Higher order receptor clustering due to the IgG3 subclass is necessary for TLR4 signaling and tolerance induction by novel human anti-TLR4 antibodies

ABSTRACT

We have previously demonstrated that an IgG3 agonistic TLR4/MD2 antibody reversed acute murine Type 1 diabetes (T1D), induced immune tolerance, and induced long-term endosomal sequestration of TLR4/MD2. We hypothesized that the IgG3 Fc was critical for agonist activity of the mouse TLR4 antibodies, due to the unique IgG3 extended hinge region and enhanced ability to form higher-order oligomeric structures. Here, we prove the essential role of the Fc region using plate-bound antibody and F(ab’)2 and F(ab) fragments, which greatly reduced or eliminated TLR4 signaling. Importantly, no agonistic TLR4 antibodies have been described for humans. We developed four novel IgG4 human agonistic TLR4/MD2 antibodies as potential therapeutic candidates for T1D. The human IgG4 anti-TLR4 antibodies failed to activate the TLR4/MD2 pathway. Switching two candidate antibodies from IgG4 to IgG3, however, resulted in robust TLR4 signaling. Cross-linking the IgG4 antibody with an IgG3 secondary antibody also induced robust TLR4 signaling. Based on this result, which suggested that increased TLR4 clustering could increase signaling, we developed tetravalent IgG3 and IgG4 anti-TLR4 antibodies. Tetravalent IgG3, but not IgG4, anti-TLR4 antibodies robustly signaled via the TLR4 pathway. Notably, however, cross-linking human IgG3 antibodies with non-IgG3 secondaries reduced TLR4 signaling, in marked contrast to activation induced by IgG4 isotypes with IgG3 crosslinker, potentially through interference with IgG3 Fc-mediated oligomerization. These results suggest that the IgG3 Fc enhances agonist function of human TLR4 antibodies via aggregation of the TLR4 receptor. Functionally, human IgG3 and IgG3 tetravalent antibodies induced tolerance in primary human monocytes, analogous to the mouse antibody. In conclusion, we developed novel human TLR4 agonistic antibodies, demonstrated that the IgG3 isotype and enhanced multivalency are necessary for their TLR4 signaling, and demonstrated their tolerogenic potential for treating inflammatory diseases.

Introduction

Previously, we showed reversal of acute Type 1 diabetes (T1D) in NOD mice by administration of a mouse anti-TLR4/MD2 agonist antibody (designated UT-18).^1–3 UT-18 treatment decreased blood glucose levels, increased insulin production in the remaining islets, and decreased islet inflammation.¹ The primary target of UT-18 was CD11b+ cells, which increased in number and phenotypically and functionally developed into myeloid-derived suppressor cells (MDSCs).³ Monoclonal antibody therapy is an effective method of targeting specific immune molecules. Given our strong data in the T1D mouse model, inducing immunosuppression through human TLR4 activation using a human monoclonal antibody might be a viable strategy to treat T1D. However, no human TLR4 agonist antibody has been developed to test this therapeutic approach. Interestingly, as an IgG3 isotype, mouse UT-18 is atypical of current therapeutic monoclonal antibodies that aim to minimize effector functions such as antibody-dependent cell-mediated cytotoxicity and complement-dependent cytotoxicity, which are generally IgG4.^4–7 Determining the functional relevance of IgG3 Fc to TLR4 agonist activity is therefore critical to informing the rational design of a human anti-TLR4 agonist antibody.

To understand the structural requirements for a successful anti-TLR4 agonist antibody it is also important to understand the structural mechanisms responsible for TLR4 receptor activation. The activated TLR4 receptor exists as two heterodimers of TLR4/MD2 complexed together to form a heterotetramer. This complex requires two molecules of lipopolysaccharide (LPS) for activation. The LPS molecule binds within the hydrophobic pocket of MD2 and draws in the Phe₁₂₆ loop, causing a conformational change in TLR4 that shifts the position of the intracellular toll-interleukin-1-receptor (TIR) domain.⁸ In the LPS-bound TLR4:MD2 heterotetramer, the intracellular TIR domains are brought into proximity, enabling a conformational change and formation of a binding site for MyD88, which initiates the signaling pathway responsible for NFkB-mediated transcription of inflammatory genes. Dimerization of the heterodimers (i.e., forming a heterotetramer) is therefore critical for the initiation of signaling; however, an ongoing question has been whether higher-order clustering of TLR4:MD2 heterotetramers are needed to form and/or stabilize the MyDDosome complex. Although there are reports of higher-level TLR4:MD2 clustering,⁹ most recent reports using super-resolution fluorescence microscopy conclude that the formation of the TLR4:MD2 heterotetramer is sufficient for the formation of a functional MyDDosome and that higher-level clustering is not observed upon ligation by LPS agonists.^10–13 Interestingly, TLR4 signaling can also be activated by aggregated protein ligands, including amyloid β fibrils or α-synuclein aggregates, leading to signaling with distinct kinetics from LPS activation.^14–16 These aggregated protein ligands could potentially induce higher-order TLR4 clustering (larger than the LPS-induced TLR4/MD2 heterotetramers).

