Omics and systems view of innate immune pathways

Abstract

Multiomics approaches to studying systems biology are very powerful techniques that can elucidate changes in the genomic, transcriptomic, proteomic, and metabolomic levels within a cell type in response to an infection. These approaches are valuable for understanding the mechanisms behind disease pathogenesis and how the immune system responds to being challenged. With the emergence of the COVID-19 pandemic, the importance and utility of these tools have become evident in garnering a better understanding of the systems biology within the innate and adaptive immune response and for developing treatments and preventative measures for new and emerging pathogens that pose a threat to human health. In this review, we focus on state-of-the-art omics technologies within the scope of innate immunity.

1 |. INTRODUCTION

1.1 |. Host-pathogen interactions in innate immune pathways

Host-pathogen interactions describe are complex and dynamic processes that occur during all stages of pathogenic infection, from invasion to dissemination. Both host cells and pathogens have evolved to adopt a wide array of strategies to interact and survive, adding to the complexity of host-pathogen interactions. Central to understanding the innate immune response was the discovery of pattern recognition receptors (PRRs) of the innate immune cells (e.g., dendritic cells and macrophages) that can recognize pathogen-associated molecular patterns (PAMPs) derived from microbes. Classes of these PRRs include the RIG-I-like receptors (RLRs), the Nod-like receptors (NLRs), and the Toll-like receptors (TLRs) [1]. TLRs can be classified as cell-surface TLRs (i.e., TLR1, TLR2, TLR4, TLR5, TLR6, TLR10, TLR11) and intracellular TLRs (i.e., TLR3, TLR4, TLR7, TLR8, TLR9, TLR10, TLR11, TLR12, and TLR13) [2–4]. Each TLR is composed of leucine-rich repeats (LRRs) that recognize PAMPs, a transmembrane domain, and a cytoplasmic Toll/IL-1 receptor (TIR) domain that triggers downstream signaling, resulting in the production of effector molecules, including cytokines, chemokines, and antimicrobial proteins, to combat the invading pathogens [5]. Signaling dynamics vary in response to the TLR ligands, TLRs, and downstream modulators. For example, TLR4 recognizes lipopolysaccharide (LPS), an integral component of the outer membranes of Gram-negative bacteria that causes endotoxic shock and activates the MyD88/IRAK signaling pathways [6]. TLR3 is localized in the endosome and recognizes double-stranded RNA (dsRNA), a viral replication intermediate, initiating the downstream signaling for interferon production [7]. TLR signaling occurs via two major pathways, the MyD88-dependent and TRIF-dependent pathways, named after the adaptor proteins required for signal transduction. MyD88 and TRIF activation causes downstream activation of NF-κB, IRFs (interferon regulatory factors) and MAPKs (mitogen-activated protein kinases) to regulate the expression of inflammatory cytokines and type I IFNs [8].

Aberrations in the PAMP recognition or mutations in molecules involved in TLR signaling can cause autoimmune, inflammatory, and/or allergic diseases. Post-translational modifications (PTMs) on the signaling molecules such as ubiquitination and phosphorylation play critical roles in the activation of TLR signaling, and hence their characterization is highly informative in decoding the mechanisms of signaling. Efforts have been made to identify the molecules (proteins, transcripts, metabolites, and lipids), through integrated approaches, at both the cell and tissue levels [9].

Here, we provide an overview of the current and emerging tools for systems level analyses of innate immunity. The following sections highlight the prominent contributions of “omics” methodologies to the understanding of innate immune signaling pathways at the systems level, with the emphasis on proteomics. Along with a description of these techniques, case studies are presented to elaborate on the suitability and applicability of each technique, the depth and breadth of the information it produces, and the analysis needed to process the data. Figure 1. presents a timeline of omics technologies and their applications. Applications of omics to COVID-19 research have accentuated the importance of these technologies and have accelerated the progress in understanding this disease. We provide examples of omics use in studies of SARS-CoV-2 pathogenicity, host response, and long-term effect.

FIGURE 1.

A timeline of major developments in the “omics” history. Created with BioRender.com.

2 |. MULTIOMICS IN INNATE IMMUNITY

2.1 |. Transcriptomics and genomics

A powerful tool for systems biology research is RNA sequencing, used to study the transcriptome of a cell or organism. The core workflow of any RNA sequencing experiment is the extraction of the RNA, followed by enrichment of the subtype of RNA to be analyzed and depletion of the RNA subtypes that are not to be included. This is followed by the preparation of an adapter ligated cDNA library, amplification of the constructed library, and high throughput sequencing of the library, typically between 10 and 30 million reads per sample. The next step is the computational analysis of the sequenced library. This involves alignment to a reference transcriptome, quantification of overlapping sequences, data normalization between samples and pre-processing, and statistical modeling which is typically done through a surplus of different coding languages and software packages. The majority of RNA sequencing experiments are done with short read sequencing instruments, but recent advances in long read sequencing and direct read sequencing offer new approaches and methodologies to tackle questions not answerable within short read sequencing alone. Each of these approaches comes with their own limitations. For short read sequencing, a major limitation involves biases that are introduced during sample preparation and downstream computational analysis. These biases can affect the quantification of gene isoforms (especially regarding longer transcripts), and how the multi-mapped reads are processed. Long read sequencing and direct read sequencing aim to overcome these limitations, yet they also have limitations of their own.

Regardless of the type of approach used, the primary application of RNA sequencing is to assess differential gene expression (DGE) [10]. DGE analysis is an effective tool for exploring epigenomic and transcriptomic changes in response to stimuli such as differing treatment conditions. For example, DGE has been used to analyze genetic reprogramming caused by stimulation of cells with LPS, which activates both the MyD88- and TRIF-dependent TLR-4 pathways to initiate an inflammatory cascade. This involves activation of key inflammatory genes such as the MAPKs, IRFs, and the master regulator nuclear transcription factor κB (NF-κB). These proteins then induce the production of both pro-inflammatory cytokines and mediators of inflammation leading to a tightly controlled and targeted inflammatory response [11, 12]. These changes can be measured using RNA-seq and analysis with bioinformatics tools and measuring DGE. Differential expression in this case refers to the changes in the expression of a gene from the reference group (unstimulated) to the treated group (LPS stimulated). Numerous software packages exist in both R and Python, to examine differential expression. Some notable examples are Limma and DESeq2 [13, 14]. These packages perform statistical tests on the data and output log fold change and p values which measure how differentially expressed a particular gene might be. Using transcriptomics data analysis methodologies to detect differential expression is a key step for many systems biology experiments, as it defines the baseline of comparison for further analysis with other omics methods and provides a comprehensive output of gene expression. DGE reveals the proteins that are translated in response to cellular signaling dynamics within the pathway of interest.

In the post-genomic era, two types of approaches have proven especially useful for studying the immune system and infectious diseases. Genome-wide association studies (GWAS) have demonstrated that rare gene variants in genes encoding key checkpoint molecules can affect susceptibility to HIV [15] M. tuberculosis [16, 17] and norovirus [18]. Another GWAS study has provided insight into how the course of hepatitis C is affected by gene polymorphisms of molecules involved in the innate immune response, such as IFNL3 (encoding interferon-λ−3) [19, 20] and its pseudogene IFNL4 [21]. Additional examples of GWAS studies in the infectious disease and immunity are described by Thorball et al. [22].

Major advancements in functional genomics, applied widely to immunology, came with the development of CRISPR screens, which have been applied to innate immunity in the studies of TLR4 [23], NLRP3 inflammasome [24], host factors modulating Salmonella [25], and Shigella [26] infection, macrophage regulation of phagocytosis [27] and T-cell activation via IFNγ-inducible MHCII [28]. Many T-cell and tumor immunology CRISPR screens provided major breakthroughs toward the understanding of the disease mechanisms [29].

