Multi-omics approach to COVID-19: a domain-based literature review

Background Omics data, driven by rapid advances in laboratory techniques, have been generated very quickly during the COVID-19 pandemic. Our aim is to use omics data to highlight the involvement of specific pathways, as well as that of cell types and organs, in the pathophysiology of COVID-19, and to highlight their links with clinical phenotypes of SARS-CoV-2 infection. Methods The analysis was based on the domain model, where for domain it is intended a conceptual repository, useful to summarize multiple biological pathways involved at different levels. The relevant domains considered in the analysis were: virus, pathways and phenotypes. An interdisciplinary expert working group was defined for each domain, to carry out an independent literature scoping review. Results The analysis revealed that dysregulated pathways of innate immune responses, (i.e., complement activation, inflammatory responses, neutrophil activation and degranulation, platelet degranulation) can affect COVID-19 progression and outcomes. These results are consistent with several clinical studies. Conclusions Multi-omics approach may help to further investigate unknown aspects of the disease. However, the disease mechanisms are too complex to be explained by a single molecular signature and it is necessary to consider an integrated approach to identify hallmarks of severity. Supplementary Information The online version contains supplementary material available at 10.1186/s12967-021-03168-8.


Background
Advances in molecular and cellular biology in last few decades have triggered tremendous growth in available experimental data, not easy to manage and interpret. Protein-protein interactions (PPI), gene expression, chromatin accessibility and epigenetic modification, signal transduction and metabolic networks are among the categories addressed in systems biology [1]. Mechanistic pathway models can link biological variations, molecular mechanisms and cellular behavior, by coupling molecular interactions in each pathway with their specific endpoint, and contextualizing database query through omics.
To represent pathophysiological mechanisms, the disease map is surely a key emerging concept. It connects bioinformatics, molecular biology and clinical research, with the potential to link conceptual domains of biomedical knowledge with clinical data, providing an  19:501 intermediate step between conceptual and executable models [2,3]. A domain can be defined as a conceptual repository, which exhaustively summarizes the variability of a specific biological field and are useful to summarize multiple biological pathways involved at different levels. The conceptual domains mirror the hierarchical structure of the corresponding biological contexts, that can be exemplified by the organization of protein domains. In fact, protein domains have specific functions into each single protein, so that they can be involved in different biological contexts, but on the other hand proteins with different biological functions can have similar domains. They may occur independently or as part of complex multidomain protein architecture with a specific function. Protein domains can therefore be viewed as a 'parts list' for biology. Since protein interactions characterize biological processes that allow the development of different levels of the biological architecture, a macro-domain can be identified as the structural biological level of defined biological processes.
One or multiple macro-domains can be identified for an infectious disease, such as COVID-19, organized on a hierarchical structure. At the upper level, the virus-host interactome based on PPIs could help to understand the disease mechanisms. However, the pathophysiology linking SARS-CoV-2 infection to its clinical phenotypes is still too complex and involves specific pathways, as well as different cell types and multiple organs. A systemic approach and perspective is necessary to untangle this complex network, by collecting mechanistic knowledge scattered across scientific literature and biological databases.
A network-based explorative method of molecular interactions, generally PPIs, can help to understand the disease mechanisms, and the virus-host interactome represents one of its main applications, as it can unravel clusters of molecular interactions, so highlighting the engagement of host pathways induced by the virus [4][5][6].
The visual exploration of diagrams on COVID-19 Disease Map repository allows to parse such clusters at mechanistic level, pointing out potential molecular interactions to be added to the existing diagrams on the repository (e.g., SARS-CoV-2 E protein interactions) [7,8].
The aim of the present study is to shed light on the molecular pathophysiology of COVID-19, linking SARS-CoV-2 infection to its clinical phenotypes, and to delineate how different and complex COVID-19 clinical courses are, depending on the involvement of pathways, as well as different cell types (alveolar cells type 1, lymphocytes, neutrophils, platelets, ect) and organs (lungs, blood, colon, liver, etc.). To achieve this goal, we carried out a scoping review of the available literature data, based on conceptual domains.

