Biochemical Journal

Research article

The C2 domain of Tollip, a Toll-like receptor signalling regulator, exhibits broad preference for phosphoinositides

Gayatri Ankem , Sharmistha Mitra , Furong Sun , Anna C. Moreno , Boonta Chutvirasakul , Hugo F. Azurmendi , Liwu Li , Daniel G. S. Capelluto


TLRs (Toll-like receptors) provide a mechanism for host defence immune responses. Activated TLRs lead to the recruitment of adaptor proteins to their cytosolic tails, which in turn promote the activation of IRAKs (interleukin-1 receptor-associated kinases). IRAKs act upon their transcription factor targets to influence the expression of genes involved in the immune response. Tollip (Toll-interacting protein) modulates IRAK function in the TLR signalling pathway. Tollip is multimodular, with a conserved C2 domain of unknown function. We found that the Tollip C2 domain preferentially interacts with phosphoinositides, most notably with PtdIns3P (phosphatidylinositol 3-phosphate) and PtdIns(4,5)P2 (phosphatidylinositol 4,5-bisphosphate), in a Ca2+-independent manner. However, NMR analysis demonstrates that the Tollip C2 domain binds Ca2+, which may be required to target the membrane interface. NMR and lipid–protein overlay analyses suggest that PtdIns3P and PtdIns(4,5)P2 share interacting residues in the protein. Kinetic studies reveal that the C2 domain reversibly binds PtdIns3P and PtdIns(4,5)P2, with affinity values in the low micromolar range. Mutational analysis identifies key PtdIns3P- and PtdIns(4,5)P2-binding conserved basic residues in the protein. Our findings suggest that basic residues of the C2 domain mediate membrane targeting of Tollip by interaction with phosphoinositides, which contribute to the observed partition of the protein in different subcellular compartments.

  • C2 domain
  • calcium
  • nuclear magnetic resonance (NMR)
  • phosphoinositide
  • Toll-interacting protein (Tollip)
  • Toll-like receptor (TLR)


Both TLRs (Toll-like receptors) and IL-1Rs (interleukin-1 receptors) provide a mechanism for host defence responses by activating the innate and adaptive immune responses [1]. These receptors are single-transmembrane proteins with ectodomains composed largely of leucine-rich repeats and with a conserved cytosolic TIR (Toll/interleukin-1 receptor) domain, which facilitates the recruitment of adaptor proteins [2]. TLRs are broadly distributed on cells of the immune system and are the best-studied immune sensors of invading pathogens. There are at least eleven human TLRs (TLR1–TLR11) [3] that use the adaptor protein MyD88 (myeloid differentiation factor 88) to signal, with the exception of TLR3 and TLR4 [4]. TLR types 1, 2, 4, 5, and 6 are expressed on the cell surface and are involved in recognizing lipopeptides and proteins. On the other hand, the antiviral TLR types 3, 7, 8, and 9 are localized in endosomes [4]. Upon activation (i.e. by microbial products), TLRs are believed to either homo- or hetero-dimerize, followed by MyD88 binding to the cytosolic TIR domain of the TLR (for a review, see [1]). This, in turn, promotes the activation of stress-activated protein kinases, including the IRAKs (IL-1R-associated kinases) 1, 2, M and 4. These kinases bind to MyD88 by their death domains, facilitating the activation of the tumour-necrosis factor-receptor-associated factor 6, which forms a protein complex with two ubiquitin-conjugating enzymes, the ubiquitin-conjugating enzyme variant 1A and ubiquitin-conjugating enzyme 13. Activation of other kinases by TLRs, including IκB [inhibitor of NF-κB (nuclear factor κB)] kinases, leads to the release of the transcription factor NF-κB from IκB and subsequent phosphorylation and ubiquitin-dependent degradation of IκB. NF-κB translocates to the nucleus, which promotes pro-inflammatory responses by mediating cytokine gene expression [5].

Tollip (Toll-interacting protein) controls IRAK function in both TLR and IL-1R signalling pathways [68]. In resting cells, Tollip regulates these pathways on two different levels. First, Tollip associates with IL-1R, TLR2 and TLR4 after lipopolysaccharide activation, inhibiting TLR-mediated cell activation [7]. Secondly, Tollip binds directly to IRAK, inhibiting IRAK autophosphorylation [6,7]. Indeed, overexpression of Tollip leads to inhibition of TLR-mediated NF-κB activation [6,7]. Following TLR signalling stimulation, the Tollip–IRAK complex associates with the cytosolic tails of IL-1R and TLR. IRAK autophosphorylates and phosphorylates Tollip [7], thus initiating downstream signalling [6]. In addition, Tollip is involved in protein sorting by its association with Tom1 (target of Myb1), ubiquitin, and clathrin [9]. Tollip is localized on early endosomes, where it is required for both degradation of ubiquitin-conjugated proteins [8] and sorting of IL-1R on late endosomes [10]. Recently, Tollip has been shown to be SUMOylated and to mediate IL-1R SUMOylation [11]. This novel function makes Tollip a putative regulator of nuclear and cytoplasmic protein trafficking.

Tollip is a modular protein containing an N-terminal TBD (Tom1-binding domain), a central C2 (conserved 2) domain and a C-terminal CUE (coupling of ubiquitin to endoplasmic reticulum degradation) domain [6]. Using a lipid–protein overlay assay, Tollip was found to bind to both PtdIns3P (phosphatidylinositol 3-phosphate) and PtdIns(3,4,5)P3 [phosphatidylinositol (3,4,5)-trisphosphate] [12]. A lysine to glutamic acid mutation located within the C2 domain (K150E) abrogates phosphoinositide binding [12], suggesting that the Tollip C2 domain is engaged in lipid binding. The CUE domain is a ubiquitin-binding module [13] that has been shown to bind to, and be phosphorylated by, IRAK proteins [7]. The tertiary structure of the CUE domain from two yeast proteins reveals a helical conformation with the Vps9p CUE domain being a dimer [14], whereas the CUE domain of Cue2 protein is a monomer [15]. Two-hybrid studies show that rat Tollip interacts with itself, suggesting that the protein forms oligomers [11]. In agreement, we have recently shown that the human Tollip CUE domain forms tight dimers that can contribute to Tollip oligomerization and ligand recognition [16].

