Reevaluating the Substrate Specificity of the L-Type Amino Acid Transporter (LAT1)

Article abstract of DOI:10.1021/acs.jmedchem.8b01007

The L-type amino acid transporter 1 (LAT1, SLC7A5) transports essential amino acids across the blood-brain barrier (BBB) and into cancer cells. To utilize LAT1 for drug delivery, potent amino acid promoieties are desired, as prodrugs must compete with millimolar concentrations of endogenous amino acids. To better understand ligand-transporter interactions that could improve potency, we developed structural LAT1 models to guide the design of substituted analogues of phenylalanine and histidine. Furthermore, we evaluated the structure-activity relationship (SAR) for both enantiomers of naturally occurring LAT1 substrates. Analogues were tested in cis-inhibition and trans-stimulation cell assays to determine potency and uptake rate. Surprisingly, LAT1 can transport amino acid-like substrates with wide-ranging polarities including those containing ionizable substituents. Additionally, the rate of LAT1 transport was generally nonstereoselective even though enantiomers likely exhibit different binding modes. Our findings have broad implications to the development of new treatments for brain disorders and cancer.

Full text of DOI:10.1021/acs.jmedchem.8b01007

Subscriber access provided by STEPHEN F AUSTIN STATE UNIV
Article
Reevaluating the Substrate Specificity of
the L-Type Amino Acid Transporter (LAT1)
Huan-Chieh Chien, Claire Colas, Karissa Finke, Seth Springer, Laura Stoner, Arik A.
Zur, Brooklynn Venteicher, Jerome Campbell, Colton Hall, Andrew Flint, Evan Augustyn,
Christopher Hernandez, Nathan Heeren, Logan Hansen, Abby Anthony, Justine Bauer,
Dimitrios Fotiadis, Avner Schlessinger, Kathleen M. Giacomini, and Allen A. Thomas
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b01007 • Publication Date (Web): 26 Jul 2018
Downloaded from http://pubs.acs.org on July 27, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Reevaluating the Substrate Specificity of the LꢀType Amino Acid
Transporter (LAT1)
a†
b†§
c†#
c†π
c†
HuanꢀChieh Chien , Claire Colas , Karissa Finke , Seth Springer , Laura Stoner , Arik A.
a�
c
c◊
c
c⊥
Zur , Brooklynn Venteicher , Jerome Campbell , Colton Hall , Andrew Flint , Evan
Augustyn , Christopher Hernandez , Nathan Heeren , Logan Hansen , Abby Anthony , Justine
Bauer , Dimitrios Fotiadis , Avner Schlessinger *, Kathleen M. Giacomini *, Allen A. Thomas *
cφ
c
cψ
c
c
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
c
d
b
a
c
a
Department of Bioengineering and Therapeutic Sciences, University of California, San
Francisco, San Francisco, CA 94158, USA
b
Department of Pharmacological Sciences, Icahn School of Medicine at Mount Sinai, New York,
NY 10029, USA
c
Department of Chemistry, University of Nebraska at Kearney, Kearney, NE 68849, USA
d
Institute of Biochemistry and Molecular Medicine, and Swiss National Centre of Competence in
Research (NCCR) TransCure, University of Bern, 3012 Bern, Switzerland
KEYWORDS
SLC7A5; Amino acid; Nutrient transport; Cancer metabolism
ABSTRACT
The LꢀType Amino Acid Transporter 1 (LAT1, SLC7A5) transports essential amino acids across
the bloodꢀbrain barrier (BBB) and into cancer cells. To utilize LAT1 for drug delivery, potent
amino acid promoieties are desired, as prodrugs must compete with millimolar concentrations of
endogenous amino acids. To better understand ligandꢀtransporter interactions that could improve
potency, we developed structural LAT1 models to guide the design of substituted analogs of
phenylalanine and histidine. Furthermore, we evaluated the structureꢀactivity relationship (SAR)
for both enantiomers of naturally occurring LAT1 substrates. Analogs were tested in cisꢀ
inhibition and transꢀstimulation cell assays to determine potency and uptake rate. Surprisingly,
LAT1 can transport amino acidꢀlike substrates with wideꢀranging polarities including those
containing ionizable substituents. Additionally, the rate of LAT1 transport was generally nonꢀ
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
stereoselective, even though enantiomers likely exhibit different binding modes. Our findings
have broad implications to the development of new treatments for brain disorders and cancer.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Introduction
The Lꢀtype amino acid transporter 1 (LAT1; SLC7A5), commonly referred to as the
“
Large Neutral Amino Acid Transporter 1”, is primarily expressed in brain, spleen, testis,
1ꢀ11
placenta, thymus, skeletal muscle and various cancer cells.
LAT1 exists as a heterodimeric
complex of 12 putative membraneꢀspanning helices (light chain) with a glycoprotein 4F2hc
(SLC3A2; heavy chain), and is responsible for importing lipophilic, neutral amino acids: leucine,
12
isoleucine, valine, phenylalanine, tyrosine, tryptophan, methionine, and histidine . Several
1
3, 14
15, 16
amino acidꢀlike drugs, including the Parkinson’s drug LꢀDOPA
and gabapentin
are
transported across the bloodꢀbrain barrier (BBB) by LAT1. The transporter is expressed at 100ꢀ
2
, 17
fold greater levels in the endothelial cells of the BBB than in other tissues.
For this reason, it
1
8ꢀ24
has been a target of interest for delivering drugs into the brain via active transport.
Lately,
in which LAT1 is upregulated
2
5ꢀ27
there has been increasing attention given to its role in cancer,
in various tumors, to meet the increased nutritional needs of growing and proliferating tumor
25ꢀ28
cells.
Additionally, LAT1 expression has been correlated with tumor aggressiveness and
2
9ꢀ34
poor prognosis.
Furthermore, LAT1 inhibitors are successful at reducing cancer cell growth
2
5, 35ꢀ41
both in vitro and in vivo.
LAT1 has been shown to transport amino acidꢀcontaining prodrugs, in which amino acids
18ꢀ24, 41, 42
were linked with drug molecules.
Surprisingly, larger entities, including nanoparticles
and liposomes may be targeted to cell membranes by attachment of amino acids, presumably
ACS Paragon Plus Environment