IgG3 is unique among IgG subclasses with respect to its highly extended hinge region and its tendency to assemble into higher-order oligomers or aggregates at lower temperatures or upon antigen binding. Human and mouse IgG3 antibodies exhibit distinct structural and functional differences. Human IgG3 has a longer hinge region with 11 disulfide bonds, compared to mouse IgG3‘s shorter hinge with fewer disulfide bonds, impacting flexibility and stability.¹⁷ In mice, IgG3 is known for its ability to form large immune complexes due to its extended hinge region and strong propensity for self-association through non-covalent oligomers, which enhances its efficacy in complement activation and agglutination of particulate antigens.^18–21 Human IgG3 has less propensity for self-assembly, but the increased flexibility from the longer hinge region allows for higher functional affinity.²² While human IgG3 may be less adept at forming large immune complexes, recent structural work has demonstrated that IgG3 binding to membrane-localized antigens leads to the formation of stable hexameric assemblies in which the Fc regions form a ‘wagon-wheel’ hexameric ring, while the extended hinge region provides sufficient flexibility to allow the F(ab) regions to self-assemble into a planar array.²³ The resulting hexameric Fc conformation facilitates binding by C1q and complements activation by antigen-bound IgG3.²³ It is not clear whether higher-order assembly of IgG3 upon TLR4 binding (whether hexamers or higher-order oligomers) might induce distinct signaling outcomes, as reported for TLR4 activation by amyloid β fibrils or α-synuclein aggregates.^14–16 Furthermore, given the differences between mouse and human IgG3 functionality, it is uncertain that anti-TLR4 agonists that cause immunosuppression in humans can be created.

The quantitative and qualitative aspects of receptor binding contribute to opposing functional outcomes in TLR4 activation. Designing human anti-TLR4 antibodies to exploit immunosuppressive signaling is key to successful therapy. Toll-like receptor 4 (TLR4) is a functionally conserved innate immune receptor primarily activated by LPS from Gram-negative bacteria. Nobel prize-winning work by Beutler and Hoffman established TLR4 as a critical receptor for sensing bacterial infections and initiating inflammation.²⁴ However, since the 1960s it has been known that prior exposure to Gram-negative bacteria or endotoxin can induce tolerance to repeated exposure²⁵ and, in 2007, Medzhitov et al.²⁶ described epigenetic mechanisms controlling tolerizeable inflammatory genes.²⁶ This dichotomy of activating and inhibitory signals from a single receptor is borne out in the NOD mouse model of autoimmune T1D. In germ-free facilities, T1D incidence is near 100%, while mice raised in dirty facilities (or if intestinal microbiota are transferred into germ-free mice) have a reduced incidence of disease.²⁷ This led Chervonsky et al. to the balanced signal hypothesis, whereby the regulatory mechanisms induced by the microbiome through TLR4 signaling to establish symbiosis also lead to tolerance of self-antigens. TLR4 signaling is unique among the toll-like receptor signaling in activating two pathways: 1) cell-surface signaling via MyD88 to activate a classic NFkB-mediated inflammatory response, and 2) an endosomal signal after internalization of the TLR4 receptor to activate the TRIF-IRF3 pathway that generates Type 1 interferons and suppressive cytokines such as IL-10.^28–30 Through a series of knockout mice, Chervonsky et al. showed that the protective response from TLR4 signaling in NOD T1D was mediated through the endosomal TRIF-mediated pathway, but only when the microbiota were present, indicating microbial products must be present to activate TLR4 for induction of tolerance.³¹ In accordance with the “balanced signal” hypothesis, it may be possible to restore balance in human T1D or other autoimmune conditions by exploiting the tolerogenic pathway of TLR4 through agonistic monoclonal therapy.

Herein, we report that the IgG3 isotype is crucial for TLR4/MD2 agonism by human TLR4-targeted antibodies, and that an antibody with the exact same epitope specificity will signal if it is an IgG3 isotype, but not signal if it is an IgG4 isotype. We hypothesize that the flexibility of IgG3 F(ab’)2 arms for stabilizing dimerization and the Fc regions’ oligomerization capabilities are necessary for optimal signaling through stabilization of the MyDDosome and/or the Triffosome. The IgG3 isotype has been overlooked for use in therapeutic antibodies because of its high complement-fixing and FcR-mediated inflammatory activity, but these studies shed light on a unique and important role IgG3 could play in therapeutic antibodies targeting TLR receptor activation.

Results

The IgG3 Fc region of mouse UT-18 TLR4-Ab is critical to TLR4 signaling

The effector functions of immunoglobulins are controlled by the Fc region. This region is cleaved during pepsin digestion to create F(ab’)2 fragments that retain their antigen-binding specificity and bivalent nature, but lack effector functions. Further digestion using papain cleaves IgG above the disulfide linkage to generate individual F(ab) fragments capable of binding antigens, but only in a monovalent manner and lacking effector functions. We generated both F(ab’)2 and F(ab) fragments of UT-18 (Figure S1) and ensured that the preparations were free from LPS contamination (Figure S2 and S3). We stimulated mouse TLR4 hEK-Blue reporter cells, which allow for the quantification of NFkB activity. UT-18 F(ab’)2 fragments showed significantly decreased activity, and UT-18 F(ab) fragments showed complete elimination of activity compared to full-length (FL) UT18 (Figure 1a). Furthermore, plate-bound UT-18 (with immobilized Fc portion and correct orientation for F (ab) binding) activity was nearly eliminated (Figure 1b). To test the hypothesis that the Fc portion of UT-18 was critical for activity in primary cells as well, thioglycolate-elicited peritoneal macrophages were stimulated with FL UT-18, F(ab’)2, or F(ab) fragments in comparison to LPS and the control anti-TLR4 antibody UT-15 (IgG1 non-agonist) (Figure 2), and tumor necrosis factor (TNF)-α protein production quantified. UT-18 F (ab) and F(ab’)2 activity were again significantly reduced compared to FL UT-18 (Figure 2b,c). FL UT-18 has an upper limit of TLR4-induced TNF-α that is reached at up to a 1000-fold decreased concentration compared to the fragments. Maximum UT18 stimulation, however, induces significantly less TNF-α than LPS (0.01 μg/ml LPS mean 1,082 pg/ml TNF-α vs 0.03 μM UT-18 mean 337 pg/ml TNF-α; p = 0.0026 by t-test) (Figure 2a). F(ab) and F(ab’)2 produce much less TNF-α (30 nM FL mean TNF-α 337 pg/ml vs 30 nM F(ab’)2 mean TNF-α 143 pg/ml vs 30 nM F(ab) mean TNF-α 89.66; p = 0.0007 by one-way Anova) and only at the highest concentration; thus, FL UT18 is approximately 2000 times more stimulatory than the fragments.

