Report

The Vaccine Adjuvant Monophosphoryl Lipid A as a TRIF-Biased Agonist of TLR4

See allHide authors and affiliations

Science  15 Jun 2007:
Vol. 316, Issue 5831, pp. 1628-1632
DOI: 10.1126/science.1138963

This article has a correction. Please see:

Abstract

The inflammatory toxicity of lipopolysaccharide (LPS), a component of bacterial cell walls, is driven by the adaptor proteins myeloid differentiation factor 88 (MyD88) and Toll-interleukin 1 receptor domain–containing adapter inducing interferon-β (TRIF), which together mediate signaling by the endotoxin receptor Toll-like receptor 4 (TLR4). Monophosphoryl lipid A (MPLA) is a low-toxicity derivative of LPS with useful immunostimulatory properties, which is nearing regulatory approval for use as a human vaccine adjuvant. We report here that, in mice, the low toxicity of MPLA's adjuvant function is associated with a bias toward TRIF signaling, which we suggest is likely caused by the active suppression, rather than passive loss, of proinflammatory activity of this LPS derivative. This finding may have important implications for the development of future vaccine adjuvants.

Immunological adjuvants are combined with noninfectious vaccine antigens to generate antibody responses that are faster, stronger, and longer lasting than responses achieved through immunization with antigen alone. Aluminum hydroxide (alum) is currently the only human vaccine adjuvant approved for use in the United States, and although it is effective at boosting antibody responses, these responses require repeated administration and tend to generate antiparasitic T helper 2 (TH2), rather than antiviral and antibacterial TH1, T cell immunity (1). As a consequence, there is much effort devoted to developing prospective adjuvants that can establish protective immunity with fewer vaccinations with less injected material, through durable antibody and TH1-dependent cytotoxic T cell activity. One of these, MPLA, is likely to be the first vaccine adjuvant to be approved for widespread use since alum because it generates clinically useful immune responses (25), and it has ∼0.1% of the inflammatory toxicity of its parent molecule, LPS (6, 7). The adjuvant effects of MPLA require TLR4 (6, 8), and although TLR signaling may not be critical for enhanced antibody responses under all conditions (9), TLR agonists do show particular promise as adjuvants of cytotoxic T cell activity. Because TLR4 is also the receptor for endotoxin, it is important to understand the mechanism(s) by which MPLA can boost T and B cell immunity, without the damaging inflammatory outcomes associated with its parent molecule.

We compared the effects of MPLA versus LPS in T cell priming using an adoptive transfer system in which ovalbumin (OVA) peptide–specific CD8+ and CD4+ T cells from OT-I (10) and OT-II (11) T cell receptor transgenic mice (C57BL/6 background, CD45.2+) were infused into major histocompatibility complex–matched recipients (B6.SJL, CD45.1+) before immunization with adjuvant plus OVA peptides 323-339 and 257-264 [SIINFEKL (Ser-Ile-Ile-Asn-Phe-Glu-Lys-Leu)] (12). Adjuvant doses, 30 μg of MPLA or 10 μg of LPS, were selected on the basis of similar induction of T cell clonal expansion in pilot experiments (fig. S1) and were used in all subsequent in vivo experiments. From 2.5 to 7 days after immunization, cells harvested from spleens and lymph nodes of treated animals revealed that MPLA and LPS had equivalent adjuvant effects on T cells, with indistinguishable patterns of clonal expansion and contraction (Fig. 1A and fig. S2). In terms of cytokine responses, interleukin-10 (IL-10) production was strong in both MPLA- and LPS-adjuvanted mice (fig. S3), although differences emerged when responses were grouped according to dependence on either of two TLR4 signaling pathways, MyD88 or TRIF, based on previous studies of genetically deficient mice (1315). Thus, MPLA appeared to be as efficient as LPS at inducing TRIF-dependent factors [Fig. 1C; P = 0.604, 0.051, and 0.058 for differences in granulocyte colony-stimulating factor (G-CSF), interferon-induced protein 10 (IP-10), and monocyte chemotactic protein 1 (MCP-1) production respectively] while only weakly stimulating MyD88-associated responses [Fig. 1B; P < 0.0005 for interferon-γ (IFN-γ), IL-1β, IL-6, and macrophage inflammatory protein 1α (MIP-1α)].