Page 3 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
43ꢀ46
through interaction with LAT1.
Though it is not the only transporter that has been
6
, 47ꢀ49
comprehensively investigated for drug delivery,
LAT1 is particularly suitable because it
possesses a large maximal transport capacity (Vmax), allowing many substrate molecules to pass
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
per unit time. It also has a relatively low K for many of its substrates (e.g. K = 11 ꢀM for
m
m
5
0
phenylalanine ), allowing such molecules to be transported at relatively low concentrations.
Moreover, it is known that LAT1 tolerates various chemical modifications to the amino acid
4
, 51
substrates that it recognizes.
Thus, amino acidꢀlike drugs or prodrugs have the potential for a
broad range of applications in treating CNS disorders and cancer.
One of the challenges for using LAT1 as a delivery tool has been the competition of the
transported drugs with endogenous amino acids substrates, which have millimolar concentrations
5
0, 52
in the plasma.
For example, it has been shown that brain uptake of LꢀDOPA decreases after a
53
meal rich in large amino acids, and LꢀDOPA decreases the concentration of other amino acids
5
4
(
i.e. tyrosine and tryptophan) in the rat brain. Potential antagonism by endogenous amino acids
5
5
has been cited as a limitation for using this transporter in drug delivery. One possible way to
overcome this problem is by increasing the affinity of drugs or prodrugs for the transporter
5
0
relative to natural amino acid substrates (K = 11ꢀ210 ꢀM). To date, there has been some
m
5
6ꢀ58
26, 51, 59
progress in developing LAT1 substrate SAR by us
and others
. For example, it is
known that substitution on the aromatic ring of phenylalanine derivatives, specifically at the
6
0
meta position, is tolerated and can allow for improved brain uptake of prodrugs. In addition, it
has been reported that LAT1 demonstrates modest stereoselectivity for some of its substrates (i.e.
Phe, Leu, and Met), while allegedly having greater stereoselectivity for others (i.e. Ile, Tyr, His,
6
1
Trp, Val). Furthermore, it has been shown that some D amino acids are taken into tumors via
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
62
LAT1, though it is presently unclear what physiological purpose there might be for LAT1 to
transport D amino acids.
Despite these observations on LAT1 substrate SAR, several relevant and potentially
misleading assumptions about the transporter have recurrently been stated in the literature:
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
. Both the amino acid and carboxylic acid functional groups are essential for
4
, 18, 47
transporter recognition.
. Large, neutral side groups are required for ligand binding.
25, 40, 63ꢀ65
2
3
1
, 66ꢀ68
. LAT1 is mostly stereoselective for L amino acids.
5
7
26
Recently, our team and Nagamori have shown that carboxylic ester and hydroxamic
acid derivatives of several amino acids are most likely LAT1 substrates based on their ability to
3
cause efflux of an intracellular radiolabeled substrate (e.g. [ H]ꢀgabapentin) via LAT1’s
5
7, 58
alternating access transport mechanism (transꢀstimulation cell assay).
Given this evidence
that assumption 1 is incorrect, we wanted to determine whether assumptions 2 and 3 about
substrate specificity might also be flawed.
Accordingly, the goal of our study was to increase our understanding of structureꢀactivity
relationships of LAT1 substrates, principally focused on the questions of whether neutral amino
acids are required for substrate specificity, and whether LAT1 translocates its substrates in a
stereoselective fashion. As part of our ongoing effort to identify potent LAT1 ligands that
expand LAT1 chemical space to nonꢀnatural amino acids that could potentially be used to design
5
6, 57
prodrugs for targeted drug delivery and/or antitumor agents,
we took an integrated approach
that includes structural modeling of LAT1 binding to putative substrates and inhibitors,
combined with testing of these compounds in cell assays. Based on the experimental results,
ACS Paragon Plus Environment

Page 5 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
models were refined. While the structure of the human LAT1 is not known, homology models
5
6ꢀ58
69
developed by us
and others have shown that LAT1 is a member of the large amino acids,
polyamines and organocations (APC) superfamily of transporters, which adopts a LeuTꢀlike fold,
7
0
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
and uses a gatedꢀpore alternating access transport mechanism. Previous models of an outwardꢀ
occluded conformation were used to guide SAR studies for this protein, but for the current work
we developed a new outwardꢀopen model based on a recently solved structure of AdiC bound to
7
1
agmatine.
Results and Discussion
Rational ligand design. We have previously built homology models of the LAT1 light
chain based on prokaryotic structures of the Lꢀarginine/agmatine transporter AdiC in the
5
6ꢀ58
outwardꢀoccluded facing conformation.
The LAT1 models based on AdiC were shown to be
41, 56, 58
useful for rational design of novel LAT1 ligands.
For example, computational modeling
and experimental testing of newly synthesized compounds guided the discovery of novel
5
7
inhibitors and substrates consisting of carboxylate bioisosteres. Both LAT1 and AdiC have a
conserved structural fold and substrate specificity (amino acids) in addition to a highly similar
58, 70
binding site.
In particular, the binding site residues interacting with the amine and carboxylic
acid groups of substrates are conserved between LAT1 and AdiC (supporting information,
Figure S1). Conversely, the residues interacting with substrate sidechains are more variable,
72
contributing to the distinct specificities between LAT1 and AdiC. Binding site differences such
as Gly 255 (Ile 205 in AdiC) and Ser 342 (Trp 293 in AdiC) result in a larger accessible surface
area in LAT1 (supporting information, Figure S1), and reveal unique subpockets termed Pockets
A, B and C (Figure 1). Furthermore, previous studies have highlighted the residues involved in
6
9
substrate binding using siteꢀdirected mutagenesis. For example, mutations of Ser 342 and Cys
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
3
35 (Trp 293 and Cys 286 in AdiC) to Gly and Ala, respectively, resulted in decreased substrate
activity. We hypothesized that targeting these regions of the binding site would allow us to
discover compounds specific to LAT1 and with a higher affinity.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Recently, the AdiC structure was solved at highꢀresolution in a unique outwardꢀopen
conformation with bound substrate, providing an excellent template for modeling LAT1 in this
71
conformation (PDB 5J4N ). Based on this unique structure of AdiC, we generated a LAT1
model in the outwardꢀopen conformation (Experimental section) to better inform ligand design.
In brief, LAT1 consists of 12 transmembrane helices with a LeuTꢀlike fold architecture,
comparable to other APC family members (Figure 1a). The outwardꢀoccluded and outwardꢀopen
models are generally similar (rmsd of 2.4 Å) exhibiting movements commonly observed in
transporters utilizing the gatedꢀpore mechanism. The most noticeable conformational change
occurs in transmembrane helix 6 (TM6), which tilts away from the binding site in the outwardꢀ
open conformation (Figure 1b). Interestingly, TM6 residue Phe252 faces alternatively inward or
outward, closing or opening the binding site, thereby acting as a gate (Figure 1bꢀc). This gate is
7
1
conserved in the AdiC structure where it is defined by Trp202, and has been shown to be
69
essential for binding. Consequently, the size and shape of the binding site is increased (Figure
1
d and supporting information, Figure S1 and Table S1). This new conformation of LAT1
highlights the flexibility of its binding site, and provides hypotheses regarding its transport
mechanism and explanation for the activity of LAT1 ligands in the current work.
ACS Paragon Plus Environment

