A Structure-based Approach to Designing Non-natural Peptides That Can Activate Anti-melanoma Cytotoxic T Cells*

Tumor antigens presented by major histocompatibility complex (MHC) class I molecules and recognized by CD8+ cytotoxic T lymphocytes (CTLs) may generate an efficient antitumor immune response after appropriate immunization. Antigenic peptides can be used in vivo to induce antitumor or antiviral immunity. The efficiency of naked peptides may be greatly limited by their degradation in the biological fluids. We present a rational, structure-based approach to design structurally modified, peptidase-resistant and biologically active analogues of human tumor antigen MAGE-1.A1. This approach is based on our understanding of the peptide interaction with the MHC and the T cell receptor and its precise degradation pathway. Knowledge of these mechanisms led to the design of a non-natural, minimally modified analogue of MAGE-1.A1, [Aib2,NMe-Ser8]MAGE-1.A1, which was highly peptidase-resistant and bound to MHC and activated MAGE-1.A1-specific anti-melanoma CTLs. Thus, we showed that it is possible to structurally modify peptide epitopes to obtain analogues that are still specifically recognized by CTLs. Such analogues may represent interesting leads for antitumor synthetic vaccines.

Tumor antigens presented by major histocompatibility complex (MHC) class I molecules and recognized by CD8 ؉ cytotoxic T lymphocytes (CTLs) may generate an efficient antitumor immune response after appropriate immunization. Antigenic peptides can be used in vivo to induce antitumor or antiviral immunity. The efficiency of naked peptides may be greatly limited by their degradation in the biological fluids. We present a rational, structure-based approach to design structurally modified, peptidase-resistant and biologically active analogues of human tumor antigen MAGE-1.A1. This approach is based on our understanding of the peptide interaction with the MHC and the T cell receptor and its precise degradation pathway. Knowledge of these mechanisms led to the design of a non-natural, minimally modified analogue of MAGE-1.A1, [Aib 2 ,NMe-Ser 8 ]MAGE-1.A1, which was highly peptidase-resistant and bound to MHC and activated MAGE-1.A1-specific anti-melanoma CTLs. Thus, we showed that it is possible to structurally modify peptide epitopes to obtain analogues that are still specifically recognized by CTLs. Such analogues may represent interesting leads for antitumor synthetic vaccines.
CD8 ϩ autologous cytotoxic T lymphocytes (CTLs) 1 specifically recognize and lyse tumor cells bearing tumor antigens in human and animal models (1). These antigens are small peptides (8 -11 amino acids (aa) in length) presented at the cell surface in the context of the major histocompatibility complex (MHC) class I molecules. In humans, several tumor antigens have been identified on the basis of gene cloning and/or sequencing of the naturally presented peptide eluted from MHC class I molecules (2)(3)(4). These tumor antigens are potential therapeutical agents for cancer immunotherapy, and some of them, under different forms, are currently assayed in vivo (5). MHC class I-restricted antigenic peptides have been tested for in vivo vaccination trials in animal models and in humans (6). The use of peptides offers a number of advantages, especially those of a highly specific response and a very low toxicity (7).
However, the results obtained with such peptide vaccines have been variable, dependent on a number of factors (6). Successful immunization was obtained after vaccination with MUT1 and MUT2 antigens expressed on murine carcinoma with regression of established metastases (8). Tumor regression was also observed after vaccination with a synthetic peptide with enhanced MHC binding derived from human melanoma antigen gp100 (9). Protective immune responses were obtained against viruses after peptide vaccination (10), but in other models peptide vaccines gave limited or no results (7,11,12).
