Modern approaches to designing novel CD80 inhibitors that mimic biological effects of Abatacept
Asha Kulkarni Almeida
T. T. Matan
Disruption of CD80-CD28 interactions is an established target for rheumatoid arthritis, transplant rejection, systemic lupus erythematosus, non-Hodgkin’s lymphoma and other inflammatory disorders where suppression of T cell activity is desirable. In this study, we developed pharmacophore models for small molecule inhibitors of this interaction, with a view to identifying novel chemical scaffolds that would necessarily block the T cell activation as induced by Abatacept. We screened the Asinex database using these models. We tested 31 of the in silico hits in vitro and identified 7 active scaffolds, which were different from those used for developing the pharmacophore models. These 7 scaffolds may provide good starting points for further medicinal chemistry efforts to develop novel and potent CD80 inhibitors.
T-cell activation plays a critical role in driving normal and pathogenic immune responses. The interaction between cell surface co-stimulatory molecules on antigen-presenting cells and the T cells is essential for modulating cell to cell communication and the immune response. Two structurally and functionally well characterized T cell surface co-stimulatory molecules are CD28 and CD152 (CTLA-4), both of which bind the ligands CD80 and CD86, on the antigen-presenting cells (APCs) (Figure 1)[2-4]. Disruption of CD80-CD28 interactions (which is involved in T cell activation) is an established target for rheumatoid arthritis (RA), transplant rejection, non-Hodgkin’s lymphoma (NHL) and other applications where suppression of T cell activity is desirable.
CD28 is constitutively expressed on naive T-cells while CD80 expression is upregulated in established disease. CD28 is believed to be involved in many activities other than just immune response and hence targeting it may result in unpredictable consequences[7,8]. The dangerous outcomes of targeting T cell activation through inhibition of CD28 were clearly seen in the Phase-1 clinical trial results with the CD28 superagonist antibody TGN1412 (developed by TeGenero) in March 2006. Treatment with the CD28 superagonist antibody TGN1412 along-side blocking of the CD28 receptors expressed on activated T cells, also induced a life-threatening cytokine-release syndrome in six volunteers recruited in this Phase I clinical trial. A retrospective investigation on the devastating effects of the massive cytokine storm generated in human volunteers who received a single infusion of the humanized monoclonal antibody TGN1412 and comparative rodent studies in mice as well as rats has suggested that TGN1412 rapidly induced a marked T cell lymphopenia by trapping T cells in the spleen and lymph nodes of mice. Possibly, a similar dramatic redistribution of T cells simulated profound T cell lymphopenia observed in human recipients of TGN1412. The abject failure of targeting CD28 receptor and the unpredictable nature of targeting T cells which are indispensable in mounting cell mediated immune responses, brings focus towards the CD28 co-stimulating receptor CD80, which is expressed on antigen presenting cells. Incidentally, galiximab (anti-CD80 antibody)-induced growth inhibition and prolongation of survival in vivo of B-NHL tumor xenografts. What further strengthens prospects of targeting CD80 is the success of Abatacept.
Abatacept is a biologic (a fusion protein comprising of CTLA4 bound to immunoglobulin) which mimics CD152 (CTLA4) the co-stimulatory molecule that down regulates expression of CD80-CD28, thus suppressing an inflammatory response. Abatacept when it binds to CD80/CD86 instead, it opposes co-stimulatory molecules namely CD80 – CD28 and prevents or suppresses T-cell activation. The mechanism of Abatacept binding to the CD80/CD86 markers downregulates the CD28 expression on T cells. It has been approved for RA in patients with an inadequate response to anti-tumor necrosis factor (anti-TNF) therapy as a second line approach. Side effects include infections such as respiratory or urine tract infections, and also tuberculosis is a potential problem similar to other anti-TNF drugs. The side effects are therefore similar to all other biologic drugs. Among multiple ongoing clinical trials, an interim 6-month analysis of clinical trial data from ACTION (Abatacept In Routine Clinical Practice); a prospective observational study assessing effectiveness, safety, and tolerability of Abatacept in patients with RA, suggests that it offers an effective and well-tolerated treatment option for patients with RA, including those who have previously failed anti-TNF treatment. In addition, higher retention rates and effectiveness outcomes were observed when Abatacept treatment was initiated earlier in the course of the disease. Further to responses in RA patients, a clinical trial to assess the efficacy and safety of a 24-week course of Abatacept in the treatment of active lupus nephritis appears promising.
CD80 inhibition, therefore, represents a logical choice of target for a small-molecule program in autoimmune disease. There is a clear, unmet medical need for a new oral therapeutic for rheumatoid arthritis, which intervenes early in the disease process, treats the causes and not the symptoms of disease, has minimal side-effects and leaves the immune system functional[11-13]. Inhibition of CD80-CD28 interaction may be a promising strategy to fulfill this need.