Page 7 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 1. Homology models of LAT1 in outwardꢀoccluded (blue) and outwardꢀopen
conformations (beige). (a) The superposed models are shown in a plane perpendicular to the
membrane. (b) The models are shown through a “clipped plane” perpendicular to the membrane
to visualize the gate, defined by Phe252 on TM6 (pink sticks). Docking pose of phenylalanine
(cyan) in the (c) occluded model and (d) open conformation. The surface of the binding site is
shown in transparence, highlighting the subpockets PA and PB.
Guided by our new, outwardꢀopen model, we have investigated the SAR at various
positions on the endogenous substrates phenylalanine and histidine as well as examined ligand
stereoselectivity (Tables 1ꢀ3). We chose phenylalanine as a template for most of our SAR
exploration due mainly to the synthetic convenience of substitution on its aromatic ring, but also,
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
61
because of its relatively high affinity for the transporter. We pursued modification of the
phenylalanine scaffold chiefly at two positions: the aromatic meta position (Table 1) and the
alpha position (Table 2). It has previously been shown by us and others that substitution at the
5
6, 60
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
meta position can provide enhanced substrate activity relative to the ortho and para positions.
Consistent with these results, our previously published LAT1 homology model in the outwardꢀ
occluded conformation showed that the binding site could accommodate large substituents at the
5
6
meta position in the subpocket termed pocket A (PA). Furthermore, our model predicted the
possibility for ligands to hydrogen bond with residues in PA (Figure 1c). Thus, we sought to
determine whether we could improve potency and/or uptake rate by meta substitution with
various polar functional groups capable of hydrogen bonding (Table 1). It has also been shown
73
that methyl substitution at the alpha position of phenylalanine is tolerated. Thus, we were
curious as to whether larger groups at the alpha position (Table 2) might improve substrate
activity through lipophilic interactions with the hydrophobic pocket C (PC, Figure 1c), proposed
by our models. In addition to modification at these two positions, we prepared both enantiomers
of selected phenylalanine analogs (i.e. compounds 79, 83, and 110, Table 1), and tested both
enantiomers of various endogenous LAT1 substrates (Table 3). Though it had previously been
6
1
reported that LAT1 was mostly stereoselective, we were interested to determine how D isomers
would perform in our cell assays, particularly in our transꢀstimulation assay.
Cell-based assays. To evaluate compounds’ activities, both cisꢀinhibition and transꢀ
stimulation assays using HEK cells engineered to overexpress human LAT1 were employed
58, 74
(Experimental section).
The cisꢀinhibition assay allows for evaluation of LAT1 ligand
binding, with IC value determinations providing a measure of binding potency relative to a
5
0
3
radiolabeled substrate, [ H]ꢀgabapentin. In some cases, IC values which are determined by
5
0
ACS Paragon Plus Environment

Page 9 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
varying a test compound’s concentration did not track with what would have been expected from
%inhibition data (e.g. compound 71), as the latter is determined from only a single concentration
of test compound. Though this cisꢀinhibition assay is beneficial for rank ordering compounds
based on the strength of their interaction with the transporter, it cannot distinguish between
substrates and inhibitors. Thus, to more directly identify substrates, we performed a transꢀ
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
7
5
16, 76
stimulation experiment which exploits LAT1’s exchanger activity
by loading cells with a
radiolabeled substrate followed by incubation with extracellular test compound. The exchange
efflux rate of the radiolabel in the presence of a test compound is compared with the efflux rate
in the absence of the test compound as well as with nonꢀsubstrates (e.g. Lꢀarginine, 54).
Compounds that are LAT1 substrates should increase the efflux rate of the radiolabeled amino
acid compared with its efflux rate in the absence of test compound or compared with the efflux
rate observed in the presence of nonꢀLAT1 substrates (e.g. Lꢀarginine, 54, Table 1). We selected
3
[ H]ꢀgabapentin as a probe substrate due to its selectivity for LAT1 relative to other membrane
1
6
transporters. Two known LAT1 substrates (Lꢀleucine and Lꢀphenylalanine) and two nonꢀ
substrates (Lꢀglycine and Lꢀarginine) served as positive and negative controls, respectively. The
range of efflux rates we have observed in our assay is generally between 0.4ꢀ4.0 fmol/min, with
nonꢀsubstrates typically exhibiting an efflux rate of 0.4ꢀ0.8 due to background levels of exported
3
[
H]ꢀgabapentin (Tables 1ꢀ3). This background efflux may be due to the presence of low
concentrations of endogenous amino acids present in the cell assay media and/or resulting from
export via an alternate transporter. To enable more facile comparisons between compounds’
efflux rates, we have normalized efflux rates for all tested compounds (Tables 1ꢀ3) to the efflux
rate of Lꢀphenylalanine 50 (2.7 fmol/min, averaged from seven different experiments). Though
transꢀstimulation assays are widely used, the assay provides indirect evidence of whether a
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
58
compound is a potential substrate of LAT1. LCꢀMS or radiolabeled compound uptake
7
7
experiments are still required for transporter kinetic parameters (i.e. K and V ). Additional
m
max
7
5, 78, 79
discussion of the transꢀstimulation assay has previously been published.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
SAR discussion. Compounds in Tables 1ꢀ3 were tested as ligands of LAT1 in both cisꢀ
inhibition and transꢀstimulation assays as described above (more details provided in the
56ꢀ58
56
Experimental section).
From our previous work with phenylalanine derivatives and in
4
, 51
agreement with SAR published by others,
we found that aromatic substitution with
hydrophobic groups (e.g. tertꢀbutyl, benzyl and phenyl as present in 68ꢀ70, respectively) at the
meta position provided increased LAT1 affinity, but with concomitant loss in substrate activity,
evidenced by the fact that these amino acids showed diminished efflux rates in our transꢀ
stimulation cell assay. However, we cannot rule out the possibility that these compounds may
simply have a slower transport rate than can be detected using our assay. Based on these
observations, we hypothesized that substitution with polar functional groups at the meta position
might allow for interactions with the transporter to improve ligand binding, without losing
substrate activity. Moreover, from our models, it appeared that there were multiple residues (e.g.
Figure 1c: Ser143, Ser338 and Asn404) near the meta position that might be capable of dipoleꢀ
dipole or hydrogen bonding interactions with polar substituents that could result in greater
transporter affinity. To test these ideas, we substituted phenylalanine’s aromatic ring at both the
meta and para positions with various polar functional groups (Table 1), selected for diversity in
forming hydrogen bonds at different angles and having a range of cLogP and tPSA values
(Figure 2).
Remarkably, the meta position of phenylalanine is considerably more tolerant of
substituents with vastly different polarities than could have been predicted based on previously
ACS Paragon Plus Environment

