28. M. B. Boxer et al., J. Med. Chem. 53, 1048 (2010).29. C. Le Goffe et al., Biochem. J. 364, 349 (2002).

Acknowledgments: We thank V. Toxavidis and J. Tigges (BethIsrael Deaconess Medical Center flow cytometry facility) forsupport with flow cytometry applications; X. Yang andS. Breitkopf for technical assistance with mass spectrometry;Ross Dickins and Scott Lowe (Cold Spring Harbor Laboratory)for the gift of tamoxifen-inducible Cre-expressing retroviralplasmid; and C. Benes, N. Wu, A. Shaywitz, B. Emerling,A. Saci, G. DeNicola, K. Courtney, A. Couvillon, S. Soltoff,A. Carracedo, A. Grassian, J. Brugge, and members of theCantley lab for helpful discussions. L.C.C. is a cofounder ofAgios Pharmaceuticals, a company that seeks to develop novel

therapeutics that target cancer metabolism. G.P. is a PfizerFellow of the Life Sciences Research Foundation. A.T.S. is aGenentech Fellow and supported by the Japanese Society forthe Promotion of Science and the Kanae Foundation forResearch Abroad. M.G.V.H. is supported by the NIH(R03MH085679 and 1P30CA147882), the BurroughsWellcome Fund, the Damon Runyon Cancer ResearchFoundation, and the Smith family. This work was supportedby grants from the NIH (R01-GM056203-13, P01-CA089021,and P01-CA117969-04 to L.C.C), the Starr CancerConsortium (L.C.C. and M.G.V.H.), the Molecular LibrariesInitiative of the National Institutes of Health Roadmap forMedical Research, and the Intramural Research Program ofthe National Human Genome Research Institute, NIH

(M.B.B., J. J., M.S., D.S.A., and C.J.T.). We apologize tocolleagues whose work we could not cite because of spacelimitations.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/science.1211485/DC1Materials and MethodsFigs. S1 to S10References

20 July 2011; accepted 24 October 2011Published online 3 November 2011;10.1126/science.1211485

Hemoglobins S and C Interfere withActin Remodeling in PlasmodiumfalciparumInfected ErythrocytesMarek Cyrklaff,1* Cecilia P. Sanchez,1 Nicole Kilian,1 Cyrille Bisseye,2 Jacques Simpore,2

Friedrich Frischknecht,1 Michael Lanzer1*

The hemoglobins S and C protect carriers from severe Plasmodium falciparum malaria. Here, we foundthat these hemoglobinopathies affected the trafficking system that directs parasite-encoded proteinsto the surface of infected erythrocytes. Cryoelectron tomography revealed that the parasite generated ahost-derived actin cytoskeleton within the cytoplasm of wild-type red blood cells that connected theMaurers clefts with the host cell membrane and to which transport vesicles were attached. The actincytoskeleton and the Maurers clefts were aberrant in erythrocytes containing hemoglobin S orC. Hemoglobin oxidation products, enriched in hemoglobin S and C erythrocytes, inhibited actinpolymerization in vitro and may account for the protective role in malaria.

Themalaria parasitePlasmodium falciparumhas exerted a selective pressure on the hu-man population that has led to the emer-gence of several polymorphismswithin the humangenome that protect carriers from severe malaria-related disease and death (1). The best-knownexamples are the structural hemoglobinopathiesS (sickle cell trait; HbS) and C (HbC), in whichglutamate at the sixth position within the b-globinchain is replaced by valine and lysine, respectively(2, 3). Protection against severe malaria correlateswith a distorted display of parasite-encoded ad-hesins on the surface of infected erythrocytes(4, 5). By reducing the cytoadhesive capacity ofparasitized erythrocytes, HbS and HbC seem tomitigate the life-threatening complications re-sulting from the sequestration of infected eryth-rocytes in postcapillary microvessels of the brainand other organs. HowHbS andHbC bring aboutthis effect is unclear. We tested the hypothesisthat HbS and HbC interfere with the machinerythat directs parasite-encoded proteins to the eryth-rocyte surface.