Figure 1.

Removing or immobilizing the IgG3 Fc portion of mouse agonistic TLR4-ab (UT-18) reduces its stimulatory activity. UT-18 was enzymatically digested to F(ab’)2 or F(ab) subunits and compared to full length (FL) UT-18 TLR4 expressing HEK-Blue NFKB reporter cells. a) mouse TLR4 expressing HEK-Blue reporter cells were stimulated at the indicated concentrations of F(ab’)2, F(ab), or FL UT-18. After 24 h supernatants were collected and assayed for alkaline phosphatase reporter activity. b) UT-18, or a non-agonist control anti-TLR4-ab (UT15), was bound to protein G plates at the indicated concentrations to immobilize the Fc portion of the antibody. Mouse TLR4 expressing HEK-Blue cells were then added, with soluble UT-18 and LPS as controls. After 24 h supernatants were assayed for alkaline phosphatase reporter activity. Data points represent three independent experiments. Statistical analysis by ANOVA *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Figure 2.

The Fc portion of IgG3 UT-18 is critical for activation of TLR4 in peritoneal macrophages. a) TNF-α production from thioglycolate elicited peritoneal macrophages 18 h post stimulation with LPS, IgG3 UT-18, or control antibody IgG1 UT-15. Enzymatic digestion of UT-18 to F(ab’)2 b) or F(ab) fragments c) show reduced TNF-α production at 18 h in these same cells. d) protein G coated plates were used to immobilize UT-18 to the surface and limit IgG3 Fc aggregating functions prior to addition of thioglycolate elicited peritoneal macrophages. Data representative of six individual mice from two different experiments. Statistical analysis by ANOVA *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

To further test the significance of Fc-mediated oligomerization, we tested plate-bound UT-18 on thioglycolate-elicited peritoneal macrophages. While protein G-bound UT-18 did not elicit a response in the reporter cells (Figure 1b), in thioglycolate-elicited peritoneal macrophages, protein G-bound UT-18 was able to elicit a TNF-α response equivalent to soluble UT-18 at the highest concentrations (Figure 2d). However, TNF-α production decreased rapidly at lower concentrations, indicating that UT-18 density on the plate surface was critical to agonist activity. Therefore, the oligomerization of soluble UT-18 through IgG3 Fc interactions is likely a key contributor to UT-18 activity.

Human IgG4 anti-TLR4Abs do not activate the human TLR4 pathway unless crosslinked by IgG3 anti-IgG4 antibodies

Human anti-TLR4 antibodies (HuTLR4Ab) were generated using an F(ab) phage library (i.e., 3D7, 8G10, 8H1, 7D11), and reactivity was confirmed by ELISA using recombinant TLR4/MD2 (Figure 3). All four clones were purified and produced as FL IgG4 antibodies. All four variants bound recombinant TLR4/MD2 as well as TLR4 alone except 7D11 (Figure 3a-b). Using a competition binding assay, soluble TLR4 competed with the ability of each HuTLR4Ab except 7D11 to interact with plate-bound TLR4/MD2, providing further evidence that 7D11 does not bind TLR4 (Figure 3d). All four variants also bound MD2 alone albeit with lower affinity than TLR4/MD2 (Figure 3c). These data indicate that 3D7, 8G10, and 8H1 bind to regions near the TLR4/MD2 primary interface where they recognize a portion of both TLR4 and MD2 in the heterodimer configuration, whereas 7D11 binds to a region specific only to MD2.

Figure 3.

Novel human IgG4 antibodies 3D7, 8G10, and 8H1 recognize TLR4 and MD2 while 7D11 recognizes only MD2. Detection of recombinant TLR4/MD2 heterodimer (a) TLR4 alone (b) or MD2 alone (c) by ELISA using a 3-fold dilution series of human IgG4 antibodies 3D7, 8G10, 8H1, and 7D11. With the concentration of TLR4Ab held constant, an increasing concentration of soluble TLR4 was added in the ELISA assay to compete with the plate bound TLR4/MD2 (d). Data represents a single experiment in duplicate.