Page 11 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
4, 59
published models of the binding site.
Though we observed little correlation between cLogP
and tPSA with a substrate’s affinity for the transporter (Figure 2), the more hydrophobic amino
acids possessing positive cLogP values (i.e. 69 and 70) were the most potent ligands, with
diminished efflux rates compared with Lꢀphenylalanine 50. Furthermore, it is unclear from the
current SAR whether there is an optimum distance away from the aromatic ring for a hydrogen
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
9
bond acceptor as a previous model had proposed. For example, amides 75 and 76 have
identical activity within the error of our assays. Upon observing that a meta methyl ester
substituent in 79 gave a modest improvement in IC relative to parent phenylalanine (IC s: 36
50
50
vs. 69 ꢀM, respectively), we were prompted to follow up with additional esters at this position
and at the para position. Unexpectedly, both a bulky, hydrophobic tertꢀbutyl group in 81 and a
polar, hydroxyethyl substituent in 82 provided the most potent substrates in the current work,
with comparable IC values of 17 and 29 ꢀM, respectively. Though these two compounds
50
demonstrated a reduction in efflux rate compared with parent phenylalanine (70% and 74%,
respectively), they were still significantly greater than negative control amino acids Gly and Lꢀ
Arg (29% and 28%, respectively), suggesting that they are substrates of LAT1. Given the
superior potency associated with hydrophobic groups at the meta position as in compounds 69
and 70, it might not be surprising that a tertꢀbutyl ester would present a higher affinity than a
methyl ester. However, this line of reasoning is contradicted by the observation that the
hydrophilic hydroxy group in 82 gave comparable activity. One possible explanation, as
suggested by our models, is that the hydroxy group in 82 may be forming a hydrogen bond with
residue Glu136 in the binding site to explain its unexpectedly good potency (Figure 3a).
Moreover, though a phenolic hydroxy group in 71 and a benzyl alcohol in 73 were tolerated
(
IC s: 130 and 83 ꢀM, respectively), they did not demonstrate a boost in activity relative to
5
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
phenylalanine as did 82, which offers further evidence of a potential specific interaction for the
latter.
Perhaps more surprisingly, compounds with aryl substituents which would be charged at
pH 7.4 exhibited substrate activity. Specifically, benzyl amine 74 and carboxylic acid 83 both
showed significant efflux rates (78% and 56%, respectively) indicating that they were LAT1
substrates. Though both benzyl amine 74 (IC : 470 ꢀM) and carboxylic acid 83 (IC : 130 ꢀΜ)
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
0
50
were less potent than parent phenylalanine (IC : 69 ꢀΜ), they do maintain the ability to bind to
5
0
1
the transporter. This is in sharp contrast to what we and others have observed for nonꢀsubstrate
amino acids containing aliphatic, basic and acidic, side chains (i.e. Arg, Lys, Asp and Glu), none
of which were LAT1 substrates in our assays (Table 1). Though the synthesis of compound 83
2
1
has been reported, apparently it was not tested for LAT1 activity. Thus, to the best of our
knowledge, this is the first report of LAT1 substrates containing sideꢀchain substituents expected
to be charged at pH 7.4. However, it should be noted that we cannot be certain whether they
would be ionized within the ligand binding site. Despite tolerating a carboxylic acid directly
attached to the metaꢀposition (i.e. 83), the SAR appears to be fairly narrow, in that analog 84
containing a oneꢀcarbon spacer separating the carboxylic acid and aromatic ring, along with
paraꢀsubstituted 87 were both nonꢀsubstrates according to our transꢀstimulation assay (24% and
2
0% relative to LꢀPhe efflux rate, respectively).
As previously noted, LAT1 has a preference for aromatic amino acids relative to aliphatic
2
3, 59
ones that is not directly correlated with lipophilicity,
in spite of the fact that it also readily
transports some bulkier, aliphatic amino acids such as leucine, valine, isoleucine and methionine
Table 3). Similarly, we found that replacing the aromatic ring of phenylalanine with an
(
unsaturated, cyclohexyl ring, as in 110 and 110-D (Table 1), led to unexpectedly good activity in
ACS Paragon Plus Environment

Page 13 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
our transꢀstimulation assay (120% and 130%, respectively), while losing potency in our cisꢀ
inhibition assay (200 ꢀM and 260 ꢀM, respectively). From our models, we cannot say whether a
pi stacking interaction is taking place to enhance ligand affinity and/or if the planarity of the
aromatic ring is optimal for binding to LAT1. However, aromaticity is clearly not required for
substrate activity, provided that a hydrophobic side chain is present (e.g. LꢀLeu 46, LꢀVal 49, and
cyclohexyl analogs 110 and 110-D).
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
As paraꢀsubstituted methyl ester 86 showed a dramatic loss in potency relative to the
meta isomer 79 (IC s: 260 vs. 36 ꢀM, respectively), we did not pursue further substitution at the
5
0
para position. This difference in SAR for para vs. meta substitution is consistent with our binding
5
9, 60
site models and those of others,
which predict less space is available near the para position.
Synthetic attempts at placing a methyl ester at the ortho position were plagued by intramolecular
cyclization via the αꢀamino group to give a lactam, thus ortho analogs were not pursued.
Additionally, we previously showed that orthoꢀmethyl substitution was inferior to metaꢀmethyl
56
substitution in terms of efflux rate (100% vs. 140%, respectively). Similarly, the D isomer of
methyl ester 79 demonstrated a ~5ꢀfold loss in activity, and was potentially not a substrate due to
its relatively low efflux rate (41%). Given the fact that both enantiomers of parent phenylalanine,
5
0 and 50-D, have comparable efflux rates and IC values, the results for 79-D suggest that its
50
methyl ester is being forced to occupy a less favorable space in the binding site compared with
7
9. However, our models do not provide an explanation for this difference in potency.
As indicated by modeling, there are two putative pockets near the aromatic ring of
phenylalanine (i.e. PA and PB, Figure 1c) that might be filled to increase binding affinity. Our
dockings suggested that one pocket was being filled by the various meta substituents of Table 1,
while the second pocket, close to the other meta position, was left empty. Thus, we hypothesized
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
that targeting both pockets simultaneously could increase substrate activity. To test this, we
synthesized methylꢀsubstituted analogs 63 and 64. Regrettably, both compounds lost activity
relative to their parent compounds 79 and 83. It is possible that our model does not account for
subtle shifts in the residues in this region of the binding site which may disfavor substitution at
both meta positions.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
As most LAT1 substrates contain hydrophobic sideꢀchains, it has been intriguing to us
that the relatively polar imidazoleꢀcontaining Lꢀhistidine 53 is also a substrate, and demonstrates
5
0, 61
high affinity.
Indeed, Lꢀhistidine appears to be an outlier in this regard, as it is considerably
more hydrophilic compared with typical substrates (e.g. cLogP/tPSA: histidine = ꢀ3.7/88;
phenylalanine = ꢀ1.6/63). We decided to explore the SAR around Lꢀhistidine 53 by preparing
benzyl histidine 59, to test whether adding a lipophilic substituent could improve ligand potency
(Table 1). Moreover, our models predicted that this Nimꢀbenzyl group should be able to
satisfactorily fill PA (Figure 3b). Though, 59 along with smaller methylꢀsubstituted 58 were both
substrates based on our transꢀstimulation assay (Table 1), they had diminished efflux rates and
IC values relative to parent 53. Thus, contrary to what we expected, Nimꢀsubstitution on Lꢀ
5
0
histidine 53 is detrimental for interaction with the binding site, though it does appear to be
tolerated.
Given the activity of Lꢀhistidine and these new histidine derivatives (i.e. 58 and 59), we
were curious as to whether pyridyl isomers 60ꢀ62 (Table 1) would also be substrates. To our
satisfaction, 60ꢀ62 all demonstrated significant efflux rates in our transꢀstimulation assay (89ꢀ
1
40% relative to the efflux rate of LꢀPhe), and comparable IC values to Lꢀphenylalanine (Table
50
1). Consistent with the SAR for substituted phenylalanine analogs, having the nitrogen in the
meta position gave the top potency (61: IC = 69 ꢀM) of the three isomers. To our knowledge,
5
0
ACS Paragon Plus Environment

