Tumor necrosis factor inhibitors

Full Text

ABBREVIATIONS

TNF – tumor necrosis factor; RA – rheumatoid arthritis; TNFR – tumor necrosis factor receptor; ER – endoplasmic reticulum; PsA – psoriatic arthritis; DC – dendritic cell; p55 – TNF receptor type I; p57 – TNF receptor type II, CRD – cysteine-rich domain; PLAD – pre-ligand binding assembly domain; IL – interleukin; GM–CSF – granulocyte-macrophage colony-stimulating factor; Th1 – T-helper cell type 1; LPS – lipopolysaccharide; CIA – collagen-induced arthritis; CPXV – cowpox virus, MPXV – monkeypox virus; VARV – variola virus.

INTRODUCTION

Tumor necrosis factor (TNF) exerts pleiotropic effects on cells and plays a central role in the regulation of host defense responses. It recruits leukocytes to inflammation sites by upregulating adhesion molecule expression, stimulates macrophages, increases vascular permeability, and activates antigen-presenting cells [1, 2]. However, TNF overproduction contributes to the development of pathological conditions associated with chronic inflammation and/or autoimmune reactions, such as rheumatoid arthritis (RA), Crohn’s disease, septic shock, and cachexia [3].

Among TNF-induced diseases, RA has a major social impact, affecting as it does approximately 0.5–2% of the adult population, particularly individuals of working age (35–55 years) [4]. The condition is characterized by both synovial and systemic inflammation; if not adequately treated, it leads to disability, degraded quality of life, work impairment, and a substantial economic burden [5]. Until the beginning of the 21st century, a third of RA patients had had to stop working within two years of the disease onset due to severe degradation [6]. Moreover, the life expectancy of patients with severe RA drops by an average of 10 years [7]. In this context, the development of effective therapeutic strategies becomes a priority.

RA management has changed substantially over the past 20–25 years: it has shifted from symptomatic treatment to attempts to arrest disease progression. This progress has been largely driven by the introduction of disease-modifying drugs, which have significantly improved outcomes for patients. TNF inhibitors represent a major class of such agents; they account for 75% of the total drug market. The most often used TNF inhibitors are the following recombinant proteins: Remicade (infliximab; Centocor Ortho Biotech/Schering-Plough), Enbrel (etanercept; Amgen/Wyeth), and Humira (adalimumab; Abbott). These antirheumatic agents have a number of drawbacks, including adverse effects that limit drug tolerability, the requirement of long-term treatment, and an increased risk of tumor development. In addition, patients often lose responsiveness to the anti-TNF therapy due to the development of an immune response, which usually requires drug replacement [3].

Therefore, there is a need for novel anti-RA agents that maintain efficacy while improving safety and cost-effectiveness. Viral TNF-binding proteins represent a promising research area. Poxviruses, the largest DNA viruses infecting mammals, encode numerous proteins that target key components of the host immune system. Orthopoxviruses, which include the variola (smallpox), monkeypox, and cowpox viruses, are of particular interest due to their ability to produce immunomodulatory proteins that bind various ligands, such as TNF, chemokines, interferons, and complement system components. As a result of virus co-evolution with humans, viral proteins can interact with the host target proteins with high affinity and specificity. This makes viral proteins attractive as candidates for the development of next-generation agents. A detailed study of a broad range of poxvirus protein homologues is required to fully illuminate their therapeutic potential. Immunomodulatory proteins of the variola virus, which are evolutionarily the most adapted to the human immune system, seem to be the most promising candidates in that regard [8].

TUMOR NECROSIS FACTOR AND ITS ROLE IN DISEASE

TNF is a pro-inflammatory pleiotropic cytokine involved in the complex regulation of inflammatory and immune processes. TNF was first described in the middle 1970s [9].

TNF-producing cells are activated in response to external stimuli (e.g., inflammation), resulting in the synthesis of a 26-kDa polypeptide. This polypeptide is expressed on the cell surface as the transmembrane form of TNF (tmTNF). The soluble form of TNF (sTNF) is generated by proteolytic cleavage between Ala76 and Val77 in TNF with involvement of the metalloproteinase TNF-α–converting enzyme (TACE). This process yields a 17-kDa sTNF and a residual cytosolic domain [10].

TNF exerts its biological function in both transmembrane and soluble forms. Transmembrane TNF initiates a cascade of biochemical reactions through direct contact between a TNF-producing cell and a receptor-exposing cell, whereas sTNF acts at sites distant from TNF-expressing cells [11].

Trimerization of both forms is required for the deployment of TNF biological activity and induction of intracellular signaling cascades. Following trimerization, TNF binds to a corresponding receptor, either TNFR1 (p55/CD120a; 55 kDa) or TNFR2 (p75/CD120b; 75kDa) [12–15].

The p55 and p75 belong to the TNF receptor (TNFR) superfamily, which comprises at least 29 structurally related type I transmembrane proteins with an extracellular N-terminal and an intracellular C-terminal domains; each protein has its distinct structural features [16]. Among these characteristics, cysteine-rich domains (CRDs) are critical for TNFR activity. Each CRD presents a pseudo-repeat of approximately 40 amino acid residues containing six Cys residues that form three disulfide bonds within a single polypeptide chain. The number of CRDs among TNFR family members ranges from one to six. The pseudo-repeats are involved in ligand binding [17]. However, not all CRDs participate directly in ligand binding, since TNF interacts primarily with the second and third CRDs from the N-terminus of p55 [18], whereas the first CRD is involved in receptor oligomerization. This domain is termed the pre-ligand assembly domain (PLAD), since it mediates receptor oligomerization at the cell surface prior to ligand binding (Fig. 1). PLAD binding is highly specific, which allows for selective receptor recognition within the TNFR superfamily [19–21].

Fig. 1. Schematic representation of TNF receptor–ligand interactions. In the absence of a ligand, TNF receptors exist as dimers; upon ligand binding, they form an oligomeric network consisting of trimeric receptor–ligand complexes. The complex formation initiates intracellular signaling [15]. CRDs – Cys-rich domains

Signal transduction following ligand binding depends on the type of cytosolic domain present in the receptor. The cytosolic domains of TNFRs are relatively short and function as docking sites for signaling molecules [18]. Receptors can be divided into three groups based on the cytosolic domain structure: death domain-containing receptors (p55), receptors lacking a death domain (p75), and soluble receptors (both p55 and p75 can exist in membrane-bound and soluble forms) [21].