Conceptual domains identification
We used previous studies from two research groups as methodological reference in order to define the conceptual domains identified for this review.
Particularly, the study by Gordon and colleagues [6] was used to define the concept of host-pathogen interaction, since it represents the first exhaustive experimental work about in vitro molecular interactions between SARS-CoV-2 proteins and human proteins involved in complex biological processes.
In addition, the studies published by Ostaszewski et al. [7,8] were selected for introducing the concept of Disease Map, since they represent theoretical and hierarchical models of complex biological mechanisms of interaction, describing the involvement of specific pathways, consistent with specific disease phenotypes.
We firstly identified two conceptual macro-domains: (1) networks of virus-human PPIs; (2) interaction clusters of pathways, involving dynamic processes at different times, not represented in a static view.
The macro-domains were grouped into three domains divided, in turn, in six sub-domains, as shown in Fig. 1.

Scoping review
An interdisciplinary team was set up with expert addressing the three domains plus the sub-domain "host signatures", that presented a conspicuous pertinent literature.
Each working group performed an independent literature review; this was conducted in compliance with international reference guidelines using the scoping review method [10,11]. For each domain, the scoping review results were processed to identify features of COVID-19 and SARS-CoV-2 infection, compared with SARS-CoV and MERS-CoV infections. In particular, the review was focused on studies that provide omics data correlated to immune response and clinical phenotypes, in order to identify knowledge gaps regarding the link among omics and clinical data, focusing on the different COVID-19 features and the on risks of clinical progression.
Protocols and tools used for the scoping review are reported in detail in Supplementary text. Articles were selected using the string specified in the Supplementary text on PubMed, in the period between January 2002 and February 2021, with the identification of 1214 articles, after removing duplicates and non-English papers. As shown in Fig. 2, in the subsequent step 691 articles were selected, catalogued in 3 domains: (1) "virus"; (2) "pathways"; (3) "phenotypes". Each working group evaluated full text article to define at best the possible subdomains, as shown in Fig. 1.
A further selection has been done in three steps, based on title, abstract and full text. This process led to a final selection of 176 articles, of which 25 were shared by two or more domains, providing 151 unique articles, listed in the table of evidences (TE). To univocally identify the mechanisms described for domains "pathways" and "phenotypes", the Reactome Pathway nomenclature and Reactome Code were used [12].

Molecular characterisation of the virus and of the entry phase into the host cells
For each domain, the omics literature was consulted in reference to available data, concerning viral genomics, proteomics and molecular interactions with the host, in light of their possible involvement in the pathogenic mechanisms of the disease, and based on information from other Coronavirus infections. The results of such analysis are summarized in Table 1, in Additional file 1:  Tables S3 B and Annex 1, and in Additional file 2: Tables  S3 A. In the next paragraphs the results of such analysis are described in more details.

Genome evolution and geographical distribution
In this field of investigation, the evolutionary history of SARS-CoV-2 was reconstructed, starting from phylogenetic comparison with related Betacoronaviruses [TE1-TE5].
Global Initiative on Sharing Avian Influenza Data (GISAID) classification of SARS-CoV-2 clades was reported together with their spread (Last up-dated on November 10th 2020, based on 175,000 genomes).

Genomic hotspots for mutation, drivers of evolution and correlation with COVID-19 pathogenesis
In SARS-CoV-2 genome, ten hyper-variable hotspots were identified. In the S gene, some regions presented signs of positive selection, i.e. dN/dS > 1, particularly within the receptor binding domain (RBD) in the S1 subunit, in the FURIN cleavage site and in the segment encoding the S2 and S2' subunits. Also ORF3a, E, ORF6, ORF7a, ORF8, N, and ORF10 presented dN/dS > 1,

Intra-host genomic variability
In SARS-CoV-2 infected host, the virus displayed small scale intra-host variation, while spatial-temporal redistribution of SARS-CoV-2 "quasispecies" in respiratory and gastro-intestinal tracts in human hosts was observed, with a higher significantly genetic diversity observed in gastrointestinal compared to respiratory tract samples [TE19-TE22].

