|
IMMUNO-DYN: Helios Biosciences
dynamic model of T lymphocytes to characterise drug targets inhibiting
the Th1 immune response.
The T-helper (Th1) lymphocyte response:
T-lymphocytes are immune system cells that play a major part in regulating immunity. Among the various T-lymphocyte responses to antigens is the Th1 response involved in anti-cellular immunity. Anti-cellular immunity is aimed at eliminating alien/ill cells from the body, such as virus-infected cells, cancer cells, grafted cells, senescent cells. Most often this immune response is adequate and protects us. Among other immune system responses are the Th2 response generating inflammation and allergy (also involving T-lymphocytes), and the antibody production by B-lymphocytes.
In some cases the Th1 response leads to pathology or deleterious
effects:
Autoimmune diseases are due to Th1 response pathologically aimed
at some sane cells in our body. By killing these cells and/or
impairing the behaviour of the corresponding organs it leads to
diseases such as diabetes, rheumatoid arthritis, kidney failure.
Graft rejection is due in great part to a normal Th1
response against the cells of the grafted organ. It leads to the
destruction of the graft which was implanted to cure the patient.
In these cases it is crucial to inhibit the Th1 response, with
immunosuppressive drugs.
Current immunosuppressive drugs:
All immunosuppressive drugs used in transplantation
non-specifically inhibit most or all immune responses (Corticoids,
Cyclosporine, Tacrolimus, Rapamycine, Azathioprine, Methotrexate,
Mycophenolate mofetil), have major side effects (Corticoids,
Cyclosporine, Tacrolimus, Rapamycine, Azathioprine, Methotrexate,
Mycophenolate mofetil, Muromonab CD3: Anti-CD3 antibody), or are used
only at early transplantation stage and need to be associated to other
drugs (daclizumab, basiliximab: anti-CD25 antibodies). In nearly all
cases the organ transplantation prognosis is chronic rejection
evolving towards the need of re-transplantation.
Some of the above quoted drugs are also used in autoimmune
diseases. Other drugs used in these indications such as gold salts,
D-penicillamine, hydroxychloroquine have numerous side
effects. Recently new drugs have been developed: Infliximab,
Adalimumab, Etanercept for rheumatoid arthritis and Crohn's disease,
Anakinra for rheumatoid arthritis. These recent drugs bring a benefit
but still have to be associated to older ones, have to be injected,
and still do not solve the therapeutic issue. Moreover many autoimmune
diseases (such as diabetes) are not yet specifically targeted by
efficient and better tolerated drugs.
Therefore despite many progresses the need for more efficient
and better tolerated new immunosuppressive drugs is huge and research
in this field remains active, in particular in pharmaceutical
companies.
To address the issue of Th1 immunosuppression, we computed a
dynamic model of the human T-lymphocyte intracellular signal
transduction network in response to an antigenic stimulation in the
context of a Th1 response.
This model includes 338 molecules, 1498 interactions between them,
338 variables and 2,174 parameters. It has a structure and a
complexity very similar to real biological networks.
IMMUNO-DYN data inputs
Two datasets were used to build IMMUNO-DYN:
The description of the signal transduction network. The
T-lymphocyte Th1 response consists of cell differentiation /
activation, proliferation and apoptosis depending on stimulus and time
after stimulus. The network we have compiled combines cascades of
general and T-lymphocyte specific signal transductions, cell cycle and
apoptosis. This network is stored in our SIMPathways database.
|
| Image of the T-lymphocyte signal transduction network
used in IMMUNO-DYN (Red dots : molecules. Blue arrows : directed
interactions) |
The description of the behavior of this signal transduction
network during the Th1 response.We used gene expression data
collected during the time-course of the Th1 response. These data were
taken from the scientific article PMID: 12195013.
We used the data describing the response to a CD3+CD28
stimulation. This stimulation mimics the presentation of antigens to
T-lymphocytes and triggers a Th1 response. The behaviors of the signal
transduction molecules were described at seven times (0h, 1h, 2h, 6h,
12h, 24h, 48h).