Current evidence indicates the existence of several major signaling pathways. One of the classifications is based on the type of receptor the ligand interacts with: there are p55- and p57-mediated pathways (Fig. 2) [22]. It has been previously considered that cytotoxic effects are primarily mediated by p55, while proliferative responses are induced by p75 [23]. However, both p55 and p57 are now known to mediate a broad range of biological functions.

Fig. 2. Intracellular signaling pathways activated by TNF binding to p55 and p75. TNF binding to p55 induces cell death and inflammatory signaling through activation of the canonical NF-κB pathway. TNF binding to p75 activates both canonical and non-canonical NF-κB pathways. Solid green and blue lines represent activating signals; and dashed lines indicate inhibitory interactions. sTNF – soluble TNF; tmTNF – transmembrane TNF; sp55 – soluble p55; tmp55 – transmembrane p55; sp75 – soluble p75; tmp75 – transmembrane p75 [22]

TNF binding to p55 initiates intracellular signaling through recruitment of adaptor molecules to the death domain of the cytosolic region of the receptor. As a result, the membrane-associated complex I is formed, which includes TRADD (p55-associated death domain protein), RIPK1 (receptor-interacting protein kinase 1), and TRAF2 (TNFR-associated factor 2) [24]. TRAF2 can be recruited to complex I in association with cIAP1 and cIAP2 (E3 ubiquitin ligases) [25]. These ligases further modify various components of the p55 signaling cascade, including RIPK1, via Lys63-linked ubiquitin chains, thus creating binding sites for the linear ubiquitin chain assembly complex (LUBAC) [26]. LUBAC further modifies RIPK1 with linear ubiquitin chains, facilitating recruitment of the transforming growth factor β-activated kinase 1 (TAK1) via TAB2 (TAK1-associated binding protein 2) and IKK (inhibitor of nuclear factor kappa B (κB) kinase complex) [25]. This cascade activates the canonical NF-κB pathway, thus initiating the transcription of various NF-κB-regulated targets (cytokine and anti-apoptotic genes). Thus, membrane-associated complex I promotes cell survival mechanisms. If dissociated from the membrane, complex I can form cytosolic complex II through association with FADD (Fas-associated death domain protein) and caspase-8. Depending on caspase-8 activity, the resulting complex can induce either apoptosis or necroptosis [24, 27].

Similar molecular cascades can be triggered in cells upon TNF binding to p75. However, a central role in p57-mediated signaling is attributed to the c-Jun N-terminal kinase, which, once activated via TRAF2 and ASK1 (apoptosis signal-regulating kinase 1), mediates NF-κB activation [28]. In addition, the p57 signaling pathway is closely linked to endoplasmic reticulum (ER) stress. The ER is sensitive to disruptions in cellular homeostasis that lead to the accumulation of misfolded proteins, thereby inducing ER stress and contributing to apoptotic cell death. Under ER stress conditions, TRAF2 associates with ER stress sensors and interacts with pro-caspase-12, promoting activation of the latter. It has also been hypothesized that ER stress induces TNF production. This suppresses the c-Jun transcription factor activation and affects cellular susceptibility to apoptosis. These findings confirm the role of the ER as an additional regulatory node in p57-mediated apoptosis [22].

Thus, the data above indicate that p55- and p75-mediated signaling pathways are closely interconnected. TNF and its receptors form a complex signaling network, as indicated by additive, synergistic, and even antagonistic interactions between the two receptor types [29].

As a key homeostasis regulator, TNF plays an important role in conditions of disrupted homeostasis by exerting its pathogenic function. Under normal conditions, TNF contributes to bone tissue regeneration by maintaining the balance between osteoblasts and osteoclasts; under pathological conditions, it promotes osteoclastogenesis [2]. TNF suppresses tumor formation under normal conditions and contributes to immune evasion by tumor cells in pathology [1]. Despite numerous examined examples, the mechanisms underlying the transition to pathological TNF signaling remain unclear. The majority of studies of TNF describe the existing pathologies and only make attempts to establish the underlying causes.

Considering the ubiquitous nature of TNF effects, the spectrum of associated diseases is broad. Major TNF-associated pathologies include rheumatoid arthritis (RA), Crohn’s disease, septic shock, psoriasis, sarcoidosis, cachexia, Sjogren’s syndrome, polymyositis, vasculitis, Behçet’s disease, atherosclerosis, and multiple sclerosis. These conditions are associated with TNF overproduction, although the etiology of the majority of them remains not fully understood [3]. In this review, we focus on three well-characterized TNF-driven diseases.

Among such pathologies, rheumatoid arthritis (RA) is the one that has most extensively been studied. RA is a chronic autoimmune disease that is primarily accompanied by synovitis, cartilage degradation, and erosive bone damage [30, 31]. RA has a significant social impact, because it is detected in approximately 0.5–2% of the adult population of working age (35–55 years) [4, 32]. The condition is characterized by both synovial and systemic inflammation and leads to disability, deteriorated quality of life, work impairment, and substantial economic burden, if not adequately treated [5]. Until the beginning of the 21st century, one third of RA patients had to stop working within two years of disease onset [6]. Moreover, the life expectancy of patients with severe RA is lower by an average of 10 years [7]. The RA pathogenesis involves a set of complex interactions of genetic, epigenetic, environmental, metabolic, immune, and microbial factors [33, 34]. TNF is a key pro-inflammatory cytokine involved in RA pathogenesis.

