A quantitative structure-activity relationship for translocation of tripeptides via the human proton-coupled peptide transporter, hPEPT1 (SLC15A1)

Abstract

The human intestinal proton-coupled peptide transporter, hPEPT1 (SLC15A1), has been identified as an absorptive transporter for both drug substances and prodrugs. An understanding of the prerequisites for transport has so far been obtained from models based on competition experiments. These models have limited value for predicting substrate translocation via hPEPT1. The aim of the present study was to investigate the requirements for translocation via hPEPT1. A set of 55 tripeptides was selected from a principal component analysis based on VolSurf descriptors using a statistical design. The majority of theses tripeptides have not previously been investigated. Translocation of the tripeptides via hPEPT1 was determined in a MDCK/hPEPT1 cell-based translocation assay measuring substrate-induced changes in fluorescence of a membrane potential-sensitive probe. Affinities for hPEPT1 of relevant tripeptides were determined by competition studies with [14C]Gly-Sar in MDCK/hPEPT1 cells. Forty tripeptides were found to be substrates for hPEPT1, having K(m)(app) values in the range 0.4-28 mM. Eight tripeptides were not able to cause a substrate-induced change in fluorescence in the translocation assay and seven tripeptides interacted with the probe itself. The conformationally restricted tripeptide Met-Pro-Pro was identified as a novel high-affinity inhibitor of hPEPT1. We also discovered the first tripeptide (Asp-Ile-Arg) that was neither a substrate nor an inhibitor of hPEPT1. To rationalise the requirements for transport, a quantitative structure-activity relationship model correlating K(m)(app) values with VolSurf descriptors was constructed. This is, to our knowledge, the first predictive model for the translocation of tripeptides via hPEPT1.

Fig. 1

a PCA score plot of 7,800 tripeptides displaying the first two components t[1] and t[2] explaining 31% and 19%, respectively, of the variance in the model. The tripeptides selected from the D-optimal onion design are shown with large dots and tested in the translocation assay. b Loading plot for the first two dimensions

Fig. 2

a Changes in fluorescence measured in the MDCK/hPEPT1 FLIPR® membrane potential translocation assay after addition of increasing concentrations of Gly-Sar (square) or Gly-Sar in combination with 1 mM Met-Pro-Pro (inverted triangle). b Changes in fluorescence measured in the MDCK/hPEPT1 translocation assay after addition of increasing concentrations of Gly-Sar (square) or Gly-Sar in combination with 1.25 mM Trp-Trp-Trp (triangle). The results are mean ± SD of three cell monolayers

Fig. 3

Plot of the PLS coefficients. On the x-axis, the VolSurf descriptors included in the PLS model are shown and the y-axis represents the PLS coefficients. A positive PLS coefficient indicates that a higher value of that VolSurf descriptor has a favourable effect on the −log K _m ^app value, whereas the opposite is true for negative coefficients. The equation for the QSAR model is as follows: −log K _m ^app = 0.15 log P + 0.13 R-OH2 + 0.12 Iw7-OH2 + 0.12 Iw5-OH2 + 0.12 Iw4-OH2 + 0.12 Iw6-OH2 + 0.12 D1-DRY + 0.11 Iw3-OH2 + 0.11 Iw2-OH2 − 0.10 BV12-OH2 − 0.11 EEFR − 0.11 HB2-N1 − 0.12 BV32-OH2 − 0.12 HB6-N1 − 0.12W8-N1 − 0.13 Cw1-OH2 − 0.13W1-OH2 − 0.13 BV22-OH2 − 0.13 HB8-O − 0.13 Cw8-OH2 − 0.14 Cw2-OH2 − 0.14 HB7-N1 − 0.14 Cw3-OH2 − 0.15 HB8-N1 − 0.15W2-N1 − 0.15 HB2-O − 0.15 Cw5-OH2 − 0.15 HB7-O − 0.15 Cw6-OH2 − 0.15 Cw4-OH2 − 0.15W8-OH2 − 0.15W6-N1 − 0.15 Cw7-OH2 − 0.15W5-N1 − 0.15W3-N1 − 0.16W4-N1 − 0.16 HB3-O − 0.16W7-N1 − 0.16W2-OH2 − 0.17W7-OH2 − 0.17W3-OH2 − 0.17 HB4-O − 0.17W5-OH2 − 0.17 HB5-O − 0.17 HB6-O − 0.18W6-OH2 − 0.18W4-OH2 − 0.18 HB1-N1

Fig. 4

Two tripeptides, Ile-Leu-Met (good substrate, K _m ^app = 0.43 mM; left) and Glu-Glu-Ser (poor substrate, K _m ^app = 28 mM; right), shown with their hydrophilic MIFs at −4.0 kcal/mol. The volume of the hydrophilic region is inversely correlated with hPEPT1-mediated transport (cf. W6-OH2; Fig. 3)

Fig. 5

a Response permutation testing plot. R ² and Q ² were calculated for 50 randomised models, i.e. the descriptors were left intact whereas the −log K _m ^app values were randomised. R ² (circles) and Q ² (boxes) are displayed as a function of the correlation coefficient between original and permutated −log K _m ^app values. The intercepts of the regression lines of R ² and Q ² were −0.003 and −0.148, respectively. b External validation plot. Measured vs. predicted −log K _m ^app values of nine tripeptides not included in the training set. The straight line indicates Y = X, and the dotted lines denote deviations of 1 logarithmic unit. The r ² of data was 0.90