Squalamine as a broad-spectrum systemic antiviralagent with therapeutic potentialMichael Zasloffa,1, A. Paige Adamsb, Bernard Beckermanc, Ann Campbelld, Ziying Hane, Erik Luijtenc,f, Isaura Mezag,Justin Julanderh, Abhijit Mishrai, Wei Quc, John M. Taylore, Scott C. Weaverb, and Gerard C. L. Wongi

aTransplant Institute, Departments of Surgery and Biochemistry and Molecular and Cell Biology, Georgetown University Medical Center, Washington,DC 20007; bInstitute for Human Infections and Immunity and Department of Pathology, University of Texas Medical Branch, Galveston, TX 77555-0609;cDepartment of Materials Science and Engineering, Northwestern University, Evanston, IL 60208-3108; dDepartment of Microbiology and Molecular CellBiology, Eastern Virginia Medical School, Norfolk, VA 23507; eFox Chase Cancer Center, Philadelphia, PA 19111; fDepartment of Engineering Sciences andApplied Mathematics, Northwestern University, Evanston, IL 60208-3125; gDepartamento de Biomedicina Molecular, Research and Advanced Studies Centerof the National Polytechnic Institute of Mexico, 07360 Mexico D.F., Mexico; hInstitute for Antiviral Research, Utah State University, Logan, UT 84322-5600; andiBioengineering Department, Chemistry and Biochemistry Department, California Nano Systems Institute, University of California, Los Angeles, CA 90095-1600

Edited* by Max D. Cooper, Emory University, Atlanta, GA, and approved August 18, 2011 (received for review May 27, 2011)

Antiviral compounds that increase the resistance of host tissuesrepresent an attractive class of therapeutic. Here, we show thatsqualamine, a compound previously isolated from the tissues ofthe dogfish shark (Squalus acanthias) and the sea lamprey (Petro-myzon marinus), exhibits broad-spectrum antiviral activity againsthuman pathogens, which were studied in vitro as well as in vivo.Both RNA- and DNA-enveloped viruses are shown to be suscepti-ble. The proposed mechanism involves the capacity of squalamine,a cationic amphipathic sterol, to neutralize the negative electro-static surface charge of intracellular membranes in a way thatrenders the cell less effective in supporting viral replication. Be-cause squalamine can be readily synthesized and has a knownsafety profile in man, we believe its potential as a broad-spectrumhuman antiviral agent should be explored.

innate immunity | hepatitis B virus | eastern equine encephalitis virus |dengue virus | yellow fever virus

Squalamine (Fig. 1A) was first discovered in the tissues of thedogfish shark (Squalus acanthias) as a broad-spectrum anti-

microbial agent in 1993 (1, 2) and later identified within thecirculating white blood cells of the sea lamprey (Petromyzonmarinus) (3). It was found to have pharmacological activity inendothelial cells, inhibiting several growth factor-dependentprocesses (such as angiogenesis, migration, and proliferation)both in vitro and in vivo (4–11). Recently, squalamine was dis-covered to enter cells and cause displacement of proteins thatare associated through electrostatic interactions with the innerface of the cytoplasmic membrane (12–14). Squalamine (3β-N-1-{N-[3-(4-aminobutyl)]-1,3-diaminopropane)-7α, 24R-dihydroxy-5α-cholestane 24 sulfate; molecular weight = 628) (2) carriesa net positive charge by virtue of its spermidine moiety andexhibits a high affinity for anionic phospholipids (15, 16); onentry into a eukaryotic cell, it neutralizes the negative charge ofthe surface to which it binds (12, 13). Surprisingly, this disruptionof electrostatic potential can occur without obvious structuraldamage to the cell membrane as measured by changes in per-meability (13). After it has gained entry into a cell, squalaminesubsequently exits over the course of hours, which was shown byits pharmacokinetic properties in numerous mammalian species,including humans (17–19).The ability of squalamine to alter the electrostatic charge in so

drastic of a fashion as to cause displacement of membrane-an-chored proteins, suggested to one of us (M.Z.) that squalaminemight have, as a consequence, antiviral properties. Many virusesenter cells through engagement of the ρ-GTPases, such as Rac1,to influence the actin cytoskeleton (20–23). Displacement of keyproteins anchored through electrostatic forces (of host or viralorigin) from the cytoplasmic face of the plasma membrane mightinterfere with entry (20, 21, 24), protein synthesis (25), virionassembly (26, 27), virion budding (28), or other steps in the viralreplication cycle (29). Certain viruses seem to require the pres-ence of the anionic phospholipid phosphatidylserine in the target

cell plasma membrane as part of the fusion process, and chargeneutralization of the anionic phospholipid could, in principle,interrupt these events (24, 25, 30–33). Because squalamine, afterparenteral administration, is cleared from the circulation overthe course of hours by the liver, from which it is excreted throughthe biliary system, we focused our initial investigations on virusesthat infect the liver.

ResultsThe charge density on squalamine enables it to have unusualinteractions with cell membranes. Because its surface chargedensity is nearly equal and opposite to the charge density of lipidcomponents of the anionic cytoplasmic surface of typicaleukaryotic membranes, squalamine should exhibit anomalouslystrong membrane binding because of the maximal recovery ofcounterion entropy (34). We show that, owing to this effect,squalamine can strongly displace membrane-bound cationicproteins such as Rac1, a ρ-GTPase recruited to the inner leafletof the eukaryotic cytoplasmic membrane for the actin remodel-ing necessary for endocytosis (35). By incubating liposomes [1,2-Dioleoyl-sn-Glycero-3-[Phospho-L-Serine] (sodium salt)/1,2-Dioleoyl-sn-Glycero-3-Phosphocholine/1,2-Dioleoyl-sn-Glycero-3-Phosphoethanolamine = 20/20/60; similar to the inner leafletlipid composition of the plasma membrane of mammalian cells](36) with Rac1 and squalamine, we obtain the diffraction sig-natures of Rac1–membrane complexes and squalamine–mem-brane complexes, respectively, using synchrotron small-angleX-ray scattering (Fig. 1B). Rac1–membrane complexes exhibita broad diffraction feature at scattering vector q ∼0.22 nm−1,which is consistent with weakly correlated bilayer membraneswith electrostatically associated Rac1 on both sides. In contrast,squalamine–membrane complexes exhibit a strong, sharp dif-fraction feature at q ∼1.19 nm−1, reflecting sheets of bilayers withnearly flat squalamine molecules sandwiched between the polarfaces of the packed bilayers at a spatial periodicity of 5.3 nm.Incubation of the membrane with both Rac1 and squalamine atthe isoelectric point shows a diffraction pattern nearly identicalto the pattern of squalamine–membrane complexes (spacing =5.44 nm), with no evidence of Rac1–membrane complexes, in-dicating that squalamine–membrane binding has occurred at the