There are at least 200 C2 domains and, after the PH (pleckstrin homology) domain, they represent the second most common lipid-binding domain [17]. Numerous C2 domains mediate signalling in a Ca2+-dependent membrane-binding manner, as is the case for the conventional protein kinase C and synaptotagmin I, in which ion binding occurs via three loops known as CBRs (Ca2+-binding regions) (for a review, see [18]). Binding of Ca2+ potentiates C2 domain recognition to acidic phospholipid membranes by changing the overall electrostatic potential at the C2 domain surface [19]. However, the C2 domain of the cytosolic phospholipase A2 binds to neutral membranes in a Ca2+-bound state [19]. A minor group of C2 domains shows weak membrane affinity and is usually engaged in protein–protein interactions or exhibits a structural function. A third group of C2 domains does not bind Ca2+ at all, but binds to membranes or participates in protein–protein interactions [2022]. The Tollip C2 domain lacks two aspartic acid residues necessary for Ca2+ binding [6]. However, recent studies of the C2 domain of synaptotagmin IV indicate that Ca2+ binding cannot always be predicted from sequence-based analysis [23].

The structure of C2 domains typically presents eight-stranded antiparallel β-sandwiches consisting of two sets of four-stranded β-sheets [18]. The C2 domain surface loops connect β-strands in two different topologies. The majority of the C2 domains present a cationic patch in the interior of the β-sandwich, which is known as the β-groove. Owing to the cationic nature of the β-groove and the Ca2+-binding loops, both are proposed to be two different lipid-binding sites [18]. The C2 domains are known to bind a wide spectrum of phospholipids without a well-defined lipid-binding site [18]. A possible explanation is that, in the absence of Ca2+, the cationic β-groove provides lipid recognition, necessary for the vesicle fusion activity of host proteins [24].

In the present study, we demonstrate that the C2 domain region is responsible for phosphoinositide recognition in Tollip. Remarkably, our findings support a broad range of specificity in which the phosphoinositide moiety is a common theme. Moreover, our results indicate that two of the most preferred ligands, PtdIns3P and PtdIns(4,5)P2 [phosphatidylinositol (4,5)-bisphosphate], interact with basic residues in the Tollip C2 domain with affinities in the low-micromolar range and that binding of Ca2+ to the C2 domain is dispensable for phosphoinositide recognition. Overall, our findings might provide an explanation for the underlying mechanism responsible for Tollip's partition into different subcellular membrane compartments.



A list of chemicals used and their suppliers follows: PC (1,2-dioleoyl-sn-glycero-3-phosphocholine), PS (1,2-dioleoylsn-glycero-3-phospho-L-serine), PA (1,2-dioleoyl-sn-glycero-3phosphate), PE (1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine), PG [1,2-dipalmitoyl-sn-glycero-3-phospho-(1′-rac-glycerol)] and PI (phosphatidylinositol) (Avanti Polar Lipids); PtdIns3P, PtdIns4P, PtdIns5P, PtdIns(3,4)P2, PtdIns(3,5)P2, PtdIns(4,5)P2, and PtdIns(3,4,5)P3 (Cayman Chemicals); Ins1P (inositol 1-phosphate), Ins(1,3)P2 (inositol 1,3-bisphosphate) and Ins(1,4,5)P3 (inositol 1,4,5-trisphosphate) (Echelon); and IPTG (isopropyl β-D-thiogalactopyranoside) (Research Products International). All other chemicals were analytical reagent grade.

Cloning, expression and purification of the Tollip constructs

The human full-length Tollip, its isolated C2 domain (residues 54–182), and the Saccharomyces cerevisiae Vam7p PX domain (residues 2–134) cDNAs were cloned into a pGEX4T3 vector (GE Healthcare) and expressed in Escherichia coli (Rosetta; Stratagene). Briefly, bacterial cells were grown in Luria–Bertani medium at 37°C until they reached a D600 of ~0.8. Induction of GST (glutathione transferase)-fusion proteins resulted from the addition of 0.1 mM IPTG followed by a 4 h incubation at 25°C. Cell pellets were suspended in ice-cold buffer containing 50 mM Tris/HCl (pH 7.3), 500 mM NaCl, 500 mM benzamidine, 0.1 mg/ml lysozyme, 5 mM DTT (dithiothreitol) and 0.1% Triton X-100. Suspensions were further processed by sonication (using a Branson sonifier 250 with a duty cycle of 30% and eight pulses of 30 s each), centrifuged (1940 g, 30 min, 4°C) and the resultant supernatants loaded on to a glutathione–Sepharose 4B (GE Healthcare) column. In some cases, fusion proteins were eluted off the beads by addition of 20 mM Tris/HCl (pH 7), 500 mM NaCl and 100 mM reduced glutathione. In other purifications, the GST tag was removed by incubation of the fusion protein with thrombin (EMD Biosciences) overnight at 4°C. Proteins were recovered in a buffer containing 20 mM Tris/HCl (pH 8) and 500 mM NaCl, concentrated using a 3-kDa-cut-off concentrator device (Millipore) and further purified by an ÄKTA FPLC system using a size-exclusion chromatography column (Superdex 75; GE Healthcare) equilibrated with 50 mM Tris/HCl (pH 8), 1 M NaCl and 1 mM DTT. Protein peak fractions were pooled, exchanged in the corresponding buffer, and further concentrated for functional and structural analysis.