Human erythrocytes lack a secretory systemand are rapidly cleared from circulation by thespleen when damaged or infected. To developwithin human erythrocytes and to avoid passagethrough the spleen, P. falciparum extensivelymodifies its host cell (6), for example, by plac-ing the disease-mediating immunovariant adhesinPfEMP1 in knob-like protrusions in the erythro-cyte plasma membrane (7). To direct PfEMP1and other determinants of virulence and pathol-ogy to the erythrocytes plasma membrane, theparasite establishes a trafficking system within thecytoplasm of its host cell, of which a prominentfeature are Maurers clefts, unilamellar membraneprofiles that serve as intermediary compartmentsfor proteins en route to the erythrocyte surface(810). To better define the elements of this ma-chinery and how they are altered when P. falcipa-rum develops in HbS and HbC erythrocytes, weapplied electron tomography to parasitized eryth-rocytes preserved by rapid freezing (fig. S1).

We initially investigated P. falciparuminfected erythrocytes (at the trophozoite stage,20 to 26 hours after invasion) containing thewild-type hemoglobin HbA (homozygous). Thetomograms revealed the erythrocyte plasmamembrane, the knobs, and the Maurers clefts(Fig. 1A). In addition, we observed an extendednetwork of long, sometimes branched filamentsthat connected theMaurers clefts with the knobs.Of the 20 knobs identified in 12 tomograms, all

were connected to Maurers clefts by filaments.The filaments were between 40 and 950 nm long(Fig. 1B) and 6.8 T 0.5 nm in diameter (Fig. 1C).Some of the filaments branched at main angles of70 T 5 and 110 T 5 (Fig. 1D).

The tomograms further revealed vesicles ofvarious sizes, ranging from 20 nm to more than200 nm in diameter (fig. S2). About 70% of theobserved vesicles (47 out of 68) were attached tofilaments (Fig. 1E), independent of their size. Theremaining vesicles were associated with Maurersclefts or appeared free in the erythrocyte cytosol.Some vesicles carried PfEMP1 (Fig. 1F) (9, 11).

Tomograms taken in areas more than 5 mmdistant from Maurers clefts also revealed vesi-cles and a filamentous network (Fig. 2A). How-ever, these filaments were shorter than thoseobserved in the vicinity of Maurers clefts (Fig.1B). Moreover, the twomain branching angles ofthe filaments were equally distributed, whereasclose to Maurers clefts the filaments preferen-tially branched at an angle of 70 T 5 in thedirection of the Maurers clefts (Fig. 1D).

The filaments had features reminiscent ofactin filaments, including diameter and branchingpattern (Fig. 1, C and D) (12). Indeed, treatingP. falciparuminfected erythrocytes (tropho-zoites) for 10 min with the actin depolymerizingagent cytochalasin D (1 mM) destroyed the fil-aments and altered the morphology of theMaurers clefts (Fig. 2B). Vesicles were also notobserved under these conditions. Immunolabel-ing of high-pressure frozen electron microscopy(EM) sections, using a gold-coupled monoclonalantibody specific for b-actin (13), provided fur-ther evidence that the long filaments contained actin(Fig. 2C and fig. S3). We noted a higher densityof actin labeling in the area between theMaurersclefts and the erythrocyte plasma membrane com-pared with areas elsewhere in the erythrocyte cyto-plasm (fig. S3), supporting the conclusion that actinfilaments connect theMaurers clefts with the eryth-rocyte plasmamembrane (13). Ankyrinmay anchorthe actin filaments to the Maurers clefts (14).

The erythrocyte owes its shape and physicalproperties to a membrane skeleton that is pri-marily composed of spectrin tetramers joined bya junctional complex that is mainly composedof actin protofilaments (15). The length of the ac-tin protofilaments is tightly regulated and is re-stricted to 14 to 16 monomers (15). Tomograms

1Department of Infectious Diseases, Parasitology, HeidelbergUniversity, Im Neuenheimer Feld 324, 69120 Heidelberg,Germany. 2Biomolecular Research Center Pietro Annigoni, 01BP 364 Ouagadougou, Burkina Faso.