IMMUNO-DYN Computation
The parameters of the model were computed by an adaptive system
method (gradient descent).
Parameter calculation: adaptive system method
 |
| pink: modeled behaviors during a 48 hours time-course,
blue: measured behaviors during a 48 hours time-course. These figures
are representative of the results obtained for all molecules of the
network. These 4 molecules are in distant areas of the
network |
HELIOS BIOSCIENCES has implemented an adaptive learning method,
which includes the mathematical framework for the calculations
(algorithms), the modelling software, the setting up of adequate
initialisation conditions to start the calculations and the setting up
of calculation constraints. This modelling platform is fully
proprietary. We use this neural network solution because it is a
reasonable compromise between modest "a priori" information on the
kinetic behaviour of the molecules involved in the signal transduction
networks and the large quantity of quantitative data potentially
available, and because the training procedure permits to "learn" in an
"a posteriori" fashion the kinetic parameters by extracting them from
time series expression data.
Our computed dynamic model fairly reproduces the Th1 response
experimental data. It also mimics the non linearity and heterogeneity
of molecule behaviors. The overall fit of the model with experimental
data is >82.5% (relative mean square error < 17.5%). This result is
very good given the complexity of the network and the heterogeneity of
the gene's expression kinetics.
Drug target identification
Our team used the IMMUNO-DYN dynamic model to identify drug targets
the inhibition of which will inhibit the th1 response. This was
performed through simulations.
During each simulation, the network was stimulated by CD3 + CD28
and let to evolve in response to the combined action of this stimulus
and the inhibition of one of its molecules. The inhibition of each
molecule of the network was successively tested (338
simulations). Each T-cell response was simulated during 48 in silico
hours. The distances were then calculated between the simulated states
reached by the network and 2 reference states: the experimental data
before stimulation (unstimulated state) and 48 hours after stimulation
without inhibition of any molecule (responsive state).
The drug targets were defined as molecules which inhibition
inhibits the T-cell response.
28 potential drug targets were selected through simulations, of
which 13 are already confirmed by literature analysis, 1 is rejected
based on expected side effects and 14 remain to be tested.
Similar results were obtained when the distances were calculated
through the whole time-courses (0hà48h). This result is highly
significant (p<10(exp-10)) and fully validates our modeling
technology.
 | SELECTION OF
POTENTIAL DRUG TARGETS
Systematic inhibition of molecules of the network :
Inhibition of T-Lymphocyte response to antigen presentation |
Helios Biosciences drug target bunch
Our company has identified 27 potential immunosuppressive drug
targets, of which 13 are already confirmed by literature analysis
based on inhibition effect on the Th1 immune response, and 14 remain
to be tested.
3 of our immunosuppressive drug targets are specifically or
un-specifically targeted by existing or currently developed
drugs.
The remaining 24 are not targeted to date, 11 of which are highly drugable (receptors, receptor ligands, kinases, phosphatases).
At least 4 of our highly drugable immunosuppressive drug targets
present an extensive freedom to operate in immunology.
| Immunosuppressive Targets |
Validated |
New Targets To Be Validated |
Total Targets |
Rejected (Expected Side Effects) |
| Total Targets |
13 |
14 |
27 |
1 |
| Receptor |
1 |
2 |
3 |
0 |
| Receptor Ligand |
0 |
1 |
1 |
0 |
| Linker |
3 |
1 |
4 |
0 |
| Kinase |
3 |
3 |
6 |
0 |
| Phosphatase |
2 |
0 |
2 |
0 |
| Transcription Factor/Cofactor |
4 |
5 |
9 |
1 |
| Other |
0 |
2 |
2 |
0 |
Examples of Helios Biosciences
immunosuppressive drug targets
All our drug targets are highly argued based on experimental
data. We provide here three examples.