Single viral protein and whole viral proteome studies
This investigation field concerned omics studies about both single proteins and whole SARS-CoV-2 proteome characterization. Some of these studies were in silico and organized in a modular hierarchical scale of virus/host PPIs, allowing to build a dynamic and integrated structure, named "SARS-CoV-2 dynamicome" [TE24-TE29].

Immune proteomics
The host-pathogen molecular mimicry was investigated on the basis of viral proteomics, also used in studies aimed at developing innovative anti-COVID-19 vaccines [TE30-TE32].

Viral RNA and host protein interactions
In silico studies predicted possible SARS-CoV-2 RNA regions of interaction with host proteins.
Experimental studies provided description of the SARS-CoV-2 RNA-protein interactome in different SARS-CoV-2 in vitro infected cell lines. Other works provided functional interrogation of the host proteins involved in such interactions, revealing that most of them regulated virus entry into host cells, protected the host from virus-induced cell death or were involved in SARS-CoV-2 pathogenicity [TE33-TE39, TE45, TE51]. Besides interaction with host proteins, some studies also identified SARS-CoV-2 genomic regions to be potential silencer RNA (siRNA) targets, or could interact with host microRNAs (miRNAs) [TE32].

Virus-host protein-protein interactions (PPIs)
Several research studies addressed PPIs between single betacoronaviruses and host proteins or further detailed, by in silico analysis, already described interactomes [T40-TE44].

Multilayer analysis of virus-host interactions (transcriptomics, proteomics)
By integrating viral-host transcriptomics and proteomics in a multilayer analysis it was possible to characterize COVID-19 phenotypes [TE45]. Omics studies, particularly relevant for pathogenesis investigations, were based on ex-vivo studies, with clinical samples derived from SARS-CoV-2 infected subjects displaying different clinical phenotypes. In fact, some of these investigations were performed by multi-omics network-biology-fueled approach to provide the principal host components affected by SARS-CoV-2 infection. They pose the basis for the constructions of a COVID 19 Disease Map. Additional file 1: Table S2 A provides a more detailed description of the above reported information [TE48, TE49, TE50, TE52, TE53].

Viral entry
Expression of host entry factors in human tissues Expression of ACE2 and TMPRSS2 in human tissues was addressed by omics approaches in several studies [TE55, TE56, TE57, TE59, TE60, TE63]. Importantly, very low or absent ACE2 expression was reported in organs/tissues considered as the main target for SARS-CoV-2 replication, including lung, bronchus, and nasal mucosa, suggesting a dynamic regulation of entry factors upon infection and a role for possible alternative receptors.

SARS-CoV-2 interaction with entry factors
Two studies explored the cross-talk between SARS-CoV-2 and host proteins during the entry and subsequent steps of viral replication. In the first paper proteins on the host cell membrane (ATP6V1A, AP3B1, STOM, and ZDHHC5) were identified that may enable binding to SARS-CoV-2 structural proteins. On other hand, several miRNAs were also identified to inhibit proteins involved in viral entry [TE73]. The second study proposed three interactomes from probabilistic modelling, using iDREM (interactive Dynamic Regulatory Events Miner): the first one, involved in creating a suitable environment for the virus, includes ATP6V1A; the second one includes PHB as alternative receptor or co-receptor; the third one, involved in sustaining viral replication, includes oxidative stress and inflammation proteins [TE48].
Regarding the viral variants of concern (VOC), B.1.351, P1 and some B.1.1.7 strains harbor, among others, mutations potentially important for pathogenesis, such as K417T/N, E484K, and N501Y. These substitutions seem to alter the interaction of S glycoprotein with ACE2 receptor, leading to increased transmissibility compared to the previous circulating strains, and especially for B.1.351 and P1 lineages, leading to reduced susceptibility to neutralizing antibodies elicited by non-variant strains and by current vaccines. In fact, viral evolution studies show that the RBD of S glycoprotein is highly variable and that immune escape mutations (i.e., E484K) may emerge independently, in multiple lineages, spreading worldwide, and leading to the further accumulation of additional changes, that may increase the risk of significant reduction of host immunity and/or of monoclonal antibody therapy, as well as of vaccine efficacy [TE69, TE70, TE71, TE72].

