Network analysis of the autophagy biochemical network in relation to various autophagy-targeted proteins found among SARS-CoV-2 variants of concern

Abstract

Autophagy is an important cellular process that triggers a coordinated action involving multiple individual proteins and protein complexes while SARS-CoV-2 (SARS2) was found to both hinder autophagy to evade host defense and utilize autophagy for viral replication. Interestingly, the possible significant stages of the autophagy biochemical network in relation to the corresponding autophagy-targeted SARS2 proteins from the different variants of concern (VOC) were never established. In this study, we performed the following: autophagy biochemical network design and centrality analyses; generated autophagy-targeted SARS2 protein models; and superimposed protein models for structural comparison. We identified 2 significant biochemical pathways (one starts from the ULK complex and the other starts from the PI3P complex) within the autophagy biochemical network. Similarly, we determined that the autophagy-targeted SARS2 proteins (Nsp15, M, ORF7a, ORF3a, and E) are structurally conserved throughout the different SARS2 VOC suggesting that the function of each protein is preserved during SARS2 evolution. Interestingly, among the autophagy-targeted SARS2 proteins, the M protein coincides with the 2 significant biochemical pathways we identified within the autophagy biochemical network. In this regard, we propose that the SARS2 M protein is the main determinant that would influence autophagy outcome in regard to SARS2 infection.

Graphical abstract

1. Introduction

Autophagy is an important cellular process responsible for degrading dysfunctional organelles, damaged cytosolic proteins, and intracellular pathogens [1,2]. Moreover, autophagy induction triggers a coordinated action involving multiple individual proteins and protein complexes [3] and is dependent on cellular energy status, nutrient abundance, amino acid (mammalian target of rapamycin), and growth factors [4,5]. Additionally, during a viral infection, the autophagy process is induced and provides an anti-virus response against the on-going infection [6,7]. In this regard, some viruses (i.e. herpes simplex virus, human cytomegalovirus, HIV-1, coronaviruses) inhibit the autophagic response by either exiting the autophagic process without being lysed or by blocking autophagic -degradation at the final stage, thereby, resulting to immune evasion [5,8,9].

Coronaviruses (CoV) are RNA viruses that are classified as: family Coronaviridae, order Nidovirales, and subfamily Othocoronavirinae [10]. Additionally, there are currently seven known human-infecting CoVs with only SARS-CoV-2 (SARS2) causing a pandemic which lead to coronavirus disease 2019 (COVID-19) [11]. Moreover, SARS2 was found to reduce autophagy induction as characterized by both SQSTM1 accumulation and LC3B-II increase which in-turn has been correlated to multiple SARS2 proteins, namely: non-structural protein 15 (Nsp15), membrane protein (M), open reading frame 7a (ORF7a), open reading frame 3a (ORF3a), and envelope protein (E) [12]. This highlights the effectiveness of SARS2 in inhibiting autophagy since there are multiple SARS2 proteins that can independently or in combination hamper the autophagy process. However, to our knowledge, it was never established which stage of the autophagy biochemical network and corresponding SARS2 proteins (Nsp15, M, ORF7a, ORF3a, E) play a significant role. Furthermore, after the detection of the original SARS2 strain, variants of concern (VOC) appeared which possessed higher virulence and transmissibility rates compared to the original strain [13]. This may likewise suggest that the autophagy process varies among the SARS2 variants. However, this was likewise not established. A common approach towards elucidating the interconnection of the various components involved in a given cellular process or pathway utilizes network analytics [[14], [15], [16]]. In this regard, this study attempted to use network analytics in order to identify significant sections within the autophagy pathway which in-turn may elucidate whether possible VOC-specific structural changes among the corresponding SARS2 proteins (Nsp15, M, ORF7a, ORF3a, E) coincide with the significant autophagy sections. A better understanding of the interconnections within the autophagy pathway and the potential correlation with the autophagy-inhibiting SARS2 proteins may help identify possible antivirulence drug targets within the autophagy pathway.