Author contributions: M.Z., A.P.A., B.B., A.C., Z.H., E.L., I.M., J.J., A.M., W.Q., J.M.T., S.C.W.,and G.C.L.W. designed research; A.P.A., B.B., A.C., Z.H., E.L., I.M., J.J., A.M., W.Q., J.M.T.,S.C.W., and G.C.L.W. performed research; M.Z. contributed new reagents/analytic tools;M.Z., A.P.A., B.B., A.C., Z.H., E.L., I.M., J.J., A.M., W.Q., J.M.T., S.C.W., and G.C.L.W. analyzeddata; and M.Z. wrote the paper.

Conflict of interest statement: M.Z. has filed a patent application that involves the use ofsqualamine for the treatment and prevention of viral infections.

*This Direct Submission article had a prearranged editor.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: maz5@georgetown.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1108558108/-/DCSupplemental.

15978–15983 | PNAS | September 20, 2011 | vol. 108 | no. 38 www.pnas.org/cgi/doi/10.1073/pnas.1108558108

expense of such complexes (Fig. 1B). This striking displacementof Rac1 from anionic membranes by squalamine is confirmed inmolecular dynamics simulations based on a coarse-grained modelof this system (Fig. 1 C–E, SI Materials and Methods, and MoviesS1, S2, S3, and S4). At concentrations high enough for eitherRac1 or squalamine to saturate the membrane, squalamine dis-places 56% of all Rac1 from the membrane. Squalamine is foundto exhibit much stronger electrostatic binding than Rac1 to a flatcharged bilayer, thus nearly doubling the counterion release thatdominates the free-energy change on binding. Together, thesedata illustrate a general property of squalamine. Because of itscharge and shape, squalamine can electrostatically displacemembrane-bound proteins and potentially influence attendingintracellular processes, with the specific interactions perturbeddependent on the cellular context. We suggest that, by transientlyaltering the electrostatic interactions of the intracellular mem-branes of a host cell, we place that cell into a state that is in-compatible with efficient replication of certain viruses.

Activity in Vitro. To determine whether squalamine exhibits an-tiviral activity in vitro, we studied the infection of a line of humanmicrovascular endothelial cells (HMEC-1) with dengue virus.We chose this model for several reasons. Dengue virus, whichinfects the microvascular and hepatic endothelium in the human

disease (37), has been recently shown to engage a Rac1-de-pendent pathway in the endothelial cell during its entry phase(22, 38); the early fusion process between the virus and theendosomal membrane is believed to involve electrostatic inter-actions between the viral E protein and anionic phospholipids(39). Our hypothesis predicts that squalamine should interruptthis infection. At concentrations between 20 and 60 μg/mL,squalamine has been shown to inhibit a broad array of growthfactor-induced, actin-dependent responses in endothelial cells,including cell migration, cell division, and vascular tube forma-tion in a 3D matrix (4–11).The HMEC-1 was inoculated with dengue virus Den V2 using

a published procedure (22). At a squalamine concentration of40 μg/mL, dengue infection was inhibited by about 60%, withcomplete inhibition observed at 100 μg/mL (Table 1). At theseconcentrations, squalamine was not cytotoxic, which was shownby phase contrast refractility, actin organization, and subsequentgrowth of uninfected cells after removal of squalamine. Atconcentrations of squalamine above 60 μg/mL, the cells detachedmore easily during the staining procedures, a consequence of thecompound’s known effect on cell adhesion (5).Next, we examined the effect of squalamine on the infectivity

of two viruses that cause human hepatitis, human hepatitis Bvirus (HBV) and human hepatitis δ-virus (HDV), on primaryhuman hepatocytes. Monolayer cultures of primary humanhepatocytes were infected with HBV using a modification ofprocedures described in the work by Taylor and Han (40).Squalamine effectively inhibited HBV replication in humanprimary hepatocytes when added either during the initial expo-sure of virus to the cells or at 24 h after infection (Table 2).There was no evidence of cytotoxicity associated with squalaminetreatment based on the morphology of individual cells and theabsence of visible damage to the physical integrity of the cellularmonolayer. A similar study was performed to evaluate the effectof squalamine on the replication of HDV. Squalamine was in-troduced at 20 μg/mL during HDV exposure, and the effectswere measured at day 7 when total RNA was extracted andassayed for HDV RNA sequences. Inhibition to 89 ± 4% wasobserved. With a higher concentration of 60 μg/mL squalamine,significant cell toxicity was observed. HDV was not inactivated byexposure to squalamine at the concentrations that exhibitedantiviral effects (Materials and Methods).

Activity in Vivo. We selected three well-characterized animalmodels of viral infection: yellow fever (YF) in the Golden Syrianhamster (41), eastern equine encephalitis virus (EEEV) in theGolden Syrian hamster (42), and murine cytomegalovirus(MCMV) in the BALB/c mouse (43, 44). In each animal, viralinfection of the liver and endothelium occurs. Both the YF (45)and MCMV (46) infection models have been used to evaluatehuman therapeutic candidates. No drug has been reported toimpact survival or viremia during EEEV infection in the hamster.Squalamine dosing regimens in mice (4–6, 9, 10), monkeys (8),

and humans (17–19, 47) have been established for the control ofpathological angiogenesis, and the dose-dependent toxicologyfor the compound is known in each of these species. The com-pound distributes widely throughout the body (excluding thebrain) and exits the bloodstream with a half-life between 1 and5 h. It is cleared by the liver and excreted through the biliary tractinto the feces; in the mouse, squalamine’s half-life in the liver isabout 12–24 h. For the antiviral studies reported here, we chosea single daily dose of 10–15 mg/kg administered over 6–8 con-secutive d as our starting regimen, a regimen that has beenproven to be well-tolerated in previous rodent studies (e.g.,cancer and ophthalmic angiopathy). In most of the studies de-scribed in this report, the compound was administered s.c.

