Title: Prediction of Collagen Stability from Amino Acid Sequence
Abstract: An algorithm was derived to relate the amino acid sequence of a collagen triple helix to its thermal stability. This calculation is based on the triple helical stabilization propensities of individual residues and their intermolecular and intramolecular interactions, as quantitated by melting temperature values of host-guest peptides. Experimental melting temperature values of a number of triple helical peptides of varying length and sequence were successfully predicted by this algorithm. However, predicted Tm values are significantly higher than experimental values when there are strings of oppositely charged residues or concentrations of like charges near the terminus. Application of the algorithm to collagen sequences highlights regions of unusually high or low stability, and these regions often correlate with biologically significant features. The prediction of stability from sequence indicates an understanding of the major forces maintaining this protein motif. The use of highly favorable KGE and KGD sequences is seen to complement the stabilizing effects of imino acids in modulating stability and may become dominant in the collagenous domains of bacterial proteins that lack hydroxyproline. The effect of single amino acid mutations in the X and Y positions can be evaluated with this algorithm. An interactive collagen stability calculator based on this algorithm is available online. An algorithm was derived to relate the amino acid sequence of a collagen triple helix to its thermal stability. This calculation is based on the triple helical stabilization propensities of individual residues and their intermolecular and intramolecular interactions, as quantitated by melting temperature values of host-guest peptides. Experimental melting temperature values of a number of triple helical peptides of varying length and sequence were successfully predicted by this algorithm. However, predicted Tm values are significantly higher than experimental values when there are strings of oppositely charged residues or concentrations of like charges near the terminus. Application of the algorithm to collagen sequences highlights regions of unusually high or low stability, and these regions often correlate with biologically significant features. The prediction of stability from sequence indicates an understanding of the major forces maintaining this protein motif. The use of highly favorable KGE and KGD sequences is seen to complement the stabilizing effects of imino acids in modulating stability and may become dominant in the collagenous domains of bacterial proteins that lack hydroxyproline. The effect of single amino acid mutations in the X and Y positions can be evaluated with this algorithm. An interactive collagen stability calculator based on this algorithm is available online. The ability to predict structure and stability from amino acid sequence is an important step in the understanding of basic protein principles and the structural consequences of pathological mutations. The vast number of amino acid sequences available from DNA data contrasts with the smaller number of high resolution protein structures and the limited experimental data on protein stability. The ability to make predictions that are in good agreement with experimental data provides insight into the stabilizing interactions within proteins. In addition, there is much interest in computing the effect of single amino acid replacements on protein stability because destabilizing effects are associated with deleterious mutations that result in clinically detectable phenotypes (1Wang Z. Moult J. Hum. Mutat. 