In early stages of disease development, initiation of the inflammatory process is accompanied by accumulation of T helper 1 (Th1) cells, macrophages, B cells, plasma cells, and dendritic cells (DCs). Macrophages and T cells represent major TNF sources at inflammation sites [35]. Together with interleukin-1 (IL-1), TNF promotes fibroblast proliferation and induces the synthesis of the pro-inflammatory mediators IL-6 and IL-8, the granulocyte–macrophage colony-stimulating factor (GM–CSF), as well as adhesion molecules and proteases (e.g., collagenases). Proteolytic enzymes degrade collagen and proteoglycans, resulting in cartilage and bone destruction, with further joint erosion. GM–CSF interacts with TNF through a positive feedback pathway and enhances the expression of human leukocyte antigen–DR isotype on antigen-presenting cells, thereby promoting T cell activation. TNF stimulates the proliferation of synovial and circulating T and B cells and upregulates adhesion molecule expression on endothelial cells, contributing to T cell adhesion and tissue migration. TNF is also associated with osteoporosis in RA; it promotes osteoclast differentiation and stimulates estrogen depletion. Altogether, these processes lead to bone tissue loss. Thus, TNF exerts pleiotropic effects on all the key cell types involved in RA pathogenesis: fibroblasts, endothelial cells, T cells, etc., which proves its central role in disease development. The crucial role of TNF is also confirmed by the clinically effective therapeutic inhibition of both tmTNF and sTNF in RA patients [36, 37].

Psoriatic arthritis (PsA) shares similar pathogenetic mechanisms. PsA typically develops in individuals with cutaneous psoriasis and commonly manifests as swelling of the fingers and toes [38]. The condition is characterized by activation of immune cells in inflamed tissues. Activated DCs and macrophages overexpress TNF and interleukin-23 (IL-23), which promote T cell priming. In PsA, activated T cells further differentiate into effector immune cells, particularly Th17 cells, which overproduce interleukin-17 (IL-17). Together with macrophages, IL-17 plays a key role in joint inflammation and destruction. Macrophages trigger molecular processes similar to those observed in RA [35, 39, 40]. IL-17 and TNF also activate keratinocytes, promoting epidermal hyperplasia and recruitment of inflammatory cells, including DCs. TNF induces keratinocyte proliferation and inhibits cell apoptosis via the NF-κB signaling pathway, contributing to microabscess formation due to the recruitment of inflammatory cells [35]. Thus, PsA is characterized by changes in the ratio of pro- and anti-inflammatory immune cells and cytokines, thus leading to inflammation and tissue damage. Numerous studies confirmed that TNF, IL-23, and IL-17 play a key role in PsA pathogenesis, since their therapeutic inhibitors exert significant therapeutic effects [38, 41].

Sepsis and septic shock represent another socially significant TNF-driven pathology and are responsible for more than 1,000,000 deaths worldwide annually. The mortality rate in septic shock is approximately one out of four cases [42]. Sepsis is a severe clinical syndrome associated with a dysregulated host response to infection. Its severity is driven by the activation of intracellular molecular cascades that lead to amplified cytokine production: the so-called “cytokine storm”, characterized by overexpression of TNF, IL-1, IL-6, IL-8, IL-12, and others. In this condition, pathogenic microorganisms are recognized by the innate immune system, which includes leykocytes, the complement system, cytokines, chemokines, and the antimicrobial peptides secreted by innate immune cells [43].

The innate immune response is activated through the recognition of pathogen-associated molecular patterns (PAMP), including highly conserved antigens, such as lipopolysaccharides and peptidoglycans, by pattern recognition receptors (PRR). PRR recognize molecules released from injured tissues (damage-associated molecular patters, DAMP). Four types of PRR are identified in vertebrates: Toll-like receptors (TLR), Nod-like receptors (NLR), RIG-like receptors (RLR), and C-type lectin receptors (CLR). Among these, TLR signaling pathways are the ones that have been most extensively studied. Activation of these signaling cascades induces the NF-κB-mediated expression of the pro-inflammatory cytokines TNF and IL-1β, interferon regulatory factors 3 (IRF3) and 7 (IRF7), as well as adaptor protein 1 (AP-1) (Fig. 3) [43, 44]. Excessive cytokine production results in generalized endothelial activation, increased expression of adhesion molecules, activation of coagulation pathways, and further production of pro-inflammatory cytokines, together ultimately leading to septic shock [45]. Current treatment strategies in sepsis involve immunomodulatory agents and antibiotics. The potential use of cytokine inhibitors such as TNF-targeting antibodies are being actively explored for disease treatment [46].

Fig. 3. Schematic representation of the activation of various TLR-mediated signaling pathways leading to cytokine storm and sepsis. Activation of these signaling cascades induces the synthesis of pro-inflammatory cytokines (IL-1, IL-6, IL-12, IL-8, and TNF), anti-inflammatory cytokines (IL-10), and type 1 interferons. Elevated levels of these inflammatory mediators lead to cytokine storm in sepsis [44]

These three diseases present overlapping molecular pathways and involve TNF overproduction. Hence, similar therapeutic approaches can be used for their treatment. Novel TNF inhibitors can be utilized to treat not only RA, PsA, and septic shock, but also other TNF-driven diseases (Table 1).

Table 1. Pathogenic role of TNF in a series of autoimmune diseases

DRUGS INHIBITING THE BIOLOGICAL FUNCTION OF TNF

The rapid development of molecular biology techniques, together with the establishment and investigation of experimental models of immune-mediated inflammatory diseases in the second half of the 20th century, enabled the establishment of the crucial role of TNF in the pathogenesis of a series of conditions, including RA, Crohn’s disease, ulcerative colitis, and septic shock.

Some of the earliest evidence for the TNF role in inflammatory arthritis was obtained in transgenic mice overexpressing human TNF (hTNF). These mice spontaneously develop chronic inflammatory polyarthritis, which was stopped by administration of hTNF-specific monoclonal antibodies [57]. It is now well established that TNF is a key effector molecule in the inflammatory cascades underlying many autoimmune diseases [58]. Advances in genetic engineering have enabled the generation of mouse models for studying the pathogenesis of such diseases as collagen-induced arthritis, autoimmune encephalomyelitis, colitis, bacterial infections, etc. The use of these models confirmed the role of TNF in disease development [59].

The results of this work served as the basis for numerous studies into the development of TNF inhibitors for the treatment of immune-mediated diseases such as RA and Crohn’s disease.

The primary objective of an anti-TNF therapy is to inhibit TNF-mediated signaling pathways and, thereby, suppress the inflammatory response. In this regard, therapeutic strategies targeting various molecules involved in TNF synthesis and processing have been developed. Potential pharmacological targets include intracellular signaling molecules such as kinases, transcription factors, and molecules involved in mRNA splicing, translation, and protein maturation [60]. However, many of these targets are pleiotropic, which limits the design of selective inhibitors [61]. Therefore, the most effective strategy seems to be to suppress the interaction between TNF and its receptors, thereby rendering the downstream signal transduction impossible [60]. To date, the strategy based on the use of TNF inhibitors remains the predominant approach.