Pathways
To hierarchically evaluate cellular processes involved in SARS-CoV-2 infection, we assigned a univocal Reactome Code to each cellular mechanism reported in the scanned literature for the pathway analysis. Of note, this approach was performed for proteomics, transcriptomics and bioinformatics studies in vitro, where the study model was represented by cell lines infected with SARS-CoV-2 ( Table 2 and in Additional file 1: Tables S4 and Annex 2). The occurrence of each mechanism was organized on the basis of Reactome Pathways (Additional file 3: Table S6). This approach allowed us to highlight several mechanisms altered by SARS-CoV-2 infection, i.e. signal transduction (R-HSA-162582), translation (R-HSA-72766), post-translation protein modifications (R-HSA-597592), immune response (R-HSA-168256), cell cycle (R-HSA-1640170), apoptosis (R-HSA-109581), autophagy (R-HSA-9612973), lipid metabolism (R-HSA-556833) and vesicle-mediated transport (R-HSA-5653656). Figure 3 shows how omics data, organized by omics technique and tissue (Proteomics, Metabolomics and Transcriptomics/CyTOF, in A, B and C respectively), contribute to highlight pathways up-or down-regulated in COVID-19 patients compared to healthy subjects, or in severe COVID-19 patients compared to mild patients.
With respect to signal transduction, mTOR pathway is strongly affected during SARS-CoV-2 infection, leading to the activation of PI3K/AKT and TNF cascades [13] [TE74, TE75, TE76, TE77, TE78]. Moreover, considering that the production of viral proteins depends on host cap-dependent translation [14], this mechanism is also altered by SARS-CoV-2 infection. While on one hand the virus usurps cellular translation machinery to promote its own reproduction, on the other hand cells attempt to reduce translation to contrast the infection. It has been extensively reported that several SARS-CoV-2 proteins bind E3-ubiquitin ligases, usurping cellular degradation machinery, thus promoting virus replication [5,6] [TE40, TE41,TE49, TE80, TE83]. Other pathways extensively addressed in COVID-19 studies are represented by both innate and adaptive immunity. Specifically, IFN, TNF and NF-kB pathways, cytokine SPP1, GRN, the receptor tyrosine kinase AXL [TE85], NLR and RIG-I signaling are strongly altered by SARS-CoV-2 [TE76], contributing to severe forms of COVID-19 [6]. Similarly, cell cycle is affected by SARS-CoV-2 infection with a rapid reshape of several host mechanisms, leading to cell cycle arrest. However, it is important to consider that most of the reported studies have been performed in cell cultures in vitro, which represent an undisputed model for infection; nevertheless, cell cycle is intrinsically altered in the in vitro culture systems, introducing a bias in COVID- 19

Host signatures
The host signature was addressed in studies evaluating the systemic profile of soluble mediators by proteomics and metabolomics approaches. Cellular immune response to SARS-CoV-2 infection was evaluated by transcriptomics, proteomics, and high dimensional single cell analysis (mass cytometry, CyTOF) in peripheral blood and tissues, reported in Table 2, and detailed in Additional file 1: Tables S4 and Annex 3.

