Towards multivalent CD1d ligands: synthesis and biological activity of homodimeric α-galactosyl ceramide analogues

Graphical abstract


Introduction
A multimeric version of a monomeric ligand can achieve higher affinity and specificity for a target receptor. [1][2][3] As a result, incorporating multiple copies of an active pharmacophore within a single molecule can increase biological activity by several orders of magnitude. These observations have opened up the development of socalled multivalent drug candidates as an active area of research. 4,5 There are now numerous examples of synthetic homodimers, which represent the simplest form of a multivalent compound, exhibiting significantly improved potency over their monomeric counterparts. [6][7][8][9][10][11] For example, dimeric derivative 2 of zanamivir (1), an influenza virus neuraminidase inhibitor, was found to be 100-fold more potent an inhibitor of influenza virus replication, both in vitro and in vivo, than its monomer analogue. 6 In a second example, certain dimeric versions (including 4) of the DNA intercalator N-[(2-dimethylamino)ethyl]acridine (DACA, 3) displayed five times the cytotoxic potency of the parent monomer. 12,13 In other cases, the dimeric molecule exhibits different properties entirely from the corresponding monomer. For example, artemisinin (5) is an antimalarial agent, whilst its homodimeric analogues 6a-c exhibit potent anti-tumour activity (Fig. 1). 14 As part of a wider research programme into the development of novel CD1d ligands, [15][16][17][18] we targeted a series of dimeric analogues of the prototypical invariant natural killer T (iNKT) cell agonist, agalactosyl ceramide (a-GalCer, KRN7000 (7), Figure 2). a-GalCer 19 7 is a synthetic glycolipid, which binds to the non-polymorphic MHC-class-I-like molecule, CD1d. The resulting glycolipid-CD1d complex is recognised by T cell receptors (TCRs) located on the surface of iNKT cells. Following recognition of the a-GalCer-CD1d complex, iNKT cells rapidly release a diverse array of both proinflammatory (Th1) and regulatory (Th2) cytokines and initiate a potent immune response. This method of activating the immune system is currently being explored as a mechanism for 'boosting' current vaccination strategies. [20][21][22][23][24][25][26][27][28][29][30][31][32] Localised clustering of CD1d molecules has been shown to play an important role in antigen presentation. Park et al. have shown that lipid rafts are essential for efficient TCR recognition of CD1d, 33 whilst Im et al. have reported that a-GalCer 7 requires intracellular loading on to CD1d molecules, which leads to the organised transport of a-GalCer-CD1d complexes into cholesterol-rich lipid rafts. 34 This localisation of CD1d molecules in membrane rafts serves to concentrate CD1d-lipid complexes in the plasma membrane. 35 We hypothesised that bivalent ligands of the a-GalCer pharmacophore might serve to stabilise these lipid rafts by forming a cross-linked matrix within the raft structure. This, in turn, might result in more stable CD1d-glycolipid complexes as a result of a chelating effect. 36 ker separating the two a-GalCer units, the dimer might be able to bind two CD1d molecules whilst still allowing space for presentation to iNKT TCRs. 39,40 The length of the linking unit in such ligands would be of key importance for optimal divalent binding; too long and it might provide too large a containment volume for the second pharmacophore and would result in the two a-GalCer moieties effectively behaving as unconnected entities; too short, 41,42 and steric hindrance might block the approach of a second CD1d molecule, although in this case, enhanced binding might still be observed as the result of a statistical re-binding effect. 10 In the event that a dimer with a shorter linker is able to bind two CD1d molecules, the presence of a second a-GalCer-CD1d complex in close proximity might still serve to block the approach of the TCR. 39,40 In this way, a dimer could function as an antagonist and yet has signal transduction applications. 11,43 We now report the synthesis and initial biological results of a range of homodimers of a-GalCer, which vary in both the length and type of linker unit.