Stimulation of a THP-1 dual monocyte reporter cell line (tracking both NFkB and IRF3 signaling) shows minimal NFkB activity by 8H1 and no activity with the other antibodies. None of the antibodies showed IRF3 activation (Figure 4a). To test the role of oligomerization, which should bring together more TLR4 signaling heterotetramers and improve activity, the IgG4 huTLR4Abs were crosslinked with either a mouse IgG3 anti-human IgG4 antibody or a rat IgG2a anti-human Fc antibody. IgG4 8H1 showed a significant increase in activity after crosslinking with the mouse IgG3 anti-human IgG4 secondary antibody in both the NFkB reporter as well as the IRF3 reporter readouts, indicating internalization of the receptor and signaling within the endosome (Figure 4b), similar to our therapeutic mouse IgG3 antibody.³ Interestingly, receptor activation was only observed with an IgG3 secondary, not with the rat IgG2 secondary, providing further data to support the hypothesis that IgG3 is important for the agonist activity of TLR4 antibodies. The other three IgG4 huTLR4Abs showed minimal activity with crosslinking (data not shown). Delayed addition of the crosslinking antibody also allowed signaling, indicating that IgG4 8H1 is likely binding but does not signal or internalize until the addition of the crosslinker (Figure S4). Titration experiments indicate that a 1:1 ratio is optimal, but a 1:10 ratio (8H1:Crosslinker) is also sufficient to signal (Figure S5). These studies further support the hypothesis that oligomerization of HuTLR4Ab is necessary for activity, but the fact that not all variants were capable of activation indicates that the location of the binding epitope on TLR4/MD2 also plays a role.

Figure 4.

IgG4 8H1 does not activate NFkB signaling, while IgG4 8H1 crosslinked with an IgG3 secondary antibody significantly activates both NFkB and IRF3 activity a) THP-1 Dual reporter cells were treated with 10 μg/ml of human IgG4 anti-TLR4/MD2 antibodies. At 48 h supernatants were collected and assayed for alkaline phosphatase (NFkB) and luciferase (IRF3) activity b) under the same culture conditions, 8H1 antibody was co-incubated with a crosslinking mouse IgG3 anti-human IgG4 or a rat IgG2a anti-human Fc antibody at a 1:1 ratio. Data points represent three independent experiments. Statistical analysis by ANOVA *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Converting the HuTLR4Ab 8H1 from IgG4 to IgG3 produces TLR4 pathway agonist activity

While crosslinking with a secondary antibody improves agonistic activity, this approach is not therapeutically viable. To test the hypothesis that the IgG3 subclass is critical to TLR4/MD2 agonism, we converted 8H1 from IgG4 to IgG3, preserving the antigen binding region, but altering the effector function Fc region. Comparing IgG3 vs IgG4 8H1 activity in THP-1 Dual reporter cells clearly shows that given identical F(ab) regions, only the IgG3 version activates TLR4/MD2 (Figure 5).

Figure 5.

Switching human IgG4 8H1 to IgG3 isotype induces TLR4 signaling activity. THP-1 Dual reporter cells were stimulated with IgG4 or IgG3 8H1 antibody. After 48 h supernatants were collected and assayed for a) alkaline phosphatase (NFkB) and b) luciferase (IRF3) activity. Only the IgG3 isotype shows significant signaling. Data points represent three independent experiments. Statistical analysis by ANOVA *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Antigen binding tetravalency induced HuTLR4Ab agonism only in IgG3, not IgG4 isotypes

The HuTLR4Ab 7D11 was unique in binding MD2 alone and not TLR4 alone. However, it failed to activate TLR4/MD2 signaling with either an IgG4 (Figure 4) or IgG3 isotype (Figure 7). As oligomerization of the receptor appears crucial for signaling, we generated tetravalent (TV) versions of 7D11 in which an additional single-chain variable fragment (scFv) was attached to the N-terminus of its heavy chain to create an antibody with four identical variable regions. These scFVs were attached to both IgG3 and IgG4-7D11 to test whether multivalency could act as a surrogate for oligomerization on an IgG4 antibody. Using THP-1 Dual reporter cells, we show that tetravalency induces TLR4/MD2 agonist activity but only in IgG3-TV7D11, not IgG4-TV7D11 (Figure 6), proving once again that IgG3 is critical for TLR4/MD2 agonism. Furthermore, crosslinking IgG4 TV 7D11 with an IgG3 mouse anti-human secondary resulted in stimulated activity similar to IgG4 8H1. However, we only observed an increase in NFkB activity (Figure 6a), not IRF3 (Figure 6b). This implies that crosslinking IgG4 TV 7D11 only allowed surface signaling through MyD88 but did not allow internalization and IRF3 pathway activation. We next attempted to increase the activity of the IgG3 TLR4 antibodies (both the bivalent and tetravalent antibodies) through higher-order clustering using an anti-human IgG3 antibody. Interestingly, unlike IgG4-8H1, crosslinking did not increase agonism in any of the IgG3 antibodies. Rather, crosslinking with either a mouse IgG1 anti-human IgG3 or a rat IgG2a anti-human IgG-Fc significantly decreased TLR4/MD2 activity of both IgG3 8H1 and IgG3 7D11 TV antibodies, and this effect was particularly pronounced in the endosomal IRF3 pathway that requires internalization (Figure 7). These results suggest the non-IgG3 crosslinking antibodies interfere with the oligomerization of IgG3 (hexameric platforms) and prevent sufficient MyDDosome stabilization and signaling.

Figure 7.

Crosslinking IgG3 8H1 and IgG3 TV-7D11 with anti-human IgG3 reduces activity. THP-1 Dual reporter cells stimulated with 5 μg/ml IgG3 8H1, IgG3 TV-7D11, or IgG3 7D11 antibody co-incubated with 5 μg/ml of IgG1 mouse anti-human IgG3 or IgG2a rat anti-human Fc. After 48 h supernatants were collected and assayed for alkaline phosphatase (NFkB) (a) and Luciferase (IRF3) activity (b). Data points represent three independent experiments. Statistical analysis by ANOVA *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Figure 6.