2001; 17: 263-270Crossref PubMed Scopus (546) Google Scholar, 2Guerois R. Nielsen J.E. Serrano L. J. Mol. Biol. 2002; 320: 369-387Crossref PubMed Scopus (1292) Google Scholar, 3Persikov A.V. Pillitteri R.J. Amin P. Schwarze U. Byers P.H. Brodsky B. Hum. Mutat. 2004; 24: 330-337Crossref PubMed Scopus (92) Google Scholar). In contrast to globular proteins, the relation among sequence, structure, and stability is simpler and better defined for the linear collagen triple helix.The collagen triple helix motif is found widely in structural proteins of the extracellular matrix and in an increasing set of non-collagenous proteins, many of which are involved in host-defense functions (4Myllyharju J. Kivirikko K.I. Trends Genet. 2004; 20: 33-43Abstract Full Text Full Text PDF PubMed Scopus (871) Google Scholar, 5Brodsky B. Persikov A.V. Adv. Protein Chem. 2005; 70: 301-339Crossref PubMed Scopus (419) Google Scholar). The close packing of three supercoiled polyproline II-like polypeptide chains in the collagen triple helix generates a requirement for Gly as every third residue (6Rich A. Crick F.H. J. Mol. Biol. 1961; 3: 483-506Crossref PubMed Scopus (590) Google Scholar, 7Ramachandran G.N. Int. Rev. Connect. Tissue Res. 1963; 68: 127-182Crossref Google Scholar, 8Bella J. Eaton M. Brodsky B. Berman H.M. Science. 1994; 266: 75-81Crossref PubMed Scopus (874) Google Scholar). The observation of such a repeating (Gly-X-Y)n sequence pattern over a stretch of residues signifies a triple helix conformation. However, the collagen triple helix is not uniform in structure or stability. Crystal structures of collagen peptides show that variation in amino acid content leads to small but significant variations in the super-helix twist (9Kramer R.Z. Bella J. Mayville P. Brodsky B. Berman H.M. Nat. Struct. Biol. 1999; 6: 454-457Crossref PubMed Scopus (281) Google Scholar, 10Kramer R.Z. Bella J. Brodsky B. Berman H.M. J. Mol. Biol. 2001; 311: 131-147Crossref PubMed Scopus (178) Google Scholar, 11Emsley J. Knight C.G. Farndale R.W. Barnes M.J. J. Mol. Biol. 2004; 335: 1019-1028Crossref PubMed Scopus (112) Google Scholar). Calorimetric results suggest the presence of multiple independent folding domains along a collagen molecule (12Privalov P.L. Adv. Protein Chem. 1982; 35: 1-104Crossref PubMed Scopus (938) Google Scholar), and the presence of regions of different stability was confirmed by recent studies on recombinant collagen constructs (13Steplewski A. Ito H. Rucker E. Brittingham R.J. Alabyeva T. Gandhi M. Ko F.K. Birk D.E. Jimenez S.A. Fertala A. J. Struct. Biol. 2004; 148: 326-337Crossref PubMed Scopus (39) Google Scholar). There are multiple binding domains in collagens (14Di Lullo G.A. Sweeney S.M. Korkko J. Ala-Kokko L. San Antonio J.D. J. Biol. Chem. 2002; 277: 4223-4231Abstract Full Text Full Text PDF PubMed Scopus (669) Google Scholar), and regions of decreased triple helix stability have been implicated in binding in some cases (15Shah N.K. Sharma M. Kirkpatrick A. Ramshaw J.A.M. Brodsky B. Biochemistry. 1997; 36: 5878-5883Crossref PubMed Scopus (47) Google Scholar, 16Chung L. Dinakarpandian D. Yoshida N. Lauer-Fields J.L. Fields G.B. Visse R. Nagase H. EMBO J. 2004; 23: 3020-3030Crossref PubMed Scopus (355) Google Scholar, 17Deprez P. Doss-Pepe E. Brodsky B. Inestrosa N.C. Biochem. J. 2000; 350: 283-290Crossref PubMed Scopus (31) Google Scholar). Self-association of type I collagen into fibrils is preceded by microunfolding of specific triple helix regions (18Leikina E. Mertts M.V. Kuznetsova N. Leikin S. Proc. Natl. Acad. Sci. U. S. A. 2002; 99: 1314-1318Crossref PubMed Scopus (443) Google Scholar, 19Kadler K.E. Hojima Y. Prockop D.J. J. Biol. Chem. 1988; 263: 10517-10523Abstract Full Text PDF