In the late 1990s, TNF inhibitors emerged as a revolutionary therapy against immune-mediated inflammatory diseases affecting joints, the gastrointestinal tract, and skin. Numerous clinical trials demonstrated the efficacy and safety of these drugs in large cohorts of patients. In 1998, the Food and Drug Administration (FDA, USA) approved the first TNF-targeted drugs based on TNFR and a monoclonal anti-TNF antibody for the treatment of RA and Crohn’s disease, respectively [62].

There are currently five major anti-TNF drugs in clinical use: the monoclonal anti-TNF antibodies infliximab, adalimumab, golimumab, and certolizumab pegol; and a TNFR-based drug, etanercept (Fig. 4) [63, 64]. Other anti-TNF agents present analogues of the above-listed drugs. The analogues have been developed primarily to improve therapeutic accessibility and reduce production costs [65, 66].

Fig. 4. Simplified representation of the molecular structures of TNF antagonists. Infliximab is a chimeric mouse/human IgG1 monoclonal anti-TNF antibody. Adalimumab and golimumab are fully human IgG1 monoclonal anti-TNF antibodies. Etanercept is a fusion protein consisting of two extracellular domains of human p75 and the Fc region of human IgG1. Certolizumab pegol is a PEGylated Fab fragment of humanized IgG1 monoclonal anti-TNF antibody [64]

Etanercept (Enbrel) was the first anti-TNF drug approved for RA treatment in 1998. It consists of a recombinant protein consisting of two extracellular domains of human p75 linked to the Fc fragment of human IgG1 (Fig. 4). The recombinant protein is produced in Chinese hamster ovary (CHO) cells [3]. The presence of the Fc region in the drug prolongs the circulating half-life of the molecule, contributing to its long-term therapeutic effect [67]. Etanercept has a broad range of clinical applications: RA, PsA, and non-radiographic axial spondyloarthritis (Table 2) [65, 68].

Table 2. Indications to the use of TNF inhibitors and dates of the first clinical approval (by the FDA and the European Medicines Agency (EMA)) [68]

Infliximab was approved by the FDA in 1998 for the treatment of Crohn’s disease. The drug represents a chimeric monoclonal antibody and consists of a human IgG1 light chain constant region and a variable region of a murine anti-TNF antibody (Fig. 4). The antibody is obtained using hybridoma technology and produced in recombinant cell systems cultured by continuous perfusion [69]. Although infliximab is highly specific and exhibits minimal adverse effects on nontarget biological pathways, the presence of a murine antibody fragment in its structure may trigger an immune response [70]. The drug is used for the treatment of Crohn’s disease, RA, PsA, ulcerative colitis, and other pathologies (Table 2) [71].

Adalimumab is a human IgG1-based monoclonal anti-TNF antibody produced in CHO cells (Fig. 4). It was approved by the FDA in 2002 for the treatment of RA and became one of the top ten most widely used monoclonal antibodies in 2018 and, subsequently, the most profitable drug worldwide [72]. Later, adalimumab was also approved to treat other immune-mediated inflammatory diseases such as Crohn’s disease, plaque psoriasis, and ulcerative colitis (Table 2).

Golimumab is another human IgG1-based monoclonal anti-TNF antibody. Unlike adalimumab, which is obtained using phage display technology, golimumab was produced using transgenic mice expressing human immunoglobulin genes. These mice were immunized with human TNF; clones producing high-affinity anti-hTNF antibodies were generated using the hybridoma technology [73, 74]. Golimumab has now been approved for the treatment of RA, PsA, and other inflammatory conditions (Table 2) [75].

Certolizumab pegol is a monovalent Fab fragment of a humanized anti-TNF IgG antibody lacking the Fc region [76]. It is the absence of the Fc region that distinguishes the drug from other TNF inhibitors. This structural feature would normally result in a more rapid drug clearance, since the interaction between the Fc region and the neonatal Fc receptor in the endosome is important in the regulation of antibody homeostasis by protecting IgG from degradation, thereby prolonging its plasma half-life. However, the plasma half-life of certolizumab pegol is already extended through the conjugation of the hinge region of the Fab fragment to two interlinked 20-kDa polyethylene glycol (PEG) moieties. Such a modification increases the drug half-life and solubility, while reducing its immunogenicity and sensitivity to proteolytic degradation [77]. The drug was first approved in 2008 for Crohn’s disease and later for such conditions as RA, PsA, and others (Table 2).

Although all TNF inhibitors neutralize the same molecular target, clinical data (summarized in Table 2) indicate that there are differences in their efficacy across diseases. For example, unlike monoclonal antibody-based TNF inhibitors (adalimumab and infliximab), etanercept has not shown any clinically significant efficacy in the treatment of the inflammatory bowel disease or uveitis [78]. Moreover, a significant variability in the individual response to TNF inhibitors has been observed. While some patients respond rapidly to the drug’s administration, others exhibit either a delayed or no response. In addition, some individuals may experience a loss of response to anti-TNF therapy over time [64].

The described differences in therapy effectiveness may be an indication of the existence of individual polimorphisms of immune-associated genes and highlight the need for further investigation into the molecular basis of disease pathogenesis to develop personalized therapeutic approaches. Disease-specific variations in the clinical efficacy of TNF inhibitors have also prompted discussion regarding potential differences in the mechanisms of action [79].

Studies using surface plasmon resonance have shown that golimumab binds sTNF with affinity comparable to that of etanercept and higher than that of infliximab and adalimumab [74]. Furthermore, the binding avidity of etanercept to sTNF is 10–20-fold higher than that of infliximab or adalimumab [80].