Fig. 1.

MPLA induces similar T cell clonal expansion kinetics as compared to LPS but shows TRIF-biased gene expression. In all experiments, OT-I (1 × 105) and OT-II (1.5 × 105) cells were adoptively transferred into B6.SJL mice. After 24 hours, mice were either untreated or immunized with SIINFEKL and OVA323–339 alone (OVA; triangles), with OVA and 10 μg of LPS (OVA+LPS; solid squares), or with OVA and 30 μg of MPLA (OVA+MPLA; open squares). *P < 0.05; n.s., not statistically significant by analysis of variance (ANOVA) and post-hoc Tukey analysis (12). (A) Every day after immunization (from days 2 to 7), cells from spleen and lymph nodes (fig. S2) were harvested; stained for CD4, CD8, CD45.1, and CD45.2; and enumerated by means of flow cytometry. Results are mean values ± SEM of triplicate mice from one of two representative experiments. (B and C) Serum samples were obtained at the indicated times and analyzed by means of multiplex analysis for cytokines and chemokines. Data were grouped as representative products of the MyD88-dependent (B) or TRIF-dependent (C) signaling pathways and plotted as mean values ± SEM. (D) Six hours after immunization, spleens were harvested and RNA was isolated. Affymetrix genechip analysis was then performed (12). Selected transcript products from the microarray data were grouped by MyD88-dependent (left) or TRIF-dependent (right) genes. COX-2, cyclooxygenase-2; Serpine1, serine protease inhibitor E1; CXCL1, chemokine (CXC motif) ligand 1; Ifit interferon-induced protein with tetratricopeptide repeats. Results are shown as mean hybridization intensity ± SEM from triplicate mice.

We next performed microarray analysis of splenocytes from mice that had been immunized as described above (Fig. 1A), which again showed that MPLA had induced strong TRIF-associated but weak MyD88-associated responses when compared with those of LPS. Some of the results from this in vivo analysis of MPLA function reproduced those previously reported (8) in that IL-1β transcription was strongly induced (fig. S4), whereas mature protein secretion was not (Fig. 1B). In the earlier study, it appeared that MPLA had failed to activate IL-1β converting enzyme (8). However, because IL-1 receptor–deficient mice remain susceptible to endotoxic shock (16), we concluded that this difference in activity was not sufficient to explain the low toxicity of MPLA.

Given the number of MyD88-associated genes that were not strongly induced by MPLA (Fig. 1, B to D), we next tested for generalized defects in MyD88 signaling. Because macrophages are necessary for LPS endotoxicity (17), we tested responses to LPS versus MPLA using bone marrow–derived monocytes (BMDMs) and thioglycollate-elicited macrophages (TGMs). Initial tests of cytokine and chemokine production showed that MPLA had the same TRIF-biased activity in BMDMs and TGMs (Fig. 2) as that seen in vivo. Thus, MPLAwas two to three orders of magnitude less potent than LPS at inducing MyD88-dependent IL-6 (Fig. 2A) but induced similar TRIF-dependent IP-10, MCP-1, and IFN-β (Fig. 3A). Stimulation of DNA binding activity by the transcription factor nuclear factor κB (NFκB) p65 was markedly slower (Fig. 2C) after exposure of wild-type (WT) BMDMs to MPLA, and phosphorylation of a component of its regulatory complex, inhibitor of NFκB kinase (IKK), was delayed and reduced when compared to phosphorylation after LPS exposure (Fig. 2B). In contrast, MyD88–/– BMDMs showed identical kinetics of IKK phosphorylation upon exposure to LPS or MPLA (fig. S5). These results reveal that MPLA's low potency in inducing IL-6 was correlated with a failure to activate MyD88-dependent events needed for proinflammatory patterns of NFκB transcriptional activation and were similar to those seen in MyD88–/– mice (13, 14, 18).

Fig. 2.

