Toll-Like Receptors
Evolutionary relationships, but functional differences, between the Drosophila and human Toll-like receptor families M. Gangloff1 , A.N.R. Weber1 , R.J. Gibbard and N.J. Gay2 Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, U.K.
Abstract The Toll receptor was first found to function in the dorsoventral patterning pathway of Drosophila embryos. ¨ It is activated by a specific protein ligand, Spatzle, generated at ventral positions in the early embryo. Drosophila Toll (dToll) also functions in innate immune responses to Gram-positive bacteria and fungi, and ¨ ¨ Spatzle is required for this response. We have shown that Spatzle is necessary and sufficient for activation of the dToll pathway, and that it probably acts by cross-linking two molecules of Toll to form homodimers. In the present paper, we contrast this mode of regulation with that proposed for the vertebrate Toll-like receptor family, which mediate analogous responses to pathogen pattern antigens. In contrast with dToll, these receptors appear to be activated by direct exposure to pathogen patterns, such as peptidoglycan and lipopolysaccharide. We discuss the evolutionary basis of this functional divergence of the vertebrate and invertebrate Toll-like receptors.
Introduction The phenotypes associated with mutations in the Drosophila Toll (dToll) gene were first described by Anderson, Jurgens and Nusslein-Volhard in the mid-1980s [1]. Recessive alleles of Toll cause embryos to lack a dorsoventral pattern, and dominant mutations have an opposite effect, imparting ventral fates to dorsal cells. At that time Toll, together with the other dorsal group genes, was viewed as a system of positional information capable of generating a diffused gradient of morphogen, consistent with the theoretical analysis of Wolpert and others dating from the 1960s [2]. It was also acknowledged that a proper understanding of dorsoventral pattern formation would require the characterization of the dorsal genes and their products. This analysis has now provided a compelling and coherent picture of this fundamental developmental process and has, to some extent, validated the hypothesis that patterning involves a diffused morphogen gradient [3]. The Toll gene was isolated and sequenced in 1988 [4], and proved to be a type 1 transmembrane receptor, suggesting that dorsoventral patterning involved spatially restricted signal transduction. In particular, the predicted ectodomain contained blocks of leucine-rich repeats (LRRs), 24-aminoacid motifs common to a superfamily of proteins, many of which bind specifically to other proteins and biological macromolecules [5]. Subsequent work identified the product Key words: Drosophila, Toll-like receptor (TLR), dToll. Abbreviations used: dToll, Drosophila Toll; LRR, leucine-rich repeat; IL(-1R), interleukin(1 receptor); NF-κB, nuclear factor κB; TIR, Toll/IL-1 receptor identity region; (d/h)TLR, (Drosophila/human) Toll-like receptor; LPS, lipopolysaccharide; Gp1b, platelet glycoprotein 1b; vWF, von Willebrand factor. 1 These authors contributed equally to this work. 2
To whom correspondence should be addressed (e-mail
[email protected]).
