Communications ChemMedChem doi.org/10.1002/cmdc.202000992 New Paradigm in NKT Cell Antigens: MCS-0208 (2- (Hydroxymethyl)phenylthio-phytoceramide) – an Aryl- Phytoceramide Compound with a Single Hydroxyl Group Stimulates NKT Cells Roser Borràs-Tudurí+,[a] Anna Alcaide+,[a] Sandrine Aspeslag,[b, c] Lorena Usero,[d] Carmen Serra,[a] Carme Roura-Mir,[d] Dirk Elewaut,[b, c] and Amadeu Llebaria*[a] Natural Killer T (NKT) cells play an important role in the immune phytoceramide lipid was found during a screening of anticancer response and can be activated by glycolipids presented by agents of marine sponge extractions showing the induction of CD1d protein. We present MCS-0208, an unprecedented promising, potent and antitumor response.[2] After this finding, arylthioether-phytoceramide able to induce potent invariant intensive efforts were directed to decipher the mechanism of NKT (iNKT) cell activation, notably when tested in human iNKT action of this compound resulting in the identification of cells. This arylsphingolipid analog has a simple phenyl group Natural Killer T (NKT) cells, a new subtype of lymphocytes with containing a single hydroxyl substituent as a surrogate of the Natural Killer (NK) cells and T lymphocytes characteristics. sugar ring. The phenylthioether structure contrasts with α- Another important finding was that the particular T cell galactosylceramide (1), the prototypical glycolipid used to receptor was restricted to CD1d-antigen complex. CD1d is a induce iNKT cell stimulation, where the galactose 2’-OH and 3’- non-classical MHC-class I that recognizes and presents lipid OH substituents are accepted as the minimal footprint and antigens and in particular glycosphigolipid derivatives, instead considered critical for NKT T cell receptor (TCR) recognition. A of peptides. The prototypical NKT cell stimulator αGalCer, computational study supports the recognition of aromatic displays a very potent activity and induces both Th1 and Th2 compound by the CD1d and TCR proteins as radically new opposite responses from NKT cells with several undesired structures for NKT cell stimulation. effects. Another undesirable effect of αGalCer was its over- stimulation leading to an unresponsiveness state of the cells. Therefore, plenty of studies were conducted to modify the structure of αGalCer in order to optimize its physicochemical During the past two decades, extensive research was done on and biological properties. Hundreds of new compounds were designing, synthesizing and evaluating new α-galactosylcera- synthesized and biologically studied concluding which part of mide (αGalCer) derivatives.[1] This galactose α-linked to a the αGalCer structure has an impact on the strength or the type of response (Figure 1). In general, most chemical efforts were focused on the galactose moiety and the lipid chains. The sugar [a] Dr. R. Borràs-Tudurí,+ Dr. A. Alcaide,+ Dr. C. Serra, Dr. A. Llebaria was extensively studied exploring both α- and β-linked Medicinal Chemistry and Synthesis (MCS) Laboratory Institut de Química Avançada de Catalunya (IQAC) glycolipids, the configuration of the hydroxylated carbons Consejo Superior de Investigaciones Científicas (CSIC) (introducing glucose, mannose or other sugars) and hydroxyl C/ Jordi Girona 18–26 substitution or fuctionalization. There is a general agreement 08034 Barcelona (Spain) E-mail: amadeu.llebaria@iqac.csic.es on the critical importance of 2’-OH and 3’-OH for CD1d [b] Dr. S. Aspeslag, Dr. D. Elewaut recognition while the modification on other parts of the Unit for Molecular Immunology and Inflammation galactose can be somehow modified to modulate the inter- VIB-Center for Inflammation Research Ghent (Belgium) action with the T cell receptor (TCR) and, as a result, modulate [c] Dr. S. Aspeslag, Dr. D. Elewaut