YF. The YF virus (YFV) is an ssRNA virus of the Flaviviridaefamily responsible for over 30,000 deaths worldwide (45). Al-though an effective vaccine exists, periodic outbreaks continue tooccur, and no antiviral agent has as yet been developed to treat

Fig. 1. Squalamine strongly displaces Rac1 from model membranes becauseof charge density matching. (A) Squalamine chemical structure. (B) Diffrac-tion patterns from lipid membrane vesicles, Rac1–membrane complexes,squalamine–membrane complexes, and membranes incubated with bothRac1 and squalamine (bottom to top). The diffraction pattern of the finalcase is nearly identical to the diffraction signature of squalamine–membranecomplexes, indicating strong suppression of Rac1–membrane binding [lipidcompositions 1,2-Dioleoyl-sn-Glycero-3-[Phospho-L-Serine] (sodium salt)/1,2-Dioleoyl-sn-Glycero-3-Phosphocholine (DOPC)/1,2-Dioleoyl-sn-Glycero-3-Phos-phoethanolamine = 20/20/60, Rac1 to lipid ratio = 1:50, squalamine to lipidratio = 1:15; all are molar ratios]. (C–E) Representative configurations frommolecular dynamics simulations of solutions of Rac1 (C), squalamine (D), andboth Rac1 and squalamine (E), respectively, in the presence of a coarse-grained membrane in which 20% of the lipids are charged. Rac1 andsqualamine are present at concentrations that yield the same net chargeon all molecules, each higher than needed to neutralize the membrane.Squalamine was found to exhibit nearly two times stronger electrostaticmembrane binding than Rac1, and it displaced 56% of Rac1 from themembrane.

Zasloff et al. PNAS | September 20, 2011 | vol. 108 | no. 38 | 15979

IMMUNOLO

GY

human YF. We conducted experiments to evaluate the efficacyof squalamine for both the prevention and treatment of YF. Inthe hamster, YFV directly infects the hepatocyte, causing a se-vere necrotizing hepatitis that reaches maximum severity arounddays 4–5 postinoculation (47). Serum concentrations of the he-patic enzyme alanine amino transferase (ALT) are used tomonitor hepatic disease and reflect in magnitude both the he-patic pathology and hepatic viral titers (48). Viral titers in vari-ous organs peak between days 4 and 6 and progressively decreaseas the adaptive immune response accelerates (48).In the prevention study, squalamine administration began

1 d before virus inoculation, and daily dosing (15 mg/kg s.c.)continued through day 6 postinoculation. By day 9, 85% of theuntreated cohort had died. Serum ALT measured on the un-treated group was elevated on day 6, indicating hepatitis (Fig.2B). Of the group administered squalamine, 100% survivedthrough day 8, 2 d after the last day of drug administration; 70%were protected from mortality and survived through the re-mainder of the experiment (day 21) (Fig. 2). Serum ALT in thiscohort was within the normal range, suggesting minimal virusinfection of the liver. We compared the effect of squalamine withribavirin, an antiviral agent that has been well-studied in thismodel (45, 48, 49). Ribavirin was administered in a two timesdaily dosing regimen at 3.2, 10, and 32 mg/kg through the normali.p. route. The highest ribavirin dose protected about 40% of thecohort, whereas the lower doses were ineffective (Fig. 2). By

raising the total daily dose of ribavirin (50–75 mg/kg), this com-pound can control infection in the hamster quite effectively (45).To determine whether squalamine can treat an existing YFV

infection in the hamster model, animals were infected with a le-thal inoculum of virus. Then, one time daily treatment withsqualamine (15 or 30 mg/kg per d s.c.) was started beginning onday 1 or 2 after viral administration and continuing until day 8or 9, respectively. Survival was monitored, and animals thatremained alive by day 21 were considered cured (Fig. 2C). Byday 11 after infection, 100% of untreated animals had died. Incontrast, 60% of the animals that had received 15 mg/kg per d(from day 1 to 8) or 30 mg/kg per d (from day 1 to 8) were cured.Serum ALT activity, measured on day 6, was lower in the treatedanimals compared with the untreated cohort (Fig. 2D). Delay oftreatment (30 mg/kg per d) until day 2 still resulted in a cure rateof 40% (Fig. 2C). Treatment with ribavirin at its optimal dosingregimen (beginning on the day of viral infection) achieved a curerate of 100% (Fig. 2C).

EEEV. EEEV is an enveloped RNA virus of the Alphavirus familygenus in the family Togaviridae for which neither an effectiveantiviral drug nor a licensed human vaccine is available (50). Thecase fatality rate is between 30% and 80% for humans and up to95% for horses, and EEEV is regarded as a potential biodefensethreat (51).EEEV causes widespread vascular disease in the hamster,

involving all of the major organs (42). Infection of the vascula-ture of the brain followed by direct infection of neuronal cellsresults in fatal encephalitis. From day 3 after peripheral in-oculation, viral titers within the brain continue to increase, de-spite rising levels of neutralizing antibodies. Because squalaminecannot enter the brain after systemic administration, we wouldnot expect it to directly influence EEEV infection of the hamsterbrain or circumvent the onset of encephalitis. However, in thehamster (but not the mouse) (52), EEEV infects the liver earlyafter viral inoculation (42). We hypothesized that, althoughsqualamine would not be expected to directly influence the viralinfection within the brain, the substance might influence theclinical course of the peripheral disease by increasing the viralresistance of organs such as the liver and perhaps, the systemicvascular endothelium.Golden Syrian hamsters were infected with a lethal s.c. in-

oculum of EEEV; 1 d before infection, animals received 10 mg/kg squalamine s.c., which continued daily for 6 d postinfection.Blood was drawn from the retroorbitus during the first 4 d tomonitor viral concentration within the vascular space. Treatmentwith squalamine extended the survival of the infected animalscompared with those animals receiving vehicle alone (Fig. 3A).Treated animals maintained body weight compared with theuntreated animals (Fig. 3B). Initial viral titers in the blood streamof squalamine-treated animals were about 100-fold lower thanthe titers in animals receiving vehicle, showing antiviral activity ofthe compound when administered systemically (Fig. 3C).