of another dorsoventral group gene, Sp¨atzle, as the likely activating ligand for the Toll receptor [6]. This work also suggested that an inactive precursor of Sp¨atzle was processed by the serine protease Easter to generate an active form only at ventral positions in the perivitelline space of the pre-cellular embryo [7]. A diffused gradient of active Sp¨atzle would then be the morphogen, establishing a continuum of fates along the embryonic dorsoventral axis. Easter is the terminal member of a protease cascade analogous to that of haemostasis, and allows the amplification of a ventral specific signal bound to the perivitelline membrane, which is a structure that overlies the plasma membrane of the embryo. Perhaps surprisingly, this elegant system of positional information is restricted to invertebrates, with higher eukaryotes apparently having a distinct cellular mechanism for specifying the dorsoventral axis. However, in the 1990s a different and unsuspected role emerged for Toll-like signalling pathways in both flies and vertebrates. First, it was found that the cytoplasmic signalling domain of Toll was conserved in the mammalian type 1 interleukin-1 receptor (IL-1R) [8,9]. IL-1R has an ectodomain with three β-trefoil motifs and signals in response to the cytokines IL-1α and -β, mediators of the acute-phase response, a part of the innate immune response to infection by bacterial and viral pathogens [10]. This finding was soon complemented by the discovery that, in Drosophila, the Toll receptor has a second function in innate immunity, and that a subset of the dorsoventral group genes, including Sp¨atzle, are also required for these responses [11]. Furthermore, components in the signalling pathway downstream of IL-1R clearly corresponded with those in the dorsoventral patterning system; for example, pelle kinase/IL-1R-associated kinase and dorsal/nuclear factor-κB (NF-κB) [12]. These findings C 2003
Biochemical Society
659
Biochemical Society Transactions (2003) Volume 31, part 3
¨ Proteolytically cleaved Spatzle binds directly to the Toll receptor, and is necessary and sufficient for signalling Although indirect evidence from genetic and modelling studies has suggested that Sp¨atzle is the activating ligand of Toll, direct binding has not been shown. Our group, in collaboration with that of Dr J.-L. Imler (at IBMC, Strasbourg, France), has now demonstrated that the mature form of Sp¨atzle activates the Toll receptor in cell culture and in vivo. Radiolabelled Sp¨atzle specifically binds to Drosophila cells, and to Cos-7 cells expressing Toll. Furthermore, in vitro experiments have shown that the mature form of Sp¨atzle binds to the Toll ectodomain (Figure 1) with high affinity (approx. 80 nM), and with a stoichiometry of one Sp¨atzle dimer to two receptors. The Sp¨atzle pro-protein is inactive in all these assays, indicating that the pro-domain sequence, which is natively unstructured, acts to prevent interaction of Sp¨atzle and its receptor. These studies show that cleaved Sp¨atzle directly binds to the Toll receptor, and that this interaction is sufficient to establish signalling. Our evidence is that homodimerization of Toll receptors by Sp¨atzle is sufficient for signal transduction: the classic mode of activation for class 1 transmembrane receptors. We also found that a type II dominant form of Toll (Toll 5B) [9], a truncated ectodomain, binds with similar affinity to Sp¨atzle as the fulllength ectodomain. Other type II dominant alleles encode C 2003
Biochemical Society
¨ Figure 1 Spatzle binds to dToll, but not to human TLR2 In surface plasmon resonance (Biacore) experiments, wild-type Toll, Toll5B and human TLR2 ectodomains were immobilized on a Biacore sensor chip, and sensorgrams were recorded upon injection of Spatzle ¨ proteins. Sensorgram (a): the Spatzle ¨ C-106 fragment shows binding to immobilized wild-type Toll (upper trace) and Toll5B (middle), but not the TLR2 ectodomain (lower curve). Sensorgram (b): the Spatzle ¨ pro-protein shows no binding to immobilized wild-type Toll, Toll5B or TLR2. Sensorgram (c): in a reciprocal experiment, Spatzle ¨ C-106 and proprotein were immobilized on the chip and a sensorgram was recorded upon injection of TollWT. TollWT binds to immobilized C-106 (upper curve), but not pro-protein (lower curve).
Relative response (RU)
150
a
100 50 0 0
250 Time (s)
500
150 Relative response (RU)
suggested that innate immunity might have a common and ancient evolutionary origin. This idea received strong support when a family of receptors with ectodomains containing LRRs as well as a cytoplasmic Toll/IL-1 receptor identity region (TIR) was identified in human expressed sequence tag (‘EST’) databases [13,14]. There are at least ten of these Tolllike receptors (TLRs) in the human genome (hTLRs), and most are expressed on cells of the immune system. Since the initial characterization of the vertebrate TLR family, work using a variety of approaches has shown that these receptors respond to pathogen pattern antigens derived not only from bacteria, but also from viruses [15,16]. For example, studies on a mouse strain hyposensitive to the Gram-negative outer membrane component lipopolysaccharide (LPS) showed that the TLR4 receptor was either deleted or had a dominantnegative mutation in the TIR domain of the receptor [17]. Similarly, knockout mice lacking TLR2 are unable to establish signalling in response to the Gram-positive pattern antigen peptidoglycan [18]. In view of the evolutionary relationships of the Drosophila and vertebrate Toll receptors, and the clear evidence of conservation in the intracellular pathway, it seemed reasonable to expect that the LRRs and associated motifs of Toll ectodomains from insects and mammals would detect pathogen patterns by similar mechanisms. In the present paper, however, we present evidence that the activation mechanism of the dToll receptor is quite different from that of the hTLRs, and has more in common with related cytokine receptors, such as IL-1R.
