Nature | Vol 578 | 13 February 2020 | 323[^14 C]d-glucose uptake by PfHT1 proteoliposomes was performed in the
presence of unlabelled epimers of d-glucose and homologous sugars
(Fig. 2c, Extended Data Fig. 5a, b). Most revealingly, d-gulose—which
differs from d-galactose in only the C3-hydroxyl orientation—displayed
fivefold-poorer competition for [^14 C]d-glucose uptake than d-galactose.
Consistently, d-allose (the C3 epimer of d-glucose) was clearly weaker
at competing for [^14 C]d-glucose uptake than either the C4 epimer
d-galactose or the C2,C4 epimer d-talose. Moreover, d-mannose and
2-deoxyglucose, which differ from d-glucose in their C2 positions, were
just as competitive for [^14 C]d-glucose uptake as d-glucose. Overall,
the C3-hydroxyl group orientation was determined to be the most
critical for d-glucose recognition, followed by the C4-hydroxy group
orientation.
In the PfHT1 structure, the C3- and C4-hydroxyl groups hydrogen
bond to Asn311 and Gln306, whereas the C1- and C2-hydroxyl groups
hydrogen bond to Gln169, Gln305 and Trp412 (Fig. 2a). Consistently,
whereas alanine substitutions of Trp412 and Gln305 residues retained
100% and 22% of wild-type activity (respectively), Gln306 and Asn311
alanine mutations were less than 10% active (Fig. 2d, Extended Data
Fig. 6a). In the observed orientation, the C1- and C2-hydroxyl groups
face the interior and the C3- and C4-hydroxyl groups face the exte-
rior, which is consistent with the binding pose that is biochemically
predicted for GLUT1^29. Because E. coli XylE also binds d-glucose in a
manner similar to that of PfHT1 and GLUT3^19 , the d-glucose binding
mode appears to be evolutionarily conserved (Extended Data Fig. 6d).
Notably, the strictly conserved Gln169 (in TM5) is the only N-terminal-
bundle residue that coordinates d-glucose and is—as expected—also
critical for transport (Fig. 2d).
To investigate PfHT1 promiscuity, we focused on the ability of PfHT1
mutants to transport d-glucose versus d-fructose, which are the most
physiologically relevant sugars and have sevenfold differences in spec-
ificity constants (kcat/KM) (Extended Data Fig. 3f, g). We substituted
Ala404 with glutamate, as this residue was the only obvious difference
between the human GLUT3 and PfHT1 sugar-binding sites (Fig. 2b).
However, the Ala404Glu mutant retained d-glucose and d-fructose
transport (Fig. 2d, Extended Data Fig. 6a, e, f ). We next investigated
whether residues peripheral to the main sugar-binding site might
influence substrate selectivity (Fig. 2e, f). Consequently, we gener-
ated His168Asn, Val314Phe and Ala439Asn single mutants of PfHT1 to
mimic human GLUT3; in these mutants, transport of both d-glucose
and d-fructose were again similarly impaired (Fig. 2f, Extended Data
Fig. 6b). As a comparison, we also assayed alanine substitutions of the
conserved residues Ile310, Phe403, Asn435 and Trp436; these substitu-
tions also led to impaired transport of both sugars—with the exception
of Asn435Ala, which selectively abolished d-fructose transport (Fig. 2f,
Extended Data Fig. 6b).
As we were unable to rationalize the sugar preferences of PfHT1 on
the basis of human GLUT3, we extended our comparison to rat GLUT5,
which revealed that the Trp412 in PfHT1 is replaced by alanine in GLUT5
(Fig. 2b, Extended Data Fig. 6g). However, the Trp412Ala mutant
retained d-glucose transport but had severely reduced d-fructose trans-
port (Fig. 2d, Extended Data Fig. 6h). Taken together, our experiments
suggest that d-fructose transport in PfHT1 requires almost the same
set of sugar-binding residues as does d-glucose transport, but that the
former shows greater sensitivity to mutagenesis—probably because
it interacts with lower affinity^3 ,^17. Indeed, the Asn435 and Trp412 ala-
nine mutations that selectively affect d-fructose transport (Fig. 2d, f)
nevertheless have reduced d-glucose turnover and are not found in
transporters that are specific for d-fructose (Extended Data Table 1,
Extended Data Figs. 1c, 6a, b). It was thus unclear how PfHT1 robustly
transports different sugars.