2.2 |. Proteomics

A commonly employed set of methodologies for systems analysis in innate immunity are proteomics-based methods. Omics strategies offer powerful biological tools, and when combined they can elucidate the complex signaling environment that occurs during host-pathogen interactions. Bottom-up proteomics refers to the digestion of proteins into peptides using suitable enzymes prior to analysis using liquid chromatography coupled to mass spectrometry (LC-MS). Bottom-up proteomics is the most widely used proteomics approach because it provides broad information on protein identity and PTMs, both qualitatively and quantitatively. Bottom-up LC-MS proteomics can be performed in numerous ways: data independent acquisition (DIA), data dependent acquisition (DDA) (e.g., shotgun MS proteomic profiling), and targeted LC-MS. DDA has often been utilized as a baseline experiment, or a first approach, in what is known as discovery level proteomics. It involves a semi-random selection of target analytes for fragmentation based on set cutoffs for relative intensities, which can be very useful for identifying unknown proteins without any prior knowledge. In DIA experiments, another method for comprehensive discovery-type proteomic experiments, non-random selection of precursor ions is performed using a wide precursor ion isolation window. This accounts for more specificity and quantification accuracy for more peptides in comparison to DDA, with the primary tradeoff being of a more challenging spectral analysis. In targeted LC-MS, the parameters describing target analytes to be fragmented are set by the user prior to LC-MS, and these include the precursor mass to charge ratio (m/z), LC elution time, and collision energy needed to fragment each analyte. Targeted MS is a useful tool for validation experiments and absolute protein quantification [30]. Targeted proteomics has two main variations: selected reaction monitoring (SRM) (also known as multiple reaction monitoring or MRM) and parallel reaction monitoring (PRM). SRM enables precise quantification and is typically used with triple quadrupole (QQQ) systems. The first quadrupole (Q1) acts as a mass filter and selectively monitors the analyte precursor, fragmentation is carried out in the second quadrupole Q2, and the third quadrupole Q3 functions as a fragment ion mass filter. Untargeted peptides and their fragment ions are discarded leaving only the fragment ions from the analyte of interest to be quantified by the detector over time [31, 32]. Like SRM, PRM also uses MS² data as input, but rather than selectively monitoring a single target analyte fragmented ions, the full MS² spectra is scanned at high-resolution, and the intensity of multiple fragment ions is monitored in parallel [32]. Bottom-up proteomics using mass spectrometry, commonly coupled with liquid chromatography, either using a shotgun or a targeted strategy, has been successfully used in many studies to assay changes in the proteomic, secretomic [33–39] and phosphoproteomic [40–44] innate immune landscape.

Bottom-up proteomics approaches allow for both relative and absolute quantification of analytes across multiple biological sample groups using different labeling techniques. Relative quantification can be achieved with a label free approach (LFQ), where LC-MS is performed on each sample separately. In contrast, labeling can be performed prior to LC-MS, such as stable isotope labeling in cell culture, and the samples can be pooled together and analyzed using a single LC-MS run to quantify analytes from multiple samples. Absolute quantification can be done by comparing analytes from biological samples to standards labeled using stable isotopes. Alternatively, chemical labels can be used to covalently modify the analyte during sample processing for relative quantification. These include tandem mass tags (TMT), other isobaric tags, ¹⁸O labeling, dimethyl labeling, and other labels [42–44]. The commercial isobaric mass tags (TMT, iTRAQ) are frequently used because they offer a high degree of flexibility. As many as 18 labels that can be analyzed together by pooling the samples into one, but they can be cost-prohibitive.

Bottom-up proteomics approaches also allow for both relative and absolute quantification of a target analyte across multiple biological sample groups using different labeling techniques. Relative quantification can also be achieved with an LFQ, where LC-MS is done on each sample separately. When labeling is done prior to LC-MS, such as stable isotope labeling, samples can be pooled together (both labeled and unlabeled) and run in a single LC-MS run to quantify the target analyte based on the ratio of labeled to unlabeled forms of the analyte. Absolute quantification can be done by comparing to known standards of the stable isotopes. There is a choice of chemical labels that can be added to the analyte post- sample processing for relative quantification such as TMT, isobaric tags, ¹⁸O labels, dimethyl labeling, and others [45–47]. The commercial mass tags (TMT, iTRAQ) are more frequently used because they give the most flexibility offering multiple labels that can be analyzed mixing the samples into one, but they may be cost-prohibitive.

For metabolic labeling-based relative quantification, stable isotope labeling by amino acids in cell culture (SILAC) using heavy isotopes of carbon and nitrogen is frequently used [48, 49], in a classical version giving a choice of three-plex (light, medium, and heavy). More recently, NeuCode SILAC has been developed, and it allows for multiplexing up to seven samples at once [50–52]. All of these labeling methods provide highly accurate and robust workflows for quantification of analytes across multiple biological sample groups [30].

Bottom-up proteomics workflows can be adapted for the detection of PTMs [53]. PTMs are chemical modifications of the side chains of amino acids within a peptide/protein. Detection methods rely on tandem MS and analysis to compare the calculated mass shift of the unmodified tryptic peptide to the modified peptide with the modification mass defined within the search software. Immune system signaling is tightly controlled by PTM changes which are often highly dynamic and the PTMs are sometimes labile. The low abundance of modified proteoforms compared to the corresponding unmodified proteoform adds another challenge to the analysis of PTMs. Consequently, enrichment of modified peptides is almost always needed for deep profiling by LC-MS. Some of the most common PTM enrichment techniques include immobilized metal affinity chromatography (IMAC) [54] and metal oxide affinity chromatography (MOAC) [55] for the enrichment of phosphorylated peptides, di-glycine remnant (KεGG) immunoaffinity enrichment for ubiquitination site identification [56], and hydrophilic interaction chromatography (HILIC) for glycopeptide enrichment [57]. Quantitative measurement of PTMs can be performed simultaneously with their identification by mass spectrometry. For example, a qualitative and quantitative phosphoproteomics analysis of immune cell signaling has been reported [58]. As with any quantification method, data analysis software is essential for making useful interpretations of the data. A comprehensive overview of the software available for proteomics analyses and the advantages and disadvantages of each application can be found elsewhere [59]. Bottom-up proteomics is widely used within systems biology for their ability to investigate functional signaling dynamics and represent a fundamental area of research in the battles against new and emerging diseases. For example, in a study conducted by Wendisch et al., using a multiomics approach including shotgun-based proteomics, it was found that monocyte-derived macrophages accumulate in the lungs during acute respiratory distress syndrome (ARDS) leading to a phenotype resembling one observed in patients with pulmonary fibrosis. Their data supports the hypothesis that SARS-CoV-2 induces a profibrotic transcriptome and proteome profile within macrophages. Usually, these damage repair pathways, when activated within macrophages, help to control inflammatory mediated tissue damage. However, when left unchecked, these pathways can also lead to dysregulated fibroproliferation as well as protracted respiratory failure [60]. This study is one example of how proteomics-based methods can further our understanding of host-pathogen interactions, and it demonstrates the need to develop more proteomics-based workflows that will broaden our knowledge of immune cell-pathogen interactions, and aid in the development of effective treatments to emerging diseases.

While it is the most often employed approach, bottom-up proteomics is not the only strategy used to quantify relative and absolute abundance of proteins. Both top-down and middle-down proteomics can provide valuable information on the higher order structure and intact structure of the target protein. Here, bottom-up falls short owing to a completely different sample processing approach. Top-down proteomics permits analysis of intact proteins and even native-like protein complexes, without the need for proteolysis, and it is useful for characterizing functional and genetic isoforms of target proteins. Since the proteins are undigested, top-down MS also preserves PTM sites and allows for the characterization of entire proteoforms [61]. With a sufficient number of useful fragments, top-down can provide complete primary sequence characterization and at the same time reveal modifications on the protein [62]. The sample preparation methods are, however, not as high throughput or standardized as in bottom-up proteomics. High-resolution mass analyzers are needed to resolve charges on high mass precursors, and the data analysis requires specialized knowledge, which keeps top-down somewhat out of reach. Bottom-up proteomics labeling techniques such as TMT and SILAC labeling have been adapted to top-down workflows and applied to quantitatively measure intact proteins [63]. The protein inference problem in bottom-up proteomics is also addressed by the top-down approach since the data is representative of the parent protein which allows for direct proteoform identification and quantification. For a more comprehensive overview of the quantitative proteomics-based methods for quantification of proteins, Neagu et al. has provided a review summarizing various MS-based approaches and applications of tandem MS for protein analysis in biomedical research [64].