2. Materials and methods

2.1. Network design and analyses of the autophagy pathway

Network design of the autophagy pathway was based on the animal autophagy pathway registered in the KEGG Pathway Database (http://www.genome.jp/kegg/pathway.html) and utilizes the Cytoscape software in order to design and connect interacting biochemical components into a single comprehensible conceptual framework [17]. Briefly, nodes represented the biochemical components involved in the animal autophagy while the directed edges (represented as an arrow) represented the direction of transition among the biochemical components.

Network analyses was conducted through centrality measurements which likewise used the Cytoscape software [17]. In this study, we performed the following centrality measurements: (1) betweenness centrality to show which biochemical component is crucial in maintaining functionality and coherence within the autophagy network; (2) stress centrality to establish how important a biochemical component is in the autophagy network; (3) closeness centrality to highlight the biochemical component that is functionally relevant to other biochemical components in the autophagy network; (4) radiality centrality to elucidate the possibility of the biochemical component to be either relevant or irrelevant within the autophagy network; (5) eccentricity centrality to emphasize the easiness of a biochemical component to be reached by other components within the autophagy network; and (6) edge betweenness centrality to determine the degree of connection between two biochemical components within the autophagy network [18]. Briefly, in all centrality measurements made, the threshold for each centrality was first determined and, subsequently, the degree of centrality based on whether the nodal or edge values were higher than the threshold was identified for all 6 centrality measurements. Thus, node and edge centrality measurements with values greater than the threshold were considered significant. Moreover, results from all centrality measurements were combined into one network design (unified network) in order to show the common significant biochemical components.

2.2. Protein model generation and model quality assessment of autophagy-related SARS2 proteins

A minimum of ten amino acid sequences (n = 10) per autophagy-targeted protein (Nsp15, M, ORF7a, ORF3a, E) per SARS2 VOC were collected from the National Center for Biological Information (NCBI) website and used to generate protein models through the Phyre2 web server [19]. The following representative amino acid sequences were utilized for protein modeling with Genebank accession number and SARS2 variant classification indicated: NSP15 (YP_009724391, original; QST05185, alpha; QXN65578, beta; UCK74385, gamma; QYC29745, delta; UPX34687, omicron BA.1; UIZ70266, omicron BA.2; UNZ13405, omicron BA.3; UPL64192, omicron BA.4; UPX84051, omicron BA.5), M (YP_009724391, original; QQP64796, alpha; QUH61239, beta; QVE57575, gamma; QYV85798, delta; UPX83240, omicron BA.1; UNR40665, omicron BA.2; URF60233, omicron BA.3.1; UPL64197, omicron BA.4; UOZ63340, omicron BA.5), ORF7a (YP_009724391, original; UFA17860, alpha; UPI39091, beta; QVK69151, gamma; QZF95006, delta; UPX83242, omicron BA.1; UPP15924, omicron BA.2; URF60235, omicron BA.3; UPU86055, omicron BA.4; UPX24774, omicron BA.5), ORF3a (YP_009724391, original; QYV40058, alpha; QTA94395, beta; QQX12070, gamma; QTW58947, delta; UHB39408, omicron BA.1; UJT21513, omicron BA.2; UNH00891, omicron BA.3; UPP22435, omicron BA.4; UOZ50987, omicron BA.5), and E (YP_009724391, original; QTN70998, alpha; QWA53293, beta; QUG11508, gamma; QVU83934, delta; UHO08789, omicron BA.1; ULG28796, omicron BA.2; URF60232, omicron BA.3; UPU03031, omicron BA.4; UOZ63339, omicron BA.5). Protein models were visualized using the Jmol applet [20].

For protein model quality assessment, available crystal structures of autophagy-targeted SARS2 proteins were superimposed with generated protein models. Protein crystal structures used were the following: Nsp15 (PDB ID: 6VWW), M (PDB ID: 8CTK), ORF7a (PDB ID: 6W37), ORF3a (PDB ID: 6XDC), and E (PDB ID: 5X29). Model:crystal superimposition was done using TM-align [21] and Root Mean Square Deviation (RMSD) values were used to establish either structural similarity or differences. For this study, we considered RMSD <1.00 to insinuate structural similarity between the generated protein model and the corresponding protein crystal, whereas, RMSD >1.00 would imply structural difference between the generated protein model and the corresponding protein crystal. Similarly, coarse grain-molecular dynamics (CG-MD) simulation using the MDWeb server [22] was performed utilizing the radius of gyration (R_gyr) of the generated SARS2 protein models in order to establish model stability. CG-MD simulation conditions were set at: 1000 ps simulation time with Δt at 0.01 ps and output frequency collected at 10 ps. All protein models that were observed to have minimal R_gyr are considered stable and structurally reliable for further analyses.