MyD88 signaling is delayed and decreased in MPLA-stimulated macrophages. *P < 0.05; n.s., not statistically significant by ANOVA. (A) BMDMs or peritoneal TGMs were cultured overnight with increasing concentrations of MPLA or LPS. Culture supernatants were assayed for IL-6 by enzyme-linked immunosorbent assay (ELISA) (12). Results are mean values ± SEM of triplicate cultures from one of two representative experiments. (B) Immunoblots of phosphorylated IKKα/β in unstimulated BMDMs and BMDMs stimulated with MPLA (1 μg ml–1) or LPS (1 μg ml–1). The membranes were stripped and reprobed for total IKKα/β as loading control, and ratios of phosphorylated/total IKKα/β were plotted. Results are from one of three representative experiments. (C) BMDMs were cultured for the indicated times with MPLA (1 μg ml–1) or LPS (1 μg ml–1). Nuclear extracts were analyzed for NFκB-binding activity; data are expressed in optical density (O.D.) units obtained with the TransAM ELISA NFκB assay performed as described in (12). (D) BMDMs were pretreated with dimethyl sulfoxide (DMSO) or with 50 nM freshly prepared wortmannin and then were cultured overnight with increasing concentrations of MPLA or LPS. Culture supernatants were assayed for IL-6 by ELISA (12). Results are mean values ± SEM of triplicate cultures from one of three representative experiments.

Fig. 3.

Equivalent TRIF-dependent signaling in response to MPLA or LPS. *P < 0.05; n.s., not statistically significant by ANOVA. (A) BMDMs or TGMs were cultured overnight with increasing concentrations of MPLA or LPS, and the culture supernatants were assayed by ELISA for IP-10, MCP-1, and interferon-β (IFN-β) (12). Results are mean values ± SEM from one of two representative experiments. (B and C) Unstimulated BMDMs or BMDMs stimulated with MPLA (1 μg ml–1) or LPS (1 μg ml–1) were analyzed by Western blot for phosphorylated IRF-3 (Ser396) (B) or phosphorylated Stat1 (Tyr701 or Y-701) (C). For loading controls, the membranes were stripped and reprobed for total IRF-3 or total Stat1, respectively. Results are representative of two independent experiments. Parameters were evaluated by ANOVA.

In studies looking at TRIF-associated signaling, several events were found to be unimpaired in BMDMs after exposure to MPLA. The time course and magnitude of phosphorylation of interferon regulatory factor 3 (IRF-3), a defining feature of TRIF-mediated signaling, were identical in both MPLA- and LPS-treated BMDMs (Fig. 3B). Signal transducer and activator of transcription 1 (Stat1) phosphorylation, which is absent in TRIF-deficient cells exposed to LPS (18) and is associated with TLR4-induced autocrine and/or paracrine production of type I interferons (19), was also the same (Fig. 3C).

It has recently been reported that phosphoinositide 5-kinase–generated phosphatidylinositol 4,5-bisphosphate (PIP2) is required for efficient MyD88 recruitment to TLR4 (20) and that proinflammatory glycogen synthase kinase 3β (GSK-3β) is inactivated through phosphoinositide 3-kinase (PI3K)–dependent protein kinase B activity (21). Because PI3K activity would be expected to decrease PIP2 levels via conversion to phosphatidylinositol 3,4,5-trisphosphate, thereby preventing recruitment of MyD88, as well as to inactivate GSK-3β, we tested whether PI3K activity might be involved in MPLA's decreased ability to induce IL-6. Pretreatment of BMDMs with the PI3K inhibitor wortmannin increased IL-6 production by MPLA but did not further increase IL-6 production by LPS (Fig. 2D), suggesting that MPLA stimulates more PI3K activity than LPS. Thus, the inefficient stimulation of MyD88-dependent signaling by MPLA may result from diminished recruitment of MyD88 to TLR4 through loss of PIP2 species and/or by PI3K-dependent inactivation of proinflammatory GSK-3β.

