Quantum chemical calculations of stability constants: study of ligand effects on the relative stability of Pd(II)–peptide complexes

Replacement of [Pd(H 2 O) 4 ] 2 ? by cis -[Pd(en) (H 2 O) 2 ] 2 ? , [PdCl 4 ] 2 - , and [Pd(NH 3 ) 4 ] 2 ? on the hydrolytic cleavage of the Ace-Ala-Lys-Tyr-Gly–Gly-Met-Ala-Ala-Arg-Ala peptide is theoretically investigated by using different quantum chemical methods both in the gas phase an in water solution. First, we carry out a series of validation calculations on small Pd(II) complexes by computing high-level ab initio [MP2 and CCSD(T)] and Density Functional Theory (B3LYP) electronic energies while solvent effects are taken into account by means of a Poisson-Boltzmann continuum model coupled with the B3LYP method. After having assessed the actual performance of the DFT calculations in predicting the stability constants for selected Pd(II)-complexes, we compute the relative free energies in solution of several Pd(II) – peptide model complexes. By assuming that the reaction of the peptide with cis - [Pd(en)(H 2 O) 2 ] 2 ? , [Pd(Cl) 4 ] 2 - , and [Pd(NH 3 ) 4 ] 2 ? would lead to the initial formation of the respective peptide-bound complexes, which in turn would evolve to afford a hydrolytically active complex [Pd(peptide)(H 2 O) 2 ] 2 ? through the displacement of the en, Cl - , and NH 3 ligands by water, our calculations of the relative stability of these complexes allow us to rationalize why [Pd(H 2 O) 4 ] 2 ? and [Pd(NH 3 ) 4 ] 2 ? are more reactive than cis -[Pd(en)(H 2 O) 2 ] 2 ? and [PdCl 4 ] 2 - as experimentally found.