*To whom correspondence should be addressed. E-mail:michael_lanzer@med.uni-heidelberg.de (M.L.); marek.cyrklaff@med.uni-heidelberg.de (M.C.)

www.sciencemag.org SCIENCE VOL 334 2 DECEMBER 2011 1283

REPORTS

on M

ay 4

, 201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

on M

ay 4

, 201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

on M

ay 4

, 201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

on M

ay 4

, 201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

Fig. 1. Structural features in the cytoplasm of P. falciparuminfected eryth-rocytes. (A) Section through a cryoelectron tomogram (left) and correspond-ing surface-rendered view (right) of a trophozoite-infected HbAA erythrocyte.Labeling and color code in all figures: PM, erythrocyte plasma membrane(dark blue); K, knobs (red); V, vesicles (cyan); MC, Maurers clefts (cyan);filaments (yellow) indicated by arrowheads. (B) Frequency histogram offilament lengths in infected (filled circles) and uninfected (open circles)erythrocytes containing HbAA, HbSC, and HbCC. T, trophozoite; R, ring.MC, close to (+), or far from (), Maurers clefts. n > 150 filaments in eachcase. **P < 0.001 compared with uninfected HbAA erythrocytes (Studentst test). (C) Mean densitogram across 30 filaments. Dotted lines indicatewidth of filament. (D) Distribution of branching angles in filaments inP. falciparuminfected erythrocytes containing HbAA (blue, close to; brown,far from Maurers clefts), HbSC (gray), HbCC (green). (E) Filament-vesiclecontacts, indicated by arrowheads. (F) Different planes of a tomogram of aTokuyasu section and surface rendered view, showing immunolabeling ofvesicles with PfEMP1 antibodies (red, 10-nm gold). Scale bars: (A) and (F),100 nm; (E), 50 nm.

Fig. 2.Organization of host actin in infected and uninfected HbAA erythrocytes.Tomograms of trophozoite-infected erythrocytes: (A) distant from Maurers clefts;(B) after treatment with cytochalasin D (1 mM). (C) Immuno-EM of infected

erythrocytes, using a gold-coupled monoclonal antibody to b-actin (see also fig.S3A). Tomograms of uninfected (D) and ring-stageinfected (E) HbAA eryth-rocytes. Cyan, vesicles and early-stage Maurers clefts. Scale bars, 100 nm.

2 DECEMBER 2011 VOL 334 SCIENCE www.sciencemag.org1284

REPORTS

of plunge-frozen uninfected human erythrocytesrevealed short filaments of 30 to 40 nm under-neath the erythrocyte plasma membrane (Figs.2D and 1B)which in terms of length, thickness(6.8 T 0.5 nm), and location are consistent withactin protofilaments. The spectrin filaments aresignificantly thinner than actin filaments (16)and were not detected in our tomograms but werevisible by classical EM in negatively stainedpreparations of the membrane skeleton (fig. S4).The hexagonal arrangements characterizing the

membrane skeleton of uninfected red blood cells(17) were not observed in membrane skeletal prep-arations from parasitized erythrocytes (fig. S4).These data suggest that the parasite remodelsthe membrane skeleton of its host cell duringintra-erythrocytic development to build an actinnetwork of its own design. Consistent with thishypothesis, the parasite-generated actin cyto-skeleton in ring-stage-infected erythrocytes (10to 14 hours after invasion) was less pronouncedin areas close to Maurers clefts (Fig. 2E), with

short filaments of 20 to 60 nm dominating (Fig.1B) (P < 0.001), suggesting that actin remodelingoccurs as the parasite develops.