Page 15 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
this is the first time these pyridyl analogs of Lꢀphenylalanine have been tested against LAT1. In
contrast to benzyl amine 74, none of these heterocyclic amino acids would be expected to carry a
full positive charge at pH 7.4. This SAR is consistent with what we observed for polarꢀ
substituted phenylalanine analogs of Table 1, in that it corroborates the argument that LAT1
readily accommodates both polar and nonꢀpolar aromatic rings.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
Table 1. Relative exchange efflux rate, uptake inhibition of [ H]ꢀgabapentin and IC values in
5
0
HEKꢀhLAT1 cells for various natural and nonꢀnatural, heterocyclic and aromaticꢀsubstituted
amino acids.
%L-Phe
Efflux
%
IC50
a
Compound
R
Phꢀ
b
c
d
Inhibition
(ꢀM)
50
100
96
85
74
68
69 ± 29
46 ± 23
68 ± 34
(
LꢀPhe)
0-D
5
Phꢀ
(DꢀPhe)
5
1
paraꢀHOPhꢀ
96
(
LꢀTyr)
5
2
59
79
160 ± 87
(
LꢀTrp)
5
3
1
40
92
49
20 ± 9
HN
N
(LꢀHis)
5
4
NH (NH)CNH(CH ) ꢀ
28
ꢀ
2
2 2
(
LꢀArg)
55
LꢀLys)
NH (CH ) ꢀ
33
27
27
85
ꢀ2.1
4.2
ꢀ2.4
70
ꢀ
2
2 3
(
5
6
HO Cꢀ
ꢀ
2
(
LꢀAsp)
5
7
HO CCH ꢀ
ꢀ
2
2
(
LꢀGlu)
5
8
190 ± 25
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
59
81
74
84
170 ± 53
160 ± 56
6
0
1
140
6
130
89
98
99
69 ± 17
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
62
100 ± 53
6
3
56
28
71
95 ± 18
64
30
680 ± 340
110
120
130
32
42
200 ± 73
110-D
260 ± 79
R = meta-X-Ph,
where X is
6
5
Meꢀ
140
59
70
44
29
29
96
73
87
ꢀ
66
CF
3
ꢀ
ꢀ
ꢀ
6
7
8
NCꢀ
67
6
tꢀBuꢀ
101
98
ꢀ
69
PhCH
Phꢀ
2
ꢀ
7.3 ± 3
6.6 ± 3
130 ± 65
70
71
72
73
74
98
HOꢀ
97
NH
HOCH
NH CH
ꢀ
100
120
78
100
99
101 ± 35
83 ± 37
2
ꢀ
2
ꢀ
27
470 ± 210
2
2
75
110
85
110 ± 42
ACS Paragon Plus Environment

Page 17 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
76
100
96
88
96 ± 32
92 ± 70
ꢀ
7
7
100
78
37
54
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
7
9
89
81
36 ± 23
79-D
41
52
40
72
99
81
69
150 ± 45
76 ± 20
17 ± 56
29 ± 10
130 ± 64
8
0
1
8
70
82
74
8
3
56
31
24
8
3-D
9.0
25
ꢀ
ꢀ
8
4
R = para-X-Ph,
where X is
85
86
87
Meꢀ
100
86
37
ꢀ
67
260 ± 90
ꢀ
20
8.9
a
Cell assay data was obtained at least in triplicate (wells) in each condition. Amino acids were
either purchased (50ꢀ62, 65, 71, 85), synthesized as previously published (68ꢀ70), or
synthesized according to Scheme 1. All compounds above are single enantiomers of L
configuration, unless indicated otherwise. Compounds were tested at 200 ꢀM for their ability to
cause efflux (fmol/min) of [ H]ꢀgabapentin from preꢀloaded HEKꢀhLAT1 cells. Efflux of [ H]ꢀ
gabapentin was calculated at 3 min after adding test compound. %Efflux was normalized relative
to LꢀPhe (50), which had an efflux rate of 2.7 ± 0.3 fmol/min, from an average of seven
experiments. Compounds were tested at 200 ꢀM for their ability to inhibit uptake of [ H]ꢀ
gabapentin into HEKꢀhLAT1 cells. Data are presented as % inhibition relative to background
56
b
3
3
c
3
d
signal in the absence of a test compound. For IC determinations, varying concentrations of
each compound were added, from 0.1 µM to 500 ꢀM. IC and standard deviation of each
compound were calculated by Graphpad Prism version 5.0. %[ H]ꢀgabapentin uptake at each
5
0
50
3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
80, 81
concentration was normalized relative to %inhibition by BCH
at 2 mM, which was set to
1
00% inhibition.
5
00
500
50
400
50
300
50
200
50
100
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a
b
4
400
3
3
00
00
R² = 0.2129
2
R² = 0.0448
2
1
100
0
50
0
-
3
-2
-1
cLogP
0
1
60
70
80
90
100
110
tPSA (Å2)
Figure 2. Scatter plots of IC values of metaꢀsubstituted Lꢀphenylalanine analogs (Table 1) vs.
cLogP (a) and tPSA (b), indicating poor correlation between ligand polarity and potency.
50
Calculated properties were obtained using ChemDraw v.16.
Amino acids D vs L. As indicated above, some D amino acids (i.e. Leu, Phe and Met)
have been shown to have substrate activity comparable to their L counterparts, while others do
61
not. We speculated that the side chains of the more potent D enantiomers were comfortable
binding in a narrow, hydrophobic pocket “C” (Figure 1c: PC), predicted by our models to be
close to the alpha carbon. In contrast, this putative PC might not be able to accommodate the side
chains of the less potent D amino acids as readily as for Leu, Phe, and Met. To summarize, we
imagined that upon binding L amino acids, the region near pockets “A + B” (PA+PB) would
typically be filled with an aromatic ring (Phe, Trp, His, Tyr) or bulky/hydrophobic alkyl group
(
Met, Leu, Ile, Val), while PC would be empty. On the other hand, we thought that D amino
acids might be filling PC with their side chains, leaving region PA+PB empty. Keeping this idea
in mind, we hypothesized that substitution at the alpha position of phenylalanine with the side
ACS Paragon Plus Environment