Introduction
Hydrolytic cleavage of the amide bond in peptides and proteins is a very important issue due to its implication in many bioanalytical and bioengineering applications such as protein sequencing, peptide mapping, protein footprinting and folding studies, protein semisynthesis, and purification of fusion proteins [1][2][3][4][5][6]. The extreme inertness of the amide linkage to be hydrolyzed has prompted the emerging of different procedures to overcome such an inconvenience [7][8][9][10]. Among them, the use of transition-metal complexes constitutes one of the most promising strategies for cleaving peptides and proteins because mild conditions can be employed [2,3,5,7,[9][10][11][12][13][14][15][16]. Besides this, the relatively small size of these complexes favors the probing of the conformation or accessibility of protein regions [3,13,17,18], which can also be important to get a better knowledge on natural metalloproteases.
Numerous experimental studies have shown that palladium (II) and platinum (II) complexes are useful reagents for cleaving peptides and proteins [15,. Particularly, the complex [Pd (H 2 O) 4 ] 2? has been studied in some detail because of its relatively high effectiveness [15,25,28]. This complex spontaneously binds to the side chains of either methionine (Met) or histidine (His) residues and regioselectively promotes the hydrolytic cleavage of the second amide bond upstream from the anchoring residue, that is, for instance, the Gly-Gly and Gly-Pro bonds in the Gly-Gly-Met (or Gly-Gly-His) and Gly-Pro-Met (or Gly-Pro-His) peptide sequences, respectively (see Scheme 1) [15,25,28]. These processes start with the complexation of the peptide, in which the metal ion anchors to the sulfur or nitrogen atoms of the Met or His residues, respectively (see 1 in Scheme 1 for the Met case). Then, the Pd (II) ion deprotonates the secondary amide group and binds to the nitrogen atom of the resulting amidate anion, thus giving the so-called hydrolytically active complex (see 2 in Scheme 1). This mechanistic proposal has gained much support in our recent computational work [42] confirming that only [Pd (peptide) (H 2 O) 2 ] ? complexes, in which the scissile peptide bond has a trans conformation, possess the adequate structure and stability for being hydrolytically active. Our calculations, which are in consonance with the most important experimental facts and their interpretation, have also uncovered the critical roles played by both the Pd-peptide dicoordination environment and one of the Pdbound water molecules during the hydrolysis reaction. In this scenario, other plausible Pd-peptide complexes lacking these two essential features, Pd-peptide dicoordination and Pd-bound water, would not be reactive.
An experimental investigation into the [Pd(H 2 O) 4 ] 2?catalyzed hydrolysis of the peptide Ace-Ala-Lys-Tyr-Gly-Gly-Met-Ala-Ala-Arg-Ala at pH 1. 8 4 ] 2? has been ascribed to a relatively slow displacement of the en ligand by the sulfur and nitrogen donor atoms of the peptide Met residue [28]. However, it has also been proposed that the more labile Pdbound water molecules are initially displaced by the peptide molecule from the cis-[Pd (en)(H 2 O) 2 ] 2? complex to form cis-[Pd(pep)(en)] ? . Then, after a relatively slow replacement of the en ligand by water in a process assisted by the acidic solvent, the Pd(II) ion would cleave the scissile peptide bond [28].
In this work, we aim at rationalizing the experimentally observed ligand effects on the hydrolytic cleavage of the Gly-Gly-Met peptide sequence. Based on the details of the theoretical hydrolysis mechanism as well as on the experimental kinetic data, the lower reaction rate in the presence of cis- From previous computational experience in metal-containing systems, the error on the relative energies computed by DFT methodologies is unlikely to exceed a range of 3-5 kcal/mol [43]. Nevertheless, exceptions exist and, therefore, the performance of the B3LYP [44][45][46] method when compared with ab initio methodologies is reassessed in this work. Thus, we performed a series of validation calculations on small palladium(II) compounds that are relevant to the coordination environment of the Pd(II) ion in the metal-peptide complexes. The gas-phase acidity of acetamide, ammonia and ethylamine is also considered in the test calculations given that binding of Pd(II) to the peptide backbone implies the deprotonation of NH groups, while the NH 3 and en ligands would be protonated at the acidic pHs required for the regioselective Pd(II)-assisted peptide hydrolysis. Molecular geometries were optimized using the DFT B3LYP method and the ab initio MP2 method. In both cases, we employed a correlation consistent basis set of triple-f quality: aug-cc-pVTZ for the non-metal atoms plus the cc-pVTZ-PP basis set with extended relativistic pseudo-potentials for Pd [47][48][49][50][51]. Given the acceptable similarity between the B3LYP and MP2 optimized geometries (see below), the nature of the critical points found was only checked by harmonic frequency calculations at the B3LYP/aug-cc-pVTZ (cc-pVTZ-PP for Pd) level.
Electronic energies were refined by performing singlepoint calculations at the CCSD(T)/aug-cc-pVTZ (cc-pVTZ-PP for Pd) level of theory (coupled cluster single and double excitation augmented with a noniterative treatment of triple excitations) [52] on the MP2/aug-cc-pVTZ (cc-pVTZ-PP for Pd) geometries. The effect of larger basis sets on the ab initio energies was also estimated by doing MP2/aug-cc-pVnZ//MP2/aug-cc-pVTZ (n = Q, 5) calculations [47][48][49][50][51]. To extrapolate our results toward the complete basis set (CBS) limit and, thereby, remove basis set truncation errors, we used the Schwarz's extrapolation formula for correlation energy: where n is the cardinal number of the basis set and E CBS and A are fitting parameters, with E CBS being the resulting estimate of the CBS limit of correlation energy. Herein, CBS extrapolations with the aug-cc-pVnZ (n = Q, 5) basis sets were used systematically on the MP2 correlation energies. The HF energies were not extrapolated, and the 5Z values were simply taken as the most accurate estimates of the HF limit. Then, the following ''composite'' formula was used to obtain a better estimate of the electronic energy: The frozen core approximation was used in all the correlated ab initio calculations. B3LYP and MP2 calculations were done using the Gaussian 03 [53] and TURBOMOLE [54] packages, respectively, while the MOLPRO code [55] was used for the CCSD(T) computations [56]. The MP2 calculations were performed in the framework of the ''resolution-of-the-identity'' approximation (RI-MP2) using the appropriate auxiliary basis set.