We next analyzed tomograms ofP. falciparuminfected erythrocytes (at the trophozoite stage)containing homozygous HbCC and heterozygousHbSC. The tomograms revealed the erythrocyteplasma membrane with knobs, vesicles of differ-ent sizes (ranging from 20 to 100 nm in diameter)(fig. S2), and large amorphous membrane con-glomerates, but no membrane profiles character-istic ofMaurers clefts (Fig. 3, A and B). However,an antiserum against the Maurers clefts markerPfSBP1 identified the amorphous membrane con-glomerates as aberrant Maurers clefts (Fig. 3C).Classical EMof chemically fixed or high-pressurefrozen cells confirmed that in HbSC- and HbCC-containing erythrocytes, Maurers clefts did notform the elaborated,multilayeredmembrane stacksseen in HbAA erythrocytes (fig. S5). Despite thealtered morphology of Maurers clefts, parasitesappeared to develop normally inHbSC andHbCCerythrocytes (figs. S6 and S7).

The tomograms further revealed actin fila-ments underneath the erythrocyte plasma mem-brane in both HbSC and HbCC erythrocytes.However, the filaments were significantly shorterthan those seen in infected HbAA erythrocytes(P < 0.001) (Fig. 1B), and the twomain branchingangles were equally distributed (Fig. 1D). Only afew actin filaments (8 of 150 in nine tomograms)

Fig. 3. Aberrant actin cytoskeleton and Maurers clefts in trophozoite-infectedHbSC and HbCC erythrocytes. Tomograms of infected HbSC (A) and HbCC (B)erythrocytes. (C) Immunolabeling of infected HbCC erythrocyte, using anantiserum against the Maurers clefts marker PfSBP1. (D) Tomogram of anuninfected HbCC erythrocyte. Scale bars, 100 nm.

Fig. 4. Inhibition of actin polymerization in vitro. (A) Actin filament length in the presence of total proteinextracts from erythrocytes containing HbAA, HbSC, and HbCC. (B) Effect of 5 mM of hemoglobin (Hb),methemoglobin (met-Hb), and ferryl hemoglobin (ferryl-Hb) on actin filament length. (C) Concentration-dependent decrease of actin filament length by ferryl hemoglobin. The percentage of ferryl hemoglobin oftotal hemoglobin (5 mM) is indicated. The means T SEM of at least four independent experiments areshown. **P < 0.01 compared with HbAA extracts and hemoglobin.

www.sciencemag.org SCIENCE VOL 334 2 DECEMBER 2011 1285

REPORTS

were attached to the degenerated Maurers clefts,and they did not interconnect them with theknobs. Only 15% (6 out of 40) of the vesiclesobserved were attached to filaments; 85% werefree in the cytoplasm.

Erythrocytes containing HbS and HbC have adysfunctional cytoskeleton among other abnor-malities (18), which we confirmed in uninfectedHbCC erythrocytes by cryoelectron tomography(compare Figs. 2D and 3D). The actin filamentsappeared less confined to the volume underneaththe erythrocyte plasma membrane (Fig. 3D) andwere significantly longer (P < 0.001) than thosein uninfected HbAA erythrocytes (Fig. 1B).

These changes in actin organization resultfrom endogenous factors and may involve oxi-dized forms of hemoglobin (19). Both HbS andHbC are unstable and readily oxidize to met-hemoglobin and finally to hemichromes, whichaccumulate more than 10-fold over normal levels(18, 20). Hemoglobin andmethemoglobin have astabilizing effect on the membrane skeleton bypromoting the self-association of spectrin dimersto tetramers (19, 21), whereas hemichromes candestabilize the erythrocyte membrane skeleton bydecreasing the spectrin-protein 4.1-actin interac-tion (19). Thus, oxidized hemoglobinsmight inter-fere with the actin reorganization brought aboutby the parasite in HbS and HbC erythrocytes.Consistent with this hypothesis, shorter actin fila-ments were formed in in vitro actin polymeriza-tion assays in the presence of total protein extracts(5 mM) from HbSC and HbCC erythrocytes com-paredwith extracts fromHbAA erythrocytes (P