PubMed Google Scholar). Thus, specific residues along the (Gly-X-Y)n sequence determine functionally important modulation of structure and stability.Experimental thermal stability data obtained from host-guest peptides is integrated here to produce an algorithm for predicting global melting temperatures of collagen triple helical peptides and short fragments and for detecting modulations in relative stability along a collagen chain. Good agreement is observed between predicted and observed stabilities of a number of collagen peptides. In cases in which the predicted Tm is significantly different from that observed, interactions involving longer range electrostatic interactions or unraveling of the ends are suggested. The variations in stability along the collagen chain appear related to known functional sites, and high stability is achieved through a combination of stabilizing imino acid and KGE/D sequences.MATERIALS AND METHODSThe Tm values of all host-guest peptides were measured under a set of standard conditions, with c = 1 mg/ml in phosphate-buffered saline, pH 7.0, and with a heating rate average of 0.1 °C/min, as previously reported (20Persikov A.V. Xu Y. Brodsky B. Protein Sci. 2004; 13: 893-902Crossref PubMed Scopus (121) Google Scholar). Small variations are seen at acid versus neutral pH, but all calculations are based on host-guest peptide data collected at pH 7.The (Pro-Hyp-Gly)n peptides for n = 6, 7, 8, and 12 were synthesized by Tufts Core Facility (Boston, MA) and purified using high pressure liquid chromatography; their identity was confirmed by matrix-assisted laser desorption ionization.To extrapolate the dependence of the Tm of the host peptides on peptide length, the experimental values for (Pro-Hyp-Gly)n and (Pro-Pro-Gly)n versus n, where n is the number of tripeptide units, were fit to the exponential decay functionTm0=Tmmax−A⋅exp(−nn0)eq.1 where Tm0(n) is defined as the base thermal stability of the repeating polytripeptide standard, Tmmax is the maximum melting temperature, and the constant n0 represents the length of the repeating peptide with Tm = 0.RESULTS AND DISCUSSIONExperimental Stabilities of Host Guest PeptidesExperimental data on host-guest triple helical peptides, using a (Gly-Pro-Hyp)8 host, have provided information on the propensities of individual residues for the X and Y positions of Gly-X-Y triplets, the interactions within the triple helix for a given Gly-X-Y sequence, and the interactions resulting from neighboring tripeptide sequences (21Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. Biochemistry. 2000; 39: 14960-14967Crossref PubMed Scopus (314) Google Scholar, 22Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. J. Mol. Biol. 2002; 316: 385-394Crossref PubMed Scopus (103) Google Scholar, 23Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. Biochemistry. 2005; 44: 1414-1422Crossref PubMed Scopus (154) Google Scholar). These data establish the basis for determining the loss of stability that will result from replacing Gly-Pro-Hyp tripeptide sequences by other Gly-X-Y sequences, essentially defining a set of rules for relating amino acid sequence and stability.Individual Residue Propensities for X and Y Positions—The propensity measurements for all 20 residues in the X position in a Gly-X-Hyp context and all 20 residues in the Y position in a Gly-Pro-Y context were determined by measuring thermal stability of host-guest peptides (21Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. Biochemistry. 