Liposome-binding assay

Stocks of phospholipids were prepared in organic solvents as described in the manufacturer's instructions. Liposomes were prepared by a weight ratio of 1:1 PC/PE, and 5% of the phospholipid under investigation was used. Controls were prepared by adjusting ratios with both PC and PE. Lipid films were generated by freeze-drying and hydrated in 50 mM Hepes (pH 6) and 100 mM NaCl to 1 mg/ml at 67°C for 30 min and freeze–thawed three times. Liposomes were sonicated (using a Branson sonifier 250 with a duty cycle of 30% and eight pulses of 30 s each), pelleted (15000 g, 20 min, 22 °C), and suspended at 5 mg/ml in the same buffer. A 10 μg portion of protein was incubated with 200 μg of total lipid for 30 min at room tempe-rature (22°C). Liposome-bound and free-protein fractions were obtained by centrifugation and analysed by SDS/PAGE (10% gels). Protein bands were quantified using AlphaEase FC software.

NMR spectroscopy

NMR protein samples contained 150 μM Tollip C2 domain, 90% H2O/10% 2H2O, 20 mM 2H11 Tris/HCl (pH 6.8), 150 mM KCl, 5 mM 2H10 DTT and 1 mM NaN3. Two-dimensional NMR experiments were acquired at 25°C using a Bruker Avance III 600 MHz spectrometer (Virginia Tech) equipped with an inverse TXI probe with z-axis pulsed-field gradients. Additional NMR experiments were performed on a Bruker Avance 800 MHz spectrometer equipped with a cryoprobe at the University of Virginia. Phospholipid head group and CaCl2 titrations into the 15N-labelled C2 domain were analysed by 1H-15N HSQC (heteronuclear single quantum coherence) experiments. Spectra were processed with NMRPipe [25] and analysed using nmrDraw [26].

CD spectroscopy

Far-UV CD spectra were generated using purified proteins (10 μM) in 5 mM Tris/HCl (pH 6.8), 100 mM KF and 0.1 mM DTT on a Jasco J-815 spectropolarimeter equipped with a Jasco PFD-425 S temperature control unit. Spectra were collected in a 1-mm-pathlength quartz cell at 25°C. Spectra were obtained from five accumulated scans from 240 to 190 nm using a bandwidth of 1 nm and a response time of 1 s at a scan speed of 20 nm/min. Buffer backgrounds were used to subtract from the protein spectra. Secondary structure content of the proteins was estimated with the online server DICHROWEB [27] using the CDSSTR algorithm [28]. Near-UV CD spectra were collected using a 0.1-cm pathlength at 20 nm/min between 340 and 250 nm with a response time of 1 s and data pitch of 0.5 nm. Owing to the weak molar absorptivity of aromatic amino acids, 100 μM protein was employed for the near-UV CD experiments. Thermal denaturation of the proteins (10 μM) was investigated in the range 4–90°C following temperature-induced changes in ellipticity at 218 nm, where the temperature was increased 1°C/min, using 1.5-nm bandwidth, an averaging time of 30 s and an equilibration time of 2 min.

Fluorescence spectroscopy

Intrinsic tryptophan fluorescence spectroscopy measurements were carried out using a Jasco J-815 spectropolarimeter at 25°C in a 1-cm-pathlength cuvette. λex was 295 nm, and the fluorescence emission spectra were recorded from 310 to 410 nm for each protein sample.

Surface plasmon resonance analysis

Measurements were carried out at room temperature in 20 mM Hepes (pH 6) and 100 mM NaCl using an L1 sensor chip on a BIAcore X100 instrument. Liposomes, containing PtdIns3P, PtdIns(4,5)P2 or PE, were made as described above with an additional extrusion step using 400-nm membranes. The surface of the sensor chip was preconditioned by injecting 40 mM N-octyl β-D-glucopyranoside at a flow rate of 5 μl/min. The first flow cell was used as a control surface, whereas the second flow cell was employed as the active surface. Both flow cells were coated with 6000 RU (resonance units) of 1 mg/ml each of the liposomes tested at a flow rate of 2 μl/min. After liposome coating, 30 μl of 10 mM NaOH at a flow rate of 30 μl/min was used to wash away any unbound liposome. Non-specific binding sites at the sensor chip surface were then blocked with the injection of 250 μl of 0.1 mg/ml fatty-acid-free BSA (Sigma) at a flow rate of 5 μl/min. A range of concentrations of protein analytes was prepared in the same buffer and injected on both flow cell surfaces at a flow rate of 30 μl/min. Sensorgrams were obtained from at least five different concentrations of each of the tested proteins. Association and dissociation times for each protein injection were set at 120 and 600 s respectively. The remaining bound protein was washed away by the injection of 30 μl of 10 mM NaOH. The sensor chip surface was regenerated using 40 mM N-octyl β-D-glucopyranoside and recoated with fresh liposomes for the next protein titration. Data were analysed using BIAcore X100 evaluation software (version 2.0).

Lipid–protein overlay assay

Lipid strips were prepared by spotting 1 μl of the indicated amount of phosphoinositide dissolved in chloroform/methanol/water (65:35:8) on to Hybond-C Extra membranes (GE Healthcare). Membrane strips containing the immobilized phosphoinositide were blocked with 3% (w/v) fatty-acid-free BSA (Sigma) in 20 mM Tris/HCl (pH 8), 150 mM NaCl and 0.1% Tween 20 for 1 h at room temperature. Then, membranes were incubated with 0.1 μg/ml protein in the same buffer without BSA overnight at 4°C. Following four washes with the same buffer, bound proteins were probed with rabbit anti-GST antibody (Santa Cruz Biotechnology) and donkey anti-rabbit-HRP (horseradish peroxidase) antibody (GE Healthcare). Protein binding was detected using enhanced chemiluminescence reagent (Pierce).