Several anti-CD80 monoclonal antibodies are in clinical trials, but the major drawbacks of these protein-based therapies are their high treatment cost, reactivation of dormant tuberculosis, induction of lymphomas on long term exposure and requirement for parenteral administration with side-effects. A series of condensed aromatic peptide inhibitors of CD80 with increased cellular activity was identified. However, these peptides showed nonspecific binding, and hence were thought to be poor lead candidates. Small molecule inhibitors of CD80-CD28 interaction have been reported, but were found to lack potency under physiological conditions, possibly due to fast dissociation of ligand from protein. Hence it was suggested that new small molecule inhibitors should have much slower dissociation rates.
Activity data for several CD80 small molecule inhibitors that belong to diverse chemical classes are currently available[6,16,17]. One of these, Rhudex, an orally bioavailable small molecule inhibitor of CD80, was in Phase II clinical trials to prevent inflammation in rheumatoid arthritis[6,18]. However, the trial was terminated due to death of a subject participating in the trial who suffered cardiac arrest. There is no other small molecule in clinical trials for this target, although there are many in biological testing. Recent developments involve collaborations between the company Medigene with Dr. Falk Pharma for further Phase II studies with Rhudex in the treatment of gastrointestinal and hepatological disorders. The aim of this study was to identify new classes of molecules for CD80 inhibition and hence, hitherto-unknown starting structures for researchers to use in designing novel inhibitors. Towards this end, we used virtual screening as a time- and cost-saving approach to shortlist compounds for biological screening. Among the different virtual screening approaches available, docking is perhaps the most popular. However, crystal structure information for interactions of CD80-CD28 and CD80-small molecule complexes are not available. Pharmacophore modeling is an alternate effective, inexpensive and fast approach to discover useful starting points in drug discovery research. This prompted us to develop three-dimensional pharmacophore models using known CD80 inhibitors for virtual screening of a large compounds database.
A “pharmacophore” is the three dimensional (3D) arrangement of atoms or functional groups essential for the molecule to bind to a specific receptor. Pharmacophore models can be used to discover or design new bioactive compounds. The screening of compounds databases using pharmacophore models is a method that has been successfully applied in many drug-discovery programs[19,20]. In this study, 3D pharmacophore model(s) were developed from known CD80 inhibitors, validated using an external test set of known inhibitors, and used for virtual screening of the Asinex database. Some of the virtual hits were tested in biological assays and found to be active. These compounds have scaffolds that are completely different from those used for developing the pharmacophore models, and may provide novel starting points for future research.
Eight molecules[6,16,17,21] with known activities (Figure 2) were used to generate the pharmacophore models.
The molecules were built using Maestro (version 8.0) 2D/3D visualizer using standard procedures. [22-26] These were loaded into the Phase (v2.5) module. A single low energy conformation was generated for each molecule.
For each of the eight molecules shown in Figure 2, we identified pharmacophoric features using the Phase program. Combined pharmacophore models containing three, four and five pharmacophoric features were then generated. Two sets of 7 molecules each provided two reasonably good pharmacophoric hypothesis, which covered all 8 molecules in the training set.
The two four-point variants were ARRR (A – Acceptor, R – Ring aromatic) and AARR. Hypothesis with the highest survival scores from each of the 2 sets (4.733 & 4.264 respectively) were selected as Hypothesis 1 and Hypothesis 2, respectively. The features and distance patterns of the two hypotheses are shown in Figure 3. We found that fitness scores were higher for Hypothesis 1 than for Hypothesis 2, and this is reflected in the survival score also. The selectivity score for Hypothesis 1 was also slightly higher than that for Hypothesis 2. Hence, Hypothesis 1 was expected to perform better than Hypothesis 2.
The molecule that contributes to the reference pharmacophore is the reference ligand having highest fitness of 3.00. The reference ligands for Hypotheses 1 and 2 are compounds 3 and 8 respectively. (Figure 3). The alignment for Hypothesis 1 appears to be better than that for Hypothesis 2 (Figure 4).
The pharmacophore hypothesis obtained were validated against a set of 39 molecules from Prous/Integrity database. We found that Hypothesis 1 was able to pick up all the molecules with fitness score above 2.00, whereas Hypothesis 2 was able to pick up only 26 molecules out of 39. This is because the aromatic ring feature attached to the tri-cyclic fused ring (Figures 2 & 3) is not present in 9 of the molecules and also the H-bond acceptor feature (present in the tricyclic ring on nitrogen) corresponds to a donor feature in the molecules that were rejected.