In addition, infliximab, adalimumab, etanercept, golimumab, and certolizumab pegol bind tmTNF expressed on the cell surface with comparable affinity [81, 82], although their affinity is lower than that of sTNF [80]. Some studies have reported that etanercept either binds tmTNF with lower affinity or does not bind it at all, compared to anti-TNF drugs based on monoclonal antibodies [83, 84]. These differences may be primarily due to the use of different cell lines with variable levels of tmTNF expression levels. Yet, another literature source indicates that etanercept forms relatively unstable complexes with sTNF, resulting in faster TNF dissociation, while infliximab binds both tmTNF and sTNF to form more stable high-molecular-weight complexes [85]. Differences in the binding stoichiometry have also been reported. Infliximab can bind both monomeric and trimeric TNF forms, resulting in stable high-molecular-weight complexes. Each infliximab molecule binds to two TNF molecules, whereas etanercept binds TNF homotrimers at a 1 : 1 ratio [86]. Taken together with the fact that infliximab forms more stable complexes with tmTNF, this may explain the drug’s clinical efficacy in the case of the inflammatory bowel disease and the absence of that for etanercept. Recent studies suggest that neutralization of tmTNF, rather than sTNF, is critical for therapeutic efficacy in the inflammatory bowel disease [87, 88].

Thus, having considered only one of the main mechanisms of action of TNF inhibitors, namely TNF neutralization, it seems obvious that differences in kinetic, stoichiometric, and other parameters of ligand-receptor interactions contribute to the variability in the therapeutic efficacy of TNF inhibitors in different diseases.

Differences in the pharmacokinetic profiles may also contribute to the variations in the clinical efficacy of TNF inhibitors. Of all these drugs, etanercept has the shortest half-life (4–5 days), whereas intact IgG1-based drugs exhibit longer half-lives: 8–10 days for infliximab, 10–20 days for adalimumab, 7–20 days for golimumab, and 14 days for certolizumab pegol. In addition, etanercept possesses the lowest steady-state concentration: 1.1 µg/mL, compared to infliximab (118 µg/mL) and adalimumab (4.7 µg/mL). Taken together, these differences suggest that monoclonal antibody-based drugs may provide a longer therapeutic effect than etanercept [64].

The pharmacokinetic properties of TNF inhibitors underpin the frequency of drug administration, which ultimately results in increased systemic exposure and production of specific anti-drug antibodies. For example, production of anti-drug antibodies has been reported in approximately 1.2% of patients receiving etanercept and 3.8% in cases of golimumab use. Infliximab was shown to be the most immunogenic (25.3% of cases), followed by adalimumab (14.1%), and certolizumab pegol (6.9%) [89]. The immunogenicity of these agents may vary depending on the disease, dosage regimen, and therapy duration [90].

Despite the substantial success of anti-cytokine therapy in the treatment of TNF-driven diseases, this therapeutic approach allows only to affect disease symptoms, rather than eliminate the underlying cause. In addition, the therapy may cause severe adverse effects due to the suppression of the cytokine protective functions [91]. Reported adverse effects include infusion and injection-site reactions, infections (in particular, reactivation of tuberculosis), autoantibody formation and development of drug-induced lupus, hepatic dysfunction, as well as hematologic and solid malignancies [92–94]. In addition, allergic reactions may emerge due to the production of antibodies to anti-TNF drugs, thus reducing treatment efficacy [95]. As noted above, immunogenicity varies depending on disease type, dosage, and therapy duration [90].

A comprehensive meta-analysis has been conducted for each TNF inhibitor to explore associations between drug administration and adverse effects; indeed, the occurrence of such effects is well documented [96]. Novel therapeutic approaches are currently under development, including modification of existing agents to reduce the adverse effects. It has been confirmed that the main limitation to the use of TNF inhibitors is related to the pleiotropic roles of the cytokine. For example, TNF production and its involvement in the formation of granulomas, which contain the infection in its inactive form, determine the host defense profile against the tuberculosis infection. In this regard, TNF inhibitors may lead to the activation of latent tuberculosis [93, 94]. Experimental studies in various disease models have demonstrated that TNF produced by myeloid cells exerts a pathogenic effect, while TNF derived from T cells has a protective role. Findings in mouse models using tissue-specific knockout suggest that, if a potential TNF inhibitor can selectively avoid inhibition of a T cell-derived TNF, it may reduce the adverse effects of the therapy [91].

A practical implementation of this hypothesis is the development of bispecific antibodies, referred to as MYSTI. These antibodies are specific to both TNF and the surface markers of myeloid cells, which enables selective neutralization of TNF produced by myeloid cells. The use of such a drug is expected to reduce the incidence of adverse effects, since earlier experimental studies in mice have demonstrated that it is the TNF derived from myeloid cells that often contributes to disease development, while T cell-derived TNF exerts a protective function. This targeted inhibition of TNF depending on the origin cell type represents a potential advantage of MYSTI-based therapeutics [97, 98].

We can conclude that the findings of the past 20 years of research on TNF-targeted therapies indicate that suppression of TNF–receptor interactions remains the most effective therapeutic strategy. In this regard, the majority of current studies aim to modify existing drugs that inhibit the interaction between TNF and its receptors, p55 and p75.

TNF RECEPTORS OF ORTHOPOXVIRUSES

For the past 10 years, there has been accumulating evidence of the potential of using viral proteins as therapeutic agents. In this context, pathogenic orthopoxviruses (genus Orthopoxvirus, family Poxviridae) are of particular interest. These viruses are among the largest DNA viruses. Their replication cycle takes place entirely in the cytoplasm [99]. These viruses encode a wide spectrum of immunomodulatory proteins [8]. It is considered that, through co-evolution with the host, these viruses can acquire the encoding sequences of various host genes in their genome and modify them, thus adapting them to viral survival and resilience in the biosphere [99, 100].

Cowpox virus (CPXV) is of low pathogenicity in humans, while having a broad host range among animals. Monkeypox virus (MPXV) is also characterized by a broad host range; in humans, it causes an infection clinically similar to smallpox and even fatal in some cases. Variola virus (VARV), which represents a rare case of a strictly human-adapted virus, is of special interest. VARV is highly pathogenic to humans; it has evolutionarily adapted to efficiently evade the human immune defense [8, 99].

Having sequenced the complete genomes of VARV [101–105], MPXV [106, 107], and CPXV [108] and performed further genomic analyses, we managed to identify the viral genes encoding proteins of the TNFR superfamily. That was the first experimental evidence that the ST-2 protein of the Shope fibroma virus and the MT-2 protein of the myxoma virus (poxviruses of the genus Leporipoxvirus), which comprise the TNFR superfamily, bind to TNF [109].