Systemic profile of soluble mediators
Proteomic studies During the early immune response, the activation of type I/III IFN response represents a key local innate immune player and is associated with chemokines [TE87] and proinflammatory mediators [TE88], leading up to a massive release of inflammatory mediators, called "cytokine storm". In particular, mild COVID  flammatory cytokine signaling are strongly expressed in severe COVID-19 (Additional file 1: Annex 3A.1). The acute phase proteins (APPs) are an additional class of mediators involved in the early phase immune response in COVID-19. These proteins are up-regulated in the severe forms of COVID-19, and can induce inflammatory cytokines, influence lipid metabolism, and induce neutrophil activation, as shown for S100A8 and S100A9 [15], thus possibly contributing to amplify the cytokine storm. Moreover, APPs can also modulate platelet aggregation and activation of coagulation cascade [16,17]

Metabolomic studies
Levels of most lipid-related molecules are altered in moderate and severe COVID-19 patients with a clear preference toward their downregulation. Main observed alterations are related to lipoproteins, Metabolites of the tricarboxylic acid cycle (TCA) and β-oxidation are reduced in COVID-19, particularly in severe patients, whereas metabolic intermediates of the glycolysis and pentose phosphate pathways are increased [TE90, TE96, TE100, TE101, TE102, TE103]. This reduction may be the consequence of declined lung functions and blood oxygen level decrease, but may also mirror a response to nutritional changes, especially in severe patients (Additional file 1: Annex 3A.4). In

Transcriptomics/CyTOF studies
Immune response in peripheral blood (BULK RNAseq) Early SARS-CoV-2 infection triggers a powerful, IFN-driven transcriptional response in peripheral blood. Type I IFN response was impaired in severe and critical COVID-19 patients: striking downregulation of ISGs, IRF-1 and STAT3, absence of circulating IFN-β in patients with all disease-severity grades and low IFN-α production in severe COVID-19 patients were reported [TE90]. Moreover, elevated levels of chemokines and chemokine receptors were detected in severe patients, exhibiting an increase in neutrophils. Downregulation of negative regulators of innate immune system and TCR signaling kinases and adaptors was observed in severe patients (Additional file 1: Annex 3B).

Innate immune cell compartment (scRNAseq/
CyTOF) The initial local respiratory SARS-CoV-2 infection elicits dynamic changes of circulating blood cells with changes in innate immunity parameters. An elevated neu-trophil/lymphocyte ratio has been identified as a sign of COVID-19 severity [TE89]. Investigation of the neutrophil transcriptomics signatures highlighted that excessive neutrophil activation is associated with severe COVID-19 more frequently than with mild disease. Moreover, Low Density Neutrophils with immature phenotype were upregulated in severe disease [TE96, TE112]. An increase of classical CD14+ monocytes, especially in convalescence stages, non-classical CD16+ monocytes and natural killer (NK) cells also with exhaustion phenotype was observed [TE89, TE105, TE112, TE113]. Impaired IFN-α production by plasmacytoid dendritic cells was also observed in COVID-19 patients. However, some ISGs were upregulated in monocytes and DC [TE89] (Additional file 1: Annex 3C).

Immune response in lung and other tissues
Major deregulation of the innate immune response has been observed in lung samples. COVID-19 patient lungs showed a compartmentalization of innate immune cells (neutrophils and monocytes), in response to the chemokine secretion [TE106, TE108]. The transcriptional profiling of the lung tissue showed the over-expression of