the NKT cell response. The other broadly studied part of the Department of Internal Medicine and Pediatrics αGalCer was the lipid chains, exploring in both tails the length Ghent University Ghent (Belgium) impact,[3,4] ramification or substitution, and the introduction of [d] Dr. L. Usero, Dr. C. Roura-Mir other groups such as aromatic rings,[5–8] polar groups[9] or Laboratori d'Immunologia Cel·lular halogens. Recently it was published a new derivative able to Institut de Biotecnologia i Biomedicina Universitat Autònoma de Barcelona bind covalently to the CD1d antigen binding groove, increasing Bellaterra, 08193 Barcelona (Spain) ligand affinity[10] and inducing a Th2 polarizing activity. [+] These authors contributed equally to this work. Our group studied non-glycosidic compounds such as Supporting information for this article is available on the WWW under aminocyclitol-ceramide derivatives HS44 and HS161, (com- https://doi.org/10.1002/cmdc.202000992 pounds 3 and 4 in Figure 2), showing high levels of cytokine © 2021 The Authors. ChemMedChem published by Wiley-VCH GmbH. This is prodution and anticancer in vivo activity.[11,12] an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, Taking advantage of our group background on the chemical provided the original work is properly cited. synthesis of a great diversity of sphingosine and phytospgingo- ChemMedChem 2021, 16, 2491 – 2496 2491 © 2021 The Authors. ChemMedChem published by Wiley-VCH GmbH Communications ChemMedChem doi.org/10.1002/cmdc.202000992 Figure 3. Chemical structure of new β-S-Glucosylceramide derivatives. candidate[15] at the time of this work was started however it was finally refused.[16,17] Immunostimulator potency of these compounds was meas- ured by IL-2 production of splenocites from Balb-6 mice in vitro. Product 6 and 8 induced low levels of IL-2 secretion only comparable to that of vehicle (Figure 4). Only compound 7 showed a slightly higher amount of cytokine secretion at the highest dose (1000 ng/mL). Considering that α-S-GalCer (2), the S-linked analogue of αGalCer (1), was inactive when tested in mice cells,[18] the low Figure 1. General overview of αGalCer chemical structure and the impact of each part in the biological response modulation. levels of cytokine production of our new compounds were not surprising. Moreover the lack of activity in the synthesized β- thioglucosylceramide analogues 6, 7 and 8 could not be entirely attributed to the β-configuration of the sugar moiety and other factors like the length of the N-acyl chain and its degree of saturation, or the fact that these analogues are S-glycosides, should be considered. Taking into account that compound 7 was more active than the O-linked analogue β-GlcCerd18:1-C24:1 (5),[16,17] we decided to explore the NKT activity of our chemical library of non glycosidic ceramide-derivatives with totally differ- ent “polar” heads non-sugar related such as compounds 9,[13] 10,[13] 11[13] and 12[19] (Figure 5). Moreover, being aware of the Figure 2. Chemical structure of some aGalCer derivatives. sine derivatives in several fields from inhibitors of sphingolipid- metabolizing enzymes[13] to NKT cell modulators,[11,12,14] some new thiosugar derivatives from our chemical collection were screened as candidates for NKT cell modulation (compounds 6,[13] 7[13] and 8[13] in Figure 3). These compounds presented a chemical structure related to or similar to β-glucosylceramide 5 Figure 4. In vitro IL-2 secretion upon NKT cell stimulation with β-S-GlcCers. (Figure 2), which was postulated as endogenous ligand Splenocites were cocultured with each compound in triplicate. Error bars indicate SD values. ChemMedChem 2021, 16, 2491 – 2496 www.chemmedchem.org 2492 © 2021 The Authors. ChemMedChem published by Wiley-VCH GmbH Communications