b
100 50 0 0
250 Time (s)
500
150 Relative response (RU)
660
c
100 50 0 0
250 Time (s)
500
even shorter truncations, and one of these (Tl84c ) leaves only part of the first block of LRRs intact. This suggests that these LRRs are critical for binding the protein ligand Sp¨atzle.
Evolutionary relationships between the Drosophila and human Toll-like receptors The genome sequence of Drosophila reveals a total of nine TLRs [19], and at least six Sp¨atzle proteins are also
Toll-Like Receptors
Figure 2 Evolutionary relationships of the Drosophila and human Toll-like receptors (A) Trees were generated using the program CLUSTAL W [30]. dToll≡dTLR. (B) Alignment of the first two LRRs in dTLRs and hTLRs. The proposed binding site of peptidoglycan in hTLR2 is shown.
encoded [20]. However, all the evidence to date suggests that only the original Toll receptor is involved in innate immunity. Null mutations in at least one other Drosophila TLR (dTLR) is embryonic lethal [21], and it seems likely that most, if not all, of the dTLRs function in embryogenesis. By contrast, none of the knockout mice lacking TLRs have embryonic defects [22]. In fact, each vertebrate TLR seems to be specialized for recognizing particular pathogen pattern antigens, and receptor activation by these patterns appears to be direct. For example, hTLR2 ectodomain binds to peptidoglycan extracted from Staphylococcus aureus [23]. Furthermore, the human genome does not contain any obvious homologues of Sp¨atzle, and thus it is unlikely that the
vertebrate TLRs are activated by dimerization induced by a Sp¨atzle-like protein ligand. As can be seen in Figure 2(a), with the exception of dTLR9, the dToll family members are more related to each other than they are to the hTLRs, regardless of whether the extracellular domain or the TIR is used in the alignment. Nevertheless, the LRR block implicated in binding of Sp¨atzle by the dTLR can be aligned with that of hTLR2 (Figure 2b), including residues 40–64, which are required for binding to peptidoglycan [24]. In Biacore experiments, we could not detect binding of Sp¨atzle and hTLR2 (Figure 1), and thus although clearly related to Toll, hTLR2 has probably lost the specific determinants in the LRRs that allow binding to Sp¨atzle. C 2003
Biochemical Society
661
662
Biochemical Society Transactions (2003) Volume 31, part 3
A number of questions arise from these observations:
How and when did the evolutionary disjunction occur between the Drosophila and vertebrate Tolls? It could be argued that dToll is more similar in function to the cytokine receptors such as IL-1R; that is, the pathway is activated by a protein ligand. In Drosophila, there are no IL1R-like genes, and so this receptor family must have evolved after the divergence of vertebrates and invertebrates, allowing the development of a two-layered innate immune response in vertebrates. The Toll family would then be free to evolve the repertoire of recognition and signalling responses to pathogen patterns that is seen today.
Figure 3 Structural modelling of LRRs from hTLR2 (A) The experimental structure of the Gp1b ectodomain (taken from [29]). (B) hTLR2 N-terminal region modelled on Gp1bα. The region shown includes a β-hairpin, which caps the Gp1b structure, and the first three units of the LRR block. The molecular surface represented is the solvent-accessible surface of the region crucial for peptidoglycan binding, as determined from mutagenesis data. The Figure was generated with an MSViewer.
