(GPCRs). This family of taste receptors is not only available throughout
the tongue but also available all over the digestive tract and upper
respiratory tract, indicating their physiological functions in taste and
taste perception. (Davaasuren et al., 2015). The GPCR family has an
extracellular Venus Flytrap Domain (VFTD) that is recognized as a taste
recognition region for the receptor. Functional expression and mouse
gene knockout studies reported that T1R1/T1R3 is considered to be the
main receptor for the umami taste (Chaudhari et al., 2009; Li, 2009;
Zhao et al., 2003). Although a complete X-ray structure of the T1R1/
T1R3 heterodimer receptor is currently unavailable, homology-modeled
structures of the heterodimer receptor have been used in many molec-
ular docking studies (Dang et al., 2014; Li, 2009; Li et al., 2002; X. Yu
et al., 2017; Zhu et al., 2021).
In addition to glutamate, certain ribonucleotides and L-aminoacids
as well as peptides and peptidomimetics were found to evoke umami
taste. Thus, there has been increasing interest in the development of
novel umami peptides (Winkel et al., 2008). It appears in literature that
an octapeptide KGDEESLA (Lys-Gly-Asp-Glu-Glu-Ser-Leu-Ala), isolated
in 1978 from papain-treated beef gravy, was reported for the rst time to
possess an umami taste, which was then named as Beef Umami Peptide
(BMP) (Yamasaki & Maekawa, 1978). Then, an ever-increasing number
of peptides have been discovered. For instance, an enzymatic hydroly-
sate of deamidated wheat-gluten was shown to give an MSG-like umami
taste. Long linear peptides are known to have more intense umami tastes
than short-chain di-/tri-peptides (Zhang, Zhao, et al., 2019) Thus far,
more than 100 synthesized umami peptides have been reported, which
were classied based on the number of amino acid residues (Li et al.,
2020; Zhang, Sun-Waterhouse, et al., 2019) Di-/tri-peptides with umami
taste contain glutamic acid (Glu) and/or aspartic acid (Asp) as well as
hydrophilic or hydrophobic amino acids (Iwaniak et al., 2019; Zhang,
Sun-Waterhouse, et al., 2019). A number of umami and/or umami taste-
enhancing peptides/peptidomimetics have been identied in many
foods with known umami taste, including cheese, onion, soy sauce, etc.
Long peptides do not seem to benet most from hydrophilic amino acid
abundance in terms of umami avor. In addition to hydrophilic and
hydrophobic amino acids, one of the key fundamental components of
umami taste is the spatial structure of peptides. However, amino acid
sequence and conformation continue to play essential roles in taste
perception (Zhang, Sun-Waterhouse, et al., 2019). As a result, amino
acids may contribute to the avor of peptides regardless of their unique
taste. Dang et al. (2019) carried out taste studies on peptides by elec-
tronic tongue and reported that the tri-peptide DED possesses a better
umami taste than the tri-peptide DEE, while the authors were unable to
make a clear correlation between peptide aa sequences and tastes (Dang,
Hao, Zhou, et al., 2019).
Docking computations allow to predict molecular interactions be-
tween a ligand and the binding site of a protein. Although the most
recent docking tools have advanced towards predicting bound con-
formers of ligands in relatively close similarity to their x-ray co-
ordinates, computationally determined bound conformers still need to
be validated by molecular dynamics (MD) computations using advanced
forceelds (Berendsen, 1999; Sousa et al., 2006; Ziada et al., 2018).
Previously, several molecular docking studies have been carried out to
assess molecular interactions between umami peptides and the umami
receptor. Dang, Hao, Cao, et al. (2019) determined by docking studies
that T1R3 may be involved in binding umami peptides. On the contrary,
Liu et al. (2019) (Liu et al., 2019) determined by MD studies that T1R1
possesses high-afnity binding sites for small molecules and the peptide
BMP, leading to controversial discussions on in silico results by Dang
et al. (Dang, Hao, Cao, et al., 2019) and Liu et al. (Liu et al., 2019).