Tetravalent 7D11 activates THP-1 Dual reporter cells but only with an IgG3 Fc region or when crosslinked with an IgG3 secondary antibody. THP-1 Dual reporter cells were stimulated with IgG3 tetravalent 7D11 antibody, IgG4 tetravalent 7D11 antibody, or IgG4 tetravalent 7D11 crosslinked with mouse IgG3 anti-human IgG4 crosslinker. After 48 h supernatants were collected and assayed for a) alkaline phosphatase (NFkB) and b) luciferase (IRF3) activity. Data points represent three independent experiments. Statistical analysis by ANOVA *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Stimulation of primary human monocytes results in tolerance induction

One of the functional outcomes of agonistic anti-TLR4 antibody treatment in mice was long-term tolerance induction.³² A murine TLR4 agonist antibody was previously shown to induce classic LPS tolerance.^32,33 We have shown that treatment with UT-18 can reverse acute T1D in NOD mice at least partly via the generation of myeloid-derived suppressor cells.^1,3 Tsukamoto et al.³⁴ described suppression of antigen (OVA)-specific T-cell responses that is mediated by myeloid cells activated by UT-12. In a first step toward testing tolerance induction by human agonistic anti-TLR4 antibodies, we performed a classic LPS tolerance assay in which primary human monocytes were primed with an initial stimulus (LPS, IgG3 8H1, IgG3 TV-7D11, or Media-only control), rested for 6 d to allow a return to baseline, and then all samples were re-challenged with LPS.³⁵ Tolerized cells should show a decrease in the production of TNF-α upon rechallenge due to epigenetic changes that limit inflammatory gene accessibility.²⁶ In Figure 8a we show a strong priming response from LPS for TNF-α production at 24 h while IgG3 TV-7D11 showed only a slight increase over baseline (and none from IgG3 8H1), but much lower compared to LPS. After 6 d, all the samples returned to undetectable levels of TNF-α (Figure 8b). When re-challenged with LPS, both the positive control (LPS primed) and IgG3 TV-7D11 primed demonstrate significant induction of LPS tolerance with inhibition of TNF-α production (Figure 8c). By contrast, IgG3 8H1 had reduced, but non-significantly reduced, TNF-α production, which likely reflects its lower stimulatory capacity compared to IgG3 TV 7D11 (Figure 7). The same assay with IgG4 antibodies showed no tolerance induction (data not shown). These experiments identify IgG3 TV-7D11 as a functional agonist in modulating primary human myeloid cells. The decreased initial cytokine release induced by IgG3 TV-7D11 compared to LPS might be a therapeutic advantage for in vivo tolerance induction.

Figure 8.

IgG3 TV-7D11 induces tolerance in primary human myeloid cells. Human CD14+ monocytes were purified by magnetic bead negative selection. 1 × 10^6 cells/well were seeded in 24-well plates and stimulated with IgG3 8H1 (10 μg/ml), IgG3 TV-7D11 (10 μg/ml), or LPS (.01 μg/ml). After 24 h supernatants were removed and tested for TNF-α by ELISA (a), followed by addition of fresh media and culture for 5 d to allow cells to return to a resting state (b). On day 6, cells were split into 96-well plates at 50,000 cells/well and allowed to rest for 24 h prior to re-challenge with LPS at .01 μg/ml. Supernatants were again collected at 24 h and assayed for TNF-α production by ELISA (c). Data points represent three individuals from two independent experiments. Statistical analysis by ANOVA *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Discussion

TLR4/MD-2 activation by LPS requires two key steps: 1) Stabilization and conformational changes of the TLR4/MD2 heterodimer to bring the two intracellular TIR domains close enough to form a MyD88 docking site and 2) binding of MyD88 followed by the formation of a large, stable complex of multiple MyD88-IRAK4-IRAK1/2 that make up the MyDDosome. The size and complexity of this MyDDosome complex determines the strength and duration of the NFKB activation signal.¹¹ However, aggregated protein ligands such as β-amyloid fibrils or α-synuclein aggregates can also activate TLR4 signaling (with distinct kinetics compared to LPS), although little is known about the mechanistic details.^14–16 It is tempting to speculate that these ligands may induce higher-order clustering of TLR4 reminiscent of TLR3 clustering by a dsRNA helix.³⁶ Our data here address similar questions about how TLR4 is activated by anti-TLR4 specific antibodies of the IgG3 subclass.

We show that the IgG3 subclass is critical to efficient TLR4/MD2 agonism, as summarized in Table 1. The IgG3 subclass has two critical traits that we suggest are responsible for its unique TLR4/MD2 agonizing potential, F(ab’)2 arm flexibility and oligomerization through hexameric platform assembly, as recently described by Abendstein et al.²³ Hinge flexibility allows for IgG3 F(ab’)2 arms to more easily bind bivalently to closely spaced antigens (as close as 3.8 nm apart) of each other, whereas the other more rigid IgG subclasses show a preference for antigen spacing between 7 and 16 nm apart.^37–40 The distance between the Phe₁₂₆ residues in the MD2 activation loops within an MD2 heterotetramer is 5.4 nm (PDB 3FXI⁴¹), making IgG3 better able to bivalently bind two heterodimers, stabilizing the heterotetramer, whereas the other isotypes are less efficient at bringing together the two heterodimers due to their more rigid F(ab’)2 arm spacing. Secondly, IgG3 bound to a multivalent antigen can form a hexameric platform in which divalent F(ab’)2s associate into a tightly bound array, thereby forming a cluster of TLR4/MD2 heterotetramers that could nucleate and stabilize MyDDosome formation.²³ In parallel, upon internalization, a higher-order cluster of TLR4/MD2 heterotetramers would presumably also be capable of nucleating and stabilizing Triffosome formation in the endosome.