ChemMedChem doi.org/10.1002/cmdc.202000992 closer structures to MCS-0208 (13), such as compounds 9, 10, 11 or 12 induced lower levels of IL-2 secretion. The main difference was on the N-acyl chain length what was in the established structural effects of the N-acyl groups required for CD1d recognition. In same line of Structure-Activity studies, the short acyl chain was related to Th2 polarized activity[2] what could explain the lower levels of IL-2 (a cytokine included among Th1 ones) of compounds 9 to 12. In order to compare the potencies of the new ArCer with the β-S-GlcCer family, we synthesized compound 14 (Figure 7), a β-S-GlcCer with the same ceramide skeleton than MCS-0208 (13), and they were evaluated together with a dose response of our new hit compound. Interestingly, MCS-0208 (13) showed consistent dose responses, and compound 14 was 2-fold less active than MCS-0208 (13) at the same dose in spite of the presence of the thiosugar (Figure 7). In view of these results, it was evident the interest in going further and test the activity of compounds 7, MCS-0208 13 and 14 in vivo. Although it was not expected to reach similar Figure 5. Chemical structure of new ArCer family. αGalCer cytokine production levels, we were more interested in evaluating the activity of the new MCS-0208 (13) compared to other weak agonists with glycosylated ceramide structures. importance of the acyl chain length, MCS-0208 (13, Figure 5) Thus, lipids were tested in vivo and were administrated intra- was synthesized with the same ceramide skeleton as αGalCer peritoneally (5 μg) in PBS to BL6 and 18-/- mice and cytokine (see Supporting Info for chemical details). To this purpose, secretion was measured on mice sera by ELISA assay (IL-4 at initial assays at the highest dose (1000 ng/mL) were performed 4 hours and IFN-γ at 16 hours) using αGalCer as a positive in vitro with splenocites from Balb-6 mice (Figure 6). control. In vivo assays confirmed our in vitro results showing an The in vitro experiments showed that most of the com- agonist profile for all three compounds, but moreover con- pounds of this set did not activate iNKT responses in a meaningful manner with the only exception of MCS-0208 (13) that showed stronger induction of IL-2 secretion, but it was not as active as αGalCer, used as reference compound (Figure 6). It is important to highlight the results of this new family of thioaryl-ceramides (ArCer). The NKT activity of MCS-0208 (13) was rather unexpected because the structure of the group attached at position 1 of the ceramide skeleton is very different from any sugar mimetic. Furthermore, the other analogues with Figure 6. In vitro IL-2 secretion induced by sphingolipid analogues at 1000 Figure 7. Comparative in vitro IL-2 secretion of ArCers and compound 14. ng/mL. Splenocites from BL6 mice were cocultured with each compound in Splenocites from BL6 mice were cocultured with each analogue in triplicate. triplicate. Error bars indicate SD values. Error bars indicate SD values. ChemMedChem 2021, 16, 2491 – 2496 www.chemmedchem.org 2493 © 2021 The Authors. ChemMedChem published by Wiley-VCH GmbH Communications ChemMedChem doi.org/10.1002/cmdc.202000992 firmed that MCS-0208 (13) was a stronger agonist than β-S- promising potency. To further explore this hypothesis, IFN-γ GlcCer’s 7 and 14 (Figure 8). Interestingly, MCS-0208 (13) was and IL-4 cytokines levels were measured. At 10ng/mL our more active than its corresponding thiosugar analogs 7 and 14, compound showed similar IFN-γ levels than aGalCer however and in all cases a Th1 polarization profile was observed, with significant lower levels of IL-4 were detected, pointing to a higher production of IFN-γ than IL-4. In addition MCS-0208 (13) marked Th1 biased response. These results were unexpected showed the strongest polarization profile. considering the low functionalization of the aromatic ring and As divergent results in mice and human NKT stimulation therefore the low number of potential contacts of the antigen have been reported for some compounds, we next tested MCS- with the contact surface of CD1d and TCR proteins. 