More recently, Yu et al. (2021) (Yu et al., 2021) implemented MD
studies for a short time (10 nsec) on interactions between 36 umami
peptides (including BMP) and the umami receptor, T1R1/T1R3. The
authors reported that T1R1 plays a crucial role in the sensation of the
umami taste. As mentioned before, Liu et al. (2019) (Liu et al., 2019)
conducted 100 nsec of MD studies with two replicas of trajectories and
reported that T1R1 possesses two adjacent binding sites for ligand
binding, which can be concurrently occupied by two ligands or sepa-
rately by a single ligand depending on the size of ligands. However, a
100-nsec MD simulation is not sufcient for such a large protein system
to reach an equilibrium state. Furthermore, even if an equilibrium state
is reached, additional production run MD simulations with three replicas
of trajectories are required to meaningfully assess the trajectories
(Knapp et al., 2018). To shed light on more meaningful and detailed
dynamic structures of T1R1/T1R3 in complex with the umami peptides,
KADEDSLA and BMP (KGDEESLA), we implemented three replicas of
200 nsec (600 nsec in total) of production MD simulations using the last
snapshot coordinates of an equilibrium MD simulation for 1000 nsec (1
μ
sec).
Thermodynamic binding characteristics of a novel octameric peptide
with an aa sequence of K
1
ADEDSLA
8
(abbreviated KADEDSLA) were
investigated in this study. The peptide was optimally designed by
mutating the beef umami peptide (BMP, with an aa sequence of
K
1
GDEESLA
8
) at two distinct AA points where amino acids with com-
parable physical properties could be substituted. In this sense, G
2
(gly-
cin) of BMP was replaced with A (alanine), both being hydrophobic
amino acid residues, and E
5
(glutamate) of BMP was swapped out for D
(aspartate), both representing negatively charged amino acids, to
generate KA
2
DED
5
SLA.
Lastly, the p.A2G, p.D5E, and p.A2G +p.D5E (BMP) peptide afn-
ities towards possible binding sites of the umami receptor T1R1/T1R3
were studied in silico by MM-PBSA (Molecular Mechanics-Poisson
Boltzmann) (Andac et al., 2021) and MAP (Mutational Afnity Predic-
tion) methods using three independent MD production trajectories (200
nsec each) of KADEDSLA in complex with T1R1/T1R3 using the last
snapshot coordinates of an equilibrium MD simulation for 1
μ
sec.
2. Materials and methods
2.1. Methods
2.1.1. Homology modeling and molecular docking
Structure coordinates for the extracellular Venus ytrap domains
(VFTD) of the hT1R1/hT1R3 complex were homology modeled by
Modeller v9.1 (licensed Linux version) (Martí-Renom et al., 2000) using
the X-ray coordinates of the T1R3 domain of the medaka sh taste re-
ceptor (PDB ID: 5X2O, chain B) (Nuemket et al., 2017), whose sequence
(aa residues 62–490) exhibits sequence identities of 35.60 % and 36.82
% with the extracellular domains of hT1R1 (Uniprot ID: Q7RTX1) and
hT1R3 (Uniprot ID: Q7RTX0), respectively; see supplementary Fig. S1.
Indeed, the hT1R1/hT1R3 receptor exists as a heterogenic dimer, whose
dimeric coordinates were determined by superposing the modeled
structure of hT1R1 and hT1R3 with the X-ray coordinates of the T1R2a
domain (chain A) and T1R3 domain (chain B) of the medaka sh taste
receptor (PDB ID: 5X2O) (Nuemket et al., 2017).
Homology-modeled coordinates of T1R1/T1R3 were further rened
by implicit solvent MD simulation. The addition of the cytine disulde
bond and parameterization of the initial structures were implemented
by the LEaP module of Amber v18 (Case, Belfon, Ben-Shalom, Brozell,
Cerutti, Cheatham, et al., 2020) using the AMBER.ff19SB force eld
(Tian et al., 2020) in a Generalized Born (GB) implicit water environ-
ment (Nguyen et al., 2013). Cystine (CYX) disulfuric bonds were
generated for T1R1 (between CYX residues 47:87, 333:339, 345:348,
387:392) and T1R3 (between CYX residues 42:83, 350:353, 390:395).
The homology-modeled coordinates of hT1R1/hT1R3 were initially
minimized over 10,000 steps by the pmemd module of AMBER v18
(Case, Belfon, Ben-Shalom, Brozell, Cerutti, Cheatham, et al., 2020),
followed by a temperature equilibration phase at 300 K over 20 psec (at
2 fsec time steps over 10 K iterations) using an innite cutoff for elec-
trostatic interactions (cut =999) and a Langevin thermostat with a
collision frequency γ =1.0 psec
−1
. Temperature-equilibrated co-
ordinates were then used for 45 nsec (at 3 fsec time steps over 15 M
C.A. Andac et al.
Food Chemistry 473 (2025) 142966
2