance [44].
The NLRP3 inflammasome is substantially related with not only T2DM but also with obesity [45] and hypertension via regulating low-grade inflammation [46]. Certain NLRP3 gene variants (for example, rs7512998) are related with hypertension [47]. Consequently, we may assume that the presence of obesity, T2DM, or hypertension exacerbates the outcome of COVID-19, as individuals with those illnesses have an active NLRP3 inflammasome, which is likely further triggered by the infection. These findings corroborate those of Lopes-Reyes et al. [48], who demonstrated that infection with SARS-CoV-2 may aggravate the pre-existing systemic inflammatory state in obese people by activating the NLRP3 inflammasome and producing pro-inflammatory cytokines. These findings also explain why the NLPR3 SNVs identified in our investigation may interact with those concomitant diseases, exacerbating critical COVID-19.
Our NN models demonstrated that age and SpO2 were the most significant input variables predicting COVID-19 mortality. After the age of 55, the chance of dying of COVID-19 infection rises linearly [49]. These effects may be a result of age-related changes in comorbid disorders such as T2DM and hypertension, as well as the effects of NLRP3-associated inflammaging, which is characterized by increased inflammatory mediators, AGE, mitochondrial dysfunction, reactive oxygen species (ROS), genomic instability, and hypoxia [49, 50].
In our PLS model, male sex had a substantial and indirect effect on the outcome of COVID-19 through the mediating effects of SSC severity and elevated inflammation. Male sex has already been demonstrated to be a predictor of death [49]. It is critical to highlight that overactivation of the NLRP3 inflammasome is associated with higher lethality in male COVID-19 patients [51]. Male COVID-19 patients had higher plasma levels of IL-8 and IL-18, as well as a greater induction of non-classical monocytes [51]. Moreover, testosterone may activate the NLRP3 inflammasome directly or indirectly via its influence on mitochondrial ROS, while progesterone and estrogen may suppress the inflammasome [51, 52].
Our results suggest that SSC is mediated in part by the NLRP3 inflammasome and that NLRP3 SNVs may influence the outcome of COVID-19 indirectly via influencing the SSC. The latter is induced by pro-inflammatory cytokines, including IL-1β and IL-6, and hence the acute phase and immune-inflammatory response during infection may have both harmful and protective effects [17]. Indeed, activation of the NLRP3 inflammasome may contribute to both positive (combat the virus) and negative (hyperinflammation) host responses [14]. The SSC has protective benefits by altering the metabolism, consequently enhancing immune cell activity, and combating acute infection [16]. The SSC is a well-preserved adaptive symptom complex that seeks to decrease the incentive to search for food to save energy for combatting infection [53]. Thus, anergy, myalgia, hyperesthesia, psychomotor retardation, libido loss, and appetite loss all serve the same aim of decreasing nutritional intake. Most crucially, decreased foraging and the subsequent metabolic reprogramming toward a negative energy balance are often beneficial to the host during inflammatory circumstances [53]. Moreover, hypermetabolism catabolizes tissues during the acute phase response in favor of cytokine-induced acute phase protein synthesis in the liver and gluconeogenesis [16].
Precision medicine ML techniques to classify critical COVID-19.
In the current study we have employed PLS path analysis to decipher the complex interactions between the many predictors that in combination lead to critical COVID-19. Using neuronal networks and SVM were able to achieve a reasonable prediction of critical disease and death due to COVID-19 with a ROC area of n the present study, we used Machine Learning algorithms to predict the risk of critical COVID-19 with AUC/ROC of 0.930 and 0.927, respectively. Using SIMCA [32] we were able to construct a digital self of mild COVID-19 which allows to authenticate people with infection as either belonging to this mild class or as an outlier, and thus a severe case with a sensitivity of 97.7% and specificity of 77.1%. Our analysis demonstrates that ML algorithms amplify the diagnostic accuracy and the discriminative efficacy of these markers, maximizing their use for predicting the risk of patients with COVID-19 to develop severe disease during the course of the disease. As more data become available, the whole procedure can easily be repeated to obtain more accurate models [32]. Although other studies have suggested predictive models for COVID-19 severity with high sensitivity and specificity [24–26, 49, 54–56], our study is one of the few that includes genetic variants combined with clinical and laboratory parameters associated with prognosis of COVID-19.
This study has some limitations, such as its retrospective single-center methodology that decreases its generalizability and, although the models have adequate validity and replicability, as tested using bootstrapping and blindfolding, holdout samples, and tenfold cross-validation. Considering the high genetic variability between different populations worldwide that influences the gene expression and individual inflammatory response, the models constructed here need prospective validation in independent samples and especially in other countries.