Solvent effects and calculation of equilibrium constants
To take into account condensed-phase effects, we used a Self-Consistent-Reaction-Field (SCRF) model proposed for quantum mechanical computations on solvated molecules [57,58]. The solvent is represented by a dielectric continuum characterized by its relative static dielectric permittivity e. The solute, which is placed in a cavity created in the continuum after spending some cavitation energy, polarizes the continuum, which in turn creates an electric field inside the cavity. This interaction can be taken into account using quantum chemical methods by minimizing the electronic energy of the solute plus the Gibbs energy change corresponding to the solvation process [59].
Among the different approaches that can be followed to calculate the electrostatic potential created by the polarized continuum in the cavity, we employed the Poisson-Boltzmann (PB) model as implemented in the JAGUAR program [60]. DG solv values were evaluated through geometry reoptimizations of all the species in solution at the PB-B3LYP/LACVP*? [61, 62] level, excepting H ? and H 2 O (see below). A relative permittivity of 80.37 was employed to simulate water as the solvent used in the experimental work. In these calculations, we used the LACVP*? basis, which comprises the 6-31?G* basis set for non-metal atoms and the Los Alamos effective core potentials for palladium, as larger basis sets including diffuse and polarization functions hardly change the aqueous solvation effects [63]. For all the species, Gibbs energies in solution (G solution ) were computed by adding the Gibbs energies of solvation (DG solv ) to Gibbs energies in the gas phase (G gas ). The G gas values were obtained by combining the B3LYP/augcc-pVTZ electronic energies with the gas-phase thermal corrections evaluated at the B3LYP/aug-cc-pVTZ(cc-pVTZ-PP for Pd) level (for the test calculations on small systems) or the B3LYP/aug-cc-pVDZ(cc-pVDZ-PP for Pd) [47][48][49][50][51] level (for the QM cluster models of the Pd-peptide complexes, see below). Thermodynamic magnitudes were computed within the ideal gas, rigid rotor, and harmonic oscillator approximations at a pressure of 1 atm and a temperature of 298.15 K. To obtain G solution values at the standard state of 1 M, it must be taken into account that the translational entropy for each species is smaller by 6.4 cal mol -1 K -1 than the entropy value obtained for the standard state of an (ideal) gas (1 atm, 298.15 K ? 0.041 M). For water, a standard concentration of 55 M was taken.
The standard free energy of a proton that is needed to compute pK a values was estimated by combining the gasphase free energy of a proton 5 [65] in the calculations of stability constants for Pd-complexes as we found that this value, which is 3.2 kcal/mol less negative than the PB-B3LYP/LACVP*? one, improves the agreement between theoretical and experimental data for the small complexes.

QM calculations on the Pd-peptide cluster models
In our previous work [42], we found that (a) the central region of the peptide molecules nearby the Pd(II) ion is conformationally rigid, and (b) QM calculations on truncated models of the peptide molecules (Ace-Gly-Gly-Met-Nme or Ace-Gly-Gly-Pro-Nme) can reproduce both the structure of the first coordination shell around the Pd(II) ion and the relative stability of various coordination modes. Assuming that exchange of water by other ligands at the Pd(II) site would mainly affect the Pd-ligand bond energies, the stability of the Pd-peptide complexes is then evaluated using again a truncated peptide model (Ace-Gly-Gly-Met-Nme).
Initial geometries for the QM cluster models were built based on the structure of the complexes studied in our previous work [42]. The augmented correlation basis sets aug-cc-pVnZ (cc-pVTZ-PP for Pd) with n = D, T, in conjunction with the hybrid density functional method B3LYP, were used to fully optimize all the critical structures. The structures were first optimized and their nature as energy minima on the Potential Energy Surface confirmed by analytical computations of harmonic vibrational frequencies at the B3LYP/aug-cc-pVDZ (cc-pVDZ-PP for Pd) level. Subsequently, the structures were reoptimized at the B3LYP/aug-cc-pVTZ (cc-pVTZ-PP for Pd) level. As the molecular geometries obtained with the double-and triple-f correlation consistent basis sets were quite similar, we did not carry out the computationally very costly frequency calculations using the larger basis set (aug-cc-pVTZ). For the sake of brevity, we only report the molecular geometries at the B3LYP/aug-cc-pVTZ (cc-pVTZ-PP for Pd) level.
All the quantum chemical computations at the B3LYP/ aug-cc-pVDZ (cc-pVDZ-PP for Pd) theory level and only those at the B3LYP/aug-cc-pVTZ (cc-pVTZ-PP for Pd) theory level involving molecular systems other than Pep and Pd(II)-Pep complexes were carried out with the Gaussian 03 series of program [53]. To perform the B3LYP/aug-cc-pVTZ (cc-pVTZ-PP for Pd) calculations on the Pd-peptide complexes, we employed the TURBO-MOLE V5.9 code [54] selecting the B3LYP version equivalent to that implemented in Gaussian 03. Gibbs free energies in the gas phase and in solution were computed using the prescriptions above described (i.e., all the structures were reoptimized in solution at the PB-B3LYP/ LACVP*? level).

