Activation and expansion of drug reactive T cells are key features in drug hypersensitivity reactions. Drugs may interact directly with immune receptors such as the human leukocyte antigens (HLA) or the T-cell receptors (TCR) itself, the pharmacological interaction [p-i] concept. To analyze whether the drug sulfamethoxazole (SMX) interacts directly with the TCR and thereby contributing to signaling and T cell activation, we analyze two SMX specific T cell clones (TCC “1.3”and “H13”). Proliferation to SMX and 11 related sulfanilamides, Ca++ influx in drug stimulated T-cells and the inhibitory effect of non-reactive sulfanilamides on SMX stimulation were analyzed. In silico docking of SMX and related sulfanilamide to the TCR were used to identify possible drug binding sites, and correlated to in vitro data to find the correct docking. In Ca++ influx assays, reactions occurred as early as 14 sec after adding SMX to TCC and APC. The broadly reactive clone (“H13”) was stimulated by 5 additional sulfanilamide, while one TCC (“1.3”) was reactive exclusively with SMX but not other sulfanilamides. Competition experiments with sulfanilamide inhibited SMX induced Ca++ influx and proliferation of the TCC1.3 ina dose dependent way. Docking experiments with SMX and related sulfanilamides confirmed and explained the in vitro data as docking localized binding sites for SMX and the 5 stimulating sulfanilamides on the CDR2β domain of the clone H13, while the 6 non-stimulatory SA failed to bind. In TCC 1.3, SMX could be docked on the CDR3α of the TCR. The other, non-stimulatory but inhibitory SA could also be docked to the same site. The combined analysis of in vitro and in silico data show that sulfanilamide can bind directly to TCRs. It shows that TCR, like other receptors, appear to be reamenable to manipulations by small molecules.
Sulfanilamides (SA) can elicit hypersensitivity reactions, resulting in various diseases [1,2]. A majority of these reactions involve the expansion of a few drug specific T-cells, which can coordinate an inflammatory immune response resulting in a clinical picture, which often involves the skin [3,4]. How the immune system is stimulated in drug hypersensitivity reactions is still a matter of debate. The hapten-concept postulates that a small molecular compound like a drug or a drug-metabolite is too small to be immunogenic per se. If the small chemical forms covalent bonds to a protein, the hapten-carrier complex is stable and remains coupled to the amino acid even after intracellular processing to small peptides. These hapten modified peptides are immonogenic when presented by the highly polymorphic proteins of the major histocompatibility complex (MHC), the so-called Human Leukocyte Antigens (HLA). The hapten/peptide-HLA complex can function as a neo-antigen for some T cells able to interact with it. Examples of haptens are penicillinG and other beta-lactam antibiotics, which via their β-lactam ring can bind directly to lysine or serine groups within proteins. Sulfamethoxazole (SMX) is an example for a pro-hapten, which is metabolized to a highly reactive SMX-NO intermediate. The latter binds as a hapten covalently to the sulfur of cysteins, however alone can also elicit immune responses [2,5]. Clinical effects of such reactions are heterogeneous, including anaphylaxis, haemolyticanaemia, or contact dermatitis, to name just as few [
A quite different, alternative explanation for immune stimulations by small molecular drugs leading to hypersensitivity reactions was recently developed by our group. It is termed “p-i-concept”, standing for “pharmacological interaction with immune receptors” [6,7]. Descriptively it postulates that a drug may interact directly with some of the highly variable immune receptors, in particular with some of the highly polymorphic HLA-molecules and TCR. This is akin to drug interactions with other receptors, where small molecular pharmaceuticals may bind to an enzymatic or ligand binding site. The p-i concept does not imply the formation of stable antigenic hapten-carrier determinants. Neither drug metabolism nor protein processing is required, as the drug itself, and not a hapten modified peptide, is binding to the receptors. All interactions involved are mediated by van der Waals forces, salt bridges, or other electrostatic interactions without any direct covalent changes. Such drug-protein interactions are common and are mostly without functional consequences. However, some drugs may fit into some of these immune receptors quite stably, giving rise to immune stimulations [4,5]. Indeed, the chance of functionally relevant drug bindings to some immune recaptors is probably related to the enormous polymorphism of specific immune receptors. One estimates that one individual harbors 109 - 1011 different T cells and distinct TCR based on CDR3 polymorphism, offering an enormous amount of variable binding sites [8-10]. Also the polymorphism of major histocompatibility complex (MHC)-alleles within the population is large, with presently more than 9300 different alleles (http://www.ebi.ac.uk/imgt/hla/stats.html). The chance that some drugs may bind to a particular immune recaptor with a rather strong affinity is thus relatively high [10, 11].