The Tollip C2 domain preferentially binds phosphoinositides

The C2 domain is a conserved module located between the N-terminal TBD and the C-terminal CUE domain in Tollip (Figure 1A). A recent report indicated that Tollip binds to both PtdIns3P and PtdIns(3,4,5)P3 using the lipid–protein overlay assay [12] and that mutation at Lys150, a residue located within the C2 domain, abrogates phospholipid binding [12]. To determine whether the C2 domain of Tollip is responsible for phospholipid recognition, we isolated this domain and investigated whether it binds phospholipids using liposomes, which closely resemble biological membranes. We found that the C2 domain preferentially bound to phosphoinositides compared with liposomes bearing other phospholipids, including PC, PA, PG, PS or PI (Figures 1B and 1C). These findings are in agreement with those observed in other C2 domains, which typically exhibit highly variable and relatively low lipid specificity [18]. Phospholipid binding by the C2 domain was at most ~60% of the total protein when compared with the Vam7p PX domain (Figure 1B), a PtdIns3P binding domain [29]. These results indicate that the Tollip C2 domain preferentially binds phosphatidylinositol species with phosphate groups at their inositol rings.

Figure 1 Lipid ligand preference of the Tollip C2 domain

(A) Schematic representation of the Tollip primary structure with the boundaries of each of the domains indicated above the diagram. (B) Phospholipid specificity of the Tollip C2 domain using liposomes containing 5% various phospholipids and compared with control liposomes (PC/PE). The Vam7p PX domain was employed as a positive control. P and S indicate pellet and supernatant fractions respectively, after centrifugation, SDS/PAGE and Coomassie Blue staining. (C) Bands were quantified using Alphaimager and normalized to the liposomes control. Both * and **, P<0.05. The ratios represent means + S.D. for three independent assays. PI, PtdIns.

To further validate our findings, next we compared phosphoinositide interactions using the head group of PtdIns3P and PtdIns(4,5)P2, the most preferred lipid ligands of the Tollip C2 domain (Figures 1B and 1C), and NMR spectroscopy analysis. Initial NMR titration experiments using PtdIns3P and PtdIns(4,5)P2 led to significant protein precipitation (results not shown). Several chemical shift perturbations of C2 domain residues were detected after the addition of the lipid head groups, Ins(1,3)P2 (for PtdIns3P; Figure 2A) and Ins(1,4,5)P3 [for PtdIns(4,5)P2; Figure 2B] in 15N-1H HSQC experiments. The rate of the association is in the fast-exchange regime and the protein exhibits a low affinity for the head group, which was anticipated, owing to the absence of a membrane interface. Most chemical-shift perturbations in each spectra are similar, suggesting that at least PtdIns3P and PtdIns(4,5)P2 head groups share some binding residues. These perturbations are absent from NMR samples of the Tollip C2 domain containing an Ins1P head group of PI, which is composed of the core of the inositol ring of phosphoinositides, but lacks the phosphate groups at the same protein/ligand ratio (Figure 2C). Thus this result serves as a control to identify specific chemical-shift perturbations with phosphorylated inositol rings. To further determine that distinct phosphoinositides share some binding residues in the Tollip C2 domain, we performed a competition lipid–protein overlay assay. In this case, the tagged recombinant protein (as a GST-fusion) was pre-incubated with Ins(1,4,5)P3 to allow binding, and this mixture was then added to a membrane-bound PtdIns3P. As shown in Figure 3(A), Ins(1,4,5)P3 significantly reduced binding of PtdIns3P to the Tollip C2 domain.

Figure 2 The Tollip C2 domain interacts with phosphorylated inositol rings at positions 3, 4 and 5

The 15N-labelled Tollip C2 domain (black) was subjected to 1H-15N HSQC analysis following titration with Ins(1,3)P2 (A), Ins(1,4,5)P3 (B) or Ins1P (C) at a protein/lipid head group molar ratio of 1:16. Tollip C2 domain residues that shift upon head-group titration in (A) and (B) are illustrated with green broken ovals. The location of these residues is also shown in (C). A representative section of the HSQC titrations is magnified in each panel for clarification.

Figure 3 Structural analysis of the free and head-group-bound Tollip C2 domain

(A) PtdIns3P (PI3P) and PtdIns(4,5)P2 compete with each other for binding to the Tollip C2 domain. The GST–Tollip C2 domain (10 μM) was pre-incubated with 50 μM Ins(1,4,5)P3 for 30 min at room temperature and the mixture was further incubated with strips containing PtdIns3P at the indicated amounts and further processed as described in the Experimental section. (B) Far-UV CD spectra of the Tollip C2 domain in the absence or presence of 16-fold of Ins(1,3)P2. MRE, mean residue ellipticity. (C) Near-UV CD spectra of the Tollip C2 domain in the absence or presence of 16-fold of Ins(1,3)P2. (D) Size-exclusion chromatography elution profile of the Tollip C2 domain monitored at 280 nm. Arrows denote the positions of the molecular-mass standards (Aprot, aprotinin; ChyA, chymotrypsinogen A; CytC, cytochrome c; Myo, myoglobin). A.U., arbitrary units. The inset represents the fractions of the protein peak loaded on an SDS/PAGE gel and stained with Coomassie Blue. (E) Thermal denaturation of the Tollip C2 domain at 218 nm. (F) Fluorescence emission spectra of the Tollip C2 domain in the absence (black line) and presence (grey line) of 6 M guanidine hydrochloride.