Original drug target:
TNF receptor-associated factor 2
TRAF2, GeneID: 7186 Our model predicts that inhibiting TRAF2
will inhibit the T-lymphocyte Th1 response. This prediction is
confirmed by literature analysis (1,2).
The TRAF2 protein is a signal transducer for the receptor of the Tumour Necrosis Factor (TNF-alpha,
TNF) (3). TNF signalling is important in the immune response
(1,2).
Three drugs that inhibit the binding of TNF on its receptor
are currently used to treat auto-immune diseases such as arthritis:
two are anti-TNF antibodies (one is infliximab from Schering-Plough,
the other is adalimumab from Abbott Laboratories), the third being a
chimeric recombinant protein combining a portion of the TNF receptor
and a portion of an antibody (etanercept from Wyeth). These drugs are
proteins so they cannot be administered orally and have to be
injected. To our knowledge no synthetic drug has been designed so far
to specifically inhibit the TRAF2 mediated signal
transduction. We believe this would indeed bring an alternative
therapeutic strategy to the global blocking of TNF signalling and
lead to a relevant and orally administered treatment for autoimmune
diseases / graft rejection, in which the Th1 lymphocyte response is
involved.
(1) Cannons JL, Bertram EM, Watts TH.: Cutting edge: profound defect in T cell responses in TNF receptor-associated factor 2 dominant negative mice, J Immunol. 2002 Sep 15;169(6):2828-31, PMID: 12218092.
(2) Lee SY, Reichlin A, Santana A, Sokol KA, Nussenzweig MC, Choi Y.: TRAF2 is essential for JNK but not NF-kappaB activation and regulates lymphocyte proliferation and survival, Immunity. 1997 Nov;7(5):703-13, PMID: 9390693.
(3) Rothe M, Xiong J, Shu HB, Williamson K, Goddard A, Goeddel DV.:
I-TRAF is a novel TRAF-interacting protein that regulates
TRAF-mediated signal transduction, Proc Natl Acad Sci U S A. 1996 Aug
6;93(16):8241-6, PMID: 8710854.
Drug targets being already used in clinics / preclinical trials:
Caspase 1, apoptosis-related cysteine peptidase
interleukin 1, beta, convertase, GeneID: 834Our model
predicts that inhibiting caspase1 will impact on the T-lymphocyte Th1
response.
Caspase1 is a protease that activates interleukin 1 beta. Interleukin
1 beta is a proinflammatory cytokine (1). The pharmacological
inhibition of caspase 1 by the specific inhibitor pralnacasan
inhibits the T-lymphocyte Th1 response in an in vivo murine model of
inflammatory colitis (2). Drugs inhibiting caspase 1 are proposed in
auto-inflammatory diseases such as Familial cold autoinflammatory
syndrome, Muckle-Wells syndrome and neonatal onset multisystem
inflammatory disease (3). As an example, VX-765, an orally active
IL-converting enzyme/caspase-1 inhibitor, is currently tested in
clinical trials in psoriasis by Vertex Pharmaceuticals Inc. Psoriasis
is recognized as a T-cell mediated autoimmune inflammatory disease
resulting in epidermal hyperproliferation (4).
Inhibiting caspase 1 is also proposed as a strategy for
neuroprotection and investigated in animal models (5).
(1) Gemma C, Fister M, Hudson C, Bickford PC.: Improvement of memory for context by inhibition of caspase-1 in aged rats, Eur J Neurosci. 2005 Oct;22(7):1751-6, PMID: 16197515.
(2) Loher F, Bauer C, Landauer N, Schmall K, Siegmund B, Lehr HA, Dauer M, Schoenharting M, Endres S, Eigler A.: The interleukin-1 beta-converting enzyme inhibitor pralnacasan reduces dextran sulfate sodium-induced murine colitis and T helper 1 T-cell activation, J Pharmacol Exp Ther. 2004 Feb;308(2):583-90. Epub 2003 Nov 10, PMID: 14610233.