One can differentiate between two forms of the p-i concept, and possibly more if other proteins of the immunological synapse are included [
A number of previous findings support the concept that direct drug binding to TCR (p-i TCR) may also play a role in drug related T cell stimulations [21,22]. Many drug hypersensitivity reactions were, in spite of intensive search, not HLA-allele restricted. Some previous investigations with SMX and lidocain specific CD4+ T cell clones (TCC), generated from patients with respective allergies, showed that a complete T cell stimulation, as measured through IL-2 secretion, proliferation, and cytotoxicity, required some interaction of the TCR with a HLA molecule, but that the allele could be exchanged [21-23]. Moreover, also the presented peptide could be exchanged [
Based on these aspects, we pursued the idea that drugs like SMX may interact directly with the TCR, and that this interaction by a non-covalently bound drug can be an essential component of T-cell activation, as it may occur in drug hypersensitivity. To support this concept, we evaluated the stimulatory capacity of SMX and of 11 structurally related sulfanilamides (SA) (
The TCC generated had the CD4 phenotype, were MHC-DRB1*1001 restricted and reacted to concentrations of SMX between 32.5 to 250 μg/ml [
According to the p-i model, the drug interacting with the TCR, HLA-molecule or both occurs immediately and independently of prior protein processing or drug metabolism. To confirm that the TCC react to SMX according to the p-i model, we analyzed the very rapidly appearing elevation of intracellular Ca++ of the two TCC upon stimulation by SMX and the indicated panel of tested SA [
In clone 1.3 the reaction was specific for SMX only. This SMX induced Ca++ influx was always very strong, similar or even higher than PHA induced influx, and started as early as 14 sec after addition of SMX to the TCC/APC cell suspension. This confirmed a p-i based mechanism (
Clone H13 gave a response to SMX and the 5 SA compounds SMT, STH, SDZ, SMR and SPD (
Proliferation assays with SMX and one other SA as competitor were conducted using a range of competitor concentrations. Toxicity assays were preformed first for all compounds, showing that most compounds started to become toxic above 350 ug/ml, with SMX and SMP having a toxicity level above 500 ug/ml before any drug induced cell death was observed (data not shown). Maximal stimulation of TCC 1.3 was reached with SMX at 250 ug/ml, and this was chosen as a maximum total drug concentration for proliferation based competition
assays. A minimal measurable response from SMX was determined to be 3 ug/ml giving a rough molar range for testing all competitors from 8:1 to 1:8 competitor to SMX. All mixtures were then tested in triplicate as 7 combined concentrations along this range. Data was then used to generate an inhibition curve, reflected as a percentage decrease in SMX induced proliferation (
compounds the standard deviation may cause curve fitting to miss small deviations.
To measure an immediate response from competition of SA an adapted Ca2++ influx was conducted at a 1 to 1 molar ratio of SMX to SA derivative. For this assay, the competitor was allowed to incubate for 5 minutes with the T cell 1.3 and APC, prior to addition of SMX. Without pre-incubation, no consistent inhibition was observed. Our adapted protocol allowed the inhibition calculated for each compound to remain the same, and no difference was observed for calculated inhibition with increased incubation times (data not shown). Influx assays were then generated for each SA against SMX, giving decreased Ca2++ influx with all compounds tested (
No inhibition was observed for H13: addition of nonstimulatory compounds did not affect the proliferation, addition of stimulatory compounds to low SMX concentrations dose dependently enhanced the proliferation. Data not shown.