MCMV. MCMV is an enveloped DNA virus of the Herpeseviridaefamily. The dynamics of infection of MCMV have been exten-sively studied in the mouse (44). When introduced into the mousei.v., MCMV initially propagates in two tissue compartments, theliver parenchyma and the vascular endothelium (44). Surprisingly,the virus that replicates in the liver does not spread throughoutthe body but likely is excreted into the feces. The virus thatpropagates within the vascular endothelium seems responsible forspread through the body. Based on the known tissue distributionof squalamine in the mouse, we hypothesized that squalamineshould influence the natural history of an MCMV infection.We inoculated BALB/c mice through the i.p. route with a

sublethal inoculum of MCMV. Squalamine was administered ata dose of 10 mg/kg daily, beginning 1 d before infection andcontinuing daily through day 6, by either the i.p. or s.c. routes.An infected cohort received only vehicle. We chose to comparei.p. and s.c. routes to determine how the pharmacokinetic profile

Table 1. Effect of squalamine on dengue virus infection ofhuman endothelial cells

Squalamine concentration(μg/mL)

Percent cells infected(percent inhibition)

0 38 ± 1 (0)10 36 ± 4 (5)20 26 ± 1 (32)40 15 ± 1 (61)60 9 ± 1 (76)100 0 (100)

Monolayer cultures of HMEC-1 cells grown on glass coverslips were pre-treated with squalamine for 2 h at the indicated concentrations. Fresh me-dium was reintroduced, and virus was added to the culture. After 48 h, cellswere fixed and immunostained for viral protein E as described in Materialsand Methods. For each concentration, the percentage of cells expressingviral E protein was scored; 10 randomly chosen fields were counted perconcentration. SD of the mean is noted.

Table 2. Effect of squalamine on HBV infection of primaryhuman hepatocytes

Squalamineconcentration(μg/mL)

Exposure tosqualamine (h)

Percent HBV replication(percent inhibition)

Recovered RNAconcentration

(μg/mL)

0 100 (0) 1872 −1 to +16 86 ± 41 (14) 1876 −1 to +16 46 ± 10 (54) 1896 +24 to +40 71 ± 18 (29) 18820 −1 to +16 16 ± 6 (84) 17720 +24 to +40 36 ± 8 (64) 171

Monolayer cultures of primary human hepatocytes were infected withHBV using a modification of procedures as described in the work by Taylorand Han (40). Zero hour is the time at which virus was first added. HBVinfection is based on the relative abundance of viral-specific sequences withrespect to the total cellular RNA recovered as measured by real-time PCR(Materials and Methods). The data are expressed relative to the controlcultures not exposed to squalamine. The errors represent the SD of themean. The experimental points were performed in at least triplicate.

15980 | www.pnas.org/cgi/doi/10.1073/pnas.1108558108 Zasloff et al.

of the dosed squalamine might influence outcome. s.c. dosing inthe mouse results in a slow release of compound from the in-jection site, reaching a peak blood and tissue concentrationwithin 5–8 h; i.p. dosing results in a rapid rise in blood and tissuelevels, peaking within 1 h of administration and generallyachieving levels 10-fold higher than the s.c. route (53). For this

study, on days 3, 7, and 14 postinfection, animals were eutha-nized, and the concentration of infectious virus present in varioustissues was measured. Because dosing ended by day 6, reductionin tissue viral titers at later times was likely to be a result of theaction of host defenses. Administration of squalamine throughthe i.p. route was themost effective, resulting in undetectable viral

Squalamine dosingbegan 1 day prior toadministra�on ofvirus

BA

DCSqualamine dosingbegan 1-2 days a�eradministra�on ofvirus

Fig. 2. Treatment of YF in Syrian hamsters. Survival of Syrian hamsters treated before (A) or after (C) viral inoculation. Dosing regimens are indicated anddescribed inMaterials and Methods. Serum ALT was sampled at day 6 postinfection for the group receiving squalamine before (B) or after (D) viral inoculation[treatment groups (n = 10) and D5W (5% dextrose) placebo (n = 20)]. All animals that had survived 14 d remained alive until day 21 and were designated ascured. [We should note that, in the treatment experiment (C), a cohort that received 15 mg/kg 1 d before virus inoculation (a control prevention study)achieved a 40% survival rate and has been omitted from the graph for the sake of clarity.]

A B

Squalamine dosing began 1

Cd i t d i iday prior rto administ a onof virus

Fig. 3. Treatment of EEEV in Syrian hamsters.Survival (A) and average body weight (B) of Syrianhamsters infected with EEEV and treated s.c. witheither squalamine (10 mg/kg) or placebo (D5W) ondays −1 to 6 after infection. (C) Viremia of Syrianhamsters on days 1–4 after infection with EEEVthat were treated with either squalamine or pla-cebo. Error bars indicate the SDs of the means. Thelimit of detection for the assay was 100 pfu/mL[treatment group (n = 10) and D5W placebo (n =6)]. P = 0.003 by one-way ANOVA.