2000; 39: 14960-14967Crossref PubMed Scopus (314) Google Scholar). The most stable tripeptide unit is Gly-Pro-Hyp (Tm = 47.3 °C). Replacing Pro in the X position leads to a decrease in stability ranging from 4 °C for Gly-Glu-Hyp (Tm = 42.9 °C) to 15 °C for Gly-Trp-Hyp (Tm = 31.9 °C). Replacing Hyp in the Y position leads to a decrease in stability ranging from almost 0 °C for Gly-Pro-Arg (Tm = 47.2 °C) to 21 °C for Gly-Pro-Hyp (Tm = 26.1 °C).Gly-X-Y Tripeptide Sequences—Direct intrachain interactions are not sterically possible between adjacent X and Y residues in the Gly-X-Y unit of a chain, but interchain interactions can take place between the Y residue in one chain and the X residue in an adjacent chain staggered by 1 residue (Fig. 1). Peptides with Gly-X-Y guest triplets were designed to model these interchain interactions. Only a restricted set of possible Gly-X-Y tripeptides are significantly populated in collagens (24Ramshaw J.A.M. Shah N.K. Brodsky B. J. Struct. Biol. 1998; 122: 86-91Crossref PubMed Scopus (280) Google Scholar), reflecting in part strong preferences for basic residues to be in the Y position and for Glu and hydrophobic residues to be in the X position and very low occurrence of Cys, Trp, and Tyr. A limited set of 41 guest Gly-X-Y sequences was selected to include the most common tripeptide sequences and to model a range of typical electrostatic and hydrophobic interactions. Because of the strong bias in collagen compositions, the selected 41 Gly-X-Y, 19 Gly-X-Hyp, 19 Gly-Pro-Y, and Gly-Pro-Hyp tripeptides cover about 80% of human fibrillar collagen sequences (22Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. J. Mol. Biol. 2002; 316: 385-394Crossref PubMed Scopus (103) Google Scholar). Although Pro residues in the Y position are post-translationally modified to Hyp in multicellular animals, collagenous domains have recently been found in bacteria and viruses where there is no hydroxylation of Pro (25Rasmussen M. Jacobsson M. Bjorck L. J. Biol. Chem. 2003; 278: 32313-32316Abstract Full Text Full Text PDF PubMed Scopus (113) Google Scholar, 26Xu Y. Keene D.R. Bujnicki J.M. Hook M. Lukomski S. J. Biol. Chem. 2002; 277: 27312-27318Abstract Full Text Full Text PDF PubMed Scopus (151) Google Scholar). To model these sequences, Gly-Pro-Pro and Gly-Ala-Pro guest triplets were also included.A complete table of the stability for all Gly-X-Y triplets was constructed using the experimental values for all frequent sequences and the predicted values for all others (Table I; experimental values are in bold). Predicted values were calculated on the basis of additivity of residues in the X and Y position (22Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. J. Mol. Biol. 2002; 316: 385-394Crossref PubMed Scopus (103) Google Scholar).TmGXY=TmGXO+TmGPY−TmGPOeq.2 Table IPredicted and experimentally observed (bold) Tm values (in °C) for all possible Gly-X-Y tripeptide units in a triple helix, based on host-guest peptide studies The rows of amino acids are listed in order of their X position propensity for triple helix formation, whereas the amino acids in columns are listed in order of their Y position propensity. Both Pro and Hyp (O) are included in the Y position.X\YOPRMIQAVETCKHSDGLNYFWP47.345.547.242.641.541.340.940.039.739.737.736.835.735.034.032.932.731.730.228.426.1E42.941.140.438.237.137.734.635.335.335.933.335.031.330.629.728.528.329.525.824.021.7A41.737.738.237.035.935.732.934.434.134.132.130.830.133.033.027.327.826.124.621.920.5K41.539.739.136.835.738.935.134.235.333.931.931.029.929.235.827.126.931.724.422.620.3R40.638.838.035.934.834.634.233.333.833.031.029.529.030.534.526.226.025.023.521.719.4Q40.438.639.535.734.634.434.033.132.832.830.832.628.828.127.126.025.824.823.321.519.2D40.138.337.135.434.334.131.632.832.532.530.530.928.527.826.825.725.524.523.021.218.9L39.037.236.434.333.235.731.231.731.431.429.431.127.426.725.724.626.923.421