Models were generated for each of the TCR indicated (
these sites appears to be blocked by other proteins, such as CD3 subunits and glycosylation [26,27]. Of interest were the additional binding sites on the CDR3α and CDR2β loops, which were found exclusively on the SMX reactive TCR H13 and TCR 1.3.
The TCC H13 and H25 proliferated to SMX and showed docking of SMX to the CDR2β loop of TCR H13 and H25. Clone H13 was chosen for a more detailed analysis at the molecular level. Using a refined approach, where parameters were set to encompass the CDR2β loop alone, it was found all compounds tested could be made to dock around the suspected site of activity (
In non-stimulatory compounds (SID, SMZ, SMP, SDM, SMD, SDX), the R substituted group size alone prevented the SO2 and NH2 groups from being properly positioned in the CDR2β binding pocket. This was by preventing the proper distance between the TYR 58 SA SO2, where no hydrogen bonding could occur (
angle with the N to S-C bond in the SA core, which has a range 110.0˚ +/− 2.5 in either direction, was the primary cause [28,29].
Overall in clone H13, SMX was able to form 6 hydrogen bonds and 1 salt bridge bound within the pocket. After energy minimization, these bonds were, SMX to residue respectively, O of SO2 to TYR 58 terminal H 2.7 Å, O of SO2 to LEU 68 Methyl H 2.8 Å, H of terminal NH2 to ASN 70 terminal N 3.2 Å, H of terminal NH2 to SER 54 terminal O 3.1 Å, N ring to ALA 56 NH backbone 1.9 Å, O ring to THR 57 NH backbone 2.9 Å. A single salt bridge is accommodated by ASP 65 terminal O and the sulfur of the SMX at 3.8 Å. Additional weak interactions with the SA core benzyl ring are re-enforced by LEU 68, and a few weak interactions between LYS 64 terminal NH3 and the pentameric ring carbons of the SMX R group. These ring carbons have a weak negative charge of 0.06 eV due to the cyclic effect of the specific R group, far lower than a hydrogen bond strength but measurable. The other docked compounds which gave a response in the Ca2++ and proliferation assays maintain all the SA core hydrogen bonding, with differences respective to the R substituted groups.
For H13, dehydrated SA showed a much wider degree of docking affinities. By removing the indicated hydrogen alone, it was possible to find an orientation in which all SO2 groups could form a bond with the TYR 58. The overall orientations placed the SA core within or 1 - 2 Å displaced from the SER 54 and ASN 70 site. Main differences were a change in the S-N-C angle from 110˚ to 127˚+/− 2, and the nitrogen charge from almost neutral to −0.4, simply from removing the hydrogen (Supplemental
Using affinity of 6.5 kCal/mol cutoff for positive versus negative docked conformations and as second criterion the involvement of at least 5 hydrogen bonds for each compound, the stimulatory SMX and all other 11 non-stimulatory sulfanilamides were found to dock into the CDR3α loopof clone 1.3. This is compatible with the ability of all 11 SA to block SMX stimulation in functional assays and suggests that competition of SA with SMX for the same binding site accounts for the blocking effect observed in Ca2++ responses, or proliferative capacity. All SA showed a range of between 5-7 hydrogen bonds per compound and direct hydrophobic interactions with PHE 92 of the CDR3α loop. A pi orbital interaction occurred with SMP and SMX in orientations where the SA core unit points to the peptide groove (
There is also an effect observable between the hydrophobic groups of the CDR3 loop, which hinder the reversed orientation for a limited set of compounds. This is apparent by looking at SMP and SMD which differ in only one additional charge (
Docking of all compounds in a dehydrated form (Supplemental
affinities, mostly by disruption of the PHE 92 hydrophobic interaction, and slight movement of the SO2. In the non-dehydrated SA the SO2 was always oriented in the CDR3α loop center with a repulsive effect from the hydrogen of the SA neck from loop backbone oxygens and positive NH backbone atoms (
The two TCC analyzed were CD4+, expressed the skin homing receptor CCR4 but not CCR10, were perforin positive and expressed CD107a upon activation by SMX (
In this study we pursued the concept of direct drug interactions with immune receptors. We used the model of SMX hypersensitivity and respective TCC to analyze a stimulatory potential of direct SA binding to certain TCR. We took advantage on previous studies of SMX specific TCC, which had already revealed their dependence on APC for optimal stimulations and that different TCC, with structurally different TCR, show quite distinct cross-reactivities with up to 11 additional SA [4,30]. In this study we focused on two distinct TCR/TCC (1.3 and H13), which showed distinct cross-reactivities to 11 SA and correlated their functional, structural, and inhibitory in vitro data with computational docking data. This combined analysis allowed to locate the SMX binding site to either the CDR3α for 1.3 and to CDR2β for H13 and to explain the different functional and inhibitory capacities of the SA on the two TCR/TCC 1.3 and H131.