Structural analysis of the Tollip C2 domain

Far-UV CD spectroscopy was used to characterize the secondary structure of the Tollip C2 domain. Figure 3(B) shows that the domain exhibits a typical spectrum of a β-sheet protein with a minimum at 218 nm. Prediction of the secondary structure content revealed that the C2 domain presents ~34% β-sheet and negligible α-helical content (~4%). After addition of 16-fold of Ins(1,3)P2, the PtdIns3P head group did not induce any significant change in the overall secondary structure of the protein (Figure 3B and results not shown). Tryptophan, tyrosine, and phenylalanine residues are the main chromophores in the CD of the near-UV region and therefore provide information about the tertiary structure of proteins [30]. The Tollip C2 near-UV spectrum exhibits a negative peak between 278 and 310 nm with a minimum at ~292 nm, which is probably due to the contribution of its tyrosine and tryptophan side chains to the tertiary structure (Figure 3C). The overall intensity of the Tollip C2 spectrum is shifted up somewhat by the addition of Ins(1,3)P2 (Figure 3B). Although the cause of this intensity difference is not clear, the shape similarities of the two spectra indicate that the head group does not induce a conformational change in the tertiary structure of the protein upon binding. Thus most of the chemical-shift changes observed in the Tollip C2 NMR spectrum upon addition of Ins(1,3)P2 (Figure 2A) are likely to be a consequence of direct contact between the protein and the lipid head group. Recently, we reported that the CUE domain forms dimers [16] and therefore it represents a contributor of Tollip oligomerization. Some C2 domains have been reported to mediate protein dimerization [3133]. Therefore we have also investigated the oligomeric state of the Tollip C2 domain by analytical size-exclusion chromatography. The protein eluted at a volume that corresponded to a 15.5-kDa globular protein (Figure 3D), a value close to its theoretical molecular mass (15.7 kDa). Mass spectral analysis of the protein revealed a prominent signal at 15481 (see Figure S1 at, confirming the monomeric nature of the protein. Thermal denaturation analysis indicated that the Tollip C2 domain shows two-state melting transitions, indicative of a highly co-operative unfolding transition between 50°C and 60°C with an apparent melting temperature of 54°C. The fluorescence spectrum of the Tollip C2 domain exhibited a maximum at 340 nm and a further red shift of ~6 nm and reduced emission intensity after denaturation with guanidine hydrochloride (Figure 3F), indicative of a relatively buried average localization for the three tryptophan residues present in the native conformation of the protein domain.

The Tollip C2 domain binds Ca2+, which is not required for phosphoinositide binding

Ca2+ is a physiological ligand for most C2 domains, whose binding is necessary to increase the affinity of the protein to lipids for membrane-docking events. The Tollip C2 domain presents three conserved aspartic acid residues involved in Ca2+ ligation found in other C2 domains [6] (see Figure 7A). Therefore we experimentally investigated whether the Tollip C2 domain binds Ca2+ using CD and NMR spectroscopy. The CD spectrum of the Tollip C2 domain does not exhibit any conformational change upon addition of up to 30 mM CaCl2 (results not shown). However, NMR spectroscopy analysis demonstrates that Ca2+ selectively induces chemical-shift perturbations in the spectrum of the protein (Figure 4A). The binding also appears to be on a fast-exchange regime, indicating a low affinity for the ligand in solution. Many C2 domains have been shown to interact with lipids in a Ca2+-dependent manner. Therefore we investigated whether Ca2+ binding contributes to phosphoinositide recognition by the Tollip C2 domain. As shown in Figure 4(B), the Tollip C2 domain did not exhibit Ca2+-dependent binding to PtdIns3P, and the binding was not further changed by the presence of EGTA, a Ca2+ chelator. PtdIns(4,5)P2 binding by the Tollip C2 domain was also Ca2+-independent (results not shown).

Figure 4 Ca2+ binding by the Tollip C2 domain and role of Ca2+ in phosphoinositide interactions

(A) Overlay of the 1H-15N HSQC spectra of the Tollip C2 domain in the absence (black) and presence (red) of 5 mM CaCl2. Perturbed resonances are labelled with black broken ovals. (B) Ca2+-independent phosphoinositide binding of the Tollip C2 domain determined by the lipid–protein overlay assay. Each spot contains 15.6–2000 pmol of PtdIns3P (PI3P). A GST–Tollip C2 domain fusion protein was incubated with lipid strips in the absence or presence of 1 mM CaCl2, 1 mM EGTA, or both 1 mM CaCl2 and 1 mM EGTA. GST was employed as a negative control.

Kinetic analysis of Tollip C2 domain–phosphoinositide interactions

The interaction of the Tollip C2 domain with phosphoinositide-enriched liposomes was kinetically examined by SPR (surface plasmon resonance), using tag-free Tollip C2 domain as the analyte and either PtdIns3P- or PtdIns(4,5)P2-enriched liposomes immobilized on an L1 sensor chip as ligands. The Vam7p PX domain binds PtdIns3P [29] and was employed as a positive control. In all cases, proteins exhibited fast association and dissociation rates. The Tollip C2 domain bound phosphoinositide-enriched liposomes following a two-state conformational change model and displayed reversible binding (Figures 5A and 5B). No fitting could be obtained when plots were analysed using the 1:1 Langmuir model (results not shown). Binding affinity and kinetic properties of the proteins and their phosphoinositide ligands are summarized in Table 1. The Tollip C2 domain bound with dissociation constants (Kd) of 4.6 and 11 μM for PtdIns(4,5)P2 and PtdIns3P liposomes respectively. These affinity values are comparable with that determined for PtdIns(4,5)P2 binding by the C2 domain of classical PKC (protein kinase C) [34], but much lower than the affinities reported for the synaptotagmin C2b and PKCθ C2 domains for the same ligand [22,24]. On the other hand, the Vam7p PX domain bound PtdIns3P following a 1:1 Langmuir model (Figure 5C). The affinity of the Vam7p PX domain for PtdIns3P liposomes was 1000-fold higher with a Kd of 9.5 nM (see Supplementary Table S1 at, a phosphoinositide affinity value similar to those determined for other PX domains [35]. The affinity differences between the proteins for phosphoinositides are reflected by plotting the log of maximum resonance units against protein concentration, in which the Tollip C2 domain exhibits a modest affinity when compared with the Vam7p PX domain (Figure 5D). However, in all titrations, phosphoinositide binding exhibited a saturable binding isotherm that is associated with specific binding (Figure 5D).

View this table:
Table 1 Liposome-binding parameters of the Tollip C2 domain determined from SPR analysis

ka1, association rate constant; ka2, forward rate constant changing complex; kd1, dissociation rate complex; kd2, reverse dissociation rate constant changing complex.

Figure 5 Kinetic analysis of Tollip C2 domain interactions with phosphoinositides

SPR sensorgrams for the binding of the Tollip C2 domain with PtdIns(4,5)P2-containing liposomes (A), and with PtdIns3P liposomes (B). As a control, the Vam7p PX domain was probed against PtdIns3P liposomes (C). Various concentrations of each of the protein domains were flowed over the liposomes attached on an L1 sensor chip for 120 s. (D) The strength of the associations is represented in the plot, in which the resonance units are on a logarithmic scale. Au, arbitrary units; PI, PtdIns.