How do non-dimeric, non-protein ligands like LPS and peptidoglycans effect high affinity binding and homo- or hetero-dimerization of the hTLRs, events presumably required to establish signalling? Another property of the hTLRs that distinguishes them from dToll is involvement of accessory proteins in the detection of the pathogen pattern. For example, CD14 was shown to have LPS-binding properties long before the discovery of the hTLRs [25], and forms a complex with hTLR4 and another peripheral membrane protein, MD-2 [26]. CD14 binds to LPS, and although not essential for signalling by hTLR4, it enhances the sensitivity of the response to LPS. CD14 is an LRR protein and also functions in conjunction with TLR2 by binding to peptidoglycan. It is also clear that responses to some microbial stimuli involve heterodimeric combinations of hTLRs; for example, diacylated lipoproteins signal through hTLR2 and hTLR6 [26]. Peptidoglycans are polymers of alternating N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) sugar residues crosslinked by pentaglycine and tetra-peptides of D-alanine. The regular arrangement of these polymeric units might allow equivalent sites on two receptors to be bound, and this might be sufficient to cross-link them into an active complex. Alternatively monotypic binding might cause a conformational rearrangement to occur in the receptor ectodomains that enables a second receptor chain to dimerize, analogous perhaps to the recruitment of the accessory protein by IL-1R on binding of IL-1α or IL-1β [27]. In the Drosophila pathway, peptidoglycan is detected by specific receptor proteins, peptidoglycan recognition proteins (‘PGRPs’) [28], and a proteolytic cascade then causes processing of Sp¨atzle and activation of Toll signalling. If the direct binding hypothesis for vertebrate TLRs is correct, two attractive features of the insect system are lost. First, the protease cascade allows an amplification of the pathogen antigen signal, and secondly such an arrangement could be self-limiting, since the terminal protease can effect a negative feedback on upstream members of the cascade. This loss of sensitivity might explain the need for non-signalling enhancer proteins such as CD14. C 2003
Biochemical Society
What structural features of the LRR allow such plasticity and rapid diversification in ligand binding? It can be seen from the alignment of hTLR2 and dToll presented in Figure 2(b) that the LRR blocks are significantly homologous and are likely to fold up into a similar structural framework. The first structure of a complex between a receptor LRR protein and binding partner – that of platelet glycoprotein 1b (Gp1b) and the A domain of von Willebrand factor (vWF) [29] – was published recently and allows us to make some general inferences about LRR–ligand interactions. The LRRs form an extended structure with each unit having a short parallel β-sheet (forming a concave surface), a turn and a region of variable secondary structure (Figure 3a). vWF binds into the concave surface, and interactions with the LRRs involve hydrogen-bonding contacts by the side chains of the first residues in the parallel β-strands of LRRs 5, 6,
Toll-Like Receptors
7 and 8. This position is highly variable in LRR sequences, but precedes a conserved leucine residue required for packing the core of the structure. In the alignment (Figure 2b), this residue is always different in hTLR and often conserved in the dTolls (results not shown). It is possible to generate a model of the first three LRRs of hTLR2 using Gp1b as a template (Figure 3b), but this does not provide any clear indication of how specificity for peptidoglycan might be achieved.
Conclusions In the past, the similarities between the innate immune systems of insects and mammals have tended to be emphasized, and there clearly is strong conservation of components in the post receptor pathway, such as dMyD88/MyD88, pelle/ IL-1 receptor-associated kinase (‘IRAK’) and dorsal/NFκB. However, it is now becoming clear that the nature of the interactions between pathogen patterns and Toll receptor ectodomains is radically different in vertebrates and invertebrates. The challenge for the future is to understand at the structural level how Toll receptors recognize pattern in both systems. This information should give insights into how the activity of hTLRs might be controlled, and might open the way to new therapeutic approaches in diseases, such as endotoxin shock and rheumatoid arthritis.