2.3. Structural comparison and model pattern recognition

Structural comparison of the 5 autophagy-targeted SARS2 proteins was made among the different SARS2 variants and subvariants using TM-align [21]. Similarly, we considered RMSD <1.00 to insinuate structural similarity among the SARS2 variants/subvariants, whereas, RMSD >1.00 would imply structural difference among the SARS2 variants/subvariants. Protein models that shared the same structural pattern (RMSD = 0) were identified and grouped together. Subsequently, identified structural patterns were likewise compared between other structural patterns.

3. Results

3.1. Autophagy network design and centrality measurements coincide with known drug modulators and putative SARS2 activity

Before any analysis can be made on the autophagy biochemical network, it is imperative that the network design is accurate. In order to establish an accurate representation of the autophagy biosynthetic pathway, both the network design and centrality measurements of the autophagy biochemical network were likewise correlated with previously proven coronavirus-specific drug modulators. As seen in Fig. 1 , the designed autophagy network was based on the mTORC1 complex as the entry point of autophagy [23]. Moreover, the following biochemical components/complexes were considered significant based on various centrality measurements: (1) betweenness (Fig. 2 A) and stress (Fig. 2B) centralities: BP1 (branch point 1) found downstream the mTORC1 complex, ULK complex, PI3K complex, C9orf72-SMCR8 complex, ATG12-ATG5-ATG16 conjugate, and biochemical components involved in autolysosome formation; (2) closeness centrality (Fig. 2C): mTORC1 complex, ULK complex, PI3K complex plus biochemical components influencing Becklin1, C9orf72-SMCR8 complex, and the ATG12-ATG5-ATG16 conjugate; (3) eccentricity (Fig. 2D) centrality: mTORC1 complex, PI3K complex and biochemical components influencing Becklin1, biochemical components influenced by the C9orf72-SMCR8 complex, STX17-SNAP29-VAMP8 complex, and biochemical components influencing the degradation of the inner vesicle; (4) radiality (Fig. 2E) centrality: all biochemical components and complexes prior to lysosome fusion to form the autolysosome; and (5) edge betweenness (Fig. 2F) centrality: BP1 transitioning to the ULK complex, ULK complex transitioning to the C9orf72-SMCR8 complex, ULK complex transitioning to the ATG12-ATG5-ATG16 conjugate, ULK complex transitioning to the PI3K complex, ATG12-ATG5-ATG16 conjugate transitioning to autolysosome formation, and STX17-SNAP29-VAMP8 complex transitioning to autolysosome formation.

Fig. 1.

Network design of the overall autophagy biosynthetic network. Solid arrows represent transition between the different biochemical components. Solid lines represent association among biochemical components. Branch points (BP) represent biochemical transitions or associations that are related to more than 2 biochemical components.

Fig. 2.