Four genes encoding TNFR proteins were identified in CPXV, while only one such gene was found in VARV and MPXV: CrmB. CrmB has been regarded as a novel, potential TNF inhibitor.

CrmB is an early viral protein (i.e. it is expressed in the host cell prior to viral DNA replication) with a molecular mass of approximately 47 kDa. The protein is composed of two domains. The N-terminal region contains the TNF-binding domain, comprising three CRDs, each containing six Cys residues [108, 110]. A PLAD is located within the first N-terminal CRD (Fig. 5) [111]. The TNF-binding domain shares approximately 37.5% to 40% sequence homology with human p55 and p75 [112]. In addition, CrmB contains a C-terminal chemokine-binding domain, referred to as the SECRET domain (Fig. 5) [113].

Fig. 5. Domain organization of the CrmB protein. Ovals represent the three N-terminal CRDs PLAD subdomain

A comparative analysis of the CrmB amino acid sequences of CPXV, MPXV, and VARV has revealed the species-specific differences that may determine the variability of their functional properties (Fig. 6) [114].

Fig. 6. Comparison of the amino acid sequences of orthopoxvirus CrmB proteins belonging to the TNFR superfamily. Two strains of each virus were compared: CPXV (GRI and MUN-85), MPXV (ZAI and CNG), and VARV (IND and GAR). Amino acid residues identical to CPXV GRI are indicated by dots; deletions are represented by dashes. Species-specific differences between VARV and MPXV/CPXV in the TNFR region are highlighted by blue and green, respectively. The SECRET domain sequence is shown in blue italics. The PLAD subdomain sequence is marked by a red line. Red vertical bars and the corresponding numbers above them denote Cys residues forming the three CRDs

We conducted extensive studies to determine the properties of the recombinant full-length CrmB proteins of VARV, MPXV, and CPXV in order to establish the species-specific differences between them. Recombinant proteins were produced using a baculovirus expression system; the resulting proteins had a molecular mass of 45–47 kDa. Despite a high degree of sequence homology (85–96%) between them, species-specific differences were identified in assays evaluating the ability of recombinant orthopoxviruses proteins to bind TNF from different animal species. VARV CrmB was shown to neutralize the cytotoxic activity of both human TNF (hTNF) and murine TF (mTNF) with similar effectiveness in vitro [115], whereas CPXV CrmB effectively neutralized only mTNF, not hTNF [114, 116]. MPXV CrmB did not exhibit any significant inhibitory effect toward either ligand [114]. Thus, the TNF-neutralizing activity of these proteins varies substantially across animal species, which is likely due to differences in their amino acid sequences.

A gel filtration analysis demonstrated elution of VARV CrmB in the fraction corresponding to a molecular mass of > 500 kDa, indicating protein oligomerization, while CPXV CrmB was primarily detected in the fraction corresponding to dimeric forms [112].

Viral immunomodulatory proteins, unlike their cellular analogues, can form stable oligomers in the absence of a ligand, which may enhance efficiency in their binding to target proteins.

It was also shown that CrmB inhibits the binding of hTNF to polyclonal anti-TNF antibodies, with inhibitory effectiveness decreasing in the following order: VARV CrmB > CPXV CrmB > MPXV CrmB [117]. The hTNF-neutralizing activity of these proteins in vitro was comparable to that of human p55 and p57, as well as the monoclonal antibody mAb MAK195. The TNF-neutralizing activity of Remicade was slightly higher than that of the two-domain VARV CrmB: at 50% cell viability, the ratio of hTNF to VARV CrmB and Remicade concentrations was 2 : 4–8 and 2 : 0.5–1 ng/mL, respectively [112].

The therapeutic efficacy of recombinant CrmB proteins from VARV, MPXV, and CPXV was further evaluated in a mouse model of lipopolysaccharide (LPS)-induced endotoxic shock. In that experiment, chimeric variants of the VARV and CPXV CrmB proteins fused to the heavy chain fragment of human IgG1 (hereafter referred to as VARV CrmB/IgG1 and CPXV CrmB/IgG1) were also tested, since such modifications increase protein avidity and the circulating half-life [118]. As a result, the VARV CrmB and VARV CrmB/IgG1 proteins demonstrated a significant therapeutic impact, unlike CPXV CrmB, CPXV CrmB/IgG1, and MPXV CrmB. Of these, VARV CrmB/IgG1 exhibited the strongest therapeutic effect [115, 116].

The obtained results indicate that even highly homologous CrmB proteins from different orthopoxvirus species can vary significantly in ligand specificity and, in particular, exhibit different biological activities towards ligands from different species.

Administration of the recombinant VARV CrmB protein to intact mice did not lead to any changes in the animals’ behavior, appearance, or histological structure of internal organs compared to the control mice, suggesting an absence of a toxic effect of the drug under the experimental conditions used [115].

In a number of additional studies, VARV CrmB also demonstrated efficacy as an hTNF antagonist, which led researchers to consider it as a potential platform for the development of novel TNF inhibitors for the treatment of TNF-driven diseases [119].

Since therapeutic agents typically require repeated administration, it is important that the drug possess low immunogenicity. In this regard, we assessed the immunogenic properties of the VARV CrmB upon repeated administration in BALB/c mice. The high immunogenicity of recombinant VARV CrmB was noted, and we speculated that it may be partially associated with the previously described fact of protein multimerization [112, 120].

To reduce the immunogenicity and establish the functional significance of individual domains, we generated recombinant deletion and chimeric variants of VARV CrmB and CPXV CrmB (Fig. 7). The genes encoding recombinant proteins were expressed using a baculovirus system in Sf-21 insect cells. It was discovered that deletion of PLAD (located in the first CRD) from VARV CrmB removed the protein’s ability to neutralize the cytotoxic effect of hTNF in L-929 mouse fibroblasts. Replacement of PLAD in VARV CrmB with that of CPXV CrmB resulted in reduced inhibitory activity towards hTNF (compared to VARV CrmB), while the reverse substitution of PLAD in CPXV CrmB yielded the opposite effect [121].