The TCC used are characterized as “p-i-clones”, meaning that their stimulation was not dependent on prior drug metabolism and presentation of a SMXmodified peptide by APC. Such p-i interactions of drugs are often very labile, and washing the cells,T-cells, APC or both, already abrogates reactivity [
Both TCC showed a rapid Ca++ influx after adding the compounds to TCC in the presence of APC. This immediate reactivity for both clones was even faster thanthe TCC reactivity of high affinity abacavir reactive TCC, which reacted after ca. 60 - 100 sec [
While both TCC reacted rapidly to SMX with immediate Ca++ influx, the two TCC H13 and 1.3, are quite different if analyzed in detail. As reported previously, using TCR transfected cell lines, which are devoid of self HLA-molecules, full activation by SMX, reflected as IL-2 secretion or proliferation, required not only the presence of the drug, but also an interaction with HLAmolecules on APC [23,31,32]. In the case of TCC H13 the HLA-DR-B1*1001 molecule, provided by autologous EBV transformed B cells alone, was a requisite. In contrast, the TCC 1.3 reacted to SMX regardless of PBMC type used, as it was self-presenting.
A clue to an in depth analysis of the two TCC was their different reactivity to 11 other SA. Their responses in the T-cell assays were only due to cross-reactivity of the originally SMX specific TCC, and not due to prior exposure to the SA, which are mostly not available as drugs or only for veterinary use. TCC 1.3 proliferated only to SMX, while the other 11 SA were not stimulatory, even if their molecular structure was rather similar (
The TCC H13 reacted with SMX and 5 additional SA, while 6 other SA failed to stimulate this TCC. The non-stimulatory compounds did not inhibit SMX reactivity in competition assays, because their intrinsic stimulatory activity was very similar to the stimulatory activity of SMX. Thus, the stimulatory compounds increased the effect of SMX in a dose dependent way (data not shown).
As drug and HLA interactions seem to be involved in full T cell stimulation, it is possible that SMX and the other SA bind first to the HLA, and are then recognized by the TCR. However, our data argue strongly against first HLA-binding (actually HLA-DR-B1*1001). All SA and SMX have shown some functional involvement, either stimulatory or blocking, and these differences cannot be explained by DR-B1*10:01 binding, which is not different between TCC 1.3and H13. E.g., the SA STH stimulates H13, but does not stimulate 1.3:If this would be explained by binding to HLA-DR-B1*1001, one has to postulate that STH has no binding sites on the TCR 1.3, while the TCR H13 binds STH. But STH was able to block SMX (1.3), and could stimulate (H13). Or the SA STH andSMTare both able to block the SMX induced stimulation of TCC 1.3. However the reaction of H13 to SMX is not blocked, and STH and SMT are either stimulatory (STH) or gives no response (SMT), a constellation which is not compatible with the hypothesis that both SA would compete with SMX binding on HLA-DR-B1*1001. Therefore, the reactivity of competing SA differs depending on the TCR, not on the HLA site, which is identical for both TCC.
[31,32] To explain the differences in the reactivity of H13 and 1.3 to SMX and related SA as well as to understand the differences in blocking SMX induced stimulation by SA we applied docking models for the SA and TCR. Initial SMX docking experiments showed that SMX and other SA might bind to a number of sites on the whole TCR. All of these global binding sites are outside the antigen-specific interaction site facing the MHCpeptide complex, and whether they are accessible to drugs is unclear. It is likely that these possible dockings occur via the sulfur moiety in the SA, which can form non-covalent interactions with a multitude of amino acids on protein surfaces without a functional consequence, as non-specific binding of this type has been shown to occur with similar compounds.