Indeed, the major factor limiting the efficiency of peptide vaccines in vivo is, in most cases, their rapid degradation in serum and other biological fluids (13,14). Peptides presented naturally by MHC class I molecules after intracellular processing of endogenous proteins were shown to be protected, by the MHC, against further protease degradation (15,16). When exogenous peptides are used for vaccination, they may be degraded by extracellular proteases before reaching, and binding to, MHC molecules. Presentation of exogenous peptides by MHC molecules might be altered by the presence of serum proteases (14,17,18). A correlation was found between peptide stability in vitro in serum and peptide persistence in vivo, in ascitic fluid as well as on the surface of cells presenting the peptide (18). The stability of peptides in biological media must therefore be taken into account in the design of synthetic peptide vaccines, and the development of non-natural peptide antigens may be of great interest. However, the major difficulty in the design of such modified peptides lies in the fact that the structural modifications introduced must not induce a loss in the antigenicity and immunogenicity of the peptide. Indeed, the activity of an antigenic peptide depends upon two critical steps: (i) its presentation by the MHC, determined by the anchor positions (19 -21) and (ii) the recognition of the peptide-MHC complex by a specific T cell receptor (TcR) (22,23). These two mechanisms depend very closely on the structure of the antigen (24). Examples of non-natural (25)(26)(27)(28)(29) or altered peptides (30) that specifically bind MHC molecules have been described. Many of these peptides were MHC blockers and did not therefore activate CTLs (25,28). In the case of cancer vaccines, peptide analogues should not only bind to MHC but must also activate specific antitumor CTLs.
In this report, we present the rational design of a nonnatural, peptidase-resistant, and antigenic analogue of human tumor antigen MAGE-1.A1. Encoded by the MAGE-1 gene expressed in 40% of melanomas but not in normal tissues (testis excepted) (2), MAGE-1.A1 (EADPTGHSY) is presented by the MHC class I molecule HLA-A1 (3). Its anchor positions are Asp (D) 3 and Tyr (Y) 9 (31). The residues of MAGE-1.A1 recognized by the different anti-MAGE-1.A1 CTL clones are located at P1 and in the P4 -P8 domain (32). Analysis of the structural parameters involved in the degradation mechanism in serum of MAGE-1.A1 and in its interactions with the MHC on one side and the TcR on the other allowed us to design stable and biologically active non natural MAGE-1.A1 analogues.
Peptide Degradation Assay-Peptides were added to preheated (15 min at 37°C before the assay) human serum to a final concentration of 10 Ϫ3 M and incubated at 37°C. Human sera were obtained by centrifugation at 2000 ϫ g of blood samples of five healthy donors. MAGE-1.A1 degradation in individual sera was qualitatively and quantitatively identical. All sera were pooled, and subsequent degradation experiments were realized in the pool. At different incubation times, an aliquot (150 l) was removed and the protease and peptidase activities were stopped by addition of 15 l of trifluoroacetic acid (TFA). Precipitated serum proteins were pelleted by centrifugation at 15000 rpm for 10 min at 4°C. The supernatants were frozen and preserved at Ϫ20°C until analysis by off-line HPLC and/or on-line HPLC/ESI-MS. In control experiments, MAGE-1.A1 and its substituted analogues did not precipitate in presence of 10% TFA and remained stable under these acidic conditions. Serum alone was precipitated by 10% TFA, and its HPLC profile was recorded to detect peaks corresponding to non-precipitated peptides present in the serum.
HPLC and HPLC/MS Analysis-Peptides and their degradation products were separated by reversed-phase HPLC and peaks corresponding to the different products were quantified. Samples analysis at different incubation times allowed the study of the degradation kinetics of the peptides, determination of their half-life as well as the formation of the different degradation products. The following HPLC conditions were used: (i) ultrasphere ODS column: 5-mm, 4.6 mm ϫ 25 cm, C 18 (Beckman); (ii) gradient: 4 -20% CH 3 CN (0.08% TFA in H 2 O, 0.08% TFA in CH 3 CN), 0 -30 min; (iii) UV detection: 200 -300 nm using a Waters 996 photodiode array detector; (iv) chromatograms at 225 and 280 nm were derived. The sequences of degradation fragments were determined by on-line HPLC/ESI-MS. Chromatographic conditions were the same as above except for the gradient (4 -15% instead of 4 -20%). MS analysis was performed with a TSQ-700 Finnigan Mat mass spectrometer (San Jose, CA). The entire HPLC flow was introduced via an atmospheric pressure ionization interface operating in ESI. The interface conditions were such that the needle voltage was maintained at 5 kV, producing a current of 70 A. The temperature of the heated capillary was 250°C. Nitrogen was used as both nebulizing gas (80 p.s.i.) and auxiliary gas (20 p.s.i.). MS data (positive ions) were collected in the full scale mode (m/z 150 -1200) at 2 ms per step.