.920.117.8V38.937.136.334.233.132.932.531.631.331.329.332.527.326.625.624.524.323.321.820.017.7M38.636.836.033.932.832.632.231.331.031.029.031.727.026.325.324.224.023.021.519.717.4I38.436.635.833.732.632.433.931.130.830.828.827.926.826.125.124.023.822.821.319.517.2N38.336.535.733.632.532.331.931.030.730.728.727.826.726.025.023.923.722.721.219.417.1S38.036.235.433.332.232.031.630.730.430.428.427.526.425.724.723.623.422.420.919.116.8H36.534.733.931.830.730.530.129.228.928.926.926.024.924.223.222.121.920.919.417.615.3T36.234.433.631.530.430.229.828.928.628.626.625.724.623.922.921.821.620.619.117.315.0C36.134.333.531.430.330.129.728.828.528.526.525.624.523.822.821.721.520.519.017.214.9Y34.332.531.729.628.528.327.927.026.726.724.723.822.722.021.019.919.718.717.215.413.1F33.531.730.928.827.727.524.126.225.925.923.923.021.921.220.219.118.917.916.414.612.3G33.231.430.628.527.427.226.025.925.625.623.626.921.620.919.918.825.317.616.119.712.0W31.930.129.327.226.125.925.524.624.324.322.321.420.319.618.617.517.316.314.813.010.7 Open table in a new tab The predicted values gave good agreement (within ±3 °C) for the GAA and for 28 other guest triplets of the 41 Gly-X-Y triplets studied. The largest deviations were observed for GKD and GRD, which were more stable than predicted by 7 °C, suggesting some interchain electrostatic stabilization. Observed Tm values of GXR sequences (GER, GAR, GKR, GQR, and GDR) were 2.5 °C smaller on average than predicted, indicating the need for a correction factor in a GXR context versus a GPR context.Interactions between Adjacent Gly-X-Y Units—Interactions between adjacent Gly-X-Y tripeptides were included in the calculations. A recent study reported stabilities of a selection of host-guest peptides including residues in two adjacent tripeptide units, Gly-X-Y-Gly-X′-Y′, covering possible direct interchain or intrachain interactions between residues that are separated by ≤3 residues in sequence (23Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. Biochemistry. 2005; 44: 1414-1422Crossref PubMed Scopus (154) Google Scholar) (Fig. 1). Significant deviations from predicted stability were seen for six hexapeptides, which suggested favorable interchain and intrachain electrostatic and hydrophobic interactions (Table II). The most dramatic difference was the electrostatic and hydrogen bonding stabilization observed when Lys is in the Y position and a negatively charged residue is in the X′ position (KGD or KGE), with observed Tm values 15.4 °C to 17.5 °C more stable than expected. The large magnitude of KGD/E interactions is comparable with the Tm spread of all X (14 °C) and Y residues (21 °C) (23Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. Biochemistry. 2005; 44: 1414-1422Crossref PubMed Scopus (154) Google Scholar) (Table II).Table IIA list of ΔTm corrections for the most significant stabilizing electrostatic and hydrophobic pairwise interactions between residues in adjacent tripeptide units (23Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. Biochemistry. 2005; 44: 1414-1422Crossref PubMed Scopus (154) Google Scholar)Sequence motifΔTm°CG_LGL_+3.8GL_GL_+7.4G_KGD_+17.5G_KGE_+15.4GE_G_K+5.6G_KG_E+7.3 Open table in a new tab Effect of Peptide Length and Blocking Groups on StabilityThe methodology for determining the global stability of a peptide (see below) employs subtraction of the relative stability of the tripeptide sequences that make up a peptide from that expected for the repeating Gly-Pro-Hyp sequence of the same length. The Tm value is seen to depend on n (the number of tripeptide units) for repeating (Gly-Pro-Hyp)n and (Gly-Pro-Pro)n peptides (27Sutoh K. Noda H. Biopolymers. 