Validation computations
Before presenting and discussing the obtained results for the Pd-peptide models, we corroborate the adequacy of the B3LYP/aug-cc-pVTZ(cc-pVTZ-PP for Pd) level for computing their molecular geometry and electronic energy by performing ab initio benchmarking calculations based on MP2/aug-cc-pVTZ(cc-pVTZ-PP for Pd) geometry optimizations followed by single-point calculations at the CCSD(T)/aug-cc-pVTZ(cc-pVTZ-PP for Pd) and MP2/ aug-cc-pVnZ(cc-pVXZ-PP for Pd) with n = Q, 5 levels of theory. To accomplish this task, we selected a series of proton dissociation and ligand exchange processes on small molecules with the aim of ranging over most of the chemical features involved in the larger Pd-peptide complexes.
In general, the MP2 and B3LYP molecular geometries of the test compounds are very similar to each other. Thus, the MP2 computed bond lengths and bond angles of the structures involved in the gas-phase acidity of acetamide, ammonia, and ethylamine differ from the B3LYP values by only (-1.0%) to (?0.4%) and (-1.0%) to (?0.2%), respectively. More significant deviations were found when comparing the B3LYP [66]. This and other MD studies have also reported that the first solvation shell of the Pd(II)-aqua complex presents four water molecules in a square-planar conformation confirming the reasonable adequacy of considering four water ligands in our computations [66][67][68]. . For the extremely weak acid (acetamide) and the stronger acids, ammonium and methylammonium, the B3LYP/aug-cc-pVTZ energies closely match the best ab initio estimate provided by the composite CCSD(T)/ aug-cc-pVTZ and MP2/CBS energies, the energy differences being lower than 0.5 kcal/mol. The gas-phase acidity of the [Pd(H 2 O) 4 ] 2? cation is slightly less accentuated at the B3LYP/aug-cc-pVTZ, but the difference between the B3LYP and ab initio energies is also quite low, 1.5 kcal/ mol. We also computed the electronic reaction energy of ligand replacement at [Pd(H 2 O) 4 ] 2? leading to the monoamin-, monochloro-, dichloro-and tetrachloro-Pd(II) complexes. In this case, we find that the discrepancy between the B3LYP and the composite ab initio reaction energies varies between 1.5 and 13.2 kcal/mol, the highest deviations corresponding to the processes involving chloride ions. These results clearly indicate that the energetics of the metal-ligand bonds is more sensitive to the correlation method and basis set effects than the deprotonation energies, particularly in the case of the H 2 O ? Clexchange processes. Overall, we conclude that the B3LYP reaction energies for deprotonation and formation of [Pd(NH 3 )(H 2 O) 3 ] 2? do not deviate significantly from the CCSD(T) and composite ones, while the relative stability of the Pd-chloride bonds tends to be underestimated by a few kcal/mol.
In addition to reliable free energies in the gas phase, the calculation of pK a 's and stability constants of metal complexes also demands accurate solvation energies for the different species involved. Hence, after having calibrated the electronic B3LYP/aug-cc-pVTZ energies, it is equally important to assess the performance of the Gibbs solvation energies for computing equilibrium constants.
To this end, we evaluated the free energies and the equilibrium constant in aqueous solution for the deprotonation and ligand exchange reactions above considered, and for which experimental data are available (see Table 2). As mentioned in the Computational Details, we employed the PB-B3LYP/LACVP*? level of theory to compute the solvation energies, excepting those of H ? and bulk H 2 O, which were chosen to give the best agreement with experimental data. Thus, by adjusting DG solvation ðH þ Þto -266.0 kcal/mol, the computed pK a 's for CH 3  are all within *1 pK unit from the experimental ones (see Table 2). Therefore, we believe that this degree of agreement is quite remarkable considering the diverse nature and acid strength of these molecules (note also that in these calculations the gas-phase Gibbs energy is obtained at the B3LYP/aug-cc-pVTZ level). We note in passing that the geometry relaxation effects accounted for by the PB-B3LYP/LACVP*? solvation energies are relatively important for obtaining reliable pK a predictions for the species studied in this work. For example, the com-  Table 2). The combination of the B3LYP/aug-cc-pVTZ electronic energies with the PB-B3LYP/LACVP*? solvation energies gives a quite reasonable value for the log b 1 (NH 3 ) constant, 10.2, which is only 0.5 log units above the experimental value. For the Pd-chloro complexes, the resulting log b constants are about 2 units below the experimental ones, what is not entirely unexpected given that the B3LYP/aug-cc-ppVTZ reaction energies deviate from the ab initio values by 2-13 kcal/mol although it also seems that the logb calculations for Clligands benefit from a partial cancelation of errors. With all these results in mind, we note that our computational protocol based on the B3LYP method would be more adequate to determine the relative stability of the complexes [Pd(L)(peptide)] (L = H 2 O, en, NH 3 ) than those including L = Cl -.