Despite the high levels of structural similarity, two antimalarial com-
pounds C3361 and MMV009085 have previously been discovered to
have 19- to 250-fold higher selectivity for PfHT1 over human GLUT1,
GLUT5, and over GLUT1–4, respectively^8 ,^11 ,^30. The compound C3361 is a
d-glucose derivative with an undec-10-en- addition at the C3-hydroxyl
position; aliphatic chain additions to the C3 hydroxyl showed the
strongest inhibition (followed by additions at the C4 hydroxyl),
whereas additions to C1, C5 or C6 positions showed no inhibition^8. In
the occluded PfHT1 structure, there is a narrow hydrophobic vestibule
that would be accessible only from the C3- and C4-hydroxyl positions
(Extended Data Fig. 7a). In the glucose-bound human GLUT3 structure,
a crystallization lipid consistently occupies this site (Extended DataOut
Inw7bcdaebPfHT1 )7*,19/96161(/<.()/'
hGLUT1 /6*,1$9)<<676,)(.$* 9
hGLUT2 )6*,1*,)<<676,)47$*,
hGLUT3 /6*,1$9)<<67*,).'$* 9
hGLUT4 /6*,1$9)<<676,)(7$* 9
hGLUT5 /6*91$,<<<$'4,</6$* 9
rGLUT5 /6*91$,<<<$'4,</6$* 9
XylE307
284
316
282
300
290
289
286 )9*,199/<<$3(9).7/*$(^310320)
TM7b
TM1/ 4 TM7/10
Gate
distance
(Å)
18
16
14
12
10
Simulation time(ns)
0 200 400 600 800 1,000
PfHT 1
GLUT 3
TM1–TM7binteractions
TM7bgating dynamics
D-Glucose
D-Glucose
TM1–TM7binteractions
D-Glucose D-Fructose
K51Q N318
A
S317A
WT K51A
S315
Y
N316
A
N48A S315AN316Y E319A
120
100
80
60
40
20
0
1
7b
10
K5 1
N4 8
S317
S315
E319
N316
N311
N318
Relative
% of WT transport
D-Glu
Gln
Gln
Gln
Gln
Gln
Gln
Gln
Gln
Trp
Trp
Trp
AlaTrp Trp
Trp
Ala
Gln
Gln
Gln
Gln
Ala
Gln
Ala
Ala
Asn
Asn
Asn
Asn
Asn
Asn
Asn
Asn
Fig. 3 | Gating helix in Pf HT1 enables substrate promiscuity. a, Cartoon
representation of the sugar-binding site in occluded Pf HT1 (green sticks)
superimposed with outward-open GLUT5 (PDB 4YBQ, blue sticks), outward-
occluded GLUT3 (PDB 4ZW9, grey sticks) and inward-open GLUT5 (PDB 4YB9
orange sticks). Asn311 (dotted ellipsoid) is the only residue that clearly
repositions during the entire transport cycle. b, TM7b sequence alignment
between human (h) GLUT1, GLUT2, GLUT3, GLUT4 and GLUT5, rat (r) GLUT5,
E. coli XylE and Pf HT1. The red box highlights that the highly conserved
occlusion-forming tyrosine residues^23 are replaced by serine and asparagine in
PfHT1. The yellow shading highlights conserved residues. c, Cartoon
representation of Pf HT1 extracellular gating interactions between TM7b
(magenta) and TM1 (blue). Potential hydrogen-bond interactions are indicated
by dotted lines and prominent residue side chains are labelled. d, Transport
activity for TM1–TM7b interacting-residue mutants for [^14 C]d-glucose (black
bars) and [^14 C]d-fructose (white bars). Residues K51, N311, N316, S317 and N318
were determined to be essential for sugar transport; their relative positions are
shown in c. Data are mean ± s.e.m. of n = 3 biologically independent
experiments. e, Gating interactions for the d-glucose-bound outward-
occluded structure of human GLUT3 (grey) (top) and the d-glucose-bound
occluded structure of Pf HT1 (bottom). Blue, NTD; magenta, CTD; shaded lines
represent the distribution of gating distances from n = 3 independent 1-μs
molecular dynamics simulations (Methods) and non-shaded lines represent
their respective mean distance. Representative side views for human GLUT3
and Pf HT1 at the start and end of the simulation time are also shown.