Zasloff et al. PNAS | September 20, 2011 | vol. 108 | no. 38 | 15981

IMMUNOLO

GY

titers in both the liver and spleen at day 14 (Fig. 4). Dosingthrough the s.c. route was also effective in reducing viral titers inthe liver and spleen but less effective than the i.p. route. Viraltiters continued to increase in the lung between days 3 and 7 afterinfection and plateaued between days 7 and 14, which is likelya consequence of the animals’ immune response. Viral infectionwithin the submaxillary salivary gland continued to increasewithout abatement.

DiscussionThe results presented here unambiguously show that squalamineis active in vitro and in vivo against a broad spectrum of humanviral pathogens, including both RNA- and DNA-envelopedviruses. The shark seems to be surprisingly immune to viral in-fection (54), despite lacking a rapidly responsive adaptive im-mune system (55). Neither an IFN gene nor its receptor has beenidentified within the recently sequenced elephant shark genome(56), leaving open the possibility that the shark lacks an IFN-based antiviral defense. Curiously, the elephant shark genomealso lacks identifiable retroviral sequences, which are abundantlyrepresented in boney fish species (57), providing additional evi-dence that the antiviral defenses of the shark are likely to beremarkably effective. We propose that squalamine might servean antiviral function in the shark.We believe that squalamine exerts its antiviral effects by al-

tering the infectivity of the tissues into which it is transportedthrough its capacity to perturb the electrostatic potential of thecellular membranes onto which it binds. On entry into a cell,squalamine would be expected to disturb the intracellular ar-rangement of membrane-associated proteins positioned by elec-trostatic forces, possibly including some of host or viral origin thatare required for the viral replication cycle. Our recently publishedstudies that relate to the mechanism proposed in this manuscriptwere conducted in various cells transiently expressing specificallydesigned probes and proteins, and they illustrate the diversepleiotropic effects that can be caused by the entry of this moleculeinto a cell; we have shown that squalamine causes displacement ofa fluorescent probe bearing the cationic tail of a Ras protein fromthe plasma membrane (12), squalamine activates an ion channel

by displacing an inhibitory cationic segment from the anionicinner face of the plasma membrane (13), and squalamine inhibitsa sodium proton exchanger by displacing a cationic control se-quence from the cytoplasmic face of the plasma membrane (14).Hence, by modifying the electrostatic interactions of a membrane,squalamine could influence many diverse intracellular processes,making that cell less capable of supporting the replication ofcertain viruses. Squalamine does not seem to be related inchemical structure or mechanism of action to any chemothera-peutic substance currently in use. By targeting host membranes,squalamine differs in mechanism from the recently describedbroad-spectrum antiviral compound LJ100, which seems to act byirreversibly damaging the membranes of enveloped viruses andinterrupting effective entry (58). If the proposed mechanism ofaction of squalamine is correct, development of viral resistancewould be expected to be unlikely. We have not yet optimizedsqualamine dosing in any of the animal models presented in thisreport, and we do not know the maximum prophylactic or ther-apeutic benefit that can be achieved in these systems. Becausesqualamine has been studied in humans in several phase II clinicaltrials for cancer and retinal vasculopathies (59, 60), can be readilysynthesized (61), and has a known safety profile in man (17–19,60), its potential as a broad-spectrum human antiviral agentshould be explored.

Materials and MethodsSqualamine (as the dilactate salt) was synthesized by the route shown in thework by Zhang et al. (61) and was greater than 97% pure. A 1-mg/mL solu-tion of squalamine in 5% dextrose and 40 mM sodium phosphate (pH 7.4) wasused for dosing. For in vitro experiments, squalamine was dissolved in 99%ethanol and diluted in water as required. Details of the scattering and simu-lation studies are presented in SI Materials and Methods. Each of the viralexperiments reported was conducted within the specific laboratory of thedesignated investigator using published protocols, with modifications as noted.

Full methods and associated references are available in SI Materialsand Methods.

ACKNOWLEDGMENTS. M.Z. wishes to acknowledge the support andencouragement of Drs. Christopher Tseng and Heather Greenstone of theNational Institute of Allergy and Infectious Extramural Antiviral Screening

Squalamine dosing began 1 dayprior to administra�on of virus

Fig. 4. Treatment of MCMV in BALB/c mice. MCMVtiters from organs harvested at 3, 7, or 14 d afterviral inoculation (n = 6 animals for each data point).The limit of detection was 10 pfu/mL. Error bars in-dicate SDs of the means. *P < 0.05, **P < 0.01, and***P < 0.001 by one-way ANOVA compared withD5W control.

15982 | www.pnas.org/cgi/doi/10.1073/pnas.1108558108 Zasloff et al.

Program. M.Z. thanks Aaron Nelson for his careful reading of the manu-script. We thank the expert assistance of M. C. Dominguez with HMEC-1/DenV2 assays. E.L. acknowledges support through National Science Founda-tion Grant DMR-1006430 and allocation of computing time on NorthwesternUniversity’s Quest cluster. The dengue studies were partially supported bya Conacyt (Mexico) grant (to I.M.). The studies on EEEV were supported by

a grant from the National Institute of Allergy and Infectious Disease throughWestern Regional Center of Excellence for Biodefense and Emerging Infec-tious Disease Research National Institutes of Health Grant U54 AIO57156 (toA.M. and S.C.W). This work was supported by National Science FoundationGrant DMR-0409769 (to G.C.L.W.) and National Institutes of Health Grant1UO1-AI082192-01 (to G.C.L.W.).

1. Moore KS, et al. (1993) Squalamine: An aminosterol antibiotic from the shark. ProcNatl Acad Sci USA 90:1354–1358.

2. Wehrli SL, Moore KS, Roder H, Durell S, Zasloff M (1993) Structure of the novel ste-roidal antibiotic squalamine determined by two-dimensional NMR spectroscopy.Steroids 58:370–378.

3. Yun SS, Li W (2007) Identification of squalamine in the plasma membrane of whiteblood cells in the sea lamprey, Petromyzon marinus. J Lipid Res 48:2579–2586.

4. Li D, Williams JI, Pietras RJ (2002) Squalamine and cisplatin block angiogenesis andgrowth of human ovarian cancer cells with or without HER-2 gene overexpression.Oncogene 21:2805–2814.