MHC Class I (HLA-A1) Binding Assay-Peptide binding to HLA-A1 was assayed using the peptide-induced MHC stabilization procedure (47) using the TAP-deficient BM36.1 cell line. Briefly, cells (2 ϫ 10 5 / sample) were incubated with or without increasing amounts of peptide (10 Ϫ9 M to 10 Ϫ4 M for 4 h in the presence of 20% inactivated fetal calf serum as a ␤ 2 -microglobulin source and peptidase inhibitors (bestatin and captopril, 10 Ϫ4 M). HLA-A1 molecules were detected at the cell surface by using the monoclonal antibody W6.32, which reacts with all human MHC class I alleles. After one wash with ice-cold culture medium, cells were incubated on ice for 1 h with 100 l of W6.32 culture supernatant or with medium alone for negative control. After one wash with ice-cold 1% bovine serum albumin-phosphate-buffered saline, cells were incubated for 1 h with fluorescein isothiocyanate-conjugated goat anti-mouse IgG secondary antibody (Sigma). Cells were then washed twice, fixed with 1% formaldehyde in bovine serum albumin-phosphatebuffered saline, and analyzed by flow cytometry (Becton Dickinson FACScan). For each peptide, a SC 50 value (stabilizing concentration 50: concentration of peptide giving half of the maximal stabilization effect) was defined, which reflects its affinity for the studied MHC allele.
CTL Assays-Lysis of target cells by CTL clones was measured by chromium release assay. Target cells were 51 Cr-labeled for 1 h at 37°C in the presence of monoclonal anti-human MHC class I antibody W6.32 and washed three times. Target cells were then incubated in V-bottom microwells (10 3 cells in 50 l) in the presence of different concentrations of peptide (50 l) for 30 min at 37°C. CTLs were then added at an effector:target ratio of 10:1 (10 4 cells in 100 l), and 51 Cr release was measured after 4 h at 37°C.
Molecular Modeling of HLA-A1-Peptide Interactions-HLA-A1 (␣1 and ␣2 domains) structure was obtained by modeling the crystallographic data of the HLA-A2 3D structure (48) (Brookhaven Protein Data Bank, Brookhaven National Laboratory, Upton, NY), both molecules sharing more than 85% sequence homology in their peptide binding grooves. The HLA-A2/Tax nonamer to HLA-A1/MAGE-1.A1 mutations were introduced manually using the program o. Side chain conformation was based on the most probable rotamer and the probable hydrogen bonds. The structure obtained was minimized by X-PLOR. Additional minimization, 1000 iterations without constraints using Steepest Descent minimization, was performed after transfer of HLA-A1/MAGE-1.A1 data to insightII (Biosym Technologies), and further studies were performed using insightII. MAGE-1.A1 residues at P2 and P8 were then substituted by their non-natural analogues using HO-MOLOGY (for D-isomers and Aib) and BIOPOLYMER (for NMe-aa) modules of insightII (Biosym Technologies) followed by minimization using VA09A (500 iterations without constraints) and STEEPEST DE-SCENT (1000 iterations without constraints) algorithms.

MAGE-1.A1 Is Rapidly Degraded in Human Serum in Vitro by
Amino-and Dipeptidylcarboxypeptidases-MAGE-1.A1 degradation fragments generated at 37°C were analyzed by online HPLC/MS. UV analysis (Fig. 1, panel a) facilitated the quantification and determination of the kinetics for the appearance and disappearance of each peak (peaks A-F). In the experimental conditions used, the MAGE-1.A1 half-life (incubation time necessary for the degradation to half of the initial quantity of peptide) was less than 30 min (Fig. 2, panel a). MS analysis (Fig. 1, panel b) of each peak facilitated the determination of the corresponding MAGE-1.A1 degradation fragment. The mass spectrum of peak D (m/z ϭ 269) is given as an example in Fig. 1b (inset). In the MAGE-1.A1 sequence, the only fragment corresponding to a m/z value of 269 is the dipeptide SY. Similar analysis of peaks A-F led to the identification of all MAGE-1.A1 degradation fragments (Table I). The kinetics studies (Fig. 2, panels a, e, and i) allowed us to establish the degradation pathway of MAGE-1.A1 (Fig. 1c); MAGE-1.A1 was sensitive to two exopeptidase activities, aminopeptidase and dipeptidyl carboxypeptidase (or angiotensin-converting enzyme, ACE). The major pathway was initiated by two successive aminopeptidase cleavages, leading to the heptapeptide DPTGHSY (peak E), then cleaved by ACE to give the pentapeptide DPTGH (peak A) and the dipeptide SY (peak D) (degraded itself to amino acids). Direct action of ACE on MAGE-1.A1 represented a minor degradation pathway.
