324 | Nature | Vol 578 | 13 February 2020
Article
Fig. 7a). Although the binding mode of MMV009085 was previously
unknown, we conclude that it also binds in the sugar-binding pocket
because—similar to cytochalasin B—PfHT1 inhibition was found to
be dependent on Trp412 (Extended Data Fig. 7b). The malarial box
inhibitor MMV009085 is a symmetric tetracyclic compound with two
butanol moieties (Extended Data Fig. 7c). Consistent with inhibition
requiring interaction with the hydrophobic-side vestibule, a compound
synthesized without the butanol moieties was unable to inhibit PfHT1
(Extended Data Fig. 7c). Thus, the occluded structure highlights how
an off-site vestibule could be targeted to improve selective inhibition
of PfHT1.
By combining the occluded PfHT1 structure with previous sugar-
porter structures, we can reconstruct what is arguably the most com-
plete MFS transporter cycle known to date (Supplementary Video 1).
During the rocker-switch alternating-access mechanism^22 , the N- and
C-terminal bundles clearly rearrange around the centrally located
substrate-binding site. Global rearrangements are further coupled with
local rearrangements of the TM7b and TM10a half-helices that gate
access to the sugar-binding site from the outside and inside, respec-
tively^4 ,^5 ,^22. Side-chain positioning of almost all sugar-binding residues
are virtually unchanged during the entire transport cycle (Fig. 3a),
which implies sugar translocation must be primarily driven by confor-
mational selection. The residue Asn311 in the extracellular substrate-
gating helix TM7b is the only residue that moves substantially during
the transport cycle. Specifically, in the transition to a sugar-bound
occluded state, Asn311 moves inward to form hydrogen bonds with
the critically important C3- and C4-hydroxyl groups (Fig. 3a, Extended
Data Fig. 8a). In human GLUT3, two highly conserved TM7b tyrosine
residues move in concert with Asn311 to occlude sugar exit^4 (Extended
Data Fig. 8a). Thus, the strictly conserved Asn311 is probably a generic
interaction site that couples sugar binding to TM7b gating.
The occlusion-forming tyrosine residues that are strictly conserved
in the GLUT proteins are replaced in PfHT1 by the polar residues Ser315
and Asn316, and the substitution of either with tyrosine abolished trans-
port (Fig. 3b, d, Extended Data Fig. 6c). Asn316 extended towards TM1
and seemed to form polar interactions with Lys51 (Fig. 3c). The muta-
tion of Lys51, which is located about 15 Å from d-glucose, to alanine or
glutamine also rendered PfHT1 non-functional (Fig. 3d, Extended Data
Fig. 6c). TM7b and TM1 interactions were further observed between the
backbone oxygen of Asn48 in TM1 and Asn318 in TM7b, which formed
hydrogen bonds with Ser317 (Fig. 3c, Extended Data Fig. 6c). Ser317Ala
and Asn318Ala (which affect TM7b) mutants also rendered PfHT1 non-
functional, whereas the Asn48Ala mutant had severely reduced activity
(Fig. 3d, Extended Data Fig. 6c). By contrast, alanine mutations of Ser315
and Glu319 residues (which point away from the TM1–TM7b interface)
retained robust transport activity (Fig. 3c, d). The functional Ser315Ala
and Glu319Ala mutants nevertheless show a reduction in turnover,
mostly for d-glucose (Extended Data Table 1). To probe TM1 and TM7b
interactions further, we compared 1-μs molecular dynamics simulations
of human GLUT3 and PfHT1 structures. In the presence or absence of
d-glucose, TM7b in human GLUT3 was found to be very mobile and it
consistently moved far enough apart from TM1 to enable the release of
d-glucose (Fig. 3e, Extended Data Fig. 8b). By contrast, TM7b in PfHT1
mostly retained an occluded conformation (Fig. 3e, Extended Data
Fig. 8c) but was somewhat more mobile in the absence of d-glucose.
For the majority of the simulation time, TM1 and TM7b contacts were
maintained between the Lys51 and Asn316 residues (Extended Data
Fig. 8d). Taken together, our findings show that TM1 and TM7b gating
interactions and dynamics appear to be of equal importance to sugar
transport kinetics as the residues in the sugar-binding site.