Table 1.

Summary of HuTLR4Ab activity in THP-1 Dual reporter cells. Green cells show conditions allowing TLR4 activation, red cells show conditions blocking TLR4 activation.

	HuTLR4Ab activity in THP-1 Dual reporter cells
	Alone	w/IgG3 Crosslinker	w/non-IgG3 Crosslinker
IgG4 8H1	No	Yes	No
IgG3 8H1	Yes	n/a	Blocks
IgG4 7D11	No	No	n/a
IgG3 7D11	No	n/a	No
IgG4 TV 7D11	No	Yes	n/a
IgG3 TV 7D11	Yes	n/a	Blocks
IgG4 8G10	No	No	n/a
IgG4 3D7	No	No	n/a

The argument for higher order clustering of the TLR4/MD2 heterotetramer is supported by Figure 4 showing that human IgG4 8H1 is unable to signal unless crosslinked with a mouse IgG3 antibody. Although an IgG4 antibody can still bind bivalently to closely spaced epitopes, such as the two sites within a TLR4/MD2 heterotetramer, it would be incapable of higher-order clustering unless it was crosslinked via an oligomerization-competent IgG3 secondary. Swapping IgG4 8H1 to the IgG3 subclass directly allows for Fc-mediated oligomerization and the resulting higher-order clustering of TLR4/MD2 heterotetramers. The human antibody 7D11 differs from the other antibodies in that it only binds to TLR4/MD2 or MD2 alone, not TLR4 alone. This signifies a binding site away from the TLR4/MD2 interaction domain. Both IgG4, IgG4 crosslinked, and IgG3 7D11 were unable to activate TLR4/MD2. However, the engineering of a tetravalent form of 7D11 in which an additional 7D11-scFv chain was added to the N-terminus of its heavy chain, thereby extending the reach of the antigen binding portion of the antibody, activated signaling. Interestingly, this was only achieved in an IgG3 TV-7D11, not an IgG4 TV-7D11, indicating that the Fc-mediated oligomerization by the IgG3 Fc region was still important to function. These data strongly suggest that an agonistic TLR4/MD2 antibody does not necessarily need to bind close enough to the Phe₁₂₆ activation loop to mimic LPS, but rather primarily needs to support bivalency to stabilize the TLR4/MD2 heterotetramer complex coupled with higher-order clustering via Fc-mediated oligomerization (Overall signaling pathway summarized in Figure 9).

Figure 9.

IgG3 isotype mediated dimerization and oligomerization of TLR4/MD2 are crucial for HuTLR4Ab activation of TLR4 signaling. a) IgG4 8H1 is unable to dimerize TLR4 due to rigid F(ab) arm angle and the distance between the F(ab) heads (12–16 nm). Addition of mouse IgG3 anti-IgG4 allows for dimerization and oligomerization through Fc interactions, aggregating TLR4 receptors, and stabilizing the MyDDosome to strengthen signaling. b) changing 8H1 to IgG3 isotype allows for signaling through dimerization and oligomerization without crosslinking due to the enhanced flexibility permitting shorter F(ab) head distance (as low as 3 nm). The bound IgG3s subsequently oligomerize. In contrast, crosslinking the IgG3 with IgG1 or IgG2a inhibits signaling by disrupting IgG3 mediated oligomerization, leaving only dimerization, resulting in a weaker signal. c) neither IgG4 nor IgG3 7D11 (an MD2 specific HuTLR4Ab) could activate signaling. Adding scFv fragment increased the reach and/or flexibility of 7D11 and enabled bivalent attachment to dimerize TLR4/MD2 again, only the IgG3 version of TV 7D11 was able to activate TLR4/MD2 and crosslinking with IgG1 or IgG2a decreased signaling.

The crucial requirement for higher-order clustering of TLR4/MD2 via Fc-mediated IgG3 oligomerization is supported by other aspects of our data. When a mouse-IgG1 or rat-IgG2a crosslinking antibody is used on IgG3 8H1 or IgG3 TV-7D11, receptor activity is reduced (opposite of crosslinked IgG4 8H1). Since the IgG1 or IgG2a secondary antibodies bind to the IgG3 Fc region, they would be likely to sterically inhibit, at least in part, the formation of the hexameric platform by the IgG3 Fc region (see Figure 9). Thus, crosslinking in this situation results in a reduction, but not elimination, of TLR4/MD2 activity. Interestingly, the maximal activity of mouse UT18 (Figure 2) and human IgG3 8H1 and IgG3 TV-7D11 was lower than that of LPS in all cells tested. This is similar to the case of β-amyloid fibrils, which induce TLR4 signaling weaker than LPS.¹⁵ Li et al. demonstrated that, compared to LPS, β-amyloid fibrils trigger the formation of larger MyDDosomes, but with slower kinetics.¹⁴ Likewise, α-synuclein oligomers and fibrils activate TLR4 signaling at lower levels and with slower kinetics compared to LPS.^14–16 We hypothesize IgG3 anti-TLR4 antibodies and aggregated proteins such as β-amyloid or α-synuclein may all share a similar mechanism of inducing signaling via higher-order clustering of TLR4 compared to LPS. The formation of larger clusters may also be involved in the mechanism of TRIF-mediated IRF3 pathway signaling.