References 1 Anderson, K.V., Jurgens, G. and Nusslein-Volhard, C. (1985) Cell 42, 779–789 2 Wolpert, L. (1969) J. Theor. Biol. 25, 1–47 3 Belvin, M.P. and Anderson, K.V. (1996) Annu. Rev. Cell. Dev. Biol. 12, 393–416 4 Hashimoto, C., Hudson, K.L. and Anderson, K.V. (1988) Cell 52, 269–279 5 Buchanan, S.G.S. and Gay, N.J. (1996) Prog. Biophys. Mol. Biol. 65, 1–44 6 Morisato, D. and Anderson, K.V. (1994) Cell 76, 677–688 7 DeLotto, Y. and DeLotto, R. (1998) Mech. Dev. 72, 141–148
8 Gay, N.J. and Keith, F.J. (1991) Nature (London) 351, 355–356 9 Schneider, D.S., Hudson, K.L., Lin, T.Y. and Anderson, K.V. (1991) Gene Dev. 5, 797–807 10 O’Neill, L.A.J. and Greene, C. (1998) J. Leukocyte Biol. 63, 650–657 11 Lemaitre, B., Nicolas, E., Michaut, L., Reichhart, J.M. and Hoffmann, J.A. (1996) Cell 86, 973–983 12 Wasserman, S.A. (1993) Mol. Biol. Cell 4, 767–771 13 Taguchi, T., Mitcham, J.L., Dower, S.K., Sims, J.E. and Testa, J.R. (1996) Genomics 32, 486–488 14 Rock, F.L., Hardiman, G., Timans, J.C., Kastelein, R.A. and Bazan, J.F. (1998) Proc. Natl. Acad. Sci. U.S.A. 95, 588–593 15 Kaisho, T. and Akira, S. (2002) Biochim. Biophys. Acta 1589, 1–13 16 Janeway, C.A. and Medzhitov, R. (2002) Annu. Rev. Immunol. 20, 197–216 17 Poltorak, A., He, X.L., Smirnova, I., Liu, M.Y., Van Huffel, C., Du, X., Birdwell, D., Alejos, E., Silva, M., Galanos, C., Freudenberg, M. et al. (1998) Science 282, 2085–2088 18 Takeuchi, O., Hoshino, K. and Akira, S. (2000) J. Immunol. 165, 5392–5396 19 Tauszig-Delamasure, S., Bilak, H., Capovilla, M., Hoffmann, J.A. and Imler, J.L. (2002) Nat. Immunol. 3, 91–97 20 Parker, J.S., Mizuguchi, K. and Gay, N.J. (2001) Proteins 45, 71–80 21 Eldon, E., Kooyer, S., Develyn, D., Duman, M., Lawinger, P., Botas, J. and Bellen, H. (1994) Development 120, 885–899 22 Schnare, M., Barton, G.M., Holt, A.C., Takeda, K., Akira, S. and Medzhitov, R. (2001) Nat. Immunol. 2, 947–950 23 Iwaki, D., Mitsuzawa, H., Murakami, S., Sano, H., Konishi, M., Akino, T. and Kuroki, Y. (2002) J. Biol. Chem. 277, 24315–24320 24 Mitsuzawa, H., Wada, I., Sano, H., Iwaki, D., Murakami, S., Himi, T., Matsushima, N. and Kuroki, Y. (2001) J. Biol. Chem. 276, 41350–41356 25 Ferrero, E., Hsieh, C.L., Francke, U. and Goyert, S.M. (1990) J. Immunol. 145, 331–336 26 Correia, J.D., Soldau, K., Christen, U., Tobias, P.S. and Ulevitch, R.J. (2001) J. Biol. Chem. 276, 21129–21135 27 Vigers, G.P.K., Caffes, P., Evans, R.J., Thompson, R.C., Eisenberg, S.P. and Brandhuber, B.J. (1994) J. Biol. Chem. 269, 12874–12879 28 Ligoxygakis, P., Pelte, N., Hoffmann, J.A. and Reichhart, J.M. (2002) Science 297, 114–116 29 Huizinga, E.G., Tsuji, S., Romijn, R.A.P., Schiphorst, M.E., de Groot, P.G., Sixma, J.J. and Gros, P. (2002) Science 297, 1176–1179 30 Thompson, J.D., Higgins, D.G. and Gibson, T.J. (1994) Nucleic Acids Res. 22, 4673–4680 Received 7 March 2003
C 2003
Biochemical Society
663