Design and synthesis
Our first consideration in the design of bivalent CD1d ligands was the selection of an appropriate position from which to append a linker to the a-GalCer molecule. The crystal structure of the CD1d-glycolipid-iNKT TCR complex 39,40 reveals that the 2-, 3and 4-hydroxyl groups of the sugar head group are all involved in hydrogen bonding to either the CD1d molecule or iNKT TCR; modifying these positions leads to reduced, or even complete loss, of activity. 44  the galactose unit and there are a number of biologically active CD1d agonists where the 6-hydroxyl group of the sugar has been replaced with another functional group. [47][48][49][50] These results are supported by the crystal structure of the CD1d-a-GalCer complex 39 and a crystal structure of a TCR-a-GalCer-CD1d complex, 40 which reveal that the 6-OH of the a-GalCer sugar head group is not directly involved in hydrogen bonding to either the CD1d molecule or the iNKT TCR. For these reasons, we selected the 6-position of the sugar head group as the most suitable site through which to link together our a-GalCer monomer units.
We have recently developed a facile route to 6 00 -azido-6 00deoxy-a-galactosyl ceramide 8, 51 which we considered a potential advanced intermediate for building a-GalCer dimers. We have shown that the azido moiety of 8 can serve as a masked amine, allowing further elaboration to amides, carbamates and ureas; however more directly, the 6 00 -azido group is also primed for click chemistry. We therefore postulated that a series of 1,2,3-triazole derivatives would be readily available through a simple Huisgen reaction with an appropriate alkyne. [52][53][54] We envisaged that a-Gal-Cer dimers could be accessed in a single step via a 'double click' reaction between 2 equiv of azide 8 and an appropriate diyne (Scheme 1). This approach would also deliver symmetrical homodimers, which would simplify characterisation and analysis.
Poly(ethylene glycol) spacers were chosen to link the a-GalCer units together. These units offer compatibility with biological environments 55 as well as increased solubility of the amphiphilic glycolipid units in both aqueous and organic solvents. 56,57 We also chose to investigate alkylene linkers as a hydrophobic comparison.
Alkyne-terminated versions of these two types of linkers are readily accessible and would allow the systematic variation of linker lengths.
In order to test the tolerance of the CD1d-glycolipid-iNKT TCR complex to these types of linker units, monomeric a-GalCer analogues 10b and 12 were first synthesised from azide 8 from alkynes 9b and 11, respectively (Scheme 2), and incubated with splenocytes from wildtype C57 BL/6 mice. The presence of IFNc was then determined by ELISA as previously described by Reddy et al. 58 Pleasingly, triazole 10b was found to stimulate iNKT cells in vitro at similar levels to a-GalCer 7, whilst triazole 12 stimulated iNKT cells, albeit to a lesser extent (Fig. 5, vide infra).
With the knowledge that the CD1d-glycolipid-iNKT TCR interaction tolerated this type of C6 derivatisation, our next consideration was the length of the spacer units to be employed in our dimer syntheses. Major histocompatibility complex (MHC) molecules, which present peptide fragments to T-cells, are structurally similar to CD1d molecules, and in a related study, Cebecauer et al. have studied MHC-peptide dimers. 59 Dimers containing short linkers, 10-30 Å in length, efficiently triggered intracellular calcium mobilisation and phosphorylation in cloned cytotoxic T lymphocytes, whereas dimers with longer linkers did not. The fact that MHC-peptide dimers containing linkers as short as 10 Å could still interact with the TCR raised the question of how the two MHC molecules align themselves. The authors describe a dimeric binding model in which two TCRs engage with their a3 domains in an anti-parallel manner, two MHC-peptide complexes facing each other. Models of adjacent MHC molecules revealed distances varying from 12 Å to 46 Å between the a3 domains. 59 Applying similar orientations to CD1d models (Fig. 3, top), we proposed a range of a-Gal-Cer dimers where the lengths of the spacer units separating the glycolipid pharmacophores could be systematically varied across a similar range. More specifically, we targeted a-GalCer dimers 13ae and 14a-c, in which the fully extended linker lengths range from 9 Å (for 13a) to 38 Å (for 13e), as estimated from models using Pro-DRG server (Fig. 3, bottom).
Diynes 16a-e and 18a-c were next employed in click reactions with azide 8. Thus, two equivalents of azide 8 and one equivalent of diyne 16a-e or 18a-c were heated in the presence of copper(II) sulfate and sodium ascorbate, to provide bis-triazole dimers 13a-e and 14a-c, respectively, in excellent yields (Scheme 4). These compounds were then tested for their ability to stimulate iNKT cells.