2.3 |. Metabolomics

A powerful tool used for systems wide analysis of, for example, innate immune pathways is metabolomics, which includes the profiling and quantification of metabolites with methods such as LC-MS and nuclear magnetic resonance (NMR) spectroscopy. A metabolite can be defined as an initial, intermediate, or terminal molecule within a metabolic pathway. Metabolites can directly affect cellular signaling dynamics in pathways, especially those that are involved in immune cell signaling. The study of the role of metabolites within the immune system makes up the field known as immunometabolism, which is emerging as a key field of study within metabolomics. It has been shown that metabolic rewiring of immune cell populations can regulate their activation. Fluxes in the metabolic cycle within a cell, as well as the metabolites themselves, can modulate cellular signaling dynamics directly and indirectly through regulation of PTMs. For example, macrophages that are stimulated with LPS have been shown to undergo metabolic reprogramming leading to a pro-inflammatory phenotype [65]. Notably, these cells underwent a switch from oxidative phosphorylation-based ATP production to glycolysis, leading to a higher reactive oxygen species production (ROS) by repurposing their mitochondria to produce ROS in a diethyl succinate dependent manner. Elevated production of succinate within the LPS stimulated macrophages also directly impacted production of interleukin-1B by modulating HIF-1a activity via ROS dependent oxidation, independent of NF-κB signaling. Elevated succinate levels inhibit the production of anti-inflammatory cytokines such as IL-1RA and IL-10, suggesting that upon macrophage activation, the accumulation of succinate directly enhances endogenous pro-inflammatory gene activation and inhibits anti-inflammatory gene expression [66]. Similarly, to a proteomics workflow, a metabolomic LC-MS experiment can be conducted in a targeted or untargeted fashion, and the experimental design is dependent on the chemical composition of the metabolites to be analyzed as well as the metabolite absolute abundance. Due to the differences in structure, chemical composition and mass of peptides and metabolites, experimental parameters required for their analysis also differ to ensure proper fragmentation of the analyte. Untargeted (global) approaches offer a wider detection window and do not require a pre-defined set of metabolites to be screened, unlike the targeted approach [67]. Targeted metabolomics is usually directed toward a subset of metabolites within a biological pathway of interest. Both NMR and LC-MS techniques can be used to perform targeted metabolomics. The spectra obtained from ¹H-NMR are usually compared to the spectra of a known chemical standard, as in a targeted LC-MS experiment. Quantification is based on the ratio of intensities of the detected metabolites that match the pre-defined standards. Untargeted metabolomics is aimed at broadly sampling metabolites with as little bias as possible. Each peak in an untargeted LC-MS run corresponds to a unique mass-to-charge (m/z) ratio and retention time known as a metabolite feature. A more comprehensive review on the analysis of metabolomics data with systems biology approaches has been published [68]. In a recent paper by Bi X. et al, in-depth metabolomic and proteomic profiling of urine compared to patient sera showed that the biomarkers of disease can be measured in urine, as 80% of the sera biomarkers were found in urine, and therefore they could serve to predict COVID-19 severity and COVID-19-induced renal injury [69].

3 |. SINGLE-CELL OMICS

The field of multiomics has seen consistent progress as new omics technologies continue to be developed. Most recently, with the development of single-cell techniques, it has become possible to explore the heterogeneity between cells of the same population. Macrophages act as the primary regulators of inflammation in the innate immune response. They can adopt stimulus induced phenotypes because of their functional plasticity. This allows macrophages to appropriately respond to diverse pathogens and aid in tissue repair following acute damage [70]. M0 monocytes represent the common progenitor lineage for the pro-inflammatory M1 phenotype and the anti-inflammatory M2 phenotype. This differentiation is dependent on induction with polarizing cytokines. The M1 phenotype is a result of stimulation with interferon gamma (IFNγ) and granulocyte macrophage colony stimulating factor (GM-CSF). M1 macrophages secrete pro-inflammatory cytokines such as IL-1β, TNF-α, and IL-12, all of which recruit immune cells to the site of infection and aid in pathogen clearance. M2 macrophages, often described as “wound-healing,” are involved in extracellular matrix repair and produce proline to assist in collagen synthesis and polyamines to induce local cell proliferation. However, this classification might be less rigid, as there are cells displaying characteristics of both, and macrophages have been reported to transition from one phenotype to another [71]. This would normally be missed by a global omics study, as the analytes originate from cells pooled together. For example, during global transcriptomic analyses, cells are generally lysed together, and the RNA extract is thus pooled. The interpretation might rely on the assumption that the cell population of interest is highly homogenous. However, that is often not the case, and with single-cell resolution, one can assess quantifiable changes from cell to cell in the population of interest [72]. Recent developments have now made it possible to extract multiple layers of omics data from a single-cell. For example, the transcriptome and a specific sub-proteome were analyzed within the same single cells in recent studies of COVID-19 using infection-induced single-cell transcriptomics and cell-surface proteomics in peripheral blood mononuclear cells [73]. In another report, V(D)J repertoire and response to tocilizumab was investigated using a similar protocol [74]. Single-cell techniques can provide insight into spatial and temporal organization, as well as the population architecture of each compartment of the cell, tissue, or the organ [75]. This is a promising tool for studying systems biology, especially in the context of innate immunity, as spatio-temporal organization within cells is imperative for relaying cellular messages, leading to activation of immune signaling pathways [76]. Isolation of single cells is a major limitation to downstream omics analyses. There is a constant battle between the loss of cellular material during the isolation process and conservation of the spatio-temporal information. Low throughput methods for single-cell sorting such as laser capture microdissection or manual micromanipulation are costly and laborious, with output typically below a thousand cells per study, but they provide the advantage of decent spatio-temporal resolution [77]. High throughput methods such as fluorescent-activated cell sorting (FACS) are less costly, automated, and provide a high yield of isolated cells. However, they do not retain spatio-temporal resolution since the tissues are homogenized during the pooling of cell suspensions. Thus, choosing the appropriate isolation method for the single-cell omics analyses is a key part of a successful experiment.

Barcoding of cells for library preparation for transcriptome analysis also represents a key component of single-cell omics experiments because it allows for libraries generated by each single-cell to be pooled and sequenced together which saves time and lowers costs. However, this requires a downstream bioinformatics analysis to distinguish barcoded single cells from one another with a high confidence interval [77]. An overview on the multiomics based single-cell approaches with detailed insights into each category has been recently published [78].

3.1 |. Single-cell transcriptomics

Analysis of the transcriptome profile of a single-cell is generally done using single-cell RNA sequencing (scRNA-seq). Most scRNA-seq workflows involve the capture of mRNA based upon separation via the 3′polyadenylated region of the transcript. However, many different protocols exist to isolate other forms of RNA, such as ribosomal RNA (rRNA), or long non-coding RNA (lncRNA), and these are available in both low and high throughput manners. Generally, library construction for scRNA-seq is relatively similar regardless of the type of RNA to be isolated. Cells must be lysed with the appropriate lysis buffer concentration and volume to expel the contents of the cell without denaturing the RNA to be isolated, followed by affinity purification of the RNA to be isolated. Traditionally, this is done by using an oligonucleotide sequence which is complementary to the poly-adenylated tail of the transcript. Because the polyadenylated region is not protein coding, this preserves the region of the transcript which would later be translated into a protein. Some of the highest throughput methods for capture of polyadenylated RNAs involve bead-based separation, where the capture bead is coated with the complementary oligonucleotide sequence, a barcode that tags the transcript with a cell-identifier, and a second barcode which serves as a unique molecular identifier (UMI) for each transcript within the cell. After library preparation, a third barcode is enzymatically added that allows for differentiation of different sample groups; this is especially important when multiple samples are sequenced together.

A study by Kong et al., investigating changes within single-cell transcriptomic profiles within monocytes after induction with Bacillus Calmette–Guerin (BCG) vaccine, found that the amount of systemic inflammation and differential response to LPS upon secondary immune stimulation were markedly reduced [79]. BCG has been used for decades to confer protection against the mycobacterium tuberculosis (TB). However, recent literature has supported the theory that BCG confers immunity against other pathogens as well, including viruses, and the phenomenon has been dubbed “trained immunity.” It has been hypothesized that the decline in the rate of BCG vaccination (owing to the lack of TB infections worldwide), might have contributed to an increase in COVID-19 mortality. In this study, most of the inflammatory mediating chemokines and cytokines, such as CCL3 and CCL4, were downregulated after treatment with BCG and before treatment with LPS, along with a strong correlation between IL-1β and CCL3 and CCL4 [80]. After training, or transcriptional reprogramming induced by BCG vaccination, this correlation was dampened or it disappeared, suggesting that the reprogramming event lowered the sensitivity of the monocytes to pro-inflammatory signals. These observations are in concordance with the hypothesis that BCG induction can lower systemic inflammation upon secondary challenge with LPS. This study represents an effective use of scRNA-seq to evaluate transcriptional reprogramming of monocytes upon BCG induction. However, the clinical relevance of the study is unclear, as it has been shown that in the mouse model, only intravenous administration of the BCG (not a route acceptable for human vaccination) can provide the COVID-19 protective effect and the subcutaneous injection is not protective [81].