1974; 13: 2391-2404Crossref PubMed Scopus (14) Google Scholar, 28Persikov A.V. Ramshaw J.A. Kirkpatrick A. Brodsky B. J. Am. Chem. Soc. 2003; 125: 11500-11501Crossref PubMed Scopus (81) Google Scholar). The sharp dependence of stability on length leveling off with increasing n can be fit to a single exponential decay (Fig. 2).Fig. 2Effect of peptide length on the melting temperatures of unblocked (POG)n (□) and (PPG)n (○). The Tm values are shown for (Pro-Pro-Gly)n for n = 12, 15, and 20 (c = 3 mg/ml, 3% HAc, equilibrium conditions) (27Sutoh K. Noda H. Biopolymers. 1974; 13: 2391-2404Crossref PubMed Scopus (14) Google Scholar); for n = 10 (20Persikov A.V. Xu Y. Brodsky B. Protein Sci. 2004; 13: 893-902Crossref PubMed Scopus (121) Google Scholar); and for n = 9 (K. Okuyama, unpublished data). The Tm values for (Pro-Hyp-Gly)n were obtained in our laboratory using the standard melting conditions. The extrapolated curve is shown as a solid line, as described by Eq. 1. The two points (▪) show the experimental Tm values for the blocked Ac-(POG)n-NH2 peptides for n = 7 and 8.View Large Image Figure ViewerDownload Hi-res image Download (PPT)The effect of blocking groups on peptide stability was also taken into consideration. Studies at different pH values and on peptides with and without blocked termini are consistent with a reduction of stability by about 2 °C when unblocked charged N termini are present and by about 3 °C when unblocked charged C termini are present, for a peptide length of n = 10 (29Venugopal M.G. Ramshaw J.A.M. Braswell E. Zhu D. Brodsky B. Biochemistry. 1994; 33: 7948-7956Crossref PubMed Scopus (170) Google Scholar). This destabilization is presumed to be due to repulsion when three charged termini are in close proximity, consistent with the unraveling of the termini observed in high resolution structures of collagen peptides (8Bella J. Eaton M. Brodsky B. Berman H.M. Science. 1994; 266: 75-81Crossref PubMed Scopus (874) Google Scholar, 30Li M.H. Fan P. Brodsky B. Baum J. Biochemistry. 1993; 32: 7377-7387Crossref PubMed Scopus (134) Google Scholar). End effects are more pronounced for short peptides than for longer ones, as seen for (Pro-Hyp-Gly)7 and for (Pro-Hyp-Gly)8 (Fig. 2).Algorithm Relating Amino Acid Sequence to Triple Helix StabilityThe relative stability of each Gly-X-Y tripeptide compared with Gly-Pro-Hyp and the interaction between adjacent Gly-X-Y tripeptides were used to derive an algorithm for predicting triple helix stability. The Tm values, rather than Gibbs free energy values, were used for calculating peptide stability. The extremely long times needed to reach equilibrium and the lack of agreement of the equilibrium curve with a two-state model presented practical and theoretical limitations to thermodynamic characterization (20Persikov A.V. Xu Y. Brodsky B. Protein Sci. 2004; 13: 893-902Crossref PubMed Scopus (121) Google Scholar). Fortunately, the use of Tm values obtained under standardized conditions has proved to be useful as an empirical measure of triple helix stability (20Persikov A.V. Xu Y. Brodsky B. Protein Sci. 2004; 13: 893-902Crossref PubMed Scopus (121) Google Scholar). Additivity of Tm values was observed for peptides with residues that cannot interact. Thus, Tm values are seen to be a good measure of relative stability, as long as standard conditions of buffer, pH, and rate of heating are maintained (20Persikov A.V. Xu Y. Brodsky B. Protein Sci. 2004; 13: 893-902Crossref PubMed Scopus (121) Google Scholar). The algorithm predicts a global Tm value for collagen model peptides between 6 and 20 tripeptides in length and predicts a relative stability for collagen sequences.The global thermal stability of homotrimeric triple helical peptides with length 6 ≤ n ≤ 20 is predicted by an algorithm consisting of the following steps.1) For the total number of triplets n in a given peptide, the base TM0(n) for (Pro-Hyp-Gly)n or (Pro-Pro-Gly)n is calculated from the length dependence (Eq. 1), including any effect of blocking groups.2) The melting temperature value is decreased for every triplet in the sequence that is not Gly-Pro-Hyp, subtracting a value of ΔTmGXY (Table I). The N-terminal and C-terminal tripeptide units are excluded from the calculation due to the staggering of the chains and the reported disorder for the peptide ends (8Bella J. Eaton M. Brodsky B. Berman H.M. Science. 1994; 266: 75-81Crossref PubMed Scopus (874) Google Scholar, 10Kramer R.Z. Bella J. Brodsky B. Berman H.M. J. Mol. Biol. 2001; 311: 131-147Crossref PubMed Scopus (178) Google Scholar, 30Li M.H. Fan P. Brodsky B. Baum J. Biochemistry. 1993; 32: 7377-7387Crossref PubMed Scopus (134) Google Scholar, 31Kramer R.Z. Venugopal M.G. Bella J. Mayville P. Brodsky B. Berman H.M. J. Mol. Biol. 2000; 301: 1191-1205Crossref PubMed Scopus (184) Google Scholar).3) The final value for the peptide melting temperature is adjusted using the ΔTmint values for interactions between neighboring tripeptides (Table II).The algorithm can be formulated as follows.Tm=Tm0−∑i=2n−1ΔTmGXY+ΣΔTminteq.3 The collagen stability algorithm is available to all users for calculation of global stability of peptides and local stability variations in collagens and collagen-like domains (rwjms.umdnj.edu/biochemistry/collagen).Prediction of Tm Values for Collagen-like PeptidesThe stability algorithm was applied to 40 synthetic collagen-like peptides whose Tm values have been experimentally determined under the same defined standard conditions (Table III). Most of the peptides are n = 10 tripeptide units in length, and some have unblocked ends, whereas others have terminal blocking groups. Excellent agreement was found between the calculated and observed Tm values for peptides with GPO tripeptide units on both ends. For instance, for the unblocked peptide T3–785, the predicted Tm value is 17.1 °C (58.8 °C - [(47.3 °C - 30.8 °C) + (47.3 °C - 38.2 °C) + (47.3 °C - 31.2 °C)]), in close agreement with the observed Tm of 18.0 °C. When KGE or KGD sequences are present, the good agreement is dependent on the inclusion of ΔTmint correction values for interactions between adjacent triplets. For instance, peptide T1–655, which has GPO caps on both ends, has an observed Tm value of 42.8 °C. If each independent triplet is considered, one would subtract 16.5 °C for GAK, 15.7 °C for GDA, and 6.4 °C for GPA, yielding 58.8 °C - 38.6 °C = 20.2 °C. However, there is a KGD sequence, which gives +17.5 °C, and an increase of 5 °C because the ends are blocked, giving a net predicted value of 42.7 °C, which is very close to the observed value of 42.8 °C. The set of peptides related to T1–892 with GPA sequences on the N-terminal ends also show excellent agreement with predictions. It is notable that the “reverse” peptide, T1–892r, which has the same tripeptide composition but in a different order, has the same Tm as T1–892, supporting the dependence of thermal stability on tripeptide unit composition when there are no interactions present (Table III) (32Buevich A.V. Silva T. Brodsky B. Baum J. J. Biol. Chem. 2004; 279: 46890-46895Abstract Full Text Full Text PDF PubMed Scopus (29) Google Scholar).Table IIIPredicted melting temperatures for test peptide set, compared with the observed valuesTmpred, predicted Tm values. Tmobs, observed Tm values.NameSequenceTmpredTmobsΔTmRef.T1–655Ac-(GPO)3-GAK-GDA-GPO-GPA-(GPO)3-GY-NH242.742.80.1aA. V. Persikov, T. Silva, A. Mohs, J. A. M. Ramshaw, and B. Brodsky, unpublished dataT1–892Ac-GPA-GPA-GPV-GPA-GAR-GPA-(GPO)4-GV-NH228.226.0–2.246Yang W. Battineni M.L. Brodsky B. Biochemistry. 