Page 19 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
chains of Phe, Leu and Met (compounds of Table 2) might allow both PA+PB and PC to be
filled simultaneously, leading to improved potency.
Disappointingly, only three of the αꢀquaternary phenylalanine derivatives tested (Table
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
) demonstrated substrate activity. Though both enantiomers of αꢀmethyl phenylalanine 92 and
2-D were substrates, they lost activity relative to parent phenylalanine. Furthermore,
9
methionine’s sulfide side chain (98) was tolerated at the alpha position; however, this compound
also showed inferior activity relative to parent phenylalanine. Its enantiomer, 98-D, lost all
activity. Based on this SAR, it is impossible to say whether putative PC is being accessed by
these αꢀquaternary analogs, or whether our hypothesis that the side chains of D amino acids are
filling PC upon binding to LAT1 is incorrect. But we can say that substitution at the alpha
position appears to be limited to relatively small groups, as both the benzyl and (1Hꢀimidazolꢀ4ꢀ
yl)methyl analogs 93 and 107, respectively, were inactive. Given the lack of activity for 107, we
decided not to prepare its D enantiomer, particularly since 107 involved a cumbersome, multiꢀ
step synthesis (see supporting information). Based on our refined models and the SAR of Table
2
, we now think that the geometry for αꢀsubstituted phenylalanine derivatives simply does not
allow for optimal interaction with LAT1.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
3
Table 2. Relative exchange efflux rate, uptake inhibition of [ H]ꢀgabapentin and IC values in
5
0
HEKꢀhLAT1 cells for αꢀsubstituted phenylalanine derivatives.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
%L-Phe
Efflux
%
IC50
a
Compound
R’
H
b
c
d
Inhibition
(ꢀM)
5
0
100
96
85
74
69 ± 29
(LꢀPhe)
0-D
DꢀPhe)
5
H
46 ± 23
(
9
2
Meꢀ
Meꢀ
100
100
23
ꢀ
ꢀ
130 ± 28
9
2-D
810 ± 350
9
3
PhCH
(CH
(CH
2
(achiral)
CHCH
CHCH
11
ꢀ
94
3
)
2
2
ꢀ
19
15
ꢀ
9
9
4-D
3
)
2
2
ꢀ
27
2.1
21
ꢀ
9
8
CH
3
S(CH
2
2
)
2
2
ꢀ
56
240 ± 63
8-D
07
CH
3
S(CH
)
ꢀ
27
ꢀ0.20
19
ꢀ
ꢀ
1
13
a
Cell assay data was obtained at least in triplicate (wells). Amino acids were either purchased
50, 93), synthesized according to Scheme 2 (92), or as previously published (98, 107
compounds above are either single enantiomers of L configuration, or achiral (93), with the
exception of the following compounds: 94-D (86% ee), 98 (78% ee), 98-D (80% ee), as
determined by chiral HPLC analysis (supporting information). Compounds were tested at 200
M for their ability to cause efflux (fmol/min) of [ H]ꢀgabapentin from preꢀloaded HEKꢀhLAT1
cells. Efflux of [ H]ꢀgabapentin was calculated at 3 min after adding test compound. %Efflux
was normalized relative to LꢀPhe (50), which had an efflux rate of 2.7 ± 0.3 fmol/min, from an
82
83, 84
(
). All
b
3
ꢀ
3
c
average of seven experiments. Compounds were tested at 200 ꢀM for their ability to inhibit
uptake of [ H]ꢀgabapentin into HEKꢀhLAT1 cells. Data are presented as % inhibition relative to
3
d
background signal in the absence of a test compound. For IC determinations, varying
50
concentrations of each compound were added, from 0.1 µM to 500 ꢀM. IC and standard
5
0
ACS Paragon Plus Environment

Page 21 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
3
deviation of each compound were calculated by Graphpad Prism version 5.0. %[ H]ꢀgabapentin
8
0, 81
uptake at each concentration was normalized relative to %inhibition by BCH
which was set to 100% inhibition.
at 2 mM,
Though it is known that LAT1 is nonꢀstereoselective for transporting Phe, Met, and
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
6
1
61
Leu, it was unclear from the cisꢀinhibition assay results described by Yanagida as to whether
the D isomers of the other endogenous LAT1 substrates were simply weak binding substrates or
not transported at all. As we are highly interested in better understanding LAT1 SAR, we
decided to test both enantiomers of several naturallyꢀoccurring amino acid substrates in our cell
assays (Table 3). To our surprise, all of the D amino acids tested, with the possible exception of
DꢀVal (49-D), appear to be LAT1 substrates based on our transꢀstimulation assay. As
anticipated, both enantiomers of Phe and Met were substrates, with the D isomers of Met and
Phe, 48-D and 50-D, respectively, having comparable IC values with their respective L
5
0
enantiomers. This result is quite different from the previously reported K values for LꢀPhe and
m
6
1
DꢀPhe, 14.2 and 121 ꢀM, respectively. Those experiments were performed using X. laevis
oocytes, which may explain this discrepancy. For amino acids Leu, Tyr and Trp (46, 51 and 52,
respectively) LAT1 showed a 2ꢀ3 fold greater selectivity for the L isomers in our cisꢀinhibition
assay. Much greater stereoselectivity (>20ꢀfold) was observed for Ile, Val and His (47, 49, and
53, respectively). Given the relatively low efflux rate for DꢀVal (41%), we are not confident that
it is a substrate. However, it clearly has poor affinity for the transporter relative to the other D
amino acids tested. From this SAR, one might deduce that the βꢀmethyl groups of DꢀVal and Dꢀ
Ile are poorly tolerated in the binding site. Likewise, the polar side chain of LꢀHis is readily
accommodated in LAT1, but with DꢀHis, our data suggest it may be forced into an unfavorable
position within the binding site.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Based on the data of Table 3, it is apparent that amino acid enantiomers (D vs. L) have a
different SAR regarding cisꢀinhibition, even though they are not that different in their ability to
cause a transꢀstimulation effect. In agreement with the comparable Vmax values previously
6
1
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
reported for each pair of enantiomers of Leu and Phe, we conclude from our assays that LAT1
is not stereoselective with regards to the rate of transporting amino acids, though the manner in
which the D and L isomers bind to LAT1 appears to be quite different, given their relative IC s.
50
Taken together, our SAR studies (Tables 1ꢀ3) indicate that the LAT1 binding site is
surprisingly promiscuous. For example, we have shown that there is little correlation between
metaꢀsubstituent polarity and ligand potency (Figure 2). However, we have also uncovered
surprising variations in SAR, such as loss in activity by adding a one carbon spacer going from
83 to 84, as well as dramatic differences in potency for some stereoisomer pairs but not others
(e.g. histidine 53 vs. phenylalanine 50). Though our models have helped us to generate
hypotheses for ligand design, and in a several cases offered a rationale for activity (e.g.
differences between meta and para substitution, and potential hydrogen bonding for polar
substituents such as in 82), they are currently unable to explain some of the divergent SAR that
we have observed, such as the differential activities of amino acid enantiomers. We are
continuing to refine and improve our models as new LAT1 SAR becomes available.
ACS Paragon Plus Environment