3.2 |. Single-cell proteomics

As scRNA-seq enables examination of the heterogeneity of the transcriptome between cells, single-cell proteomics (SCP) enables quantification of proteomic heterogeneity. In contrast to scRNA-seq, where RNA can be amplified, proteins cannot be, which leads to less material in SCP experiments. There are many methodologies that attempt to overcome this limitation by absolute quantification of a small number of proteins or through multiplexed measurements [77]. Due to the wide variety of methods, there is no general workflow for exploring the single-cell proteome. Instead, the workflow chosen by the researcher should be specific to the research question being addressed. As such, the field of SCP can be broken down into two subsections: “absolute quantification” which includes targeted and untargeted methods, and “multiplexing methods” to carry out multiple protein measurements within the same experiment. As is the case with all quantification methods, there are tradeoffs in SCP as well. While one technique may have a wider dynamic range, it can suffer from low-resolution, or a versatile multiplexing assay may lose sensitivity for the protein targets. Quantification with SCP using LC-MS eliminates the reliance on high affinity antibodies to capture the target protein using a traditional immunoassay like ELISA, especially considering non-specific antibody binding as well as antibody availability. However, many immunoassays have been multiplexed and miniaturized. Microfluidic technologies have emerged that now allow for experiments to be performed in the nanoliter (nL) and picoliter (pL) range, which is essential for detecting the low protein concentrations released from a single-cell [82]. For example, Hughes et al. have developed a single-cell western blotting (scWestern) method that has been multiplexed for 11 protein targets and supports up to 1000 concurrent blots in only four hours [83]. When a low starting number of cells is used (<200), FACS sorting can be applied in tandem to improve resolution for single cells. The overall workflow of a typical scWestern blot uses a 30 μm thick photoactive polyacrylamide gel which rests atop a glass microscope slide with an array of 6270 wells. Every step of the procedure from single-cell sorting, cell lysis, target protein capture, and fluorescence detection are performed within this apparatus. The study concluded that scWesterns represent a viable singe cell protein assay capable of high throughput quantitative analysis coupled with multiplexing ability. This technique can target molecular masses via protein electrophoresis, as well as subsequent high affinity antibody binding. The information when combined, generates a high confidence protein identification and specificity, representing a powerful diagnostic tool for SCP analysis.

Likewise, a miniaturized version of ELISA was developed by Shirai et al., where they created a single-molecule ELISA apparatus using micro/nanofluidic technology. The device can use sample volumes in the picoliter range and has the capability to chemically process and capture single molecular targets [84]. Both techniques are significant developments in the fields of microfluidics and SCP with vast applications in medicine and systems biology research. Even though most immunoassays rely upon fluorescent-based readouts, there are new methods such as single-cell barcode chip (SCBC) which do not require fluorescence [85]. Multiplexing in antibody-based methods has been traditionally limited by spectral overlap of fluorophores that are conjugated to the detection antibodies used for reading out data in a highly multiplexed experimental workflow. This limitation is bypassed using antibodies conjugated to metal isotopes which do not have spectral overlap, in a technique known as mass cytometry (CyTOF) [82]. CyTOF can be applied to different questions, including SCP as shown by Orecchioni et al., where the effects of graphene oxide on fifteen types of immune cells were examined using 30 cell markers on single cells [86]. CyTOF also has uses also in spatial proteomics, where a tissue sample region is vaporized using a laser beam, and the ejection or plume released is aerosolized, atomized, ionized, and enters into a time-of-flight mass spectrometer to quantify the isotope abundance [87]. Using the quantified isotope abundance, the “spots” of the vaporized material can then be mapped back to the original coordinates on the tissue section, with the generation of a high dimensional protein map. This provides targeted proteomic information mediated by the detection antibodies while maintaining spatial resolution. The technique can be used with up to 40 unique isotope labeled antibodies to detect up to 40 different proteins. Since multiplexed mass cytometry-based imaging (IMC) provides information of cell composition, phenotype, and spatial organization, all of which dictate cellular signaling, IMC represents an important area of research for systems biology.

Mass spectrometry-based global SCP bypasses the need for pre-defined protein targets and their corresponding high affinity antibodies, which most immunoassay techniques rely upon. Similar to the miniaturization of immunoassays to accommodate single-cell methods, bottom-up proteomics workflows have also been adapted to the single-cell scale with different considerations during the workflow [88]. For example, when extracting the proteins from single cells, it is imperative to limit the non-specific absorption to materials that are housing the sample. Another consideration is related to the most commonly used protease, trypsin. Because trypsin follows Michaelis–Menten Kinetics, the digestion rate is significantly hampered when using small substrate amounts (i.e., total protein within a single-cell) when compared with the amount of substrate in a global bottom-up proteomics workflow (i.e., total protein from many cells pooled together). The ramifications of both these factors on the quantitative readout can be controlled by minimizing sample volumes. In-depth reviews of sample processing and emerging technologies in nanoliter-scale LC-MS proteomics have been recently published [89–91].

Specht et al. describe using SCoPE-MS, a method they developed (and later improved and termed SCoPE2) which takes advantage of multiplexing with TMT labeling, and increases MS signal using one TMT-channel dedicated to a carrier sample, and they were able to detect underlying heterogeneity within macrophage populations differentiated from monocytes using the agonist phorbol-12-myristate-13-acetate (PMA) [92]. They concluded that macrophage heterogeneity exists within each cell’s proteome independent of the cytokine induced differentiation processes. The group assessed whether cell type can be assigned based on the abundance of proteins specific to monocytes or macrophages by color-coding cells based on the median abundance of differentially abundant proteins. They also assessed protein fold changes by averaging the fold changes of protein abundance in single cells (monocytes and macrophages) and comparing these values to the fold changes from analyses of bulk samples for each cell type. This demonstrates the utility of SCoPE2 as an accurate method for determining protein fold changes in single cells. Their findings also support the known biological functionality of M1 and M2 polarized macrophages cell types suggesting that SCoPE2 provides an excellent framework for quantifying relative protein abundance at the single-cell level and facilitates identification of cell-type specific proteins to assess population heterogeneity. This is one example of how SCP can provide meaningful biological insight into heterogeneity, which is an important aspect of any systems biology investigation as cellular heterogeneity is sometimes not measured or accounted for.

3.3 |. Single-cell metabolomics

Single-cell metabolomics (SCM) is the evaluation of the metabolic profiles of cells with single-cell resolution. Like other single-cell omics approaches, SCM provides a means of assessing heterogeneity between single cells [89]. The majority of SCM pipelines rely on mass spectrometry-based methods to identify and quantify metabolomic profiles [93]. As with all single-cell methodologies, the key step is to isolate single cells. This can be accomplished while keeping the morphology intact using techniques such as FACS or microfluidic arrays, or by atomic force microscopy (AFM) which completely isolates the single cell’s metabolites using a probe. For large cells, the successful use of capillary electrophoresis coupled to mass spectrometry has been demonstrated for Xenopus oocytes [94] and neurons [95], but not yet for much smaller immune cells. The next step is to quench all metabolic activity within the single cells. This can be accomplished with the use of organic acids or solvents to denature enzymes and impede further conversion of metabolites, or by a technique known as snap freezing using liquid nitrogen to halt metabolic activity and promote membrane lysis. Snap freezing eliminates the use of reagents that could be contaminants or otherwise problematic during the downstream MS analysis, but this technique requires further work to obtain pure metabolic profiles. The use of organic solvents in the quenching phase is beneficial because it also aids in metabolite extraction from the lysate. Depending on the type of MS used, different combinations of organic solvents are utilized. Guo et al. provide an excellent review on the appropriate use of solvents for each type of MS [93]. After quenching metabolic activity and extracting the metabolites, the samples are ionized. Ionization can be broken down into two distinct mechanisms, vacuum based or ambient methods. A recent review by Liu et al. provides an excellent summary of sub-variations including applications for each technique and data preprocessing/analysis strategies for the obtained MS spectra [96].