Docking revealed two unique sites on either TCC H13 or 1.3, characterized by 5 - 7 hydrogen bonds, which were absent on the TCR without SMX specificity. They were identified on the CDR2β of TCR H13 and on CDR3α of TCR 1.3. How can one be sure that these sites are relevant? As we used TCC instead of TCR transfected cell lines, which had a better discriminatory power than TCR transfected cell lines, we took the approach to use distinct SA to define the possible binding site. This proved to give rather clear results, and helped discriminate HLA or TCR interactions. The CDR2β binding site for SMX on the H13 TCR was capable of binding the 5 additional stimulatory SA, while the 6 non-stimulatory SA did not bind. Thus, the docking data were in agreement with the functional data and could thus explain the selective stimulation by SMX and the 5 SA.
With TCR 1.3, docking revealed 7 hydrogen bonds to SMX on the CDR3α region, with almost an antibody like recognition of the compound [
For clone H13 two separate characteristics of the compounds and their ability to bind in specific orientations seem to be important, which may also shed light on more detailed functioning of the TCR at the amino acid level. In the 6 non-stimulatory SA (SID, SMZ, SMP, SDM, SMD, and SDX), the size of the R group hinders the SA core to positioning itself in the orientation depicted in
Our data suggest that drugs like SMX and other SA may interact directly with TCRs and that they can induce T cell activation if an additional interaction with the HLA molecule is provided [23,30]. Actually, the SMX binding may alter the interaction of the TCR with pMHC, as shown by computational analysis of TCR H13§. The drug binding has to occur at certain, functionally relevant sites of the TCR, such as the CDR3 or CDR2 sites. Structurally related molecules may also bind: they may signal as well, or may just bind, without signaling, or may block signaling of a competing SA. This means that computational studies may reveal a possible binding site, but must be followed by functional analysis, as drug binding does not necessarily lead to signal transmission.
The consequences of these findings are multifold. With regard to drug hypersensitivity, the starting point of our investigations, our data underline the p-i concept and their two subforms. In one model, p-i (HLA), drugs preferentially interact with certain HLA molecules [13, 17-19]. These reactions are exemplified by strong HLAallele restricted hypersensitivity reactions. Immune stimulation in this model occurs by the reactive T cell, which itself is not altered by the drug, but possesses a TCR able to interact with the MHC-drug-peptide complex. As shown in the carbamazepine model, only if such T-cells are present in the circulation, the carbamazepineMHC complex is harmful [
In the p-i (TCR) model, an initial interaction of the drug with a peculiar TCR occurs.It is a more direct effect of the drug on the reactive cell, the T cell. The details of the drug induced TCR signaling are not yet clear: The MHC/peptide may act as scaffold and/or the SMX binding to TCR may result in some change of the TCR which results in better binding to pMHC§.
The immunological consequences of our data are also multifold. The data illustrate that the drug binding regions of the TCR show an extreme fine specificity and fine tuning for signaling. Our findings may provide further tools to decipher this still enigmatic area of TCR signaling and specificity. These studies also show that TCR-MHC interactions might be amenable to usual pharmacological tools like blocking of a receptor by a competing ligand.
The p-i TCR concept implies that the T cell stimulated by a drug in adverse reactions also have a peptide specificity, which is mostly unknown. Preliminary data with alloand simultaneously drug reactive T cells have indeed revealed that an allo-reactivity can be blocked by a drug (unpublished). As the data always refer to only one single TCC/TCR, and another TCC behaves differently, a therapeutic intervention (e.g. blocking a certain TCR) could be imagined only on clonal expansions, since other TCC would react differently to the same drug. An example is the SA STH: it is inhibitory for SMX stimulation by TCC 1.3, but the same SA further enhances the SMX triggered stimulation of H13, which comes from the same individual.
In conclusion, drug hypersensitivity reactions may not only develop because drugs are binding to HLA but also on drug binding to TCR. These are highly interesting models for “unusual” T cell stimulations and may open new possibilities of pharmacological interventions with the specific immune system. In addition, drug interactions may prove a means to further understand functionality of immune receptors.