Page 23 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
3
Table 3. Relative exchange efflux rate, uptake inhibition of [ H]ꢀgabapentin and IC values in
5
0
HEKꢀhLAT1 cells for L and D isomers of endogenous LAT1 substrates and various nonꢀ
substrate amino acids.
O
O
R''
R''
OH
OH
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
NH2
NH2
L isomers
D isomers
a
%L-Phe
Efflux
%
d
Compound
R’’
b
c
IC (ꢀM)
5
0
Inhibition
4
Gly)
46
LꢀLeu)
6-D
DꢀLeu)
5
H (achiral)
29
33
73
56
ꢀ
ꢀ
(
(
CH ) CHCH ꢀ
120
85 ± 21
220 ± 47
140 ± 27
>3,000
3
2
2
(
4
(CH ) CHCH ꢀ
100
93
3
2
2
(
4
7
(
S)ꢀCH CH (CH )CHꢀ
3 2 3
(
LꢀIle)
47-D
(DꢀIle)
(
S)ꢀCH CH (CH )CHꢀ
78
18
ꢀ
3
2
3
4
8
CH S(CH ) ꢀ
89
170 ± 23
120 ± 33
68 ±21
3
2 2
(
LꢀMet)
8-D
DꢀMet)
4
CH S(CH ) ꢀ
110
56
43
11
85
74
3
2 2
(
4
9
(
CH ) CHꢀ
100
41
3
2
(LꢀVal)
9-D
DꢀVal)
4
(
CH ) CHꢀ
>50,000
69 ± 29
46 ± 23
3
2
(
(
5
0
Phꢀ
Phꢀ
100
96
LꢀPhe)
0-D
DꢀPhe)
5
(
5
1
paraꢀHOPhꢀ
paraꢀHOPhꢀ
96
85
68
35
68 ± 34
(
LꢀTyr)
5
1-D
(DꢀTyr)
380 ± 73
52
(LꢀTrp)
5
8
9
1
79
160 ± 87
5
2-D
DꢀTrp)
29
92
380 ± 130
20 ± 9
(
5
3
140
HN
N
(
LꢀHis)
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
5
3-D
78
33
49
460 ±170
HN
(
DꢀHis)
N
54
NH (NH)CNH(CH ) ꢀ
28
33
27
27
ꢀ
ꢀ
ꢀ
ꢀ
2
2 2
(
LꢀArg)
5
5
NH (CH ) ꢀ
ꢀ2.1
4.2
ꢀ2.4
2
2 3
(
LꢀLys)
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
6
HO Cꢀ
2
(
LꢀAsp)
57
HO CCH ꢀ
2
2
(LꢀGlu)
a
Cell assay data was obtained at least in triplicate (wells). Amino acids were purchased from
commercial vendors. All compounds above are single enantiomers of L configuration, unless
indicated otherwise. Most of these amino acids (except 54ꢀ57) were converted and tested as their
HCl salts to improve water solubility. Compounds were tested at 200 ꢀM for their ability to
cause efflux (fmol/min) of [ H]ꢀgabapentin from preꢀloaded HEKꢀhLAT1 cells. Efflux of [ H]ꢀ
gabapentin was calculated at 3 min after adding test compound. %Efflux was normalized relative
to LꢀPhe (50), which had an efflux rate of 2.7 ± 0.3 fmol/min, from an average of seven
experiments. Compounds were tested at 200 ꢀM for their ability to inhibit uptake of [ H]ꢀ
gabapentin into HEKꢀhLAT1 cells. Data is presented as % inhibition relative to background
signal in the absence of a test compound. For IC determinations, varying concentrations of
each compound were added, from 0.1 µM to 500 ꢀM. IC and standard deviation of each
compound were calculated by Graphpad Prism version 5.0. %[ H]ꢀgabapentin uptake at each
concentration was normalized relative to %inhibition by BCH
00% inhibition.
b
3
3
c
3
d
5
0
5
0
3
80, 81
at 2 mM, which was set to
1
ACS Paragon Plus Environment

Page 25 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3. Docking poses of two newly discovered substrates. (a) Compound 82, shown in green
sticks is docked in the outwardꢀopen conformation (beige). (b) Compound 59, shown in cyan
sticks, is docked in the outwardꢀoccluded conformation (light blue). While both compounds
occupy PA, as predicted, compound 82 establishes additional hydrogen bonds with Glu136 and
Ser342.
Chemistry
Arylꢀsubstituted phenylalanine analogs of Table 1 were prepared according to Scheme 1,
5
6
using similar methodology as we previously employed. Conversion of protected iodoalanine
8
5
5 to organozinc 6 using conditions which have been shown to maintain the chirality of the
8
6, 87
88
starting amino acid moiety as developed by Jackson
and later modified by Huo was
conducted in DMF, using catalytic I to activate zinc prior to sonication and/or vigorous stirring
2
with 5. We found that using THF as solvent gave considerable acrylate byꢀproduct (~50%) due
to elimination, presumably prior to forming 6. Nonetheless, acrylate remained a commonly
1
observed byꢀproduct (10ꢀ30% in H NMR of the crude reaction mixture) using DMF. We
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
89
confirmed the results of Jackson that there was no loss in stereochemistry during formation of
organozinc 6 and subsequent Negishi coupling in forming compounds 8, by preparing BOC/tertꢀ
butyl protected Lꢀphenylalanine (i.e. R = Ph) using the current methodology (see supporting
information). The resulting optical rotation of this compound matched what was previously
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
90
reported in the literature for enantiomerically pure material prepared by a different route.
Though aryl iodides have been more commonly reported as coupling partners than aryl
8
6, 87, 89, 91ꢀ94
bromides in Negishi reactions with analogs of 6,
we preferred to prepare the desired
phenylalanine analogs using the latter due to commercial availability. To determine the best
1
catalyst system for formation of 8, we compared H NMR spectra of crude Negishi product
mixtures after aqueous workup for various catalyst combinations using bromobenzene as a
8
5
model coupling partner. Based on results by Paju, we tested Pd (dba) /QPhos alongside other
2
3
commonly used catalyst combinations, including PdCl (dppf) and Pd(Ph P) . We found that the
2
3
4
latter two were vastly inferior, giving negligible amounts of desired 8. Combination of Pd (dba)3
2
9
5
96
with ligands previously used in Negishi couplings of other substrates, SPhos and CPhos , gave
1
97
significantly more desired 8 (20ꢀ30% of crude by H NMR), but were inferior to QPhos and
P(oꢀtolyl) , which both gave ~50% of 8 present in the crude reaction mixture. Byꢀproducts
3
included acrylate from elimination of 5, along with BOCꢀprotected alanine tertꢀbutyl ester, likely
from protonation of uncoupled 6, and 1ꢀ2 other unidentified compounds. As the latter two
catalysts gave comparable results and due to the significantly lower cost of P(oꢀtolyl) , we
3
proceeded to perform all Negishi reactions with this catalyst. Isolated yields (17ꢀ51%) for
formation of each coupling product 8a-8o are given in Table S3 (supporting information).
The primary reason for choosing the protecting groups depicted in Scheme 1 (i.e. tertꢀ
butyl ester and BOC), was that they could be conveniently and efficiently removed in a single
ACS Paragon Plus Environment