M1 macrophages undergo metabolic reprogramming from oxidative phosphorylation to glycolysis upon phenotypic differentiation [97]. Sustained M1 macrophage activation leads to sustained inflammation and can cause tissue damage if not appropriately regulated. The M2 phenotype is acquired after polarization with IL-4 and macrophage colony stimulating factor (M-CSF). M2 macrophages mediate the inflammatory response by secreting anti-inflammatory cytokines such as IL-10 and TGFβ. This gives them an important role in tissue repair after the host inflammatory response. Unlike the M1 phenotype, M2 macrophages rely upon the citric acid (TCA) cycle to support the production of ATP through oxidative phosphorylation. Phenotypical assays to assess macrophage population heterogeneity rely upon detection of cytokines or membrane bound surface markers. However, cytokines and surface markers are expressed in both phenotypes, thus presenting the need to develop a more specific phenotypical classification assay. To reliably quantify phenotypical differences, metabolic profiling with an LC-MS approach can be used. For example, time of flight SIMS MS (TOF-SIMS) coupled with an Orbitrap analyzer (3D OrbiSIMS), allows for MS/MS of metabolites. TOF-SIMS provides highly localized spatial resolution of cell surfaces and does not require extensive sample preparation compared to most other LC-MS techniques. Historically, TOF-SIMS approach has not been reliably used to document endogenous metabolic profiles due to the poor mass resolving power. However, when coupled with the high mass resolving power and mass accuracy of an Orbitrap, characterizing metabolic profiles at the single-cell level is possible. Using a targeted approach to analyze the lipid palette (matching lipid ion peaks to the LIPID MAPS database), it was found that M1 macrophages had the highest lipid counts and different lipid composition compared to M2 or M0 [70]. The amino acid composition and other metabolites showed notable differences as well. Overall, this study represents a novel approach for in situ characterization of metabolic profiles to assess phenotypical differences between closely related cell types with single-cell resolution.

4 |. HIGH THROUGHPUT IMAGING

High throughput imaging (HTI) is a robust set of methods to generate information on cellular morphology, and it includes large-scale automated sample preparation and image analysis [98]. All HTI workflows are based upon a targeted approach. A dye, a fluorescent reagent, a fluorophore conjugated antibody/oligonucleotide, or a genetic construct which expresses a fluorescent protein is used to label a component of the cell. These can include proteins, nucleic acids, or organelles within a cell. A typical experiment begins with perturbing the cells by the addition of ligands to activate cellular signaling pathways and thus inducing DGE or by utilizing short hairpin RNA (shRNA) and clustered regularly interspaced short palindromic repeats (CRISPR/Cas9) which can respectively knockdown and knockout genes in a targeted manner. The ligand used to disrupt the cellular steady-state condition determines the cellular pathways that are involved.

Traditionally, drug discovery methods are target based, they rely on pre-selection of a target molecule (e.g., an interacting protein or receptor), and assess the structural dynamics of “hit/lead” compounds that can bind to the target and modulate its function [99]. This approach typically involves screening the target against a library of compounds predicted to have high binding affinity potential based on knowledge of the binding pocket and/or molecular docking simulations, and assessing the mechanism of action for each compound downstream using experiments [100]. Structure-based drug discovery is rapidly developing as artificial intelligence and deep learning algorithms progress [101]. It should be noted that there is inherent bias with target-based screening, such that hit compounds designed for an individual target may have off-target effects (polypharmacology). To resolve this bias, phenotypic based drug discovery (PDD) methods start by looking at a cellular phenotype and attempt to correlate hit compounds that can modulate the phenotypic readout. Lin et al. provide a review that encompasses the details of workflow design and data analysis for each of these screening methodologies in greater detail [99].

HTI can also be used to profile compounds or libraries in a multiparameter fashion based on statistical analysis and clustering of compounds that elicit a similar phenotypic readout [102]. For example, a multiplexed image-based assay termed “Cell Painting” can measure ~1500 morphological features pertaining to different combinations of size, shape, texture, staining intensity, etc. in response to multiple perturbations [103]. These perturbations can be chemically induced or be a result of genetic manipulation (knockdown/knockout). The technique is capable of sensing subtle phenotypical differences, and it groups perturbing agents (compounds/genes) based on similar pathways that they affect, thus shedding light on the markers of disease. The basic workflow involves genetic or chemical perturbation of a target cell line, staining cells with fluorescent dyes that label 8 different compartments within the cell (nucleus, F-Actin, plasma membrane, mitochondria, etc.), microscopy imaging, and image analysis where the morphological features taken from the images correlate to a profile that reflects the phenotypic state of the cell. Comparing profiles taken at different time points or with different perturbing agents can elucidate the mechanism of action and cellular signaling pathways pertaining to a particular phenotype.

Lastly, HTI methodologies can be classified into a third category, used for deep imaging which combines the use of fully automated high-resolution microscopy and sophisticated computational analysis [104]. Due to the massive amount of cellular image input, this technique can assess rare cellular phenotypes [105]. This niche application of HTI is relatively unexplored, perhaps due to the limited availability of cost prohibitive high-throughput imaging systems. However, even with the limited number of studies utilizing the deep imaging approach, the potential of HTI to identify rare cellular phenotypes in response to a small subset of perturbing agents represents an important area of study in systems biology, that is, phenotypical profiling of biomarkers of disease pathogenesis for rare and understudied host-pathogen interactions.

Nuclear factor kappa-light-chain enhancer of activated B cells (NF-κB) is a family of transcription factors that regulate innate and adaptive immune responses, cellular differentiation, proliferation, and apoptosis [106]. The mammalian NF-κB family is dimeric in nature and includes five different protein monomers (p65/RelA, RelB, cRel, p50/105, and p52/100) that form either homo or heterodimers, and each dimer differentially binds to DNA. All of the monomers have a conserved N-terminal domain known as the Rel homology domain (RHD), which is essential for DNA binding, dimerization, nuclear localization, and inhibitor binding [107]. NF-κB p105 and p100 proteins contain an IκB inhibitory domain which contains multiple copies of the ankyrin repeat (ANK). Both NF-κB subfamily proteins (p105 and p100) undergo proteasome-dependent proteolytic processing to generate their active DNA binding forms (p50 and p52, respectively) [108]. NF-κB dimers reside in the cytoplasm bound to inhibitor proteins of the IκB family, where upon degradation of the inhibitor (phosphorylation of IκB-by-IκB kinase (IKK) followed by ubiquitylation and proteasomal degradation), NF-κB is translocated into the nucleus where it binds DNA and stimulates transcription of target genes. Importantly, one of the target genes is the inhibitor itself which provides a mode of NF-κB signaling regulation via a negative feedback loop. With constant stimulation, the degradation of the inhibitor as well as NF-κB re-synthesis leads to oscillations of NF-κB nuclear translocation [109]. Oscillations in NF-κB translocation are signal dependent. For example, sustained stimulation of cells with tumor necrosis factor a (TNFα) leads to oscillations, whereas a short one-time stimulation with TNFα leads to only one sharp peak of NF-κB translocation/activation [110]. Induction of cells with LPS leads to different kinetics, where NF-κB has been found to translocate in either one cycle of translocation, persistent translocation, or oscillations [111]. Since the translocation of NF-κB has been found to be stimulus-dependent, it is important to consider the kinetic and spatio-temporal landscape of NF-κB when studying pertinent signaling dynamics. These studies highlight the importance of high throughput imaging techniques for studying cellular signaling dynamics with real time quantification and showcase how HTI workflows can be utilized to elucidate the mechanisms underlying host-pathogen interactions.

5 |. COMPUTATIONAL SIMULATION AND MODELING

Computational biology involves the assessment of complex biological systems through the development of computational models and simulations which can be used to develop predictive models of the factors involved in disease pathogenesis. The field is rapidly progressing with developments in computer hardware, software, and experimental methods, lowering the computational efforts required to produce these models. Computational models can be separated into two subgroups, quantitative and logical models. A quantitative model utilizes sets of differential equations to define the dynamics of the model which are sometimes non-linear. It requires pre-defined knowledge of details regarding the pathway or cellular event under study and is thus often limited to modeling small portions of a well classified pathway. A logical model is based upon a Boolean system and qualitatively defines the dynamics of the model. It does not require a pre-defined set of knowledge about the system to be analyzed, and it can be applied to larger systems.