Point Amino Acid Modifications Can Confer Peptidase Resistance to MAGE-1.A1-A series of MAGE-1.A1 analogues modified at positions engaged in one or both of the peptidasesensitive bonds Glu 1 -Ala 2 and His 7 -Ser 8 was synthesized and then studied for their degradation properties (Table II). Peptides substituted at either position 1 or 2 (analogues 1 to 5) showed increased half-lives but this effect remained poorly effective since it varied from very weak (analogue 5) to moderate (analogue 3) in a range of 5-25 min increase as compared with MAGE-1.A1 half-life. However, all these substitutions were efficient against aminopeptidases, since none of the fragments 2-9 and 3-9 expected from aminopeptidase cleavage of the nonamer were detectable by UV (see for example analogue 5 in Fig. 2, panel f) or on-line HPLC/MS (data not shown) analysis. In contrast, in all cases, the fragment 1-7, generated from ACE cleavage, was more intense than for unmodified MAGE-1.A1 degradation, showing that NH 2 -terminal protection of the peptide rendered its COOH terminus more sensitive (see analogue 5 in Fig. 2, panel j). This explains the poor overall gain in half-life for these analogues. At the COOH terminus, substitutions at positions 7 or 8 (analogues 6-8) did not significantly improve MAGE-1.A1 half-life, as expected, since ACE cleavage represents the minor degradation pathway. However, they efficiently protected MAGE-1.A1 against ACE, since fragments 1-7 and 8 -9 were totally undetectable among the degradation products of the COOH terminus-protected analogues by UV (see analogue 8, Fig. 2, panel k) or by on-line HPLC/MS (data not shown) analysis. However, analogues were degraded by aminopeptidases giving fragments 2-9, 3-9 and 5-9 (see for example analogue 8 in Fig. 2, panel g). As protection of one extremity sensitized the opposite one to peptidase activity, we synthesized four analogues (9)(10)(11)(12) in which substitutions at NH 2 and COOH terminus were combined. All four disubsti-tuted peptides showed no detectable degradation even after a 24-h incubation in serum (Table II) (analogue 10 is shown in Fig. 2, panels d, h, and l). This shows that disubstitution conferred long term resistance to both aminopeptidase and ACE and, importantly, that MAGE-1.A1 was not sensitive to other peptidase activities in serum.
The HLA-A1 Binding Groove Tolerates Structural Modifications in the MAGE-1.A1 Sequence-Binding of the structurally modified analogues to HLA-A1 was then tested and compared with that of the parent peptide MAGE-1.A1 (examples are shown in Fig. 3, panels a, d, g, and j; results are presented in Table II, right column). The effect of a substitution on MHC binding depended on both the nature of the chemical modification and its position in the peptide sequence. Introduction of Aib at position 2 (analogue 5), a D-isomer at position 7 (analogue 6), or a NMe group at position 8 (analogue 8) did not alter dramatically HLA-A1 binding of MAGE-1.A1 (Table II). When two of these substitutions were introduced in the same sequence, their effects accumulated; the disubstituted analogues 9 and 10 bound to HLA-A1 with lowered but still good affinities (Fig. 3, panels a and d; Table II). Unlike the previous substitutions, D isomerization at position 2 (analogue 3) or 8 (analogue 7) dramatically decreased HLA-A1 binding. Consequently, the disubstituted [D-Ala 2 ,D-Ser 8 ] peptide (analogue 12) did not bind to HLA-A1. Finally, the monosubstituted analogues 1 and 4 and disubstituted analogue 11 exhibited intermediate to weak (i.e. 30 -100 times lower than that of MAGE-1.A1) binding to HLA-A1 (Fig. 3, panel g; Table II). Thus, out of the 12 non-natural MAGE-1.A1 analogues, 2 (analogues 9 and Amount of the parent peptide or its degradation fragments was determined at different incubation times in serum by measuring the surface of the corresponding HPLC peak. Peptide degradation kinetics are shown in upper panels (a-d).
Fragments resulting from aminopeptidase and ACE degradation are presented in middle (e-h) and lower (i-l) panels, respectively.