Biological results
All of the a-GalCer dimers were found to be active following incubation with splenocytes from wildtype C57 BL/6 mice (  Fig. 4); 58 PEG-linked dimers 13a-e stimulated iNKT cells in vitro at similar levels to a-GalCer 7 at high concentrations (>10 nM), although there was no obvious sensitivity to linker length. At lower concentrations (<1 nM), dimers 13a and 13b, both of which possess short linkers between the a-GalCer residues, were less active than dimers 13c-e, which contain longer linkers. Alkylene-linked dimers 14a-c also stimulated iNKT cells in vitro, but to a lesser extent than a-GalCer 7. No detectable IFNc was observed when compounds 13a-e and 14a-c were incubated with splenocytes from CD1d À/À (iNKT cell-deficient) mice (data not shown), indicating that the observed iNKT cell stimulation by our a-GalCer dimers is CD1d-dependent.
In order to test the ability of the iNKT TCR to recognise the CD1d-glycolipid complex, in vitro binding experiments were performed ( Fig. 6). 60 Dimers 13c-e were comparable to a-GalCer 7 at high concentrations (>10 nM), whilst dimers 13a,b again proved to be considerably weaker than a-GalCer 7. At lower concentrations (<5 nM), dimers 13c-e, which contain the longer linker units, displayed greater iNKT-cell recognition than a-GalCer 7. Dimer 13d, which contains a penta(ethylene glycol) linker ($26-27 Å fully extended length), proved optimal. Alkylene-linked dimers 14a-c showed weaker CD1d binding than a-GalCer 7 (Fig. 6) and we tentatively postulate that the lower activity of this series of dimers might be a result of the hydrophobic alkylene linker coiling to minimise contact with the aqueous medium. This coiling will serve to decrease the effective length of the linker and may account for the observed activity of these dimers more closely resembling that observed for the ethylene glycol dimers containing shorter linkers (i.e., 13a,b).
In order to investigate whether or not the trends observed with dimers 13a-e were a consequence of the presence of the second pharmacophore or due to the presence of the linker unit alone, we targeted a series of monomers 10a-e each containing a PEG tail length which corresponded to the same linker lengths separating the a-GalCer residues in dimers 13a-e.   readily accessed via click reactions between azide 8 and alkynes 9a-e. Alkyne 9a was available commercially, whilst alkynes 9b-e were readily synthesised from alcohols 19a-d (Scheme 5).
Each of the monomers 10a-e was tested for its ability to stimulate iNKT cells ( Fig. 5). a-GalCer 7 served as a reference compound, enabling us to compare the data for monomers 10a-e with the data for the corresponding dimers 13a-e. At high concentrations (>10 nM), the level of iNKT-cell stimulation for all of the monomers 10a-e was not significantly different to that observed for the corresponding dimers (relative to a-GalCer 7 reference).
Surprisingly, at lower concentrations (<1 nM), monomer 10d proved to be a more potent activator of iNKT cells than its dimeric equivalent (13d), relative to a-GalCer 7, whilst 10a, the control for 13a, was again inactive at low concentrations in vitro. The fact that dimers 13a-e do not show any enhanced biological activity over their corresponding monomers 10a-e suggests that a bivalent effect is not taking place in vitro. A comparison of the data for monomer 12, containing an octyl chain, and the alkylene-linked dimers 14a-c also provides little evidence for a bivalent effect. The effect of lipid rafts and clustering effects on iNKT-cell activation might be expected to be more apparent in vivo, 61 and the in vivo results were indeed more interesting. PEG-linked dimers 13c-e, which had proven to be the most active from the in vitro studies, were injected intravenously into C57 BL/6 WT mice. Blood serum was taken at 18 h and the presence of IFNc determined by ELISA (Fig. 7). 62 iNKT-cell stimulation was found to increase with linker length, up to the point where the dimer with the longest octa(ethylene glycol) linker, 13e, had almost the same activity as   linkers (13a-e) stimulated iNKT cells to a similar extent as a-GalCer 7 for longer linker lengths (13c-e), whilst those with shorter lengths (13a,b) were less-effective CD1d agonists. We considered the possibility that this trend might be due to the linking poly(ethylene glycol) unit itself rather than as a result of a bivalent effect and therefore synthesised and tested a series of monomers (10a-e) containing a poly(ethylene glycol) tail length corresponding to each dimer linker length. These monomers showed a similar trend to the corresponding dimers, with 10a,b being the least active. Surprisingly, monomer 10d was more active than its corresponding dimer, 13d. Dimers based on alkylene linkers proved to be less active than a-GalCer 7 in vitro, and showed no obvious sensitivity to linker length. Overall, the in vitro results show little evidence for a bivalent effect. Preliminary in vivo experiments show an increase in activity with increasing linker length, with the octa(ethylene glycol)-linked dimer 13e being the most potent activator of iNKT cells, having a similar activity to a-GalCer 7. Future work will involve producing dimers and multimers with longer linker lengths and further probing the multivalent effects using different techniques.