After confirming their reliability, we employed Hypotheses 1 & 2 as 3D search queries on the Asinex database (containing 127,281 molecules) which was already Lipinski-filtered and had pre- generated 3D conformations and pharmacophoric sites of compounds. Fitness score, which measures how well the conformer matches the pharmacophoric hypothesis, was used to select “hits”. Thus, 61,791 hits were obtained for Hypothesis 1 and 65,654 hits for Hypothesis 2. Then we applied a further fitness score filter, to select only the molecules that have fitness greater than or equal to 2.00. The number of hits then reduced to 200 and 207 for Hypothesis 1 and 2, respectively. Interestingly, there was no common molecule between the two sets of virtual screening hits. This indicates that the two hypothesis are independent and compounds missed by one may be picked up by the other. Hence, it is important to use both the hypothesis to increase chances of picking up all true positives, particularly those with novel scaffolds not used in the training set.
For the purpose of biological testing, 31 compounds were randomly selected from the 407 potential hits. Interestingly, all of these compounds had scaffolds that were different from those of compounds in the training set, demonstrating the validity of virtual screening using pharmacophoric models for finding novel active scaffolds from a database of molecules.
The selected compounds were screened in a human peripheral blood mononuclear cell (hPBMC) assay at a concentration of 100µM, and the functional end-point of T-cell proliferation inhibition was measured (see Figure 1). PBMCs were isolated from the peripheral blood of healthy human volunteers as previously reported. The PBMC were treated with the compounds for 30 minutes prior to stimulation. T-cell proliferation was induced in presence of phytohemagglutinin and phorbol ester. Finally, the extent of proliferation was measured by BrdU labeling of proliferating T-cell DNA. Dexamethasone (Sigma) and anti-human CD80 monoclonal antibody (Southern Biotech, USA) were used as standards in the biological testing.
Sixteen compounds showed 50% or higher inhibition of T-cell proliferation (Figure 5). The effect was equivalent to the effect of monoclonal antibody to human-CD80 at 25µg/ml (62.2±12.4% inhibition). We used dexamethasone as a standard small molecule inhibitor of CD28 – CD80 interactions because it has been reported to downregulate CD80 and CD28 receptors on APCs and T cells respectively[29,30]. Dexamethasone at 10µM showed 72.3±5.9% inhibition of T-cell proliferation in parallel biological experiments.
The 31 tested compounds could be classified into 11 scaffold classes, of which 5 contained more than 1 compound. 3 out of 4 compounds in scaffold class 1 (SC1) showed more than 50% inhibition in T-cell proliferation assay. Only 2 of the 6 singleton scaffolds showed more than 50% inhibition, giving a total of 7 active scaffolds.
In summary, the work described here identified two four-point pharmacophore models for disrupting the interaction of CD80 with CD28. Further, preliminary results of biological testing of virtual screening hits, provided new hits from 7 different scaffolds which were hitherto unknown for this activity, and may provide good starting scaffolds for the synthesis of novel and potent CD80 inhibitors by researchers. Recent advances with abatacept describe the success of this antibody to the T-cell co-stimulatory molecule CD80, in inducing remission in five patients with focal segmental glomerulosclerosis (FSGS) resistant to rituximab and glucocorticoids (one patient with primary FSGS and four with recurrent FSGS after transplantation). The rationale for using Abatacept was that CD80 is induced in podocytes in primary and recurrent FSGS as well as in patients with membranous nephropathy. Similarly reports from clinical trials with patients diagnosed for rheumatoid arthritis are available. The clinical data suggests that Abatacept plus methotrexate (MTX) demonstrated robust efficacy compared with MTX alone in early RA, with a good safety profile. The achievement of sustained remission following withdrawal of all RA therapy suggests an effect of abatacept’s mechanism on autoimmune processes. Interestingly, withdrawal from the antibody treatment has not generated undesirable effects but has also sustained remission suggesting that the immune mechanisms targeted by the antibody successfully achieve the desired remission and end points desirable in RA patients whereas in contrast the patients exposed to the drug MTX do not exhibit similar remission patterns. This further underscores the significance of the target in immune modulatory responses with abatacept. In context with our results the role of CD28-CD80 interaction is once again prominently highlighted and is therefore suggested as a good target for novel drug designing.
We would like to thank Dr. Jyothi Subramanian for her suggestions and Dr. Periyasamy Giridharan for his excellent technical assistance. We are grateful to Dr. Somesh Sharma for many useful discussions and directions.
The authors have no conflict of interest to declare.
- Sørensen, P.; Kussmann, M.; Rosén, A.; Bennett, K. L.; Thrige Dda, G.; Uvebrant, K.; Walse, B.; Roep- storff, P.; Björk, P. J. Immunol. 2004 172, 6803.
- Chambers, C. A.; Allison, J. P. Curr. Opin. Immu- nol. 1997 9, 396.
- Tremoulet, A. H.; Albani, S. Expert Opin. Inves- tig. Drugs 2006 15, 1427.