Phenotypes
We selected the studies with clear stratification of patients by disease severity; multi-omics data from these studies which highlighted significant differences between healthy controls and COVID-19 patients, and between mild and severe clinical presentation. We then classified the selected studies in four topics: (1) Key Genes and Proteins in SARS-CoV-2-host interactions and pathogenesis in the Lung; (2) DEG and DEP analysis in other organs and tissues; (3) Hub genes and pathways of innate immune response; (4) Comorbidities, further subclassified in comorbidities COVID-19-associated not sharing COVID-19 pathogenesis, and comorbidities associated and related to COVID-19 pathways, reported in Table 3, Additional file 1: Tables S5 and Annex 4.
As shown in Table 3A, our review confirms that SARS-CoV-2 RNA is highly localized in cells that express TMPRSS2, especially ciliated and secretory cells in the airway epithelium, and Alveolar Type 1 (AT1) cells in the lung [TE115]. As reported before, transcriptomics and proteomics analysis show the pathways more involved in patients with severe disease [TE107, E108, TE110, TE116, TE117]. Both mild and severe COVID-19 patients present elevation of chemokines associated with lung inflammatory disorders, such as acute respiratory distress syndrome, asthma, and pulmonary fibrosis [TE88].
Genomics highlights that chromosome 3 is significantly associated with respiratory failure, since in its loci genes functionally interacting with ACE2 are located [TE130].
In Table 3B, pathways most involved in severe COVID-19, highighted by transcriptomics and proteomics data on gastrointestinal, genital and neurologic departments, are reported. The potential susceptibility of these tissues to SARS-CoV-2 entry is due to the high co-expression of ACE2 and TMPRSS2 [TE122-TE127]. However, in these organs SARS-CoV-2 does not determine the damage observed in the respiratory tract, suggesting that ACE2/ TMPRSS2 expression alone is not sufficient to mediate the tissue injury.
As shown in Table 3C The inflammation described in severe COVID-19 is also reflected by metabolomics and lipidomics data, which show imbalanced homeostasis of glycolysis, lipogenesis, heme and ketone biosynthesis, gluconeogenesis, fatty acids oxidation, and cholesterol biosynthesis, through the activation of β-oxidation pathways [TE109]. Metabolomics and lipidomics characterize the difference between mild and severe forms in quantitative terms, and are mostly found in COVID-19 phenotypes associated with comorbidities. The strong association between inflammation and metabolic alterations allows to identify two groups of comorbidities: (1) COVID-19-associated diseases that increase patient frailty by COVID-19-independent pathogenic mechanisms (e.g., chronic heart disease); (2) COVID-19-associated diseases that increase patient frailty by COVID-19-dependent pathogenic mechanisms (e.g., diabetes) (Table 3D, E).
Genomics studies suggest that ACE2 polymorphisms might be associated with cardiovascular and pulmonary conditions by altering the AGT-ACE2 interactions, and transcriptomics data confirm the upregulation of the gene encoding ACE2 receptor in lung tissue in several comorbidities associated with Table 3 Pathogenic mechanisms in COVID-19 phenotype: SARS-CoV-2-host interactions in the lung. (A), DEG and DEP analysis in other organs and tissues (B) Hub genes and pathway of innate immune response (C), Comorbidities COVID19 associated not sharing COVID19 pathogenesis (D), Comorbidities associated and related to COVID-19 pathway (E) severe COVID-19, such as COPD or PAH, and even in people who smoke. Diabetes is the best described co-morbidity related to COVID-19 pathways. This complex metabolic disease is able to complicate COVID-19 by several mechanisms: (1) presence of bone marrow changes, predisposing to excessive proinflammatory response and contributing to insulin resistance, reducing vascular repair and worsening function of heart, kidney, and systemic vasculature; (2) increased circulating FURIN levels, that cleaves the S glycoprotein; (3) dysregulated autophagy, that may promote replication and/ or reduce viral clearance; (4) gut dysbiosis, leading to widespread systemic inflammation, increased glucose and sodium absorption, and reduced absorption of tryptophan needed for glucose homeostasis [TE142].

Phenotypes
COVID-19 patients showed relevant changes in serum levels of lipoprotein subclasses and their components [TE94, TE100, TE102], mainly reflecting the metabolic pathways of lysine degradation, metabolism of taurine, hypotaurine, alphalinolenic acid, glycerophospholipid, arginine, proline, and arginine biosynthesis [TE145]. Severe patients are characterized by pathways listed in Table 3E, possibly linked to a reduced hepatic capacity to oxidize acetyl-CoA in the mitochondria, consistent with serum glucose elevation [TE104].
The technique of transcriptomics per single cell (scRNAseq) contributed also to better understand the etiology of COVID-19 neurological sequelae, although further analyses are needed [TE124].