0208 (13) in human NKT cell stimulation assays. MCS-0208 (13) To further understand the activity of this unprecedented induced a TNFα production similar to that of aGalCer (Figure 9 family, some computational studies were done to explore how A) suggesting a similar NKT cell stimulation ability and a this compound could interact with CD1d and the TCR. Initial docking studies were carried out using crystal structure of our potent aminocyciltol HS44 (3) and mouse proteins (PDB code: 3RTQ).[14] Protein was prepared following Protein Preparation[20] protocol from Schrödinger,[21] the corresponding grid was centered to the polar head of the aminocyclitol assuming a similar binding mode of the long lipid chains. Due to the high number of rotatable bonds and to speed up the calculation, truncated ligands and positional constraints were used to properly locate each compound into the CD1d cavity and leaving the aromatic moiety unconstrained (or the aminocyclitol moiety of HS44 (3) as positive control). This experiment was repeated 10 times to have a good average of potential preferable binding mode. These experiments suggest a binding mode for MCS-0208 (13) with the aromatic ring in parallel to the carbohydrate ring however slightly closer to CD1d than the sugar moiety. Regarding to the only hydroxyl group from MCS- 0208 (13), we postulate that it could occupy a similar space to those occupied by the 2’OH and 3’OH of the aminocyclitol (or Figure 8. In vivo results from BL6 mice injected i. p. Cytokine levels were measured by ELISA assays. Error bars indicatge SD values. the sugar mimetic in any case). This binding mode could allow the interaction of this exclusive polar group with two residues essentially involved in the ternary complex formation: Gly96 on the TCR a-chain and Asp153 on CD1d α-chain. However, a single H-Bond would be fromed with Asp153CD1d instead of two that could be observed by the glyco-mimetic moieties (Fig- ure 10). To reinforce this hypothesis, a short Molecular Dynamic (MD) simulation was run to explore the system relaxation. The best pose obtained for MCS-0208 (13) was selected as the starting point to run 20 ns MD (300 K, periodic boundary Figure 10. Docking results for MCS-0208 (13) with PDB: 3RTQ crystal Figure 9. A) In vitro TNFα production of human NKT cells. B) In vitro IFN-γ structure, A) Superposition of 10 poses obtained after iterative docking jobs and IL-4 production of human NKT cells. Purified iNKT from human blood B) representative image of the best pose. CD1d carbons are coloured in were cocultured with pre-loaded transfected B-cells with each antigen in green while TCR α-chain carbons are coloured in pink. H-Bond interactions triplicate. Error bars indicate SD values. are coloured in black dashed lines. ChemMedChem 2021, 16, 2491 – 2496 www.chemmedchem.org 2494 © 2021 The Authors. ChemMedChem published by Wiley-VCH GmbH Communications ChemMedChem doi.org/10.1002/cmdc.202000992 conditions, NPT ensemble) in explicit water, using the Desmond manner, in our previous docking experiments MCS-0208 (13) software.