Pd-peptide cluster models
As mentioned in the Introduction, our previous calculations have revealed that [Pd(H 2 O) 4 ] 2? can promote hydrolysis of the Ace-Ala-Lys-Tyr-Gly-Gly-Met-Ala-Ala-Arg-Ala peptide through an external attack of a water molecule to the scissile Gly-Gly peptide bond in trans conformation (see Scheme 2) [42]. To analyze the changes in the stability of the hydrolytically active Pd-peptide complex upon ligand  4 ] 2? complexes, are substituted by en, NH 3 and Cl -. For these processes to occur, an acidic environment is assumed in order to match more closely the experimental conditions. This means that the en and NH 3 ligands should be considered to be in their protonated forms as enH 2 2? and NH 4 ? , respectively, when they are not bound to the Pd(II) ion. The chemical equations and the computed energy data are summarized in Table 3 while Fig. 1 displays the optimized geometries of the Pd-bound peptide complexes at the B3LYP/aug-cc-pVTZ (cc-pVTZ-PP for Pd) level of theory. Unless otherwise stated, we will discuss in the text the Gibbs energies in solution.

Influence of ethylenediamine ligand
The chelate effect associated to the bidentate en ligand is well reproduced by our calculations. Thus, the cis-[Pd(en)(H 2 O) 2 ] 2? complex is predicted to be 11.2 kcal/mol more stable than [Pd(H 2 O) 4 ] 2? . This free energy difference assumes a 1 M standard concentration of H ? in aqueous solution, so that water replacement by en takes place with the release of two H ? from the doubly protonated enH 2? species. In Fig. 1, we see that the equilibrium geometry of the corresponding complexes in the presence of the Pep ligand, [Pd(Pep)(H 2 O) 2 ] ? and [Pd(Pep)(en)] ? , is quite similar in terms of the Pep-moiety conformation and the intramolecular H-bonds between the Pd-water/amino groups and the backbone carbonyl group. We note in passing that this structural similarity gives support to the use of these QM cluster models for estimating the relative stability of the Pd-peptide complexes. Interestingly, the computed free energies indicate that water replacement by en stabilizes the Pd-Pep complex by 3.1 kcal/mol, that is, the presence of the Pep ligand diminishes the chelate effect of the en ligand in such a way that the resulting energy difference could be small enough for establishing an equilibrium between the [Pd(Pep)(H 2 O) 2 ] ? and [Pd(Pep)(en)] ? complexes.

Influence of ammonia ligands
According to our calculations in acid aqueous solution, the predicted stability of the Pd-ammonia complexes with respect to [Pd(H 2 O) 4 ] 2? is rather low in consonance with expectation. For example, [Pd(NH 3 ) 2 (H 2 O) 2 ] 2? is slightly more stable than the Pd(II)-aqua complex by 0.9 kcal/mol. Moreover, when the experimentally used [Pd(NH 3 ) 4 ] 2? is taken into account in the calculations, the stability is reversed by 7.1 kcal/mol (see Table 3). From a structural point of view, ammonia and water ligands result in very similar [Pd(Pep)(L) 2 ] ? complexes (see Fig. 1

Influence of chloride ligands
According to our results and taking as reference the [Pd(H 2 O) 4 ] ? species, the replacement of two and four water ligands by Clions largely stabilizes the Pd-complex by 10.7 and 18.1 kcal/mol, respectively. Similarly, when comparing the relative stability of the Pd(II) complexes coordinated to the Pep fragment, the [Pd(Pep)(Cl) 2 ]complex is much more stable than the precursor water complex by 17.2 kcal/mol. These figures can be understood in terms of the larger strength of the metal-ligand bonds formed between the Pd(II) and Clions. However, they should be taken with caution, particularly for the Pd-peptide structures. First, our validation calculations suggest that the quality of the relative stability calculations involving chloro-complexes would be lower than those for the other ligands. Secondly, the use of a tripeptide model instead of the full decapeptide can result in larger truncation effects in the [Pd(Pep)(Cl) 2 ]complexes. In effect, as seen in Fig. 1 Fig. 1). The resulting structure is a model of complex 3 in Scheme 1, so that the computed change in the standard free energy in solution, which amounts to -2.7 kcal/mol, can give some indication about the stability of the reactive complex 2 with respect to the acidity of the solution.