.A1 and its substituted analogues
Part a, half-lives were determined as follows: after different incubation times in human serum, the parent peptide was separated from its degradation fragments by HPLC, and the corresponding peak was quantified (Fig. 2, upper panels)  sponding disubstituted analogues are shown in Fig. 3, middle and right panels).
Activation of CTL clone 82/30 tolerated some (panels b or h) but not all (panel k) structural modifications at the carboxylterminal side of MAGE-1.A1. Analogues 6 and 8 activated CTLs within a concentration range close (1 order of magnitude or less) to that of the parent peptide (panels b and e and Table  III). Analogue 7 had a low biological activity (panel k and Table  III), which could be explained by its low HLA-A1 binding affinity (see Table II). For NH 2 terminus-modified MAGE-1.A1 analogues, out of four peptides with a good HLA-A1 binding affinity, analogue 2 retained a good biological activity (Table  III), analogues 1 and 5 had intermediate activities (panel e and  Table III), and analogue 4 had a weak activity (panel h and Table III). As expected, analogue 3, a weak HLA-A1 binder (Table II), activated weakly CTL (Table III). Disubstituted analogues 10 and, to a lesser extend, 9 showed good capacities to activate clone 82/30 (panels b and e and Table III), whereas analogues 11 and 12 exhibited only weak or no biological activity (panels h and k and Table III).
Since the disubstituted analogue 10 was highly stable in human serum (half-life Ͼ 24 h), had a preserved good HLA-A1 binding affinity (see Table II), and activated efficiently one of two tested anti-MAGE-1.A1 CTL clones (Table III), we then tested its capacity to activate a third anti-melanoma MAGE-1.A1-specific CTL clone, 253/47, that recognizes Glu 1 in addition to Gly 5 , Thr 6 , and His 7 (32). Analogue 10 activated efficiently anti-MAGE-1.A1 CTL clone 253/47 (Table IV). DISCUSSION In a perspective of design of synthetic peptide vaccines, we showed in this study that non-natural, peptidase-resistant tumor antigen analogues might specifically activate antitumor CTLs. The rational design of these analogues was based on: (i) the understanding of the interactions of the antigenic peptide with the MHC and the TcR and (ii) the precise knowledge of the degradation mechanism of the peptide.
On-line HPLC/MS analysis led to the identification and quantification of degradation fragments and to the determination of the peptidase-sensitive bonds. Despite the diversity of peptidases present in serum, we found that MAGE-1.A1 was degraded by exopeptidases (mainly aminopeptidases, then by ACE, which are both exopeptidases). Indeed, short peptides are often degraded by exopeptidases rather than endopeptidases (13,33). This structure-based approach combined with the knowledge of the interactions of MAGE-1.A1 with HLA-A1 and TcR obtained in the present work and from previous structureactivity relationship studies (32, 34) allowed us to introduce precise and local substitutions at the sensitive site as opposed to systematic modifications of all of the peptide positions. Indeed, this point is of crucial importance for the use of biologically active peptides in the MHC-peptide-TcR system (Ref. 24 and see below). The non-natural residues we used (D-amino acids, N-methyl-amino acids, and Aib) are efficient inhibitors of exopeptidase (aminopeptidase and ACE) activities (35). We may reasonably postulate that the disubstituted analogues should also be peptidase-resistant in vivo (18).