Discussion
This review highlights that omics technologies could significantly contribute to increase knowledge of COVID-19 pathophysiology. However, it also highlights that disease mechanisms are too complex to be explained by a single molecular signature, even if identified by a multi-omics approach, and an integrated approach is necessary to identify hallmarks of severity. In fact, specific omics techniques are useful to study specific organs and/or tissues and diseases phases, but a single omics cannot explain the complexity of pathways, as shown in Fig. 3, where the same pathway is reported to be eventually up-or downregulated in SARS-CoV-2 infection and in the same severity profile.
The first step in COVID-19 understanding was the genomics and proteomics characterization of the new infectious agent in relation to other coronaviruses. Omics data helped to decipher the perturbation induced by SARS-CoV-2 in miRNA, highlighting possible interactions between viral genomes and host miRNAs. Moreover, they allowed to identify small non-coding viral RNA participating in the pathogenesis of lung disease and to identify regions of SARS-CoV-2 genome as potential sites for RNA silencing.
To reconstruct pathological interactions between virus and host, close to what happens in vivo, it is necessary to combine in silico and in vitro data, integrating as much as possible different kinds of omics data. The integrated multi-omics approach categorizes the most important molecular interactions, allowing the creation of a disease map, able to elucidate pathogenetic interactions between pathogen and host. In this view, results deriving from ex vivo studies should be preferred as compared to those coming from in vitro studies, since the latter may have several limitations, and are usually performed using immortalized cell line, derived in some cases from nonhuman hosts.
Besides virus structure and entry mechanisms, the omics approach allows to identify systemic effectors, and to identify pathways and host signatures correlated with both stage and severity of the disease, mirroring new aspects of COVID-19.
Particularly interesting were the findings regarding changes in oxidative pathways and in lipid metabolism. Alterations of the levels of tricarboxylic acid cycle (TCA) metabolites indicate a metabolic response to declining lung functions and to decreasing blood oxygen, but also reflect a response to nutrition changes, especially in severe patients, leading to glucose elevation as compensatory increase of gluconeogenesis [TE100]. The dysregulated metabolism of glyoxylate and dicarboxylate is also indicator of energy metabolic dysfunction. In moderate and severe patients, most of the lipidomics studies show alterations suggesting an increase in adipose tissue lipolysis [TE101, TE146].
Omics approach highlights that the early phase of the infection elicits a strong activation of innate immunity and an increase of APPs, followed by the "cytokine storm". The upregulation of APPs induces inflammatory cytokines and neutrophil activation in the early phase, but also contributes to activate pathways involved in the subsequent stages of the disease, such as amplification of the cytokine storm, platelet aggregation, activation of coagulation cascades and complement system, and alteration of lipid metabolism.
Most of the studies analyzed in this review reveal that severity appears closely related to dysfunction of platelet degranulation and of coagulation cascades [TE92, TE95, TE96, TE97]. This was demonstrated by: elevation of components and regulators of the complement system; increased expression of thrombotic pathway genes and SERPINE1; decreased expression of PROS; increased levels of D-dimer and fibrinogen degradation products; decreased expression of F13A1 [TE96, TE97]. These evidences are consistent with several non-omics-based clinical studies, reporting a correlation between increased D-dimer concentration and poor prognosis. Other markers of coagulopathy, such as platelet counts and prothrombin time, were viewed as progressively dysregulated with worsening patient conditions [TE147, TE148].
In accordance with those findings, autoptic analysis on fatal COVID-19 cases highlighted high incidence of pulmonary macroemboli and occlusion of alveolar capillaries, revealing severe endothelial injury, increased angiogenesis and microemboli [TE149, TE150]. Only few evidences reported that early stages of COVID-19 are featured by downregulation of complement and coagulation cascades, compared to other viral infections [TE89].