5. Williams JI, et al. (2001) Squalamine treatment of human tumors in nu/nu mice en-hances platinum-based chemotherapies. Clin Cancer Res 7:724–733.

6. Schiller JH, Bittner G (1999) Potentiation of platinum antitumor effects in human lungtumor xenografts by the angiogenesis inhibitor squalamine: Effects on tumor neo-vascularization. Clin Cancer Res 5:4287–4294.

7. Yin M, Gentili C, Koyama E, Zasloff M, Pacifici M (2002) Antiangiogenic treatmentdelays chondrocyte maturation and bone formation during limb skeletogenesis.J Bone Miner Res 17:56–65.

8. Genaidy M, et al. (2002) Effect of squalamine on iris neovascularization in monkeys.Retina 22:772–778.

9. Higgins RD, Sanders RJ, Yan Y, Zasloff M, Williams JI (2000) Squalamine improvesretinal neovascularization. Invest Ophthalmol Vis Sci 41:1507–1512.

10. Higgins RD, Yan Y, Geng Y, Zasloff M, Williams JI (2004) Regression of retinopathy bysqualamine in a mouse model. Pediatr Res 56:144–149.

11. Sills AK, Jr., et al. (1998) Squalamine inhibits angiogenesis and solid tumor growthin vivo and perturbs embryonic vasculature. Cancer Res 58:2784–2792.

12. Yeung T, et al. (2008) Membrane phosphatidylserine regulates surface charge andprotein localization. Science 319:210–213.

13. Sumioka A, Yan D, Tomita S (2010) TARP phosphorylation regulates synaptic AMPAreceptors through lipid bilayers. Neuron 66:755–767.

14. Alexander RT, et al. (2011) Membrane surface charge dictates the structure andfunction of the epithelial Na+/H+ exchanger. EMBO J 30:679–691.

15. Selinsky BS, Smith R, Frangiosi A, Vonbaur B, Pedersen L (2000) Squalamine is nota proton ionophore. Biochim Biophys Acta 1464:135–141.

16. Selinsky BS, et al. (1998) The aminosterol antibiotic squalamine permeabilizes largeunilamellar phospholipid vesicles. Biochim Biophys Acta 1370:218–234.

17. Hao D, et al. (2003) A Phase I and pharmacokinetic study of squalamine, an amino-sterol angiogenesis inhibitor. Clin Cancer Res 9:2465–2471.

18. Bhargava P, et al. (2001) A phase I and pharmacokinetic study of squalamine, a novelantiangiogenic agent, in patients with advanced cancers. Clin Cancer Res 7:3912–3919.

19. Herbst RS, et al. (2003) A phase I/IIA trial of continuous five-day infusion of squal-amine lactate (MSI-1256F) plus carboplatin and paclitaxel in patients with advancednon-small cell lung cancer. Clin Cancer Res 9:4108–4115.

20. Pelkmans L, et al. (2005) Genome-wide analysis of human kinases in clathrin- andcaveolae/raft-mediated endocytosis. Nature 436:78–86.

21. Pelkmans L, Helenius A (2003) Insider information: What viruses tell us about endo-cytosis. Curr Opin Cell Biol 15:414–422.

22. Zamudio-Meza H, Castillo-Alvarez A, González-Bonilla C, Meza I (2009) Cross-talkbetween Rac1 and Cdc42 GTPases regulates formation of filopodia required fordengue virus type-2 entry into HMEC-1 cells. J Gen Virol 90:2902–2911.

23. Harmon B, Campbell N, Ratner L (2010) Role of Abl kinase and the Wave2 signalingcomplex in HIV-1 entry at a post-hemifusion step. PLoS Pathog 6:e1000956.

24. Mercer J, Helenius A (2008) Vaccinia virus uses macropinocytosis and apoptoticmimicry to enter host cells. Science 320:531–535.

25. Ahola T, Lampio A, Auvinen P, Kääriäinen L (1999) Semliki Forest virus mRNA cappingenzyme requires association with anionic membrane phospholipids for activity. EMBOJ 18:3164–3172.

26. Chukkapalli V, Hogue IB, Boyko V, Hu WS, Ono A (2008) Interaction between thehuman immunodeficiency virus type 1 Gag matrix domain and phosphatidylinositol-(4,5)-bisphosphate is essential for efficient gag membrane binding. J Virol 82:2405–2417.

27. Chukkapalli V, Oh SJ, Ono A (2010) Opposing mechanisms involving RNA and lipidsregulate HIV-1 Gag membrane binding through the highly basic region of the matrixdomain. Proc Natl Acad Sci USA 107:1600–1605.

28. Stansell E, et al. (2007) Basic residues in the Mason-Pfizer monkey virus gag matrixdomain regulate intracellular trafficking and capsid-membrane interactions. J Virol81:8977–8988.

29. Agirre A, Barco A, Carrasco L, Nieva JL (2002) Viroporin-mediated membrane per-meabilization. Pore formation by nonstructural poliovirus 2B protein. J Biol Chem277:40434–40441.

30. Coil DA, Miller AD (2004) Phosphatidylserine is not the cell surface receptor for ve-sicular stomatitis virus. J Virol 78:10920–10926.

31. Coil DA, Miller AD (2005) Enhancement of enveloped virus entry by phosphati-dylserine. J Virol 79:11496–11500.

32. Coil DA, Miller AD (2005) Phosphatidylserine treatment relieves the block to retro-virus infection of cells expressing glycosylated virus receptors. Retrovirology 2:49.

33. Soares MM, King SW, Thorpe PE (2008) Targeting inside-out phosphatidylserine asa therapeutic strategy for viral diseases. Nat Med 14:1357–1362.

34. Liang H, Whited G, Nguyen C, Stucky GD (2007) The directed cooperative assembly ofproteorhodopsin into 2D and 3D polarized arrays. Proc Natl Acad Sci USA 104:8212–8217.

35. Ridley AJ, Paterson HF, Johnston CL, Diekmann D, Hall A (1992) The small GTP-bindingprotein rac regulates growth factor-induced membrane ruffling. Cell 70:401–410.