The PBMC, EBV B cell lines and T cells were derived from a patient with exanthema, malaise and well documented SMX specific T cell proliferation in vitro [4,37]. Cloning was performed by limiting dilution as described previously. To test specificity of the clones, T-cells were analyzed by means of proliferation, with 250 μg/ml SMX. The specific TCC were expanded further by PHAstimulation (1 µg/mL; LaboratoiresEurobio, CortaboeufCedex B, France) in combination with allogeneic, irradiated PBMCs every 14 days. IL-2 was added after 24 hours to a concentration of 100 IU final volume, and then every 48 hours for 14 days. Both TCC required interaction with HLA and H13 required autologous APC like EBV-B-LCL and did not react to addition of allogeneic APC. TCC 1.3 was also stimulated weakly without adding APC in prior studies, and robustly with the addition of non-autologous APC, or HLA-DR-B1*1001 transfectants.
DNA sequences obtained from sequencing of the reactive clones were translated using DNA to Protein from EMBL translation tool (http://www.ebi.ac.uk/Tools/st/). These were then used to obtain the IMGT (http://www.imgt.org) nomenclature used by the NIH blastp protein sequence searching (http://blast.ncbi.nlm.nih.gov) using the IMGT database as a template. Further sequence alignments were made with all respective clone sequences for the α and β domains of each sequenced TCR using clustalW multiple sequence alignment.
Briefly, 25,000 TCC and 30,000 irradiated PBMCs or EBV-B-LCL were incubated together per well with 250 ug/ml of the indicated compounds, medium alone (CM) or PHA (1 ug/ml). 3H-thymidine was added after 48 to max 72 hours for exactly 12 hrs and proliferation measured via scintillation counting (Top Count, Perkin Elmer). Averages of at least three and up to six runs for each experimental condition were collected and all means along with statistical differences were calculated. The SA compounds are listed in
To test for inhibition of proliferation, the total of two compounds, SMX and competitor, were added at the same time, with a standard proliferation assay as above. All compounds tested in proliferation assays were also used in the competition assays. Both compounds together were not exceeding 250 μg/ml in total. The range of competitive testing was 8:1 - 1:8 (molar ratios). The averages of at least three independent runs were plotted per molar ratio across seven different ratios within the stated range, for each point plotted.
TCC were incubated for 10 minutes, and spun down twice in RPMI w/o Phenol red and then incubated with 2 μg/ml Fluo-4 AM (Invitrogen, Carlsbad, CA) in RPMI w/o Phenol red (Sigma Chemicals Co, Buchs, Switzerland) at 37˚C for 30 min. After incubation, cells were washed twice with and resuspended in RPMI w/o Phenol red. Cells were plated in half-area clear bottom 96 well plates (Corning, LifeSciences, CA, USA) at 1 × 105 cells per well (5 × 104 TCC and 5 × 104 irradiated APC or EBV) spun for 1 minute at 1500 RPM’s, and further incubated for 20 min at 37˚C in dark to allow cells to equilibrate and uniformly cover the plate surface. Measurement was performed on a Synergy-4 instrument (BioTek Instruments Inc., Highland Park, VT, USA) at 37˚C with an excitation band of 485/20 nm and fluorescence was measured at 528/20 nm. Baseline signal (F0) was recorded during 5 min before the addition of antigen solutions and subsequent fluorescence measurements were performed for 60 min. PHA stimulation (1 ug/ml final well concentration) or ionomycine at 1 ug/ml served as positive control. The results are shown as normalized fluorescence (F/F0)-1. Averages of at least three independent runs for each compound were tested per T cell clone used with similar results.
In competition assays, the final combined concentrations of each competitor and SMX were 250 ug/ml to avoid toxicity, at a 1 to 1 ratio. To reduce or eliminate variation observed without pre-incubation in calculated inhibition, the competing compounds were added with a multi-pipette at the start of the calcium reading, and SMX added 5 minutes later.