Severity seems also related to changes in PBMCs: expression of neutrophil surface receptors (CXCR1, SINGLEC5, and CD177), neutrophil granule contents (DEFA1) and transcriptomics signatures related to neutrophil activation, chemotaxis, degranulation and migration, are dramatically upregulated in severe compared to mild patients. Other hallmarks of COVID-19 severity are elevated neutrophil/lymphocyte ratio [TE89] and HLA-DR down-modulation in monocytes [TE90].
Transcriptomics studies on local mucosal immune response at upper and lower respiratory tract demonstrate the induction of genes related to immune modulatory functions, including inflammatory response, IFN-α response, IL6/JAK/STAT3 signalling, neutrophil activation, and generation of NETs [TE99, TE110, TE111].
Transcriptomics studies on other districts (e.g., colon) suggest that SARS-CoV-2 infection might influence host responses in body districts not directly hosting the COVID-19 infection, sustaining the inflammatory The heterogeneity of COVID-19 phenotypes remains one of the key questions, just partially explained, and the omics approach might contribute to a better defining the specific mechanisms, characterizing and determining clinical phenotypes.
Like other respiratory infections, SARS-CoV-2 initially activates inflammatory mechanisms in the respiratory tract. Subsequently, the inflammation spreads to other organs and tissues. ACE2 expression in lung cells surely represents an entry factor for SARS-CoV-2, but it is not enough to explain why the damage of SARS-CoV-2 mainly occurs at lung level. Transcriptomics at the lung level optimally characterizes severe forms, showing the activation of several pathways such as Type I INF signalling, histone modifications, neutrophil degranulation, and disorders of transmembrane transporters, and particularly a global dysregulation of immune-related pathways in severe patients.
Metabolomics and lipidomics well characterized in quantitative terms the difference between mild and severe forms, and are particularly useful to study the role of comorbidities. Several amino acid profiles observed in COVID-19 are also observed during acute hepatic failure [24], insulin resistance and increased risk of Type-2 diabetes [25]. In addition, the increase of creatine, creatinine, polyamines spermidine and acetyl-spermidine could suggest renal dysfunction [26] [TE94, TE103, TE104]. These characteristics are well known features of increased risk of cardiovascular disease, diabetes, liver and renal diseases, and they could therefore explain the increased severity of COVID-19 in patients with these co-morbidities.
As discussed above, COVID-19 severity is also linked to dysfunction of coagulation cascades. Based on clinical courses and coagulation parameters, three stages of clinical COVID-19 coagulopathy can be defined [TE151]: -Stage 1: mild systemic inflammation and coagulopathy, with patients having mild symptoms and no need for respiratory support; -Stage 2: progressive pulmonary inflammation, coagulopathy in pulmonary alveoli and microthrombosis, with patients developing more severe symptoms and often requiring additional oxygen supply; -Stage 3: strong pro-inflammatory reaction, development of local and systemic coagulopathy, characterized by high D-dimer and fibrinogen concentration, prolonged prothrombin time, reduced platelet counts, and high incidence of deep vein thrombosis (DVT) or pulmonary embolism (PE); patients' conditions deteriorate, requiring organ support, in particular mechanical ventilation, including extracorporeal membrane oxygenation (ECMO) [TE108].
This sequence of events brings to the death of pneumocytes, along with pulmonary thrombotic microangiopathy (TMA), causing severe clinical conditions with high need for oxygen supply. In this context, the increased D-dimer and enhanced platelet activation levels are strongly linked with pathogenic pro-coagulant phenotype. In turn, the cytokine storm triggers inflammatory processes, bring to systemic endothelitis, cellular and organ dysfunction (i.e. acute respiratory distress syndrome, ARD), and activation of coagulation cascade [27][28][29][30].

Limitations
Limits of this review are intrinsically related to the limits of the omics techniques. In fact, the current omics platforms have several technical limits and are not standardized for the translational use in a clinical context. Many aspects of pathogenetic mechanisms can be investigated by multi-omics approach, but biological mechanisms not yet identified by omics analysis may exist.
Moreover, we have considered the integration between different omics layers as a whole, without carrying out more in-depth analyses, which would have allowed the identification of further suitable markers of disease severity.
Even if clinical studies eventually show good correspondence with the observed omics data, their number is still quite small and some of them are editorials