General methods
Infra-red spectra were recorded neat as thin films between sodium chloride discs, or as potassium bromide discs. The intensity of each band is described as s (strong), m (medium) or w (weak) and with the prefix v (very) and suffix br (broad) where appropriate. The poor solubility of all a-GalCer dimers at rt prevented us from obtaining reliable optical rotation data. 1 3 OD, d C 49.0. The term 'stack' is used to describe a region where resonances arising from non-equivalent nuclei are coincident, and multiplet, m, to describe a resonance arising from a single nucleus (or equivalent nuclei) but where coupling constants cannot be readily assigned. Mass spectra were recorded on a LCT spectrometer utilising electrospray ionisation with a methanol mobile phase, or electron impact ionisation, and are reported as (m/z (%)). HRMS were recorded on a LCT spectrometer using a lock mass incorporated into the mobile phase. Melting points were determined using open capillaries and are uncorrected.

Chemicals
All reagents were obtained from commercial sources and used without further purification unless specified otherwise. Tetrahydrofuran was freshly distilled under nitrogen from sodium benzophenone ketyl. All solutions are aqueous and saturated unless specified otherwise. NaH (60% w/w in mineral oil, 3 equiv (for bis-propargylation) or 1.5 equiv (for mono-propargylation)) was added to a solution of appropriate diol 15ba-d or alcohol 19a-d (1 equiv) in anhydrous THF (0.2 M) at 0°C under an N 2 atmosphere. A catalytic amount (spatula tip) of Bu 4 NI (0.05 equiv) and propargyl bromide (3 equiv (for bis-propargylation) or 1.5 equiv (for mono-propargylation)) was added sequentially. The mixture was stirred at rt overnight, before being concentrated under reduced pressure. Purification of the residue by flash column chromatography afforded diynes 16b-e and alkynes 9b-e as pale yellow oils.

General procedure for lithium acetylide addition to dibromoalkanes 17a and 17b (synthesis of diynes 18b and 18c)
Lithium acetylide-ethylenediamine complex (2 equiv) was added to a solution of dibromoalkane 17a or 17b (1 equiv) in DMSO (0.2 M) at 0°C. The mixture was allowed to warm to rt and stirred overnight. H 2 O (5 Â volume) and hexane (5 Â volume) were added and the phases were separated. The aqueous phase was extracted with hexane (5 Â volume). The combined organic extracts were washed with brine (5 Â volume), dried (MgSO 4 ), filtered and concentrated under reduced pressure. Purification of the residue by flash column chromatography (hexane) afforded diyne 18b or 18c both as white solids.

Soluble human NKT TCR binding assay
Soluble iNKT TCR tetramers were prepared according to the protocol described by McCarthy et al. 60 C1R-hCD1d cells were pulsed with lipids or vehicle at various concentrations overnight and following washes incubated with fluorescently labelled iNKT TCR tetramer. iNKT TCR-CD1d-lipid complexes were detected by flow cytometry on a FACScalibur device using CellQuest software.