1997; 36: 6930-6935Crossref PubMed Scopus (57) Google ScholarT1–892rAc-(GPO)4-GPA-GPA-GPV-GPA-GAR-GPA-GV-NH228.226.0-2.232Buevich A.V. Silva T. Brodsky B. Baum J. J. Biol. Chem. 2004; 279: 46890-46895Abstract Full Text Full Text PDF PubMed Scopus (29) Google ScholarT1–892(P26A)Ac-(GPA)2-GPV-GPA-GAR-GPA-(GPO)2-GAO-GPO-GV-NH222.624.11.547Xu Y.J. Hyde T. Wang X. Bhate M. Brodsky B. Baum J. Biochemistry. 2003; 42: 8696-8703Crossref PubMed Scopus (22) Google ScholarT1–892(O24A)Ac-(GPA)2-GPV-GPA-GAR-GPA-GPO-GPA-(GPO)2-GV-NH221.823.21.447Xu Y.J. Hyde T. Wang X. Bhate M. Brodsky B. Baum J. Biochemistry. 2003; 42: 8696-8703Crossref PubMed Scopus (22) Google ScholarT1–892 unblGPA-GPA-GPV-GPA-GAR-GPA-(GPO)4-GY23.220.6–3.2aA. V. Persikov, T. Silva, A. Mohs, J. A. M. Ramshaw, and B. Brodsky, unpublished dataT1–904Ac-GAR-GPA-GPQ-GPR-GDK-GET-(GPO)4-GV-NH238.930.8–8.146Yang W. Battineni M.L. Brodsky B. Biochemistry. 1997; 36: 6930-6935Crossref PubMed Scopus (57) Google ScholarT1A2–697Ac-GFO-GAA-GRT-GPO-GPS-GIS-(GPO)4-GV-NH21.6〈4+aA. V. Persikov, T. Silva, A. Mohs, J. A. M. Ramshaw, and B. Brodsky, unpublished dataT2–508Ac-GSO-GAQ-GLQ-GPR-GLO-GTO-(GPO)4-GV-OH21.125.03.9aA. V. Persikov, T. Silva, A. Mohs, J. A. M. Ramshaw, and B. Brodsky, unpublished dataT3–505Ac-GGK-GDA-GAO-GER-GPO-GLA-(GPO)4-GV25.320.9–4.415Shah N.K. Sharma M. Kirkpatrick A. Ramshaw J.A.M. Brodsky B. Biochemistry. 1997; 36: 5878-5883Crossref PubMed Scopus (47) Google ScholarT3–508Ac-GDA-GAO-GER-GPO-GLA-GAO-(GPO)4-GV26.623.2–3.415Shah N.K. Sharma M. Kirkpatrick A. Ramshaw J.A.M. Brodsky B. Biochemistry. 1997; 36: 5878-5883Crossref PubMed Scopus (47) Google ScholarT3–511Ac-GAO-GER-GPO-GLA-GAO-GLR-(GPO)4-GV21.325.94.615Shah N.K. Sharma M. Kirkpatrick A. Ramshaw J.A.M. Brodsky B. Biochemistry. 1997; 36: 5878-5883Crossref PubMed Scopus (47) Google ScholarT3–514Ac-GER-GPO-GLA-GAO-GLR-GGA-(GPO)4-GV6.916.59.615Shah N.K. Sharma M. Kirkpatrick A. Ramshaw J.A.M. Brodsky B. Biochemistry. 1997; 36: 5878-5883Crossref PubMed Scopus (47) Google ScholarT3–517Ac-GPO-GLA-GAO-GLR-GGA-GPO-(GPO)4-GV6.915.88.915Shah N.K. Sharma M. Kirkpatrick A. Ramshaw J.A.M. Brodsky B. Biochemistry. 1997; 36: 5878-5883Crossref PubMed Scopus (47) Google ScholarT3–520Ac-GLA-GAO-GLR-GGA-GPO-GPE-(GPO)4-GV15.417.52.115Shah N.K. Sharma M. Kirkpatrick A. Ramshaw J.A.M. Brodsky B. Biochemistry. 1997; 36: 5878-5883Crossref PubMed Scopus (47) Google ScholarT3–772Ac-GPO-GAO-GPL-GIA-GIT-GAR-GLA-(GPO)4-GG-NH2–4.5〈4+15Shah N.K. Sharma M. Kirkpatrick A. Ramshaw J.A.M. Brodsky B. Biochemistry. 1997; 36: 5878-5883Crossref PubMed Scopus (47) Google ScholarT3–785PO-(GPO)2-GIT-GAR-GLA-(GPO)4-G17.118.00.915Shah N.K. Sharma M. Kirkpatrick A. Ramshaw J.A.M. Brodsky B. Biochemistry. 1997; 36: 5878-5883Crossref PubMed Scopus (47) Google ScholarT3–997Ac-GPR-GNR-GER-GSE-GSO-GHO-GQO-GPO-GPO-GAO-GV-NH21.4〈4+aA. V. Persikov, T. Silva, A. Mohs, J. A. M. Ramshaw, and B. Brodsky, unpublished dataT7–2031Ac-GLA-GEO-GKO-GIO-GLO-GRA-(GPO)4-GV-NH223.325.42.1aA. V. Persikov, T. Silva, A. Mohs, J. A. M. Ramshaw, and B. Brodsky, unpublished dataT7–2058Ac-GER-GER-GEK-GER-GEQ-GRD-(GPO)4-GV-NH236.323.2–13.1aA. V. Persikov, T. Silva, A. Mohs, J. A. M. Ramshaw, and B. Brodsky, unpublished dataAchE-HG-C1Ac-GPO-GPO-GPO-GKR-GKO-GPO-GPO-GPO-GG-NH233.332.3–1.048Doss-Pepe E. Deprez P. Silva T. Inestrosa N.C. Kirkpatrick A. Ramshaw J.A. Brodsky B. Biochim. Biophys. Acta. 2004; 1698: 187-195Crossref PubMed Scopus (13) Google ScholarAchE-HG-C2Ac-GPO-GPO-GRO-GKR-GKO-GPO-GPO-GPO-GG-NH226.626.90.348Doss-Pepe E. Deprez P. Silva T. Inestrosa N.C. Kirkpatrick A. Ramshaw J.A. Brodsky B. Biochim. Biop