Implications for peptide hydrolysis
The mechanistic proposal that the [Pd(Pep)(H 2 O) 2 ] ? complex is the hydrolytically active species regardless the complex as the latter one is only about 3 kcal/mol less stable in terms of standard free energy. In this way, the hydrolytic delay experimentally observed can be attributed to a slow displacement of the bidentate en ligand by two water molecules instead of by the N and S donor atoms in peptide as experimentally suggested [28].
As indicated before, the first-order rate constant for the cis-[Pd(en)(H 2 O) 2 ] 2? and [Pd(Cl) 4 ] 2complexes reacting with the peptide molecules is nearly identical (0.14 and 0.13 h -1 ) [28]. Our calculations show again that [Pd(Pep)(Cl) 2 ]is more stable than [Pd(Pep)(H 2 O) 2 ] ? , what is in qualitative agreement with the observed kinetic effect for the Clligands. However, the calculated DG solution value, -17.2 kcal/mol favoring [Pd(Pep)(Cl) 2 ] -, is clearly too large in absolute value for explaining the similar rate constants experimentally reported for the en and Clligands. As previously noticed, two computational facts can justify this discrepancy: a poorer theoretical description of the Pd-Cl bond energies and larger truncation effects due to the use of a cluster model in our QM calculations. We believe that truncation effects play a more important role since the stability constants of the small [Pd(H 2 O) 2 Cl 2 ] and [PdCl 4 ] 2complexes are actually underestimated according to our benchmarking calculations (see Table 2). Nevertheless, the possibility that the peptide hydrolysis reaction could also take place at the [Pd(Pep)(Cl) 2 ]complex through the attack of an external water molecule cannot be ruled out.
The rate constant for the regioselective peptide hydrolysis induced by the Pd(II) complexes increases with acidity. On the basis of NMR experiments, it has been demonstrated that at pH 2. reaction is not large, -2.7 kcal/mol, and thereby an acidbase equilibrium between the two Pd(II) complexes representing 2 and 3 can be expected: the lower pH, the larger concentration of 2.

Summary and conclusions
In this work, we present benchmark ab initio calculations on small Pd(II) complexes that together with a careful comparison with experimental data (pK a and stability constants) allow us to propose the combination of the B3LYP/aug-cc-pVTZ level of theory with the PB-B3LYP/ LACVP*? solvation energies in order to estimate reliable free energies in solution for Pd-peptide complexes with different ligands: H 2 O, NH 3 , en and Cl -. This computational protocol, which could be useful to analyze other problems related with the relative stability of metal-peptide complexes, allow us to study the ligand effects on the kinetics of cleavage of Ace-Ala-Lys-Tyr-Gly-Gly-Met-Ala-Ala-Arg-Ala promoted by different Pd(II) complexes. Thus, our calculations confirm that complexes in which the Pd(II) ion carries labile ligands such as [Pd(H 2 O) 4 ] 2? and [Pd(NH 3 ) 4 ] 2? would be more effective than those containing a bidentate ligand, such as cis-[Pd(en)(H 2 O) 2 ] 2? , or anionic ligands, such as [PdCl 4 ] 2-, as promoters of the cleavage. Besides, we propose a more detailed explanation of the experimental facts on the basis of the computed relative stability of the various Pd-peptide complexes with respect to the hydrolytically active complex [Pd(peptide)(H 2 O) 2 ] ? and on the assumption of an initial formation of the respective peptide-bound complexes, which subsequently evolve to give [Pd(peptide)(H 2 O) 2 ] 2? . The greater or lower labile character of the ligands en, Cl -, and NH 3 compared to H 2 O coordinated to Pd seems to control the relative stability of the Pd-peptide complexes investigated in this work and, as a consequence, the global rate constant of peptide cleavage.