36. Vance JE, Steenbergen R (2005) Metabolism and functions of phosphatidylserine.Prog Lipid Res 44:207–234.

37. Jessie K, Fong MY, Devi S, Lam SK, Wong KT (2004) Localization of dengue virus innaturally infected human tissues, by immunohistochemistry and in situ hybridization.J Infect Dis 189:1411–1418.

38. Wang JL, et al. (2010) Roles of small GTPase Rac1 in the regulation of actin cyto-skeleton during dengue virus infection. PLoS Negl Trop Dis 4:e809.

39. Stauffer F, et al. (2008) Interaction between dengue virus fusion peptide and lipidbilayers depends on peptide clustering. Mol Membr Biol 25:128–138.

40. Taylor JM, Han Z (2010) Purinergic receptor functionality is necessary for infection ofhuman hepatocytes by hepatitis delta virus and hepatitis B virus. PLoS One 5:e15784.

41. Tesh RB, et al. (2001) Experimental yellow fever virus infection in the Golden Hamster(Mesocricetus auratus). I. Virologic, biochemical, and immunologic studies. J Infect Dis183:1431–1436.

42. Paessler S, et al. (2004) The hamster as an animal model for eastern equine enceph-alitis—and its use in studies of virus entrance into the brain. J Infect Dis 189:2072–2076.

43. Cavanaugh VJ, Deng Y, Birkenbach MP, Slater JS, Campbell AE (2003) Vigorous innateand virus-specific cytotoxic T-lymphocyte responses to murine cytomegalovirus in thesubmaxillary salivary gland. J Virol 77:1703–1717.

44. Sacher T, et al. (2008) The major virus-producing cell type during murine cytomega-lovirus infection, the hepatocyte, is not the source of virus dissemination in the host.Cell Host Microbe 3:263–272.

45. Julander JG, Shafer K, Smee DF, Morrey JD, Furuta Y (2009) Activity of T-705 ina hamster model of yellow fever virus infection in comparison with that of a chemi-cally related compound, T-1106. Antimicrob Agents Chemother 53:202–209.

46. Cardin RD, et al. (2009) Amphipathic DNA polymers exhibit antiviral activity againstsystemic murine Cytomegalovirus infection. Virol J 6:214.

47. Xiao SY, Zhang H, Guzman H, Tesh RB (2001) Experimental yellow fever virus infectionin the Golden hamster (Mesocricetus auratus). II. Pathology. J Infect Dis 183:1437–1444.

48. Julander JG, Morrey JD, Blatt LM, Shafer K, Sidwell RW (2007) Comparison of theinhibitory effects of interferon alfacon-1 and ribavirin on yellow fever virus infectionin a hamster model. Antiviral Res 73:140–146.

49. Sbrana E, et al. (2004) Efficacy of post-exposure treatment of yellow fever with ri-bavirin in a hamster model of the disease. Am J Trop Med Hyg 71:306–312.

50. Wang E, et al. (2007) Chimeric Sindbis/eastern equine encephalitis vaccine candidatesare highly attenuated and immunogenic in mice. Vaccine 25:7573–7581.

51. Arrigo NC, Watts DM, Frolov I, Weaver SC (2008) Experimental infection of Aedessollicitans and Aedes taeniorhynchus with two chimeric Sindbis/Eastern equine en-cephalitis virus vaccine candidates. Am J Trop Med Hyg 78:93–97.

52. Vogel P, Kell WM, Fritz DL, Parker MD, Schoepp RJ (2005) Early events in the path-ogenesis of eastern equine encephalitis virus in mice. Am J Pathol 166:159–171.

53. Zasloff M (1997) US Patent 6,143,738.54. Villarreal LP (2005) Viruses and the Evolution of Life (ASM Press, Washington, DC).55. Adelman MK, Schluter SF, Marchalonis JJ (2004) The natural antibody repertoire of

sharks and humans recognizes the potential universe of antigens. Protein J 23:103–118.56. Venkatesh B, et al. (2007) Survey sequencing and comparative analysis of the ele-

phant shark (Callorhinchus milii) genome. PLoS Biol 5:e101.57. Basta HA, Cleveland SB, Clinton RA, Dimitrov AG, McClure MA (2009) Evolution of

teleost fish retroviruses: Characterization of new retroviruses with cellular genes.J Virol 83:10152–10162.

58. Wolf MC, et al. (2010) A broad-spectrum antiviral targeting entry of enveloped vi-ruses. Proc Natl Acad Sci USA 107:3157–3162.

59. Pietras RJ, Weinberg OK (2005) Antiangiogenic steroids in human cancer therapy.Evid Based Complement Alternat Med 2:49–57.

60. Connolly B, Desai A, Garcia CA, Thomas E, Gast MJ (2006) Squalamine lactate forexudative age-related macular degeneration. Ophthalmol Clin North Am 19:381–391.

61. Zhang XH, et al. (1998) Synthesis of squalamine utilizing a readily accessible sper-midine equivalent. J Org Chem 63:8599–8603.

Zasloff et al. PNAS | September 20, 2011 | vol. 108 | no. 38 | 15983

IMMUNOLO

GY

Correction

IMMUNOLOGYCorrection for “Squalamine as a broad-spectrum systemic anti-viral agent with therapeutic potential,” by Michael Zasloff, A.Paige Adams, Bernard Beckerman, Ann Campbell, Ziying Han,Erik Luijten, Isaura Meza, Justin Julander, Abhijit Mishra, WeiQu, John M. Taylor, Scott C. Weaver, and Gerard C. L. Wong,which appeared in issue 38, September 20, 2011, of Proc NatlAcad Sci USA (108:15978–15983; first published September 20,2011; 10.1073/pnas.1108558108).The authors note that the affiliation for John M. Taylor should

instead appear as Fox Chase Cancer Center, Philadelphia, PA19111. The corrected author and affiliation lines appear below.The online version has been corrected.