[22] Two main parameters were analyzed from MD formed this interaction, at least intermittently. A correlation simulations: ternary complex stability in terms of root-mean- between H-Bond interaction and the binding affinity together square deviation (RMSD) and ligand’s H-Bond formation with the potency of NKT cells activation is an attractive capacity all along simulation. In this sense, interactions with key hypothesis which is beyond the scope of this preliminary study. residues Asp80CD1d, Thr156CD1d, Asp153CD1d, Asn30αTCR, and As it could be expected, the number of interactions Gly96αTCR were monitored. Ternary complex structure was mediated by “head” group of new thioaryl-ceramide ligand mostly stable in all cases with RMSD in the same range of MCS-0208 (13) decreased due to low level of functionality reference HS44 (3). H-Bond analysis showed conserved inter- compared to the aminocyclitol scaffold. To further understand actions with Asp80 and Thr156 of CD1d in a similar frequency why this compound seems to increase its potency when tested as HS44 (3), only one H-Bond was observed with Asp153 versus with human cells, preliminary docking studies with human the two observed with HS44 (3) (Figure 11). This result was CD1d-TCR proteins were done (Figure 12). The results point to a expected as our ligand has a single hydroxyl group. Moreover, new π-π stacking interaction of the phenyl ring of MCS-0208 interactions with Gly96 on the TCR were lost for MCS-0208 (13) and the unique tryptophan 153 of human CD1d (which is a as well as for HS44 (3); this result was rather unexpected as this glycine in mice protein). This interaction could increase the interaction was in the crystal structure analysis of HS44 (3)[14] ArCer-CD1d complex stability and therefore stabilizing the and was also observed on our docking experiments. In a similar recognition by TCR. Overall, these results point to this compound as promising candidate to be further study. More extensive structure-activity studies could put some light on this new family potential and deeper structural are needed to corroborate the above mentioned structural hypothesis. Conclusions Here we have presented a new NKT cell recognized antigen, MCS-0208 (13), with an unprecedented structure having an ortho-hydroxymethyl substituted phenylthioether as a sugar mimetic in the glycolipid. These results open the possibility of further development for new families of compounds capable to stimulate NKT cells. Considering the low functionalization of this hit compound, we are optimistic of the potential of this new family of NKT cell modulators with a more controllable activity and selectivity. In summary, we have tested a small but representative group of sphingolipid analogues as NKT cell stimulators in vitro and the most potent antigens were tested in vivo. Surprisingly, the aromatic compound MCS-0208 (13) showed the highest stimulation capacity for both mice and Figure 11. H-Bond monitoring of HS44 (3) (black dots) and compound 13 human NKt cells among all tested compounds. Although all of (blue dots) during 20 ns MD simulation. them were weaker antigens than the reference αGalCer, MCS- 0208 (13) showed a better Th1 polarizing profile. Taking into account that one of the drawbacks of αGalCer (1) is the overstimulation of NKT cells, the immmuno-stimulatory proper- ties showed by MCS-0208 (13) could be an interesting starting point for the search of new, selective and adjustable molecules as antigens for CD1d-restrictred NKT cell stimulation with more controlled potency and less disadvantages. Acknowledgements Financial support by the Ministry of Economy, Industry and Figure 12. Docking results for MCS-0208 (13) with PDB: 4NE3[23] crystal Competitiveness (CTQ2014-57020-R and CTQ2011-29549-C02-01 structure. Two different views are shown. CD1d carbons are coloured in green while TCR α-chain carbons are coloured in pink. H-Bond interactions grants) is gratefully acknowledged. Pre-doctoral FPU Spanish are coloured in yellow dashed lines and π-π interactions are coloured in research and teaching fellowship from the Ministry of Education, blue. ChemMedChem 2021, 16, 2491 – 2496 www.chemmedchem.org 2495 © 2021 The Authors. ChemMedChem published by Wiley-VCH GmbH Communications ChemMedChem doi.org/10.1002/cmdc.202000992 Culture and Sport (FPU2013-01783 for R. Borràs-Tudurí grant and [12] Y. Harrak, C. M. Barra, A. Delgado, A. R. Castaño, A. Llebaria, J. Am. Chem. Soc. 2011, 133, 12079–12084. AP2007-01826 for A. Alcaide grant) are also acknowledged. [13] A. Alcaide, A. Llebaria, J. Org. Chem. 2014, 79, 2993–3029. [14] J. Kerzerho, E. D. Yu, C. M. Barra, E. Alari-Pahisa, E. Girardi, Y. Harrak, P. Lauzurica, A. Llebaria, D. M. Zajonc, O. Akbari, A. R. Castaño, J. Immunol. 