Conserved basic residues in the Tollip C2 domain play a critical role in phosphoinositide recognition

Since phosphoinositides are negatively charged, several conserved basic residues of the Tollip C2 domain (see Supplementary Figure S2 at were mutated to alanine and analysed for binding to PtdIns3P, one of the most preferred Tollip C2 domain ligands (see Figures 1B and 1C and [12]), using the lipid–protein overlay assay. As expected, Tollip bound to PtdIns3P in a lipid-mass-dependent manner (Figure 6A). Mutation of the non-conserved Lys150 residue (Figure S2) to glutamic acid in Tollip has been shown to abrogate lipid binding [12]. However, we found that a mutation of the same residue to a neutral amino acid reduced, but did not eliminate, PtdIns3P binding; rather, mutation of the conserved Lys162 residue to alanine almost completely abolished PtdIns3P binding (Figure 6A). Curiously, the isolated C2 domain consistently showed a higher binding signal when compared with the full-length Tollip protein (Figure 6A). This observation suggests that neighbouring regions may modulate phosphoinositide recognition by the C2 domain. We further screened for PtdIns3P-binding residues by site-directed mutagenesis in the Tollip C2 domain. We found that mutations to alanine at Arg78, Arg123, His135, Arg157, and Lys162 almost eliminated PtdIns3P binding, whereas mutations at Lys102 and Lys150 mirrored the strength of binding of the wild-type Tollip C2 domain for the same phosphoinositide (Figure 6A). Overall, these results indicate that specific conserved basic residues in the Tollip C2 domain play a critical role in phosphoinositide recognition. We then functionally and structurally investigated the K162A mutant of the C2 domain for simplicity. Consistent with our findings, the inability of the Tollip C2 K162A to bind PtdIns3P was evident using the liposome-binding assay (Figure 6B). Likewise, PtdIns3P binding by the Tollip C2 K162A mutant was also negligible using SPR (results not shown). Mutation at Lys162 did not alter the secondary structure, the tryptophan fluorescence emission, or the stability and compactness of the Tollip C2 domain (Supplementary Figure S3 at and results not shown), indicating that the mutation specifically disrupts PtdIns3P binding. Our studies also demonstrate that Tollip bound PtdIns(4,5)P2 and that Tollip K162A exhibited a slightly reduced, but not abolished, binding to this phosphoinositide (Figure 6C). Mutation at Lys150 showed no major differences from Tollip for binding to PtdIns(4,5)P2 (results not shown). Whereas residues Arg78, Arg123 and His135 in the isolated C2 domain seemed to be critical for PtdIns(4,5)P2 binding, the Lys162 mutation exhibited only a minor reduction in binding to the same lipid (Figure 6C). PtdIns(4,5)P2 binding by K102A and R157A C2 domains were indistinguishable to that observed for the wild-type protein (Figure 6C). Thus our findings indicate that unique Tollip C2 domain residues are able to discriminate the position of the phosphorylation(s) at the inositol ring in phosphoinositide ligands.

Figure 6 Identification of the Tollip residues critical for PtdIns3P and PtdIns(4,5)P2 binding

(A) Lipid–protein overlay assay of immobilized PtdIns3P at the indicated amounts and GST-fusion Tollip, C2 domain and mutants in basic residues within the C2 domain. GST was used as a negative control. (B) Liposome-binding assay of wild-type Tollip C2 domain or its K162A mutant with liposomes without or with PtdIns3P. P and S represent pellet and supernatant fractions respectively after centrifugation, SDS/PAGE and Coomassie Blue staining. PI, PtdIns. (C) Lipid–protein overlay assay of immobilized PtdIns(4,5)P2 at the indicated amounts and GST-fusion Tollip, C2 domain and mutants in conserved basic residues within the C2 domain.


Tollip is a key regulator of TLR-dependent signalling pathways [36]. To exert this function, Tollip binds to the tail of TLR proteins in association with adaptor proteins at the boundaries of the cytosolic face of the plasma membrane and endosomes to inhibit IRAK function. However, the determinants by which Tollip associates with these proteins at the membrane boundaries still remain unclear. In the present paper, we show for the first time that Tollip preferentially binds to phosphoinositides by its conserved C2 domain. Phosphatidylinositols represent less than 15% of the total phospholipids present in eukaryotes, with PtdIns4P and PtdIns(4,5)P2 being the most abundant phosphoinositides in mammalian cells [37]. Each of the seven phosphoinositides found in these cells exhibits distinctive subcellular membrane localization. For example, PtdIns(4,5)P2 and PtsIns(3,4,5)P3 are enriched at the plasma membrane, whereas PtsIns3P is found exclusively in endosomes [37]. Remarkably, the subcellular localization of these phospholipids correlates with the reported subcellular localization of Tollip. Tollip is localized on early endosomes, multivesicular bodies [8] and the Golgi apparatus [12]. Thus the subcellular localization of Tollip and its ability to bind phosphoinositides and ubiquitin suggest that it may be involved in the recognition and transport of ubiquitinated proteins in endocytic pathways. TLR proteins signal from both plasma membranes and endosomes [2]. Thus, given the broad phosphoinositide specificity of its C2 domain, it is conceivable that Tollip is phosphoinositide-dependently partitioned in different subcellular membrane pools to control TLR function through MyD88-dependent pathways. We speculate that the broad specificity of the Tollip C2 domain to phosphoinositides may also increase the membrane affinity of Tollip or it may help to properly orient Tollip at the membrane.