(3) Stack JH, Beaumont K, Larsen PD, Straley KS, Henkel GW, Randle JC, Hoffman HM.: IL-converting enzyme/caspase-1 inhibitor VX-765 blocks the hypersensitive response to an inflammatory stimulus in monocytes from familial cold autoinflammatory syndrome patients, J Immunol. 2005 Aug 15;175(4):2630-4, PMID: 16081838.
(4) Chow S, Rizzo C, Ravitskiy L, Sinha AA.: The role of T cells in cutaneous autoimmune disease, Autoimmunity. 2005 Jun;38(4):303-17. Review, PMID: 16206513.
(5) Cao Y, Gu ZL, Lin F, Han R, Qin ZH.: Caspase-1 inhibitor Ac-YVAD-CHO attenuates quinolinic acid-induced increases in p53 and apoptosis in rat striatum, Acta Pharmacol Sin. 2005 Feb;26(2):150-4, PMID: 15663890.
Lymphocyte-specific protein tyrosine kinase
LCK, GeneID: 3932
Our model predicts that inhibiting LCK will inhibit the T-lymphocyte
Th1 response.
This prediction is confirmed by literature analysis (3, 4, 5,
6).
LCK tyrosine kinase is a proximal signal transducer for the T-Cell
Receptor (TCR) (1). The TCR is a membrane receptor that recognizes
antigens and triggers the subsequent T-lymphocyte response. LCK is
also involved in mutations that cause T-cell acute lymphoblastic
leukemia (2). Circulating T cells from the LCK knockout mice have a
significantly decreased TCR response in terms of proliferation,
calcium mobilization, and IL-2 production (3, 4), they do not produce
IL-2 or proliferate in response to CD3/CD28 stimulation (5, 6). LCK
inhibitors have thus been proposed as a therapeutic approach to
autoimmune diseases and transplant rejection (7). Abbott Laboratories
develops a new drug, A770041, which is a specific inhibitor of LCK
and prevents heart allograft rejection in rodents (8). Another drug,
Imatinib from Novartis, is an inhibitor of kinases ABL, ARG, PDGFR
{alpha}and {beta}, and c-KIT, used in the treatment of chronic
myeloid leukaemia, Philadelphia chromosome\x{2013}positive acute
lymphocytic leukaemia, myleoproliferative disorders due to
chromosomal rearrangements in the PDGF-R locus and gastrointestinal
stromal tumors with mutations in c-KIT. It has been recently shown
that Imatinib also directly inhibits LCK and TCR signalling with
powerful suppressive effects on T-lymphocyte proliferation and
function (9, 10).
(1) Ahmed Z, Beeton CA, Williams MA, Clements D, Baldari CT, Ladbury JE.: Distinct spatial and temporal distribution of ZAP70 and Lck following stimulation of interferon and T-cell receptors, J Mol Biol. 2005 Nov 11;353(5):1001-10. Epub 2005 Sep 27, PMID: 16219325.
(2) Burnett RC, Thirman MJ, Rowley JD, Diaz MO.: Molecular analysis of the T-cell acute lymphoblastic leukemia-associated t(1;7)(p34;q34) that fuses LCK and TCRB, Blood. 1994 Aug 15;84(4):1232-6, PMID: 8049439.
(3) Straus DB, Weiss A.: Genetic evidence for the involvement of the lck tyrosine kinase in signal transduction through the T cell antigen receptor, Cell. 1992 Aug 21;70(4):585-93, PMID: 1505025.
(4) Trobridge PA, Levin SD.: Lck plays a critical role in Ca(2+) mobilization and CD28 costimulation in mature primary T cells, Eur J Immunol. 2001 Dec;31(12):3567-79, PMID: 11745376.
(5) Legname G, Seddon B, Lovatt M, Tomlinson P, Sarner N, Tolaini M, Williams K, Norton T, Kioussis D, Zamoyska R.: Inducible expression of a p56Lck transgene reveals a central role for Lck in the differentiation of CD4 SP thymocytes, Immunity. 2000 May;12(5):537-46, PMID: 10843386.