Fig. 7. Schematic representation of chimeric and truncated variants of the CrmB protein produced using the baculovirus expression system. Regions corresponding to VARV CrmB and CPXV CrmB are shown in blue and orange, respectively. 1 – VARV CrmB lacking the SECRET domain. 2 – CPXV CrmB lacking the SECRET domain. 3 – VARV CrmB lacking the PLAD subdomain. 4 – VARV CrmB containing the PLAD subdomain of CPXV CrmB. 5 – CPXV CrmB containing the PLAD subdomain of VARV CrmB

We also demonstrated that deletion of the SECRET domain did not impair the ability of the truncated CrmB protein (hereafter referred to as the TNF-binding domain (TNF-BD)) from VARV to inhibit the cytotoxic activity of hTNF [121].

To evaluate VARV TNF-BD as a potential agent for TNF-targeted therapy we needed further study into its TNF-binding properties in both in vitro and in vivo experiments.

We compared the ability of the VARV TNF-BD and two-domain VARV CrmB proteins to inhibit the TNF interaction with specific cellular receptors. For this, we evaluated their capacity to neutralize the cytotoxic effect of hTNF and mTNF in L-929 cells. The experiments demonstrated that both proteins inhibit the cytotoxic effect of hTNF and mTNF with comparable efficacy [121].

The biological activity of VARV TNF-BD in vivo was evaluated in a mouse model of LPS-induced endotoxic shock. In mice receiving LPS, mortality reached 90% within 72 h after injection. In contrast, administration of recombinant proteins (both VARV TNF-BD and VARV CrmB) resulted in a survival rate of 62.5% [121].

To minimize the potential cost of VARV TNF-BD production, a E. coli expression system was developed. The biological activity of the recombinant VARV TNF-BD protein was assessed based on its ability to neutralize the cytotoxic effects of hTNF and mTNF in L-929 mouse fibroblast cells. The protein produced in the prokaryotic system was shown to retain biological activity and effectively neutralize TNF-induced effects in vitro [121].

Furthermore, a comparative analysis of the effects of VARV CrmB and VARV TNF-BD on the TNF-induced oxidative and metabolic activity of blood leukocytes in intact mice showed that removal of the chemokine-binding domain in VARV CrmB does not impair the ability of the truncated, non-glycosylated VARV TNF-BD to neutralize the biological actions of TNF [122].

The obtained results support the potential of the non-glycosylated TNF-BD protein derived from VARV, produced in a bacterial expression system, as a candidate for anti-TNF therapy. In this regard, we further evaluated the immunogenic properties of the truncated TNF-binding protein. Deletion of the chemokine-binding domain in VARV CrmB resulted in a reduced humoral immune response compared to the full-length protein, which may favor the therapeutic application of the truncated protein TNF-BD [120].

Since the truncated VARV TNF-BD protein exhibits reduced immunogenicity, it can be considered a promising candidate for the development of novel anti-TNF therapeutics. Therefore, the affinity of its interaction with TNF was further investigated.

First, protein–protein interactions between truncated CrmB and TNF were analyzed using bioinformatic modeling [123]. VARV TNF-BD and CPXV TNF-BD proteins were chosen as receptors, and hTNF and mTNF were used as ligands. We selected truncated CrmB proteins for the experiments, because their full-length analogues, despite high, 87% sequence homology, exhibit species-specific differences in ligand binding [114].

The three-dimensional structures of VARV TNF-BD, CPXV TNF-BD, and TNF were predicted using the Swiss-model (Fig. 8). The structure of the TNF-binding domain of the human p75 receptor (pdbid: 3ALQ) in a complex with mutant hTNF was used as a template for modeling the three-dimensional protein structures [123].

Fig. 8. Predicted three-dimensional structure of the complex of hTNF homotrimer with a CPXV TNF-BD molecule, shown in side (A) and top (B) views. CRDs are indicated by square brackets. Colors and labels correspond to different complex subunits (R – CPXV TNF-BD; A, B, and C – individual units of hTNF homotrimer). The structure is presented as a ribbon diagram [123]

Next, the efficiency of protein–protein interactions was evaluated using surface plasmon resonance (SPR) analysis. Dissociation constant (K_D) values were determined for both the full-length proteins VARV CrmB and CPXV CrmB produced in Sf-21 cells and their truncated variants VARV TNF-BD and CPXV TNF-BD generated in E. coli.

The K_D values obtained for the complexes CPXV TNF-BD/hTNF and VARV TNF-BD/hTNF aligned with the results of molecular dynamics simulations using the MM-GBSA implicit solvation model: the free energy of complex formation was higher for VARV TNF-BD/hTNF compared to CPXV TNF-BD/hTNF (Table 3) [123].

Table 3. SPR analysis of the effectiveness of interaction between hTNF, mTNF, and viral receptors

In terms of immunogenicity and hTNF-binding efficiency, VARV TNF-BD demonstrates potential advantages relative to currently available anti-TNF drugs and may become a prospective candidate that can compete with foreign alternatives based on antibodies and TNFR (Table 4). The binding affinity of VARV TNF-BD for hTNF is either comparable or even higher than that of available alternatives. The lower molecular weight of TNF-BD contributes to its reduced immunogenicity and may allow for usage of lower therapeutic doses. VARV TNF-BD production in E. coli also allows for significantly reduced manufacturing costs compared to protein expression in eukaryotic cells.

Table 4. Comparison of anti-TNF agents

To further curb the immunogenicity, we used an alternative strategy for protein delivery. For this, we constructed a recombinant plasmid (VARV pcDNA/TNF-BD) encoding VARV TNF-BD to enable its expression in mammalian cells. The therapeutic potential of this construct was evaluated in a rat model of collagen-induced arthritis (CIA). Injection of the recombinant plasmid in CIA rats resulted in impeded histopathological joint damage compared to the control animals receiving the pcDNA vector [124]. These findings suggest the potential of using gene therapy approaches to treat RA based on local injection of a recombinant plasmid to ensure VARV TNF-BD expression.

Repeated administration of a therapeutic dose of VARV pcDNA/TNF-BD induced a specific immune response against the expressed protein in mice; however, the response was significantly more muted than that observed upon repeated injection of the recombinant protein [125]. This observation further supports the potential advantages of this anti-TNF approach.

The collected experimental data indicate that VARV TNF-BD represents a promising platform for the development of novel TNF inhibitors.

Further study of viral TNFR homologues may significantly contribute to the improvement of existing anti-TNF therapies, primarily through expanding our knowledge of the mechanisms of ligand-receptor interactions.