2012, 188, 2254. Conflict of Interest [15] P. J. Brennan, R. V. Tatituri, M. Brigl, E. Y. Kim, A. Tuli, J. P. Sanderson, S. D. Gadola, F. F. Hsu, G. S. Besra, M. B. Brenner, Nat. Immunol. 2011, 12, The authors declare no conflict of interest. 1202–1211. [16] P. J. Brennan, R. V. Tatituri, C. Heiss, G. F. Watts, F. F. Hsu, N. Veerapen, L. R. Cox, P. Azadi, G. S. Besra, M. B. Brenner, Proc. Natl. Acad. Sci. USA Keywords: NKT cells · alfa-galactosylceramide · aryl-ceramide · 2014, 111, 13433–13438. [17] P. J. Brennan, T. Y. Cheng, D. G. Pellicci, G. F. M. Watts, N. Veerapen, D. C. immunostimulation Young, J. Rossjohn, G. S. Besra, D. I. Godfrey, M. B. Brenner, D. B. Moody, Proc. Natl. Acad. Sci. USA 2017, 114, 8348–8353. [18] A. E. Hogan, V. O’Reilly, M. R. Dunne, R. T. Dere, S. G. Zeng, C. O’Brien, S. [1] X. Laurent, B. Bertin, N. Renault, A. Farce, S. Speca, O. Milhomme, R. Amu, P. G. Fallon, M. A. Exley, C. O’Farrelly, X. Zhu, D. G. Doherty, Clin. Millet, P. Desreumaux, E. Henon, P. Chavatte, J. Med. Chem. 2014, 57, Immunol. 2011, 140, 196–207. 5489–5508. [19] A. Alcaide, A. Llebaria, Tetrahedron Lett. 2012, 53, 2137–2139. [2] A. Bendelac, P. B. Savage, L. Teyton, Annu. Rev. Immunol. 2007, 25, 297– [20] Schrödinger, Schrödinger Suite 2014 Protein Preparation Wizard; Epik, 336. Schrödinger, LLC, New York, NY, 2014; Impact, Schrödinger, LLC, New York, [3] K. Miyamoto, S. Miyake, T. Yamamura, Nature 2001, 413, 531–534. NY, 2014; Prime, Schrödinger, LLC, New York, NY, 2014. [4] R. D. Goff, Y. Gao, J. Mattner, D. Zhou, N. Yin, C. Cantu 3rd, L. Teyton, A. [21] Schrödinger, Schrödinger Suite 2014–3, Schrödinger, LLC: New York, NY, Bendelac, P. B. Savage, J. Am. Chem. Soc. 2004, 126, 13602–13603. 2014. [5] M. Fujio, D. Wu, R. Garcia-Navarro, D. D. Ho, M. Tsuji, C. H. Wong, J. Am. [22] Schrödinger, Desmond Molecular Dynamics System, D. E. Shaw Research, Chem. Soc. 2006, 128, 9022–9023. New York, NY, 2014. Maestro-Desmond Interoperability Tools, Schrödinger, [6] J. J. Park, J. H. Lee, K. C. Seo, G. Bricard, M. M. Venkataswamy, S. A. New York, NY, 2014. Porcelli, S. K. Chung, Bioorg. Med. Chem. Lett. 2010, 20, 814–818. [23] J. López-Sagaseta, J. E. Kung, P. B. Savage, J. Gumperz, E. J. Adams, PLoS [7] Y. J. Chang, J. R. Huang, Y. C. Tsai, J. T. Hung, D. Wu, M. Fujio, C. H. Biol. 2012, 10, e1001412. DOI 10.1371/journal.pbio.1001412. Wong, A. L. Yu, Proc. Natl. Acad. Sci. USA 2007, 104, 10299–10304. [8] T. Toba, K. Murata, K. Nakanishi, B. Takahashi, N. Takemoto, M. Akabane, T. Nakatsuka, S. Imajo, T. Yamamura, S. Miyake, H. Annoura, Bioorg. Med. Chem. Lett. 2007, 17, 2781–2784. [9] S. Inuki, T. Aiba, N. Hirata, O. Ichihara, D. Yoshidome, S. Kita, K. Maenaka, K. Fukase, Y. Fujimoto, ACS Chem. Biol. 2016, 11, 3132–3139. [10] J. Kishi, S. Inuki, E. Kashiwabara, T. Suzuki, N. Dohmae, Y. Fujimoto, ACS Manuscript received: December 23, 2020 Chem. Biol. 2020, 15, 353–359. Revised manuscript received: March 10, 2021 [11] Y. Harrak, C. M. Barra, C. Bedia, A. Delgado, A. R. Castaño, A. Llebaria, Accepted manuscript online: April 6, 2021 ChemMedChem 2009, 4, 1608–1613. Version of record online: June 17, 2021 ChemMedChem 2021, 16, 2491 – 2496 www.chemmedchem.org 2496 © 2021 The Authors. ChemMedChem published by Wiley-VCH GmbH