A limitation of this study is the incomplete examination of the full repertoire of not only IgG isotypes but others as well. We have identified that clustering of TLR4 by antibody is important to agonist activity and that IgG3 can accomplish this. However, it has not yet been determined if IgG3 huTLR4Abs can still perform this function in-vivo and leaves to question the developability of this monoclonal antibody strategy for therapeutic interventions. Potential alternatives to the IgG3 isotype that could aggregate TLR4 may include pentameric IgM, as well as engineered HexaBodies or Stellabodies,^42–44 and antibody conjugate-based nanoparticles.^45,46 Notably, while our control mouse antibody (UT15) binds to TLR4, it is IgG1 and does not signal. UT-15 is also proposed to bind at a different site on TLR4/MD2 than UT-18, potentially at the interaction domain, while UT-18 binds closer to the dimerization domain.³³ Switching the respective isotypes on both UT-18 and UT-15 in future experiments, as well as full epitope mapping of both mouse and human TLR4/MD2 antibodies will greatly increase our understanding of the factors responsible for agonist activity and improve the effectiveness and developability of anti-TLR4/MD2 monoclonal therapy.

We have previously published the remarkable effect of UT-18 on the reversal of T1D in NOD mice. Further investigation revealed that the mechanism behind this reversal was through the generation of MDSC. Similarly, Tsukamoto et al. showed antigen-specific suppression of OVA-positive OT-II T-cells when mice were treated with a similar IgG3 TLR4/MD2 agonist (UT-12) before immunization, ascribed to UT-12-mediated induction of MDSC.³⁴ MDSCs are generated in response to chronic, low-grade inflammatory conditions, as first described in cancer.⁴⁷ Pure LPS is a poor inducer of MDSC as it is cleared rapidly from circulation (within minutes) and is unable to effectively/chronically stimulate TLR4/MD2 at necessary concentrations in that short timeframe. Furthermore, there is a risk of endotoxic shock in an outbred human population with enhanced LPS sensitivities.⁴⁸ IgG3 antibody-based TLR4/MD2 agonism, on the other hand, provides a low-grade inflammatory response that is only limited by the half-life of IgG3 and the rate of uptake by the receptor, providing chronic stimulation for MDSC generation with a very minimal initial cytokine release (Figure 7). These qualities and proven effectiveness in reversing T1D in mice, a chronic inflammatory condition, make our novel human IgG3 TLR4/MD2 agonist antibodies a potentially viable therapeutic strategy for the treatment of human T1D and suppression of myeloid-based inflammatory conditions.

Materials and methods

Enzymatic digestion of mouse anti-TLR4 antibody (UT-18) to F(ab’)2 and F(ab) subunits

UT-18 hybridoma was kindly provided by Hiroki Tsukamoto³³ and grown in serum free PFHM-II media + L-Glutamine. Supernatant was harvested and UT-18 purified by anion exchange chromatography. Purified UT-18 was then digested using pepsin F(ab’)2 and papain F(ab) preparation kits followed by cleanup up with Protein A to remove Fc fragments according to manufacturer’s instructions (Pierce, #44688 and 44,985). Purified preparations (including full-length UT-18) were then assayed for purity by SDS-PAGE and for LPS contamination using a chromogenic endotoxin quant kit (Pierce, #A39552).

Determination of IgG3 Fc contributions to mouse anti-TLR4 antibody (UT-18) using reporter cells

TLR4 agonist activity of F(ab’)2, F(ab), and FL UT-18 were tested using HEK-Blue™ mTLR4 reporter cells that stably express mouse TLR4, MD2, and CD14, as well as SEAP (secreted embryonic alkaline phosphatase) under control of NFkB (Invivogen, #hkb-mtlr4). Cells were seeded at 120,000 cells per well in a 24-well plate and stimulated with F(ab’)2, F(ab), or FL UT-18 at the indicated concentrations in 1 ml of media for 24 h at which point supernatants were collected and assayed for alkaline phosphatase activity using Quantiblue solution according to the manufacturer’s instructions (Invivogen, #rep-qbs). The importance of the Fc portion of UT-18 was also tested by immobilization to a protein G-coated 96-well plate (Pierce, #15131) for 1 hr, followed by washing 3× with phosphate-buffered saline (PBS), and then the addition of 100,000 HEK-Blue mTLR4 reporter cells. Supernatants were collected at 24 h and assayed for TLR4 activity using Quantiblue solution. Results were normalized using the formula (treatment group OD) minus (PBS-treated control OD).

Determination of F(ab’)2, F(ab), FL, and protein G-immobilized UT-18 on primary mouse macrophages

Healthy NOD mice with blood glucose < 200 mg/dL were injected IP with 1 ml of 3% Thioglycollate w/v (Sigma, #T9032). On the fourth day post injection mice were euthanized and macrophages collected by peritoneal lavage. Macrophages were seeded 100,000 per well in a 96-well plate and tested for activity with the indicated concentrations of UT-18 F(ab), F(ab’)2, FL, or protein G immobilized identical to experiments in the HEK-Blue reporter cells. Supernatants were collected at 18 h and assayed for TNF-α by ELISA (R&D Systems, #DY410–05).