The formation of the occluded state is an important intermediate
for understanding substrate coupling. A statistical comparison of the
PfHT1 structure supports its designation as an occluded conformation
that links previously determined sugar-porter states (Fig. 4 , Extended
Data Fig. 9a–c). For a sugar to be a substrate, it not only has to bind but
must also induce formation of the occluded state, which is a prerequi-
site for alternating access. In the transition from the outward-occluded
to the fully occluded conformation, TM7b breaks and extends closer to
TM1, adopting the position seen in inward-facing structures (Extended
Data Figs. 4d, 8a). The fact that the occluded state can be observed by
crystallography implies this state is likely to be more stable in PfHT1
than it is in GLUT proteins, consistent with the comparative molecular
dynamics simulations and the additional polar interactions observed
between the TM7b and TM1 helices. Rather than modifying the chemis-
try of the sugar-binding pocket, we conclude PfHT1 has evolved TM7b
substrate-gating dynamics so that it can transition into the occluded
state more easily. In this way, the PfHT1 is a more robust and promiscu-
ous sugar transporter than the GLUT transporters, as it is less sensitive
to a specific type of sugar being bound. Certainly, substrate promiscuity
would be an advantage to P. falciparum, which is able to use d-glucose
or d-fructose as a sole source of energy^2 ,^3 ,^8.
Although TM7b substrate-gating interactions and dynamics might be
exaggerated in PfHT1, we think they are also of functional importance
to GLUT proteins. Indeed, the QLS motif—which (prior to structural
information) was thought to confer d-fructose specificity by acting as
a selectivity filter—is not located in the main sugar-binding pocket, but
is instead juxtaposed to the TM7a and TM7b breakpoint^4 ,^31. Likewise,
an isoleucine-to-valine mutation in GLUT7 that abolishes d-fructose
transport (while leaving d-glucose transport unaffected) is not located
in the sugar-binding pocket, but is instead between TM7b and TM10
half-helices^32. The importance of fine-tuned sugar-binding and gating
would further explain why XylE binds d-glucose in a manner similar to
that of PfHT1 and human GLUT3 (Extended Data Fig. 6d) but is incapable
of transporting the sugar. To conclude, PfHT1 highlights that substrate-
gating dynamics is probably a greater determinant for evolving sugarOutward-open
(GLUT5)eennOutward-
occluded
(GLUT3)
Inward-open
(GLUT5)Inward-
occluded
(XylE)Occluded
(PfHT1)PC1 (global rocker switch)..... .....Occluded Inward-occludeXylE (4JA3) d
PfHT 1(^20) TR angle (°) 34
0 –2 –4 –6 –8 –50 –55 –60 –65 –70
Outward-open
XylE (6N3I)
GLUT5 (4YBQGLUT3 (5C65, 4Z) WC)
Outward-
occluded
XylE (4GBZ, 4GBYGLUT3 (4ZWB, 4ZW9)) Inward-open
GLUT5 (4YB9)
GLUT1 (4PYP)
GLUT1 (5EQI, 5EQH, 5EQG)
XylE (4JA4)
Fig. 4 | The conformational-selection-driven rocker-switch mechanism for
facilitative sugar transport. The fully occluded conformation of Pf HT1 is the
last remaining state to be observed within the rocker-switch alternating-access
mechanism of MFS transporters that belong to the sugar-porter subfamily. The
observed structural states are shown as surface transversal cross-sections and
clockwise from the top left: outward-open rat GLUT5 (PDB 4YBQ), outward-
occluded human GLUT3 (PDB 4ZW9), fully occluded Pf HT1, inward-occluded
XylE (PDB 4JA3) and inward-open bovine GLUT5 (PDB 4YB9). In the forward or
reverse direction, the attainment of the occluded intermediate represented by
Pf HT1 is required. The principal component (PC) analysis from the conserved
MFS ensemble core (n = 17 structures from 16 PDB codes; Methods) is shown
below the structures; this analysis yields a major PC1 component (65% of the
total structural variance) that tracks the rocker-switch global motion.
Projections are coloured according to the angle between tandem repeats (TR)
(Methods).