One such study was performed in 2019. In it, modification of the primary structure of etanercept, in particular the introduction of a Glu–Phe–Glu amino acid motif into one of the structural loops, resulted in a 3- and 60-fold reduction in its biological activity against lymphotoxin (LT) and TNF, respectively. Modulating the ability of this potential TNF inhibitor to bind to LT should reduce the incidence of infectious diseases during therapy, since LT is involved in inducing immune responses in conditions of suppressed TNF signaling. In addition, the role of LT in inflammatory diseases remains uncertain; for instance, studies of RA patients have revealed no therapeutic effect from LT-targeted antibodies [126, 127].

At the same time, the reduced anti-TNF activity in the modified etanercept variant may adversely affect its therapeutic efficacy. Such limitation can be mitigated in clinical settings through increasing the dose. Thus, the modified protein may serve as a basis for the development of safer etanercept-like therapeutics.

A distinctive aspect of the work by Pontejo et al. [127] is that the authors originally investigated viral TNFRs. The proposed modification of etanercept represents an application of fundamental knowledge about viral proteins. In particular, it has been shown that the poxvirus protein CrmD can inhibit hTNF, mTNF, and murine LT, but not human LT. The study of the molecular basis of the differing inhibitory activity of the protein in relation to species-specific ligands demonstrated that its inability to inhibit human LT is due to the presence of a Glu–Phe–Glu motif in one of its structural loops. The importance of this motif in determining the receptor’s ability to bind to human LT was further confirmed in studies of etanercept modification [127].

Thus, anti-TNF agents engineered on the basis of orthopoxvirus proteins may become a promising alternative to existing TNF inhibitors with a similar mechanism of action (Fig. 9).

Fig. 9. Schematic representation of the cellular response induced by the interaction of TNF and LT with their cellular receptors, and the molecular targets of clinically used TNF inhibitors [128], as well as orthopoxvirus TNF-binding proteins

About the authors

T. V. Tregubchak

Email: snshchel@rambler.ru
Russian Federation, Koltsovo, Novosibirsk region, 630559

S. N. Shchelkunov

Author for correspondence.
Email: snshchel@rambler.ru
Russian Federation, Koltsovo, Novosibirsk region, 630559

References

Supplementary files

Supplementary Files

Action

1. JATS XML

Download

2. Fig. 1. Schematic representation of TNF receptor–ligand interactions. In the absence of a ligand, TNF receptors exist as dimers; upon ligand binding, they form an oligomeric network consisting of trimeric receptor–ligand complexes. The complex formation initiates intracellular signaling [15]. CRDs – Cys-rich domains

Download (152KB)

Indexing metadata

3. Fig. 2. Intracellular signaling pathways activated by TNF binding to p55 and p75. TNF binding to p55 induces cell death and inflammatory signaling through activation of the canonical NF-κB pathway. TNF binding to p75 activates both canonical and non-canonical NF-κB pathways. Solid green and blue lines represent activating signals; and dashed lines indicate inhibitory interactions. sTNF – soluble TNF; tmTNF – transmembrane TNF; sp55 – soluble p55; tmp55 – transmembrane p55; sp75 – soluble p75; tmp75 – transmembrane p75 [22]

Download (477KB)

Indexing metadata

4. Fig. 3. Schematic representation of the activation of various TLR-mediated signaling pathways leading to cytokine storm and sepsis. Activation of these signaling cascades induces the synthesis of pro-inflammatory cytokines (IL-1, IL-6, IL-12, IL-8, and TNF), anti-inflammatory cytokines (IL-10), and type 1 interferons. Elevated levels of these inflammatory mediators lead to cytokine storm in sepsis [44]

Download (470KB)

Indexing metadata

5. Fig. 4. Simplified representation of the molecular structures of TNF antagonists. Infliximab is a chimeric mouse/human IgG1 monoclonal anti-TNF antibody. Adalimumab and golimumab are fully human IgG1 monoclonal anti-TNF antibodies. Etanercept is a fusion protein consisting of two extracellular domains of human p75 and the Fc region of human IgG1. Certolizumab pegol is a PEGylated Fab fragment of humanized IgG1 monoclonal anti-TNF antibody [64]

Download (151KB)

Indexing metadata

6. Fig. 5. Domain organization of the CrmB protein. Ovals represent the three N-terminal CRDs PLAD subdomain

Download (74KB)

Indexing metadata

7. Fig. 6. Comparison of the amino acid sequences of orthopoxvirus CrmB proteins belonging to the TNFR superfamily. Two strains of each virus were compared: CPXV (GRI and MUN-85), MPXV (ZAI and CNG), and VARV (IND and GAR). Amino acid residues identical to CPXV GRI are indicated by dots; deletions are represented by dashes. Species-specific differences between VARV and MPXV/CPXV in the TNFR region are highlighted by blue and green, respectively. The SECRET domain sequence is shown in blue italics. The PLAD subdomain sequence is marked by a red line. Red vertical bars and the corresponding numbers above them denote Cys residues forming the three CRDs

Download (463KB)

Indexing metadata

8. Fig. 7. Schematic representation of chimeric and truncated variants of the CrmB protein produced using the baculovirus expression system. Regions corresponding to VARV CrmB and CPXV CrmB are shown in blue and orange, respectively. 1 – VARV CrmB lacking the SECRET domain. 2 – CPXV CrmB lacking the SECRET domain. 3 – VARV CrmB lacking the PLAD subdomain. 4 – VARV CrmB containing the PLAD subdomain of CPXV CrmB. 5 – CPXV CrmB containing the PLAD subdomain of VARV CrmB

Download (75KB)

Indexing metadata

9. Fig. 8. Predicted three-dimensional structure of the complex of hTNF homotrimer with a CPXV TNF-BD molecule, shown in side (A) and top (B) views. CRDs are indicated by square brackets. Colors and labels correspond to different complex subunits (R – CPXV TNF-BD; A, B, and C – individual units of hTNF homotrimer). The structure is presented as a ribbon diagram [123]

Download (180KB)

Indexing metadata

10. Fig. 9. Schematic representation of the cellular response induced by the interaction of TNF and LT with their cellular receptors, and the molecular targets of clinically used TNF inhibitors [128], as well as orthopoxvirus TNF-binding proteins

Download (235KB)

Indexing metadata