A subfield of computational biology utilizes both modeling approaches and resides at the intersection of systems biology and traditional bioinformatics, known as systems bioinformatics [112]. The field of systems bioinformatics can be defined as the framework for integrating the multiomics landscape traditionally used in systems biology approaches, to provide insight into each individual omics layer and the cumulative interactions between them. In this way, systems bioinformatics provides methodologies capable of assessing the biological mechanisms of the entire interwoven system rather than the summation of each individual component or omics layer. This field is based around systems theory which is holistic in nature, and its use in systems bioinformatics is dependent on graph theory, network science, and other mathematical approaches which facilitate the analysis of complex networks derived from a system of interest.

Mathematical models are fundamental for analyzing network topology and kinetics. As multiomics based quantification methods advance in both high throughput ability and sensitivity, more accurate parameters can be fed into models to produce more accurate quantifiable simulations of signaling dynamics. In general, networks can be applied to qualitative models of pathway modeling. Many open access platforms are available for this purpose. These software programs work by taking an input list of gene symbols or protein names and assessing their annotations such as gene ontology (GO) annotations to map them. For example, Cytoscape facilitates the visualization of complex biological networks with annotated gene symbols, gene annotations, and gene expression data [113]. Reactome enhanced pathway visualization is an alternative [114]. There are several pathway databases which enable the visualization of signaling components based upon, for example, GO terms. A review comparing some of the most widely used databases has been published [115]. Another detailed review of the construction and analysis of biological pathways has been can be found here [116].

The construction of networks to showcase biological pathways can incorporate both mathematical modeling functions and qualitative visualization to help researchers understand the molecular dynamics involved in cell signaling. One such software program is Simmune, which can generate computational models incorporating spatially resolved reaction-diffusion networks. Simmune utilizes rule-based approaches to lower computational complexity of simulations. This involves pre-programming of the simulation with fundamental signaling components and their pair-wise interactions, which allows Simmune to assemble the complexes that constitute the signaling network [117]. This rule-based approach was incorporated as a response to one of the most traditional challenges in pathway modeling: combinatorial explosion. Combinatorial explosion can occur when there is an excessively high number of alternative interactions arising from a network consisting of many different signaling components, or individual components that have multiple binding sites and thus many possible interactions. Simmune works by generating a local network in a multi-step fashion. The first step involves the construction of a non-spatial network that includes every possible molecular interaction for each fundamental signaling component. This “template network” is then adjusted to reflect the local molecular environment which lowers computational extensivity of the simulation. Simmune is also able to account for morphologically dynamic models which usually requires rebuilding the network every time the cellular morphology changes, to account for spatial constraints concerning the receptor ligand interactions during membrane fluctuation. In-silico approaches for assessing pathway dynamics represent an important stepping-stone in systems bioinformatics and all related disciplines to either validate or predict experimental findings. In a study by Manes et al., the chemosensing pathway of sphingosine-1-phosphate (S1P) was explored using a combinatorial approach of RNA sequencing, targeted proteomics, and Simmune based modeling [118, 119]. This highlights the importance of in silico based computational models in providing insights into molecular mechanics of complex signaling networks and validating the experimental findings.

Protein folding simulations such as AlphaFold2 are also an informative tool for predicting and assessing protein structure and can be used to diagnose a protein’s function [120]. AlphaFold2 can construct a 3-dimensional representation of how a protein will fold, based upon its primary sequence. The software uses a deep learning-based algorithm with multi-sequence alignment which incorporates both physical and biological knowledge regarding protein structure [121]. In addition to the software’s abilities, the AlphaFold team has now released accurate structure predictions for the proteomes of numerous species in a freely available database [122]. The availability of a database with highly accurate protein structures that are continuously updated is a major step forward for the field of structural biology as it takes away the burden of generating these structural models from scratch [123]. As AlphaFold2 evolves along with the database, we can expect to see more structural predictions that are publicly available and provide researchers with tools that can be exploited for drug discovery, investigating protein-protein interactions, and creating simulations of pathway dynamics where each component of the signaling pathway includes a highly accurate 3-dimensional model of its native conformational shape [124]. This will greatly benefit the field of systems biology as better structural predictions of the human proteome can help researchers assess all of the possible functions of a protein and build more complex models of their kinetics.

6 |. INTEGRATING OMICS APPROACHES

The integration of omics approaches can be performed in two ways: by (1) combining technologies and discovering biological information that can be inferred from correlations between these technologies and (2) using one or more technologies to simultaneously probe interacting populations, such as host and pathogen or host and microbiome.

An example of the first approach directly pertaining to the innate immune system has been recently provided in a study by Qie et al. where the authors characterized the proteome and transcriptome of macrophages in mouse tissue using mass spectrometry and bulk RNA sequencing in 10 primary macrophage populations from seven mouse tissues, bone marrow derived macrophages, and the commonly used monocyte-macrophage cell line RAW264.7, allowing for construction of regulatory networks revealing cell-specific and tissue-specific differences between the populations [125]. This study serves as an atlas of mouse macrophage proteomes and transcriptomes in a given population in the basal state.

The microbiome is an important system involved in both gutmucosal and systemic immunity. 70%–80% of all immune cells are present in the gut. There are intricate interplays associated with the intestinal microbiota and local mucosal immune system which shape the immune response to invasive pathogens [126]. The gut microbiota influences many aspects of host physiology such as the development of the immune system, drug metabolism, regulation of inflammatory diseases, overall nutritional status, and drug metabolism. For assessing the impact of the microbiota on host physiology, it is important to have a model that facilitates drawing meaningful biological conclusions. One such model is the use of germ free (GF) mice which have dramatic impairments on their metabolism and immune system [127]. In the study, proteomic and transcriptomic expression profiles within the terminal ileum (the end of the small intestine, which has a high concentration of commensal microbes) of GF and conventional mice were compared. The germfree status was treated as the perturbed state. The work compared significantly affected genes based on germ status in both datasets to assess their statistically significant correlations. The presence of the microbiota in conventional mice was linked to upregulation of immune system pathways for the gene subsets which were transcriptome-proteome concordant. The group hypothesized that this was due to migration of immune cells to the ileum as shown in previous studies [128]. Also, metabolic pathways were downregulated in conventional mice. Due to transcriptomic-proteomic discordance caused by the microbiota, neither of the layers alone is sufficient to produce a complete picture of the complex influence the microbiota has on the host’s physiology, emphasizing the need for integrative omics-based approaches to fully elucidate the host-microbiota interactions and their molecular underpinnings. An overview of the discussed “omics” approaches and their application to analyzing biomolecules (transcripts, proteins, metabolites) within a cell is presented in Figure 2.

FIGURE 2.

An overview of the various applications of omics approaches to analyzing biomolecules within a cell as well as interactomics and morphology and at both the cell and tissue levels. Created with BioRender.com

Metatranscriptomics investigations of human microbiota in different organ communities have led to a better understanding of commensal microbial communities in healthy humans [129] and in diseases such as diabetes [130], inflammatory bowel disease [131] (where metatranscriptomics was elegantly combined with metabolomics and antibody profiling), bacterial vaginosis [128], cystic fibrosis—where metatranscriptomics was elegantly combined with metabolomics and antibody profiling into a multi-omics study, bacterial vaginosis [132], cystic fibrosis [133] periodontitis [130], and persistent wounds or periodontitis [134], as well as persistent wounds [135]. More examples can be found in a recent review [136].

Another area where integrative approaches have been successfully developed is antiviral innate immunity, where proteomic analyses of virus-host protein-protein interactions have been explored. In one example, Dybas et al. used a DNA-centric affinity purification combined with mass spectrometry (AP-MS) approach to identify novel antiviral restriction factors with functions during adenovirus infection by comparing wild type virus and a mutant strain lacking immunomodulatory effectors [137]. Another study of protein-protein interactions using mass spectrometry-based proteomics provided the first interactome between SARS-COV-2 and host cells and revealed pan-viral disease mechanisms [138].