Generation of human anti-TLR4 antibodies

Vik Biotinylated, recombinant Human TLR4-MD2 complex (R&D Systems, #3146-TM-050), was immobilized on streptavidin beads and incubated with human F(ab) phage library. After three rounds of phage panning, ~800 clones screened, ~20 positive clones were identified, resulting in six unique sequences. Antigen binding of four clones was confirmed with purified F(ab), which was then produced as FL IgG4. ExpiCHO and Expi293 cells were grown and maintained in culture at 37°C, 8% CO₂, and 125 rpm for 7–10 d prior to transfection. Once the cells had sustained a viability of >90% with a consistent doubling time of 20–24 h, transfections were performed following the ExpiCHO or Expi293 Transfection System standard protocols (ThermoFisher Scientific, #A29133 and #A14635). A standard amount of 0.75–1 μg/ml of plasmid DNA was used for each transfection.

The supernatant was collected by pelleting the cells at room temperature and 1800 RCF for 10 min. The Protein A column was equilibrated with binding buffer, PBS pH7.4. A ratio of 0.25 mL of Protein A resin per 100 mL of sample was used. The clarified supernatant was applied to the column and washed with binding buffer, 30× column volumes. Antibodies were eluted with 0.1 M glycine pH 2.5. The elute was immediately neutralized with 1 M Tris-HCl pH 9.0 and concentrated with a buffer change to PBS using a Millipore Ultra filter (Millipore Sigma, #UFC9010). Endotoxins were removed with the Pierce High-Capacity Endotoxin Removal Spin Column as per manufacturer direction (Pierce, A39552). Cleanup was repeated until a level of 0.1 EU/mg was achieved. Residual endotoxin was quantified using a Chromogenic Endotoxin Quant Kit (Pierce, #A39552). Quality of the protein was assessed via SDS-PAGE and Western blot.

The binding of TLR4/MD2, TLR4, or MD2 was characterized by ELISA. Clones of interest were produced as IgG3 Isotype to test Fc functionality. We further generated tetravalent versions of 7D11 in which two additional single-chain fragment variables (scFv) are attached to the N-terminal variable region to create an antibody with four identical variable regions.

Measuring HuTLR4Ab activity with THP-1 Dual reporter cells

THP1-Dual™ MD2-CD14-TLR4 cells (Invivogen, #thpd-mctlr4) stably express MD-2 and CD14, as well as overexpress TLR4. In addition, these cells express two inducible reporter genes: SEAP under the control of NFkB, and Lucia luciferase under the control of IRSE. Cells were seeded at 100,000 cells/well in a flat bottom 96-well plate and then stimulated with HuTLR4Abs or LPS as a positive control at the indicated concentrations. After 48 h, supernatants were collected and assayed for SEAP using QUANTI-Blue Solution or Lucia Luciferase using QUANTI-Luc Reagent 4 according to the manufacturer’s instruction (Invivogen, #rep-qbs and #rep-qlc4lg2, respectively). The activity of NFkB (SEAP) and IRF3 (Lucia luciferase) were measured by optical density at 650 nm and relative luminescence units, respectively.

Induction of tolerance in human monocytes

Human CD14+ monocytes were purified by magnetic bead negative selection (Miltenyi, #130–117–337). 1 × 10^6 cells/well were seeded in 24-well plates and stimulated with IgG3 8H1 (10 μg/ml), IgG3 TV-7D11 (10 μg/ml), and LPS (0.01 μg/ml). After 24 h supernatants were removed and tested for TNF-α by ELISA (R&D Systems, #DY210–05), followed by the addition of fresh media and culture for 5 d to allow cells to return to a resting state. On day 6, cells were split into 96-well plates at 50,000 cells/well and allowed to rest for 24 h prior to re-challenge with LPS at 0.01 μg/ml. Supernatants were again collected at 24 h and assayed for TNF-α production by ELISA (R&D Systems, #DY210–05).

Statistical Analysis: Statistical analysis was performed using GraphPad Prism v10. Figure 1 used a 2-way ANOVA to evaluate the main effects of treatment group on NFkB activity with Tukey’s post hoc multiple comparisons. Comparisons of multiple treatment groups to a single control group (media) using one-way ANOVA with Dunnett’s multiple comparisons test are shown in Figures 2, 4, 5, 6, and 8. Figure 7 uses one-way ANOVA with Dunnett’s multiple comparisons test as well, but by using non-crosslinked HuTLR4Ab as control with mouse IgG1 and rat IgG2a crosslinked as treatment groups. Comparisons between maximal stimulation of UT-18 and LPS were done by independent T-test, while comparisons of maximal stimulation between UT-18 fragments and FL UT-18 were done by one-way ANOVA and Dunnett’s multiple comparisons with FL UT-18 as the control. Statistical p-values are reported in all graphs as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Funding Statement

National Institutes of Health, Grant/Award Number: [1R01DK136815-01]; National Institute of Diabetes and Digestive and Kidney Diseases.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Compliance and ethics in mouse and human subjects

All studies in NOD mice were performed with the prior approval of the UC Davis Institutional Animal Care and Use Committee (IACUC). Human subjects were recruited and provided with a written informed consent form that was signed and dated prior to peripheral blood collection under authorization of the UC Davis Institutional Review Board (IRB ID: 1723319–1).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2025.2515415