Our observation that weak stimulation of MyD88-dependent IL-6 by MPLA was correlated with efficient T cell adjuvant function appears to contradict a previous report concluding that LPS failed to boost T cell priming in MyD88-deficient mice because IL-6 expression was impaired (22). One explanation for this discrepancy might be that MPLA induces low levels of MyD88-mediated signaling that stimulate sufficient production of IL-6 (Fig. 1B). Another explanation is that TRIF, rather than MyD88, may be needed for TLR4-induced adjuvant effects. To distinguish between these possibilities, we returned to the adoptive transfer system to measure T cell priming effects in recipient mice that were either MyD88- or TRIF-deficient. The TLR2 agonist N-palmitoyl-S-[2,3-bis(palmitoyloxy)-(2R,S)-propyl]-(R)-cysteinyl-seryl-(lysyl) (3)-lysine (PAM3CSK4), which signals only through MyD88 (23), was used as negative control to demonstrate the absence of MyD88-dependent effects in MyD88-deficient recipients. Four days after immunization, spleens were harvested and tested for adjuvant effects on T cell proliferation. In these experiments, MyD88 appeared completely dispensable for the MPLA- or LPS-induced adjuvant effects on OT-I and OT-II cells, at least during early clonal expansion (Fig. 4A). Neither T cell population showed evidence of PAM3CSK4-mediated adjuvant effects in MyD88–/– mice. These results confirm that MyD88 was not required for TLR4-mediated adjuvant effects on early T cell priming.

Fig. 4.

MyD88 is not required for full adjuvant effects on T cell priming. WT (C57BL/6), MyD88-deficient (MyD88–/–) (A), or TRIFLps2/Lps2 (B) mutant mice received 1 × 105 OT-I.SJL and 1.5 × 105 OT-II.SJL cells by adoptive transfer and were treated as in Fig. 1A with the combined OVA peptides in the absence and presence of LPS (10 μg), MPLA (30 μg), or PAM3CSK4 (PAM) (50 μg). T cell expansion in the spleens was determined after four days. Results are mean values ± SEM from triplicate mice for fold increases of OVA plus adjuvant-treated mice versus OVA-treated mice and are representative of two experiments. Statistical analysis was performed with the use of the nonparametric Mann-Whitney test; all comparisons between OVA+MPLA and OVA+LPS groups showed statistically insignificant differences (U >0.05).

The adjuvant effects of MPLA versus LPS were also compared in TRIFLps/Lps2 mice, which express a truncated form of TRIF that is inactive (18). These studies revealed that adjuvant-boosted OT-I T cell priming was severely impaired in TRIFLps2/Lps2 hosts, whereas OT-II priming was less strongly affected (Fig. 4B). Responses of TRIFLps2/Lps2 mice to control PAM3CSK4 were not diminished, as was expected because TLR2 is known not to require TRIF for its activity (23). Hence, costimulation of OT-I and OT-II T cell proliferation by TLR4 agonists showed substantially greater dependence on TRIF than on MyD88 in this model system. This result indicates that although MPLA is inefficient with respect to stimulation of TLR4/MyD88-induced gene expression, it has fortuitously retained TLR4/TRIF-associated activities, such as induction of type I interferon (Fig. 3A), that may be especially important for T cell clonal expansion (24).

Clinically relevant adjuvants other than MPLA, such as CpG oligonucleotides, also induce type I interferon but in a MyD88-dependent (25), rather than a TRIF-dependent, manner. Thus, different TLRs can reach the same immunostimulatory endpoints without requiring the same signaling adaptors. In the case of the endotoxin receptor TLR4, it is MyD88 that is most associated with proinflammatory outcomes, perhaps because of involvement of the MyD88 coadapter, Mal (MyD88 adapter–like) (26, 27). The mechanistic basis for MPLA's failure to stimulate full MyD88 signaling, with associated proinflammatory effects, remains to be fully elucidated. However, we propose that an important component is the acquisition, relative to LPS, of an anti-inflammatory function that requires PI3K activity (Fig. 2D). This would be distinct from the loss of a proinflammatory activity, such as a simple failure to recruit Mal and/or MyD88 to TLR4. Discovering the precise mechanism of MPLA's ability to function as a low-toxicity adjuvant may permit improvements in the design of future TLR4-dependent vaccine adjuvants.

Supporting Online Material

www.sciencemag.org/cgi/content/full/316/5831/1628/DC1

Materials and Methods

Figs. S1 to S5

References

References and Notes

View Abstract

Navigate This Article