In COVID-19, mutated variants of the SARS-CoV-2 genome are responsible for enhancing the pathogenicity of the virus and thus the overall impact of the pandemic. Using a multiomics approach, Thorne et al. demonstrated that SARS-CoV-2 affects more than just the adaptive immune response; the innate immune system is implicated as well in the severity and transmission of the virus [139]. They first assessed the multiplicity of infection for the first wave isolates (early lineage) and Alpha variant isolates in Calu-3 human epithelial cells by measuring intracellular copies of envelope (E), Nuclear capsid (N), and virion production. It should be noted that viral dsRNA is classified as a PAMP which are recognized by RNA sensing adaptors such as mitochondrial antiviral-signaling protein (MAVS). Interestingly, they observed that at 6 h post infection (hpi), total dsRNA decreased for the Alpha isolates even though replication was comparable between alpha and first wave isolates. They hypothesized that this could be due to two factors. One being that the Alpha N protein can contribute to innate immune evasion by sequestering dsRNA and thus induce epitope masking; the other being transposable elements contributing to the reduction of endogenous dsRNA production. It has been shown that Alpha infection leads to lower expression and secretion of interferon-β (IFNβ)—a prominent marker for innate immune activation [140]. At an early timepoint (24 hpi), Alpha was shown to induce less expression of IFNβ and interferon stimulated genes (ISGs) when compared to an early lineage variant B.1.13 hCoV-19/England/IC19/2020 (IC19) using RNA-Seq. This suggested enhanced innate immune evasion potential of the Alpha variant. To further characterize innate immune antagonism by Alpha, they compared global host responses using mass spectrometry-based protein abundance and phosphopeptide enrichment as well as total RNA-seq in Calu-3 cells (10 and 24 hpi). Notably, changes in RNA abundance and protein phosphorylation seemed to be infection driven yet the changes in overall protein abundance were less significant. It was also noticed that there was no clear correlation between protein/mRNA abundance and the levels of protein phosphorylation, indicating that enhanced phosphorylation may be driven by another mechanism independent of the levels of protein abundance. This study demonstrated how multiomics approaches can be used to study changes across SARS-COV-2 variants, and that the enhanced innate immune evasion is an important mechanism for enhanced pathogenicity of the Alpha variant.

The COVID-19 disease has been known to lead to immune dysregulation and potentially to ARDS. ARDS can be caused by hypercytokinemia, otherwise known as a “cytokine storm.” This is characteristic of a hyperactivated and unregulated pro-inflammatory cytokine response, which can lead to inflammatory induced tissue damage for the host. Typically, viral infection is mediated by type-I IFN gene induction leading to the expression of ISGs that exhibit anti-viral mechanisms of action. SARS-CoV-2 has developed mutations that are structural changes in the proteome and can antagonize the host IFN response. Thus, studying the host immune response to SARS-CoV-2 represents a fundamental step for diagnosing phenotypes indicative of a dysregulated immune response and may provide insight on how to treat the disease. In the study put forth by Zhou et al., bronchoalveolar lavage fluid (BALF) was collected from 8 COVID-19 patients (SARS2), 146 community-acquired pneumonia patients (CAP), and 20 healthy controls (Healthy) [141]. Meta-transcriptomic analysis resulted in alignment to the human genome among all samples, and it also captured the transcriptomic information for the microbes found in the samples. The differentially expressed gene (DEG) population for the SARS2 versus healthy (SARS2-H) comparison was markedly higher than the rest, showing that SARS-CoV-2 infection causes significant perturbations from homeostasis in the host lung tissue. Some key upregulated DEGs in SARS2-H included pro-inflammatory cytokine and chemokine genes (IL-1β, CXCL17, CXCL8, CCL2), anti-viral ISGs (IFIT, IFITM family genes), and calgranulin genes (S100A8, S100A9, S100A12). Downregulated DEGs included genes involved in morphogenesis and cellular migration (NCKAP1L, DOCK2, SPN, DOCK10). Among the cell signaling pathways, “interferon signaling” was most enriched in the SARS2 group and was also enriched to a lesser extent in the Virus-like CAP samples. This indicates an IFN-driven response to SARS-CoV-2 infection. Many of the enriched pathways for the SARS2 group are classified as innate immune pathways such as NF-κB, TNF, and the IL-17 pathways. Network analysis showed a densely connected subnetwork between the ribosome and chemokine signaling pathway. Upon assessment of the cytokine profiles, it was observed that the pro-inflammatory cytokine expression dissipates over time, suggesting that inflammation during COVID-19 is resolved over time and unquenched inflammation may lead to detrimental outcomes regarding disease progression. CXCL17, which has a role in neutrophil recruitment in the lung, was the most upregulated chemokine for SARS2. Monocyte attractors, such as CCL2 and CCL7 were upregulated as well. Data indicated a correlation between viral load and chemokine production, where higher viral load corresponds to higher chemokine gene expression. IL1RN and ILB interleukin genes were enriched specific to SARS2, validated by their quantification of cognate protein products (IL-1Ra and IL-1β respectively) in COVID-19 patient plasma. IL-1Ra is the inhibitor to IL-1β, and IL-1β has been previously reported to be the driving factor of the proinflammatory response during ARDS, and this suggests a potential biomarker for COVID-19 severity based upon the ratio of IL-1Ra and IL-1β. The IFN response was then examined where SARS2 showed elevated expression of ISGs (83 significantly upregulated) compared to CAP subgroups which diminished over time for each patient. These included IFIT and IFITM genes with broad antiviral functionality (IFIT1,2,3 and IFITM2,3); these have also been shown to inhibit viral cellular entry of SARS-CoV-2. Also, two key innate immunity associated transcription factors, IRF7 and STAT1, were markedly upregulated, which could further potentiate the IFN response. SARS2 also resulted in higher neutrophil populations compared to the pneumonia group, and less diversity in the T and B cell populations when compared to the innate cell populations. In a previous study, the neutrophil to lymphocyte ratio was characteristic of disease severity in COVID-19 cases, indicating yet another biomarker for disease severity [142]. This study showcases the capacity of transcriptomics to detect phonotypes associated with disease severity at different expression levels, using DGE analysis of cytokines/chemokines, ISGs, and broad cell population determination. Several potential crosstalk points between the innate and adaptive immune system also revealed the innate immune pathways which affect the pro-inflammatory response to SARS-CoV-2 pathogenesis. A study of host-pathogen interactions at both the innate and adaptive immune levels is required to decode the early and late responses associated with pathogenicity and virulence and to develop effective treatment options in the clinical setting.

Unterman et al. used single-cell multiomics (gene expression profiling and cell lineage protein marker analysis), highlighting the abnormal MHCII/LAG3 interaction between myeloid and T cells and alterations to the TCR and BCR repertoire, and measured the effects of tocilizumab treatment on different peripheral blood immune cells to demonstrate the dyssynchrony of the innate and adaptive immune system interactions in patients with progressive COVID-19, which may contribute to the delayed virus clearance [74]. With all the above examples, it is evident and important to note that with the increasing complexity of integrative multi-omics studies it is imperative to develop computational data analysis tools. Software packages and complex approaches have been developed recently. Examples include software for the analysis of the DIA mass spectrometry and metaproteomics datasets, which was demonstrated using a gut microbiota metaproteomics dataset [143]. Additionally, the transkingdom network analysis protocol (TransNet) [140] can be used [144] to integrate and interrogate multi-omics data to identify causal relationships between diverse datasets, and this has been essential to the exploration of the relationship between Lactobacilli and hepatocytes [145], and between the microbiota and adipocytes in diabetes [146].

7 |. CONCLUSIONS

Multiomics approaches are powerful tools for systems biology such as studying host-pathogen interactions to generate a broad view of the innate immune system. Transcriptomics and proteomics are being employed to generate valuable data of several types such as DGE, protein absolute abundance, site of PTM, proteoform profiling, higher order structure, and protein interactions, both qualitatively and quantitatively, and with high accuracy. Considering the role of metabolites in the regulation of immune cell signaling, immunometabolism has recently developed into an area of intense research. The mechanisms of how precise changes in the metabolites involved in a signaling pathways shape the immune response can be predicted. Single-cell measurements have made it possible to observe and quantify the heterogeneity between cells of the same population, whether it be the heterogeneity between transcriptome, proteome, or metabolome or the spatial architecture of a population of cells. Single-cell omics have completely revolutionized the field as a tool to study systems biology. As is the case with any technique, limitations regarding cost, highly specialized instrumentation, trained personnel, and availability remain for omics methods. Table 1 summarizes prominent features, limitations, and executability of the major omics technologies.

TABLE 1.

A summary of important features for each “omics” technique.