Page 27 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
98
step using TFA plus anisole as a carbocation scavenger. In some cases, poor yields of the
desired phenylalanine analogs were obtained (<20%) not necessarily because of the deprotection
step, but rather due to carry over of byꢀproducts from the Negishi coupling reaction. All final
products were obtained in high purity (>95% LCꢀMS) after preparative HPLC purification,
which easily separated amino acids from anisole and other byꢀproducts that were not removed
after the Negishi coupling step. Saponification of aryl methyl esters in 63, 79, 80 and 86 readily
provided the corresponding carboxylic acidꢀsubstituted analogs 64, 83, 84, and 87 in good yield
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(57ꢀ82%) after preparative HPLC purification.
Scheme 1. Synthesis of compounds 63ꢀ64, 66ꢀ67, 72ꢀ84, 86ꢀ87. Reagents and conditions: (a) Zn,
I , DMF, sonication or vigorously stirred at rt, 18h; (b) RBr: aryl bromides 7aꢀ7o, Pd (dba) , P(oꢀ
tolyl) , DMF, rt, 18h, 17ꢀ51%, flash chromatography; (c) TFA, DCM, anisole, rt, 18h, 15ꢀ72%,
2
2
3
3
preparative HPLC; (d) LiOH, water, rt, 18h, 57ꢀ82%, preparative HPLC.
Alphaꢀsubstituted amino acids of Table 2 were prepared using “selfꢀregeneration of
9
9, 100
chirality” methodology developed by Seebach,
which has been shown to give high
diastereoselectivity (>97% de) for alkylation of enolates derived from either imidazolidinones or
1
01
oxazolidinones (i.e. Step b, Scheme 2). At the outset, we selected the oxazolidinone system
102
because of the ease of deprotection under relatively mild conditions. Conversion of CbzꢀLꢀ
1
03
phenylalanine to oxazolidinone 90 was performed using a method developed by Karady and
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 58
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
104
later modified by Cheng . Crystallization of 90 gave material that was a single diastereomer by
1
104
H NMR, with cis configuration as previously assigned. Unfortunately, alkylation of 90 was
problematic. We obtained <5% of the desired 91 when using isobutyl iodide as an electrophile.
Presumably the enolate of 90 decomposed more quickly than the alkylation event, as there was
no evidence of 90 in the crude after allowing the reaction to slowly warm from ꢀ78 °C to rt over
several hours, followed by aqueous workup. As ester enolates are known to be unstable giving
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
05
rise to reactive ketene intermediates, this should not have been too surprising. Fortunately, we
found that by increasing the reactivity of the electrophile by using isobutyl triflate, we were able
1
to obtain a good yield (72%) of 91 that appeared to be a single diastereomer by H NMR.
1
02
Deprotection using KOSiMe as developed by Coe simultaneously removed the oxazolidinone
3
and Cbz protecting groups in reasonable yield (55% of 94 after preparative HPLC purification).
The synthesis of Scheme 2 was repeated starting with the D isomer of phenylalanine to give the
opposite enantiomer 94-D. To determine enantiopurity of the final products, their amino groups
were acylated with acetic anhydride and chiral HPLC analysis was performed on both
enantiomers (supporting information). Though compound 94 was obtained in 98% ee, its isomer
9
4-D was obtained in only 86% ee using the same synthesis. It is unclear to us how this
reduction in stereochemical integrity took place, given the fact that oxazolidinone precursor 90-
1
D appeared to be a single diastereomer by H NMR. Nonetheless, our conclusions concerning
SAR were not affected by this discrepancy. To our knowledge, this is the first synthesis, racemic
or enantioselective, of αꢀisobutyl phenylalanine, probably due to the challenges associated with
forming such a stericallyꢀcongested molecule. Also, we found that alkylation of 90 with
isopropyl triflate could be performed, albeit with significantly poorer yield (17%). However,
ACS Paragon Plus Environment

Page 29 of 58
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
purification issues at the final step and the lack of activity for 94 and 94-D (Table 2) deterred us
from completing the synthesis of the planned αꢀisopropyl phenylalanine analogs.
Though the chemistry of Scheme 2 provided the desired isobutylꢀsubstituted
phenylalanine analogs 94 and 94-D, we were unable to apply this methodology to more
functionalized derivatives (e.g. compounds 98 and 107). The analogous triflates, 2ꢀ
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(methylthio)ethyl triflate and protected versions of (1Hꢀimidazolꢀ4ꢀyl)methyl triflate, are not
known in the literature, likely due to the presence of internal nucleophiles which would not be
compatible with the highly reactive alkyl triflate group. In fact, when we attempted to prepare
Tritylꢀprotected (1Hꢀimidazolꢀ4ꢀyl)methyl triflate from the corresponding alcohol and triflic
anhydride using pyridine as base, we obtained a highly complex mixture. Thus, we explored
taking an alternate approach for introduction of the αꢀsubstituent by preparing the phenylꢀ
substituted oxazolidinones (as present in 90) of both methionine and histidine with the idea of
alkylation using either benzyl iodide or benzyl triflate to prepare the desired αꢀquaternary amino
acids 98 and 107, respectively. However, this approach was also problematic. We were unable
to separate diastereomers of the methionine version of 90, and we failed to obtain the histidineꢀ
derived analog of 90 using the given reaction conditions, probably due to incompatibility of Cbzꢀ
Tritylꢀprotected histidine with BF ꢀEt O. Rather than spend more time exploring various
3
2
conditions solely for the sake of using the same methodology as for 94, we opted to prepare
compounds 98 and 107 according to their previously described stereoselective syntheses from
the literature. These published routes involved diastereoselective alkylation with benzyl iodide of
82
a methionineꢀderived tertꢀbutylꢀsubstituted oxazolidinone and a histidineꢀderived tertꢀbutylꢀ
8
3, 84
substituted imidazolidinone
for 98 and 107, respectively (see supporting information for full
experimental details). The enantiomeric purities of compounds 98 and 98-D (78% ee and 80%
ACS Paragon Plus Environment