Centrality analyses of the overall autophagy biosynthetic network. Significant biochemical components based on (A) betweenness, (B) stress, (C) closeness, (D) eccentricity, (E) radiality, and (F) edge betweenness centralities are indicated. Significant nodes are colored green. Significant edges are marked by solid red arrow lines. Threshold for each centrality measurement is shown on the upper left. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Considering the unified network (Fig. 3 ), two biochemical networks within the autophagy network were putatively found to be holistically significant when the whole autophagy biochemical network was considered: (1) BP1 found after the mTORC1 complex transitioning to the ULK complex which subsequently transitions to both the C9orf72-SMCR8 and PI3K complexes, and (2) phosphatidylinositol 3-phosphate (PI3P) transitioning to WIP1 which similarly transitions to the ATG12-ATG5-ATG16 conjugate and ATG3. Previous works have shown that both nitazoxanide and repamycin/sirolimus reduced SARS2 infection by inhibiting mTORC1 thereby activating autophagy, whereas, both chloroquine and hydroxychloroquine block endocytosis-mediated cell entry (except in lung cells) of SARS2 thereby impeding autophagy function [[24], [25], [26], [27]]. This is consistent with our results (Fig. 3) showing that the activity of previous autophagy modulators used to treat SARS2 infection coincide with the significant nodes and edges highlighted in the unified network analysis of the autophagy biochemical network. Earlier publications have emphasized the dual nature (beneficial and detrimental) of autophagy, whereby, autophagy can serve as a host defense against viral infection through xenophagy (beneficial) and, likewise, autophagy can be manipulated by the viral pathogen to produce replication organelles (detrimental) [[28], [29], [30]]. In the case of SARS2, it was reported that SARS2 disrupts the ULK complex formation thereby functionally impairing autophagy (particularly the degradative capacity possibly found in xenophagy) [31] and, similarly, SARS2 was also shown to utilize the autophagy biochemical network to promote viral replication via PI3P [30]. These points are likewise consistent with our results (Fig. 3). Taken together, we postulate the following: (1) one of the putative significant biochemical network (via the ULK complex) within the autophagy network that is responsible for the beneficial nature of autophagy is possibly the same biochemical network stimulated by certain autophagy modulators (nitazoxanide and repamycin/sirolimus) and targeted by SARS2 to impair autophagy action involved in host defense; and (2) the other potential significant biochemical network (via PI3P) within the autophagy network that is manipulated by certain viruses for viral replication is the same biochemical network used by SARS2 for organelle formation.

Fig. 3.

Unified network highlighting common nodes and edges established from centrality measurements. Significant nodes are colored green. Significant edges are marked by solid red arrow lines. Branch points (BP) represent biochemical transitions or associations that are related to more than 2 biochemical components. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.2. SARS2 M protein is putatively the main protein determinant that would influence autophagy outcome during a SARS2 infection

Prior to further downstream protein structural analyses, it was previously suggested that protein structures regardless of being obtained theoretically (i.e. computer-based) or experimentally (i.e. crystallized) should undergo a quality assessment [32]. In this regard, to elucidate whether the generated protein models are ideal for further downstream protein structural analyses, protein structural superimpositions of both generated protein models and known crystal structures of autophagy-targeted proteins (Nsp15, M, ORF7a, ORF3a, and E) were performed. For this study, we considered superimpositions with RMSD < 1.00 to be ideal for further downstream protein structural analyses. We first generated the Nsp15, M, ORF7a, ORF3a, and E protein models (Suppl. Fig. 1) and, afterwards, found that all model:crystal superimpositions of autophagy-targeted proteins have RMSD < 1.00 (Suppl. Fig. 2) and, based on CG-MD simulation (Suppl. Fig. 3), protein models showed minimal R_gyr. These results would mean that the generated protein models are potentially suitable for further structural analyses.

SARS2 genome regularly undergoes mutations [33] which may suggest that viral proteins (in this case autophagy-targeted proteins) may have differed within the different SARS2 variants and subvariants. To determine whether any structural variations occurred among the autophagy-targeted proteins found among the original SARS2 variant and the different VOC (both variants and subvariants), protein superimposition and RMSD score comparison were performed. Interestingly, we observed that Nsp15 (Suppl. Fig. 4), M (Suppl. Fig. 5), ORF7a (Suppl. Fig. 6), ORF3a (Suppl. Fig. 7), and E (Suppl. Fig. 8) proteins have no structural difference (RMSD = 0) among the original SARS2 variant and the different VOC (both variants and subvariants). This may insinuate that throughout SARS2 evolution, these particular autophagy-targeted proteins were conserved which in-turn could mean that the role of these proteins in autophagy manipulation remained constant. Interestingly, among the autophagy-targeted SARS2 proteins, only the M protein coincides with the 2 significant biochemical pathways we previously identified (Fig. 3) [7]. SARS2 M protein is a glycosylated structural protein that plays a significant role in virion assembly and, likewise, inhibits TBK1-related innate antiviral immune response [34]. In this regard, we postulate that the SARS2 M protein is the main protein determinant influencing autophagy outcome during a SARS2 infection. Admittedly, additional experimental work is needed to further prove this suspicion.