Phospholipid-binding domains, including C2, PX, FYVE, PH and ENTH domains, are modules engaged in membrane trafficking by recruiting signalling peripheral proteins to specific cell membrane surfaces. These domains exhibit diverse affinity and specificity to negatively charged phospholipids found at the membrane and present a strong positive potential on their surface that promotes these associations. Of these, members of the PH and C2 domains are most commonly found in eukaryotes. Despite its ubiquitous presence, the C2 domain, unlike other lipid-binding modules, does not exhibit a well-defined lipid-binding site nor a conserved cationic patch [18]. Moreover, not all C2 domains can be easily identified by sequence homology. Another level of complexity is that C2 domains are present in soluble and transmembrane proteins, they bind lipids either in a Ca2+-dependent or -independent fashion, and they can either bind strongly or weakly to biological membranes [18]. Therefore, sequence homology is unlikely to predict the biological function of an uncharacterized C2 domain on the basis of studies performed with the same module found in unrelated proteins. From our present studies, the Tollip C2 domain exhibited the highest preference of binding to liposomes enriched with the phosphoinositides PtdIns3P, PtdIns(4,5)P2, and PtdIns(3,4,5)P3 (Figures 1B and 1C). Therefore we investigated further the Tollip-binding properties of two of these ligands [PtdIns3P and PtdIns(4,5)P2], given their strength of binding as well as the different locations of the phosphate groups at their inositol rings. Titration of the Tollip C2 domain with the head groups of PtdIns3P and PtdIns(4,5)P2 showed chemical-shift deviations for restricted sets of NH signals (Figure 2). We were unable to assign the residues associated with these NMR chemical shifts due to the insolubility of the Tollip C2 domain at high concentrations. However, comparison of chemical-shift perturbations of the protein NMR spectrum by the addition of Ins(1,3)P2 and Ins(1,4,5)P3 suggest that phosphoinositide ligands share some residues in their binding sites. Competition analysis further confirmed this observation (Figure 3A). These results are not surprising given the intrinsic nature of C2 domains in recognizing a broad range of phospholipids [18]. Some C2 domains exhibit a dual lipid-recognition mechanism, in which multiple lipid ligands must be present at the membrane to achieve high affinity and to properly localize intracellularly [17]. However, our experimental results suggest that two Tollip C2 domain lipid ligands, PtdIns3P and PtdIns(4,5)P2, overlap their binding sites (Figures 2 and 3A).

Far-UV CD spectroscopy analysis of the Tollip C2 domain support the presence of β-sheet secondary structure and that the protein lacks any major conformational change upon Ins(1,3)P2 binding (Figure 3B and results not shown). In addition, a co-operative unfolding transition is observed in the protein with a relatively low-temperature midpoint (~54°C). Similar properties are displayed by the synaptotagmin I C2A domain [38]. The near-UV spectrum of the Tollip C2 domain is unaffected by the presence of the lipid head group, indicating that no major changes in the tertiary folding of the protein are induced upon ligand binding. This is in agreement with the observation of selected NMR cross-peaks perturbed by the addition of phosphoinositide head groups (Figure 3).

A proposed Ca2+-binding role has been shed by the observation that CBRs are usually surrounded by positively charged residues [39,40], amino acids that are also found near the second aspartic acid-rich region in Tollip (see Figure 7A). However, the presence of only three of the five aspartic acid residues required for Ca2+ binding suggests that Tollip may not bind Ca2+ (Figure 7A). Nonetheless, we experimentally demonstrate that the Tollip C2 domain binds Ca2+, but that the ion seems to be dispensable for phosphoinositide binding. This result is not surprising, since similar observations were reported for the C2A domain of dysferlin, which contains four negatively charged residues of five required for Ca2+-binding proteins [41]. Intriguingly, the synaptotagmin IV C2B domain contains all five aspartic residues, but is still unable to bind Ca2+ [23]; therefore Ca2+-binding properties cannot be predicted from sequence analysis. As a Ca2+-independent lipid-binding protein, basic residues at the surface of the Tollip C2 domain could make direct membrane contact, as proposed for other C2 domains [18]. A distinct C2 family, related to the phosphoinositide 3-kinase C2 domain, has recently been proposed to interact with negatively charged lipids in a Ca2+-independent manner through a patch of basic and hydrophobic residues located at the concave surface of the upper β-sheet that makes a PtdIns(4,5)P2-binding and membrane-penetration site [42]. However, this region, located between β-strands 5 and 6 in C2 domains, does not exhibit a conserved basic and hydrophobic signature in Tollip (Figures 7A and S2). The CBR of Ca2+-independent lipid-binding C2 domains has been proposed to play a role in lipid binding and specificity [4345]. Further investigation is required to determine the role of Ca2+ binding by the Tollip C2 domain, which is beyond the scope of the present study.

Figure 7 Sequence alignment, structural properties and location of critical basic residues in the modelled structure of the Tollip C2 domain

(A) Sequence alignment of the C2A domain of human synaptotagmin I (Syt I) (GenBank® accession number NP_005630) , the C2 domain of human PKCα (GenBank® accession number NP_002728) , the C2 domain of the human Rab11 family of interacting proteins (FIP) (GenBank® accession number NP_079427) , the C2 domain of human Tollip (GenBank® accession number CAG38508), the C2 domain of human PKCϵ (GenBank® accession number CAA46388) and the C2 domain of human PKCθ (GenBank® accession number NP_006248) constructed from the Biology WorkBench database ( Boxes indicate the conserved aspartic acid residues engaged in Ca2+ ligation in both the synaptotagmin I and PKCα C2 domains. The secondary-structure content determined for the synaptotagmin I C2 domain is depicted above the sequence alignment. Mutated Tollip C2 domain residues from the present study are labelled in grey. (B) Two views of the predicted tertiary structure of the Tollip C2 domain constructed from the AL2TS database ( using the C2A domain of synaptotagmin I as a template and depicted using PyMol ( Experimentally determined PtdIns3P-binding residues are labelled in red on the predicted Tollip C2 domain tertiary structure.