- Carreno, B. M.; Carter, L. L.; Collins, M. Curr. Opin. Pharmacol. 2005 5, 424.
- Bluestone, J. A.; Bour-Jordan, H.; J. Clin. Immu- nol. 2002 22, 1.
- Huxley, P.; Sutton, D.H.; Debnam, P.; Mathews, I.R.; Brewer, J. E.; Rose, J.; Trickett, M.; Wil- liams, D. D.; Anderson, T. B.; Classon, B. J. Chem. Biol. 2004 11, 1651.
- Daikh, D.; Wofsy, D.; Imboden, J. B. J. Leukoc. Biol. 1997 62, 156
- Suntharalingam, G.; Perry, M. R.; Ward, S.; Brett, S. J.; Csatello-Cortes, A.; Brunner, M. D.; Pan- oskaltsis, N. N. Engl. J. Med. 2006 355, 1018.
- Hünig, T. Nature Reviews Immunology 2012 12, 317-318
- Hariharan, K., Chu, P., Murphy, T., Clantoni, D., Berquist, L., Molina, A., Ho, S. N., Vega, M. I. and Bonavida, B. Int. J. Of Oncology 2013 43, 670-676,
- Shimoyama, Y.; Nagafuchi, H.; Suzuki, N.; Ochi, T.; Sakane, T. J. Rheumatol. 1999 26, 2094.
- Van der Merwe, P.A.; Bodian, D.L.; Daenke, S.; Linsley, P.; David, S.J. J. Exp. Med. 1997 185, 393.
- Mackenzie, N.M. Drug Discov. Today 2006 11, 952.
- Kuchroo, V.K.; Das, M.P.; Brown, J.A.; Ranger,A.M.; Zamvil, S.S.; Sobel, R.A.; Weiner, H.L.; Nabavi, N.; Glimcher, L.H. Cell 1995 80, 707.
- Hegde, V.R.; Puar, M.S.; Dai, P.; Patel, M.; Gullo, V.P.; Chan, T.M.; Silver, J.; Pramanik, B.N.; Jenh, C.H. Bioorg. Med. Chem. Lett. 2003 13, 573.
- Erbe, D.V.; Wang, S.; Xing, Y.; Tobin, J.F. J. Biol. Chem. 2002 277, 7363.
- Green, N.J.; Xiang, J.; Chen, J.; Chen, L.; Davies, A.M.; Erbe, D.; Tam, S.; Tobin, J.F. Bioorg. Med. Chem. 2003 11, 2991.
- Howat, D.W.; Blades, M.; Bombardieri, M.; Baker, R.; Pitzalis, C.; Millar, A.; Jakobsen, B.; Mackenzie, N.M. Ann. Rheum. Dis. 2006 65, 693.
- Milne, G.W.; Wang, S.; Nicklaus, M.C. J. Chem. Inf. Comput. Sci. 1996 36, 726.
- Langer, T.; Wolber, G. Drug Discov. Today: Tech- nologies 2004 1, 203.
- http://integr ity.prous.com/integr ity/s er v let/xmlsl/
- Schrödinger Inc., LLC, New York, NY, 2006.
- Dixon, S.L.; Smondyrev, A.M.; Knoll, E.H.; Rao, S.N.; Shaw, D.E.; Freisner, R.A. J. Comput Aided Mol. Des. 2006 20, 647.
- Dixon, S.L.; Smondyrev, A.M.; Rao, S.N. Chem. Biol. Drug Des. 2006 67, 370.
- Halgren, T.A. J Comput. Chem.1996 17, 520.
- Schrodinger Inc., MacroModel 2.0 User Manual, LLC, New York, NY, 2006.
- Wolf, A.M.; Wolf, D.; Rumpold, H.; Ludwiczek, S.; Enrich, B.; Gastl, G.; Weiss, G.; Tilg, H. Proc. Natl. Acad. Sci. U S A. 2005 102, 13622.
- Zhu, J.W.; Field, S.J.; Gore, L.; Thompson, M.; Yang, H.; Fujiwara, Y.; Cardiff, R.D.; Greenberg, M.; Orkin, S.H.; DeGregori, J. Mol. Cell Biol. 2001 21, 8547.
- Agarwal, S.K.; Marshall, Jr. G.D. J. Interferon Cy- tokine Res. 2000 20, 927.
- Mogensen, T.H.; Berg, R.S.; Paludan, S.R.; Øster- gaard, L. Infect. Immun. 2008 76, 189.
- Alachkar, N., Carter-Monroe, N. and Reiser, J. N Engl J Med. 2014 370, 1261-1266
- Emery, P. et al. Ann. Rheum. Dis. 2014 0, 1–8. doi:10.1136/annrheumdis-2014-206106
- Nüßleinetal. BMC Musculoskeletal Disorders 2014 15, 14