All fluorescent conjugated or unlabeled monoclonal antibodies were purchased from BD (Becton, Dickinson and Company. NJ, USA) or Biolegend (San Diego, CA, USA). For all standard staining, cells were plated in V bottom, 96 well plates at 200 K TCC, or 200 K TCC and 200 K PBMC as indicated. Cells were dyed for 30 minutes at 4˚C in dark, and washed 3 times all on ice with chilled PBS before reading. All data points are at least 3 readings from independent wells, on 1 - 2 separate days, using a BD FACSCantoTM. They were analyzed using FACS DIVA software.
In CD107a expression up-regulation kinetic experiments, cells were plated at 200 K TCC and 200 K PBMC per well in normal RPMI supplemented with heat inactivated 10% FCS. Either the indicated compound, PHA 1 µg/mL or nothing as a negative control were added [
All protein models were generated using automated model building via Swiss-modeller web site (http://swissmodel.expasy.org/) [
As controls for docking, random poly alanine TCR α and β sequences or TCR from a T-cell non-reactive to SMX, were generated as above. A second control, we used a single TCR α and β from an M. musculis found in Uniprot A2NTU7 and Q7TND8, with models generated as described. Sequences are shown in supplemental
Small molecules used in all models, Docking or Molecular Dynamics were generated using simple sketch and conversion software. In short, the compound structure was drawn into ACDlabsChemsketch (Advanced Chemistry Development, Toronto, Ontario, Canada) to generate a MDL mol file. These files were then converted to PDB files with 3D coordinates using OpenBabel (GNU public license) freeware. As these were small molecules, energy minimization prior to using them in Docking runs was conducted using simple Chimera Software on a PC (UCSF Chimera package from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco).
All docking was performed using AutoDock or AutoDockVina, used as cross comparisons [42,43]. The later software is an optimized version of AutoDock for parallel processor execution. In all Docking runs, at least 6 rigid and 6 flexible docking runs were conducted for the final analysis. Each run constitutes 10 individual dockings of each compound used. An initial approach was to conduct Docking runs using a spars matrix approach encompassing the lower 2/3 of the TCR, and then limiting runs to regions of the TCR where clustering was observed. Clustering included 3 or more docked conformations in the same region or area within 3 - 4 Å of each other from at least 3 sparse matrix runs. These limiting runs were then further refined to determined sites showing clear affinities, and at least 6 limited runs were conducted. For flexible docking, amino acids in the CDR3 or CDR2 loops were allowed to move freely, including backbone flexibility as allowed by AutoDock or AutoDockVina. For this later, at least 6 additional flexible docking runs were conducted. For each TCR model, the compound library containing all SA was screened against the finalized sites of interest. A cutoff for positive versus negative docked conformations for all molecules was generated for this work alone. The cutoff is based on underestimation of full hydrogen bonding energies, and takes into account only an arbitrary mean based on docking output scores, total hydrogen bonds along with initial visual inspection. The resulting scoring cutoff is 6.5 kCal/mol +/− 0.2, which corresponds to at least 5 hydrogen bonds [42,44].
All final docked ligand models, or model comparisons were analyzed using multiple software to check for validity. Software used including Pymol ((Schrodinger, LLC (2010) The PyMOL Molecular Graphics System, Version 1.3r1)), Gromacs and AutoDock Tools, all of which have internal analysis tools to measure atomic distances and energies produced from Docking software, and internal hydrogen bonding or salt bridging based on distances and angles [
Statistics were performed using qtiplot (Copyright © 2004-2011 Ion Vasilief and Stephen Besch). Briefly, all proliferation, proliferation inhibition and FACS data were fed into the software using spreadsheets, and standard error, means and standard deviations plotted for individual points in all graphs. One way ANOVA was performed on plotted data, and significant differences scored as <0.05 >0.001, *, or <0.001, ** for respective calculated P values.
We would like to thank Michael Lüthi for obtaining the TCR sequences from clones listed in sup.
SMX: sulfamethoxazole, SA: sulfanilamides, p-i concept: pharmacological interaction concept, TCC: T cell clones, TCR: T cell receptor for antigen, MHC: major histocompatibility complex, pMHC: peptide-MHC, CDR: complementary determining region