4. Discussion

Autophagy is involved in multiple physiological processes (i.e. cell survival, cell metabolism, and host defense) while also being associated with several diseases (i.e. cancer and metabolic diseases) [35,36]. Currently, there are three autophagy types, namely: microautophagy wherein lysosonal membrane protrusions are utilized for cargo capture; chaperone-mediated autophagy wherein membrane structure are not utilized to secure cargo; and macroautophagy wherein cargo is sequestered away from lysosome [37]. In addition, SARS2 has many viral proteins that are capable of manipulating autophagy (particularly macroautophagy) at different stages [7]. Throughout this study, we attempted to identify significant biochemical pathways within the autophagy biochemical network and, likewise, correlate these significant biochemical pathways to the different autophagy-targeted SARS2 proteins.

Network data is a source of information found within complicated patterns such as biochemical pathway networks with network analytics being utilized to provide a holistic analyses which in-turn allows for the integration of complementary data thereby giving additional new insights [18,38]. One common approach in performing network analytics is centrality analysis which involves ranking and identifying network components into significant elements based on multiple centrality measurements [18,39]. Subsequently, in order to avoid major limitations of centrality analysis such as: centrality measures are unable to capture the information flow of the overall network, rankings of each centrality measure differ across all measures, and centrality measures only capture the localized information and do not take into consideration the overall global network [40,41], all centrality measurements are considered. In this regard and based on our network analytics results, we believe that the autophagy network design and analyses were accurate since the results were consistent with the dual nature of autophagy and known SARS2 drug modulators [[24], [25], [26], [27], [28], [29], [30]]. This would mean that the designed autophagy network can potentially be utilized for further downstream analyses. Subsequently, considering the multiple autophagy-targeted SARS2 proteins that affect different stages of autophagy, our results suggest that all autophagy-targeted SARS2 proteins (Nsp15, M, ORF7a, ORF3a, and E) may have no altered function since all 5 protein structures were consistent in all SARS2 VOC. This would putatively mean that the mechanism of autophagy manipulation associated to Nsp15, M, ORF7a, ORF3a, and E are conserved among SARS2 VOC which in-turn would emphasize that viral manipulation of the autophagy biochemical network is independent from structural variations associated with autophagy-targeted SARS2 proteins. Similarly, considering our autophagy biochemical network and analyses, the different stages of autophagy targeted by the Nsp15, ORF7a, ORF3a, and E proteins [7] were putatively not found to be significant, whereas, the 2 significant autophagy biochemical pathways we identified are putatively correlated with the M protein. This would insinuate that among the autophagy-targeted SARS2 proteins, the M protein potentially plays a vital role in influencing SARS2 pathogenicity and immunosuppression [34].

In general, viruses evolved multiple strategies to counteract selective autophagy like xenophagy, mitophagy, aggrephagy, lipophagy, ferritinophagy, and ER-phagy [42]. More specifically, viruses have developed counteraction strategies that may include resisting, escaping, subverting, and hijacking autophagy to enhance viral replication [43]. Considering the 2 significant autophagy biochemical pathways we identified are putatively correlated with the M protein, we hypothesize that a possible counteraction strategy associated with SARS2 pathogenicity and immunosuppression [34] involves the SARS2 M protein subverting the ULK complex formation to functionally impair autophagy response [31] and, likewise, the SARS2 M protein hijacking the PI3P complex to promote viral replication [30]. Our hypothesis is consistent with previous autophagy modulators used to treat COVID-19 [[24], [25], [26], [27],44].

In summary, we determined 2 significant biochemical pathways (one starts from the ULK complex and the other starts from the PI3P complex) within the autophagy biochemical network. Additionally, we established that all autophagy-targeted SARS2 proteins (Nsp15, M, ORF7a, ORF3a, and E proteins) are structurally conserved, thus, SARS2-linked autophagy manipulation is not associated with structural variation among the autophagy-targeted SARS2 proteins. Moreover, we found that the M protein is correlated to the 2 significant biochemical pathways we identified within the autophagy biochemical network which potentially emphasizes the vital role of the SARS2 M protein in autophagy manipulation.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.