(6) Seddon B, Legname G, Tomlinson P, Zamoyska R.: Long-term survival but impaired homeostatic proliferation of Naive T cells in the absence of p56lck, Science. 2000 Oct 6;290(5489):127-31, PMID: 11021796.
(7) Kamens JS, Ratnofsky SE, Hirst GC.: Lck inhibitors as a therapeutic approach to autoimmune disease and transplant rejection, Curr Opin Investig Drugs. 2001 Sep;2(9):1213-9, PMID: 11717807.
(8) Stachlewitz RF, Hart MA, Bettencourt B, Kebede T, Schwartz A, Ratnofsky SE, Calderwood DJ, Waegell WO, Hirst GC.: A-770041, a novel and selective small-molecule inhibitor of Lck, prevents heart allograft rejection, J Pharmacol Exp Ther. 2005 Oct;315(1):36-41. Epub 2005 Jul 12, PMID: 16014572.
(9)Seggewiss R, Lore K, Greiner E, Magnusson MK, Price DA, Douek DC, Dunbar CE, Wiestner A.: Imatinib inhibits T-cell receptor-mediated T-cell proliferation and activation in a dose-dependent manner, Blood. 2005 Mar 15;105(6):2473-9. Epub 2004 Nov 30, PMID: 15572591.
(10)Fabian MA, Biggs WH 3rd, Treiber DK, Atteridge CE, Azimioara MD, Benedetti MG, Carter TA, Ciceri P, Edeen PT, Floyd M, Ford JM, Galvin M, Gerlach JL, Grotzfeld RM, Herrgard S, Insko DE, Insko MA, Lai AG, Lelias JM, Mehta SA, Milanov ZV, Velasco AM, Wodicka LM, Patel HK, Zarrinkar PP, Lockhart DJ.: A small molecule-kinase interaction map for clinical kinase inhibitors, Nat Biotechnol. 2005 Mar;23(3):329-36. Epub 2005 Feb 13, PMID: 15711537.
Potential side effects
The potential indications of our drug targets are severe
diseases. Graft rejection and many autoimmune diseases engage the
vital prognosis of patients. In such situations side effects may be
accepted. However we searched for potential major side effects that
might be expected when inhibiting these molecules. One potential drug
target was excluded from the final selection due to expectation of
unacceptable side effects (the inhibition of this target may be
oncogenic).
Technological validation
A first validation of the IMMUNO-DYN dynamic model is its fitting
with the Th1 response experimental data (>82.5%). However the overall
validation comes from its predictive power.
The drug target validation ratio of the IMMUNO-DYN dynamic model is
72%. Its prediction sensitivity is estimated between 20% and 40% and
its prediction specificity is 94%.
Although this model may be improved, in its current state of art it
predicts innovative drug targets with a sensitivity, specificity and
target validation ratio well above current industrial standards.
More specifically 33 drug targets were predicted to inhibit the Th1
response when they are inhibited,
13 drug targets (39.4%) are validated in the literature with
regard to the expected effect: validated immunosuppressive drug
targets.
5 (15.1%) are prediction errors (3: no biological impact, 2:
inverse biological impact)
Model predictions:
| Immunosuppressive Targets Predicted By Immuno-Dyn |
Total Drugs Targets |
Validated Targets |
Prediction Errors |
Validation Ratio |
| 33
| 28 (84.9%)
| 13 (39.4%) |
5 (15.1%) |
72% |
Technological benchmark
Our extensive benchmark analyses within the framework of the
Th1-lymphocyte response show that selecting the genes which expression
is the most modulated is inefficient: in the first 30 most modulated
genes during the CD3+CD28 response no validated immunosuppressive
target is selected, and it is required to select 237 genes to identify
6 validated targets: yield of target detection: 2.5%.
Comparatively the target detection yield of Helios Biosciences is
46%, and dynamic modeling would require the testing of only 13 genes
to validate 6 targets. This may partially explain the high current
attrition rate in drug targeting (selection of inefficient
targets).
|