The main successful examples of antigenic mimicry and cross-reactivity between natural peptide antigens and their non-natural analogues (retro-, retro-inverso-, or D-peptides) have been described for antibody recognition (36,37). A major difficulty encountered in the design of biologically active analogues of MHC-restricted epitopes is that their activity depends upon the accomplishment of two critical steps: MHC binding on the one hand and TcR signal triggering on the other. Several examples of MHC-restricted epitope analogues have been described. All-retro-inverso analogues (peptides in which all bonds have been retro-inverted) of MHC class II-restricted epitopes lost their MHC binding affinity (38). In other studies, binding to MHC class I was altered for some peptides and not for others following partial modifications, retro-inverso or re- duced bonds, depending on the position of the modification (26,27). However, in two studies on MHC class II-restricted peptides, even when only partial structural modifications that did not alter MHC binding were introduced, recognition of nonnatural analogues by CTL was dramatically affected (28,29). Despite the strong structural analogies between the Fab antibody fragment and the V␣-V␤ TcR domain (39), the rather common cross-reactivity found between parent peptides recognized by antibodies and their non-natural analogues (36) appears to be much less applicable to MHC-restricted TcR ligands. In fact, this observation is likely to be explained by the higher stringency of the TcR-epitope interaction in comparison with the antibody-epitope, since the TcR has dual functions that must both be fulfilled: (i) the specific recognition of the antigen and (ii) the activation of the lymphocyte functions. Indeed, structurally altered ligands, partial agonists or antagonists, that can only fulfill the first TcR function, but not the second one, may then lead to partial activation or anergy of the lymphocyte (24). Thus, great care must be taken in the development of antitumor vaccine strategies to avoid specific inactivation or anergy of antitumor CTL. Of the three stabilized analogues (9, 10, and 11) with good HLA-A1 binding, only analogue 10, [Aib 2 , NMe-Ser 8 ]MAGE-1.A1, activated CTL in the same range of concentrations as the parent MAGE-1.A1 peptide. In this analogue, the side chain orientations of the anchoring residues Asp 3 and Tyr 9 and, more importantly, that of the residues pointing out toward the TcR (mainly Glu 1 , Pro 4 , Thr 5 , His 7 , Ser 8 ) were left unmodified by the structural modifications at P2 and P8 (see Fig. 4). This observation has a dual consequence in terms of MHC binding and CTL activation. Considering MHC binding, the drop in affinity confirms that non-anchor residues, particularly P2 for HLA-A1-restricted peptides (40), may influence antigen presentation (21). Considering CTL activation, the remarkable correlation observed between the stable analogue 10 and the natural antigen in their ability to sensitize target cells to lysis by CTL clone 82/30 (and to lesser extend CTL clone 253/47) and their MHC affinity indicates unambiguously that the lower CTL activation properties of analogue 10 are a strict and direct consequence of its lower MHC presentation and not at all of alteration in TcR recognition. This point, in perfect accord with the proposed structural model (see Fig. 4) is critical since it indicates that, once bound to the MHC, analogue 10 is able to engage the MAGE-1.A1-specific TcR as the natural ligand does. We can thus postulate that this synthetic peptide can fully activate the anti-melanoma CTL functions and therefore acts as a full agonist, an important parameter in terms of successful vaccine strategies (see above). Analogue 10, stable after 24 h in serum and able to activate two of three known MAGE-1.A1specific CTL clones, may therefore be considered as a potential activator of anti-melanoma CTL response in vivo. To overcome the ethical problems encountered when testing non-natural molecules such as ours in human cancer vaccine trials, an alternative and attractive approach to follow would be to use these molecules in ex vivo or in vitro procedures to generate potent MAGE-1.A1-specific CTL (41).
Even though promising, T cell-based immunotherapies may face limitations. First, unlike B cell (antibody)-based vaccines, T cell immunotherapeutics are not global but are restricted, by definition, to a fraction of individuals sharing a given HLA allele (for example, HLA-A1 is carried by 26% of the caucasian population). Second, the use of peptide vaccines to induce antitumor or antiviral protection may result in a wide range of effects, ranging from effective protection, through non-response and even to the induction of tolerance. The factors influencing the in vivo immune response and leading to protection or tolerance are not yet clearly understood. Priming or tolerance toward a viral antigen could be induced by injecting the peptide subcutaneously or intraperitoneally, respectively (42). The induction of protection seemed favored by the slow liberation of the peptide when administrated subcutaneously. However, tolerance was induced toward an Ad5E1A-transfected tumor by a single subcutaneous injection of epitope Ad5E1A234 -243 (43). In this case, tolerance depended on the peptide used rather than on the administration protocol. Immunization of mice with tumor peptides MUT-1 and MUT-2 injected subcutaneously or intradermally in IFA led to regression of established metastasis (8). The use of stabilized analogues may thus be of particular interest when administrated subcutaneously in order to be liberated slowly. They will not be degraded in situ before reaching MHC molecules on professional antigen presenting cells. It will be important, however, to verify that the use of highly stable peptides will not mimic the use of high doses of peptide that lead to CTL tolerance. Tumor antigens that are not derived from viral proteins, as opposed to Ad5E1A234 -243, have low MHC binding affinity (44), and the use of stabilized analogues in these cases might be of high interest.