Technology	Prime features	Limitations	Ease of execution
Proteomics	Quantitative Ability to identify 1000s of proteins in single experimental run. Native (e.g., protein complexes) and non-native (e.g., peptide level) states can be examined. Lower sample amounts required (nanograms). Ability to assess PTMs and their distribution. Characterization of binding sites that modulate protein effector function.	Data analysis is customized based on the experiment and need knowledge of a variety of software. Isobaric labeling approaches for targeted proteomics may suffer from incomplete incorporation in cell culture. Membrane proteins are hard to isolate and digest due to difficulty of precipitation because of aggregation.	High-resolution instruments are cost prohibitive; specialized personnel needed to perform advanced workflows. Proteomics based MS workflows are well documented in literature, which makes execution of these methods easy to follow.
Transcriptomics	Accurate quantification of transcript abundance. Alignment to a reference genome allows for identification of target genes. Can be used to assess differential gene expression after perturbation from homeostasis.	cDNA library construction can lead to dimerization of primers making datasets noisy. During cDNA library construction spurious second strand cDNA artefacts can be generated, which can confound sense vs. antisense transcripts.	Next generation sequencing instruments are expensive. However, sending samples to specialized sequencing facilities is cost effective and often a better alternative. Protocols for sequencing, alignment, and statistical analysis are well defined making RNA-Seq practical for the masses.
Metabolomics	Metabolic profiling in quantitative manner. Ability to assess metabolic flux in response to perturbation from homeostasis.	Real time metabolic profiling is difficult. Lipids can be particularly hard to ionize using MS based approaches.	MS based approaches for metabolite quantification are well defined in the literature. Difficult to execute without access to a high-resolution Mass spectrometer.
Single-cell omics	Deeper analysis of rare cell types. Recent advances have increased throughput and reproducibility. Phenotypical classification by assessing intrapopulation heterogeneity at each omics layer.	Sample preparation, isolation of single cells and throughput are limitations for all single-cell approaches. Targeted approaches are reliant on antibody availability. With new improvements coming up rapidly, reproducibility could be an issue.	Due to the abundance of methodologies for both targeted and untargeted approaches, it is now relatively easy to incorporate single-cell workflows into any workflow. Some approaches rely on expensive instrumentation (MS, NGS), others, for example, miniaturized immunoassays for scProteomics are more affordable.
High throughput imaging	Localization of surface molecules, subcellular localization of organelles and macromolecules. Phenotypical characterizations based on expression of specific markers in response to homeostatic perturbations. Accurate characterization of cellular morphology. Assessment of intrapopulation heterogeneity based on expression of specific markers or overall cellular morphology.	The availability of antibodies specific to the markers of interest.	High-resolution imaging instruments are expensive.
Computational modeling	Ability to profile the kinetics of molecular interactions. In silico validation of experimental findings can help predict the outcome of experiments. Highly accurate 3-dimensional modeling of protein structures. Crucial for early stages of drug discovery.	Specific parameters need to be set from experimental results for the simulation to mimic endogenous conditions. Assumptions are required at the software end that may not accurately describe the environment, for example, assuming the reaction volume to be homogenous. Alphafold multimer does not accurately account for PTMs and not all PPIs are accurately mapped.	Most software packages that can be used to model molecular dynamics, pathways, or protein modeling are open source and do not require a commercial license. This makes in-silico computational models highly accessible and easy to use. Some modeling software are computationally costly and require the use of high-performance clusters.

The development of advanced instrumentation has empowered omics approaches and has reduced reliance on high affinity antibodies for exploring the immune system. High throughput imaging is highly informative in exploring cellular morphology and has been used to profile large compound libraries for drug discovery. Assessing molecular interactions over time to predict disease pathogenesis has become possible with the use of computational models and molecular dynamics simulations. Deep learning-based tools such as AlphaFold have revolutionized structural prediction and investigating protein-protein interaction and can support simulating pathway dynamics. Multi-omics analysis platforms have accelerated the diagnosis and the understanding of COVID-19 pathogenesis, as well as aided in the identification of potential therapeutic treatments. Taken together, omics tools are highly powerful for investigating systems biology holistically and multidimensionally, providing a larger and more complete picture of the innate immune landscape and the whole immune system.

Abbreviations:

AFM

atomic force microscopy

ANK

ankyrin repeat

ARDS

acute respiratory distress syndrome

ATP

adenosine triphosphate

BALF

bronchoalveolar lavage fluid

BCG

Bacillus Calmette-Guerin

CAP

community-acquired pneumonia

CCL

chemokine (C-C motif) ligand

cDNA

copy DNA

COVID-19

coronavirus disease 2019

CRISPR

clustered regularly interspaced short palindromic repeats

CXCL

chemokine (C-X-C motif) ligand

CyTOF

mass cytometry

DDA

data dependent acquisition

DIA

data independent acquisition

DGE

differential gene expression

DNA

deoxyribonucleic acid

GF

germ free

GM-CSF

granulocyte-macrophage colony stimulating factor

GWAS

genome-wide association studies

HIF1A

hypoxia inducible factor 1 subunit alpha

HILIC

hydrophilic interaction chromatography

HIV

human immunodeficiency virus

HTI

high-throughput imaging

IFIT

interferon induced protein with tetratricopeptide repeats

IFITM

interferon induced transmembrane protein

IFN

interferon

IL

interleukin

IKK

IκB kinase

IMC

imaging mass cytometry

IRAK

interleukin-1 receptor-associated kinase

IRF

interferon regulatory factor

ISG

interferon-stimulated gene

LFQ

label-free quantification

lncRNA

long non-coding RNA

LPS

lipopolysaccharide

LRR

leucine-rich repeat

MAPK

mitogen-associated protein kinase

MAVS

mitochondrial antiviral-signaling protein

M-CSF

macrophage colony stimulating factor

MHC

major histocompatibility complex

MOAC

metal oxide affinity chromatography

mRNA

messenger RNA

MyD88

myeloid differentiation primary response 88

NF-kB

nuclear factor kappa light chain enhancer of activated B cells

NLR

NOD-like receptor

PAMP

pathogen-associated molecular pattern

PDD

phenotype-based drug discovery

PRR

pathogen recognition receptor

QQQ

triple-quadrupole

RHD

Rel-homology domain

RNA

ribonucleic acid

rRNA

ribosome RNA

SARS-CoV2

severe acute respiratory syndrome coronavirus 2

sc

single cell

SCBC

single cell barcode chip

SCM

single cell metabolomics

SCoPE-MS

Single Cell ProtEomics by Mass Spectrometry

SCP

single cell proteomics

shRNA

short hairpin RNA

STAT

signal transducer and activator of transcription

TB

tuberculosis

TCA

citric acid cycle

TIR

Toll/IL-1 receptor

TLR

Toll-like receptor

TMT

tandem mass tag

TNF

tumor necrosis factor

TRIF

TIR-domain-containing adapter-inducing interferon-β

Biographies

Dr. Aleksandra Nita-Lazar received her Ph.D. in biochemistry in 2003 from the University of Basel for studies performed at the Friedrich Miescher Institute for Biomedical Research, where she analyzed atypical protein glycosylation using mass spectrometry and protein biochemistry methods. After postdoctoral training at Stony Brook University and Massachusetts Institute of Technology (Ludwig Cancer Foundation Fellow), where she continued to investigate post-translational protein modifications and their influence on cell signaling, she joined the Program in Systems Immunology and Infectious Disease Modeling, now the Laboratory of Immune System Biology, DIR, NIAID, NIH, in April 2009, as an independent investigator and Chief of the Cellular Networks Proteomics Unit. Dr. Nita-Lazar was granted tenure in December 2018 and she now continues her work as Senior Investigator and Chief of the Functional Cellular Networks Section. Her main research interests are protein expression and modification changes regulating the Toll-like receptor signaling and macrophage activation.

Dr. Deepali Rathore is a Visiting Fellow at the National Institute of Allergy and Infectious Diseases (NIAID)/NIH in the Functional Cellular Networks section, under the mentorship of Dr. Aleksandra Nita-Lazar. She obtained her PhD from the University of Nebraska-Lincoln, focused on integrating tandem MS with ion mobility MS for multidimensional protein complex analysis. Her current research includes understanding the role of RNA binding proteins as modulators of innate immune signaling, role of linear ubiquitination in Toll-like receptor pathway activation and signaling and investigating phosphorylation dynamics in antibiotic resistance.

Matthew Marino received his Bachelor’s degree in Biomolecular Science from Central Connecticut State University with a minor in chemistry. He is currently a post baccalaureate fellow at the National Institute of Allergy and Infectious Diseases (NIAID)/NIH in the Functional Cellular Networks section headed by Dr. Aleksandra Nita-Lazar. His research interests include multi-omics approaches for studying innate immunity, Toll-like receptor signaling, and the RNA bound interactome within mouse macrophages. He is looking to pursue a Ph.D. in chemical biology at the end of this year.

DATA AVAILABILITY STATEMENT

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.