SPR experiments allowed us to characterize the kinetics of Tollip C2 domain binding to phosphoinositides. The protein binds to PtdIns3P and PtdIns(4,5)P2 with Kd values of 11 and 4.6 μM respectively, following a conformational change model. This may be the general mechanism by which Tollip binds to TLRs through a first association step with phosphoinositide-enriched membranes accompanied by a conformational change of the protein that further enhances binding to TLRs. For comparison purposes, we also measured the affinity of the Vam7p PX domain to PtdIns3P, which resulted in 1000-fold higher affinity than that of the C2 domain and the mode of binding followed a one-to-one interaction model. Despite the fact that both Tollip and Vam7p proteins are found in endosomes, differences in affinity can be explained by the low phospholipid specificity exhibited by the Tollip C2 domain. Our investigation indicates that the isolated C2 domain binds phosphoinositides more strongly than Tollip, suggesting that other domains (i.e. TBD and CUE) in the protein may exert a modulatory function. The C2 domain of the cytosolic phospholipase 2α binds more prominently than the full-length protein, and a local effect on the C2 domain was proposed [22]. Likewise, the PtdIns(4,5)P2-binding ubiquitin ligase Smurf2 protein exhibits auto-inhibitory properties by intramolecular interaction between its C2 and HECT domains, leading to a reduction of its enzymatic activity and the stabilization of the protein levels in the cell [46].

In the present work, we have identified for the first time conserved basic amino acids critical for Tollip C2 domain interactions with PtdIns3P and PtdIns(4,5)P2. Many structures of Ca2+-dependent C2 domains have been solved in detail, and they exhibit a common fold with a well-defined cationic β-groove and with the CBR in the protein engaged in lipid binding [18]. A phosphoinositide-binding site has also been mapped in the C2B domain of rabphilin-3A, which shows that its polar C-terminal region, as well as its β-strands 3, 4, 6 and 7, are critically involved in lipid recognition in a Ca2+-independent manner [47]. Our mutational analysis indicates that the Tollip C2 domain could bind phosphoinositides in a similar fashion (Figure 7B). Mutation of Lys150 to glutamic acid has been shown to abolish phospholipid binding by Tollip [12]. We mutated this residue to alanine, since this amino acid would not cause a change in the overall charge of the protein. Both Tollip and C2 domain K150A mutants exhibited indistinguishable binding to phosphoinositides (Figure 6A and results not shown). This result is supported by the fact that the residue is not conserved among Tollip proteins (Figure S2). Also, we found that some mutations in basic residues in the Tollip C2 domain abolished PtdIns3P, but not PtdIns(4,5)P2, binding. However, these lipids compete with each other for binding to the Tollip C2 domain. Therefore it is conceivable that phosphoinositide-binding sites overlap in Tollip, but that distinctive residues provide broad specificity of the protein for phosphoinositides. Binding between the Tollip C2 domain and the evaluated phosphoinositides is with moderated affinity, which can be explained from the proposed shallow, surface-exposed, PtdIns3P-binding residues in the modelled structure of the protein (Figure 7B). Overall, we propose that, together with Ca2+, a variety of hydrophobic and electrostatic forces can contribute to the Tollip–phosphoinositide interactions with biological membranes necessary for Tollip to modulate TLR signalling. Given the role of Tollip in recruiting adaptor proteins at the cytosolic tail of TLRs by its CUE domain, formation of such complexes may be further enhanced by initial Tollip interactions with phosphoinositides and membranes by its C2 domain. Further molecular studies are necessary to address how Tollip modulates TLR signalling. We are currently investigating the co-ordination between the C2, TBD and CUE domains for their molecular interactions, which will provide new insights into understanding the multimodular nature of Tollip.


Gayatri Ankem performed the purification of proteins and carried out liposome-binding assays, NMR titrations, CD measurements and protein–lipid overlay assays. Sharmistha Mitra carried out size-exclusion chromatography analysis, lipid–protein overlay assays and SPR experiments. Furong Sun and Anna Moreno purified proteins. Boonta Chutvirasakul designed constructs and purified proteins. Hugo Azurmendi carried out NMR experiments. Liwu Li provided reagents. Daniel Capelluto conceived the study, designed experiments, analysed data and wrote the manuscript.


This work was supported by the American Heart Association [grant number 086077E (to D.C.)].


We are grateful to Dr Kae-Jung Hwang and Rebecca Lehman for their contribution in the initial phase of this work, to Jeff Ellena for his assistance during the NMR titration experiments performed at the University of Virginia, and to Dr Carla V. Finkielstein and Dr Janet Webster for thoughtful comments on the manuscript.

Abbreviations: C2 domain, conserved 2 domain; CBR, Ca2+-binding region; CUE, coupling of ubiquitin to endoplasmic reticulum degradation; DTT, dithiothreitol; GST, glutathione transferase; HSQC, heteronuclear single quantum coherence; IκB, inhibitor of nuclear factor κB; IL-1R, interleukin-1 receptor; Ins(1,3)P2, inositol (1,3)-bisphosphate; Ins(1,4,5)P3, inositol (1,4,5)-trisphosphate; IPTG, isopropyl β-D-thiogalactopyranoside; IRAK, IL-1 receptor-associated kinase; MyD88, myeloid differentiation factor 88; NF-κB, nuclear factor κB; PA, 1,2-dioleoyl-sn-glycero-3-phosphate; PC 1, 2-dioleoyl-sn-glycero-3-phosphocholine; PE, 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine; PG, 1,2-dipalmitoyl-sn-glycero-3-phospho-(1′-rac-glycerol); PH, pleckstrin homology; PI, phosphatidylinositol; PKC, protein kinase C; PtdIns3P, phosphatidylinositol 3-phosphate; PtdIns(4,5)P2, phosphatidylinositol (4,5)-bisphosphate; PtdIns(3,4,5)P3, phosphatidylinositol (3,4,5)-trisphosphate; PS, 1,2-dioleoyl-sn-glycero-3-phospho-L-serine; SPR, surface plasmon resonance; TBD, Tom1-binding domain; TIR, Toll/interleukin-1 receptor; TLR, Toll-like receptor; Tollip, Toll-interacting protein


View Abstract