Michael Zasloffa,1, A. Paige Adamsb, Bernard Beckermanc,Ann Campbelld, Ziying Hane, Erik Luijtenc,f, Isaura Mezag,Justin Julanderh, Abhijit Mishrai, Wei Quc, John M. Taylore,Scott C. Weaverb, and Gerard C. L. Wongi

aTransplant Institute, Departments of Surgery and Biochemistry andMolecular and Cell Biology, Georgetown University Medical Center,Washington, DC 20007; bInstitute for Human Infections and Immunity andDepartment of Pathology, University of Texas Medical Branch, Galveston,TX 77555-0609; cDepartment of Materials Science and Engineering,Northwestern University, Evanston, IL 60208-3108; dDepartment ofMicrobiology and Molecular Cell Biology, Eastern Virginia Medical School,Norfolk, VA 23507; eFox Chase Cancer Center, Philadelphia, PA 19111;fDepartment of Engineering Sciences and Applied Mathematics,Northwestern University, Evanston, IL 60208-3125; gDepartamento deBiomedicina Molecular, Research and Advanced Studies Center of theNational Polytechnic Institute of Mexico, 07360 Mexico D.F., Mexico;hInstitute for Antiviral Research, Utah State University, Logan, UT 84322-5600; and iBioengineering Department, Chemistry and BiochemistryDepartment, California Nano Systems Institute, University of California,Los Angeles, CA 90095-1600

www.pnas.org/cgi/doi/10.1073/pnas.1115667108

18186 | PNAS | November 1, 2011 | vol. 108 | no. 44 www.pnas.org/cgi/doi/10.1073/pnas.1115667108

Supporting InformationZasloff et al. 10.1073/pnas.1108558108SI Materials and MethodsComputer Simulation (Laboratory of E.L.). All modeling was per-formed by means of molecular dynamics simulations using theLAMMPS package (1). To increase computational efficiency, thelipid molecules, squalamine molecules, Rac1 proteins, and theircounterions were coarse-grained and represented by sphericalbeads or assemblies of beads that interacted through genericshort-range potentials. In addition, all Coulombic forces werefully accounted for by Ewald summation.Coarse graining. The system was set up to simulate the interactionbetween squalamine molecules and Rac1 proteins with the outersurface of a lipid bilayer membrane. This surface was modeled by400 beads of uniform size of 8.5 Å, which were confined to the x–yplane. Using the small-angle X-ray scattering setup, we used amembrane in which 20% of the beads carried a charge of −1e.Coarse-grained models of the Rac1 and squalamine moleculeswere obtained using the VMD Shape-Based Coarse-Graining(SBCG) tool (2, 3). To accurately and efficiently reproduce theshape of a molecule, the SBCG tool generally uses a distributionof bead sizes. However, because the distance of closest approachbetween two beads determines the electrostatic binding, weopted for beads of uniform size, equal in diameter to the 8.5-Åbeads representing the lipid head groups. This uniformity wasachieved by adjusting the number of coarse-grained beads so thatthe average bead size was as close to 8.5 Å as possible, and then,all beads were set to that uniform size. For squalamine, thisuniformity required three beads, with a separation of 9.6 Å be-tween the head and body beads and a separation of 12.4 Å be-tween the body and tail beads. The charge on each bead wasassigned manually based on the coarse-grained model (i.e., −1efor the acidic head, 0 for the hydrophobic body, and +3e for thebasic tail). For the coarse graining of the Rac1 protein, we iso-lated the Rac1 molecule from the Protein Data Bank file 2RMK(4) and then separated the protein into a tail domain and a bodydomain. The polybasic tail was coarse-grained with the same 8.5-Å building block as the squalamine and the lipid head groups toensure that the electrostatic interaction strength between the tailand the membrane was modeled in a manner consistent with thesqualamine–membrane attraction. The total charge on the tailwas +5, which was divided into partial charges of +0.38, +1.53,+2.09, +1.09, +0.86, −0.04, and −0.91, respectively. For theweakly charged body of Rac1, steric interactions dominate, andtherefore, a larger bead size (11.6 Å) could be used to increasecomputational efficiency while maintaining an accurate repre-sentation of the excluded volume. For both the body and the tail,the VMD SBCG tool was used to obtain the number of beads,their relative positions, equilibrium bond lengths between beads,and individual charges on each molecule.Bead interactions. The interaction between bonded beads wasmodeled by a harmonic potential (Eq. S1):

Ubond ¼ 200εðr− r0Þ2; [S1]

where r0 was the equilibrium bond length and ε was the Lennard–Jones (LJ) unit of energy. For squalamine, a bond angle potentialwas introduced to model its stiffness (Eq. S2):

Uangle ¼ 4ε ðθ− θ0Þ2; [S2]

with θ0 = 180° of the equilibrium angle between the two bonds.All nonbonded beads interacted through a short-range, purelyrepulsive-shifted and truncated LJ potential that was truncated

at the LJ energy minimum rcut = 21/6 σ, where σ is the LJ unit oflength that is shifted by ε so that both energy and force vanishedat rcut. The cutoff rcut was chosen as the effective diameter of theparticle and equated to the bead diameter chosen in the coarse-graining procedure. Electrostatic energies and forces werecomputed using Ewald summation with a relative accuracy of10−4. Using information in the work by Stevens and Kremer (5)and customary lengths in coarse-grained simulations of electro-static complexation phenomena, the Bjerrum length was set to3σ. The solvent was modeled as a homogeneous dielectric me-dium, with Brownian effects represented by a Langevin ther-mostat with damping time 10τ, where τ is the LJ unit of time(Eq. S3):

τ ¼ffiffiffiffiffiffiffiffiffimσ2

ε

s; [S3]

with m as the LJ unit of mass. The equations of motion wereintegrated using the velocity Verlet algorithm. The temperaturewas controlled by the Langevin thermostat as well, and it was setto T = 1.2ε/kB, where kB is Boltzmann’s constant.Simulation setup. The total number of squalamine molecules andRac1 proteins was determined from the ratio of the total net(positive) charge on all molecules of either species to the totalnegative charge on the membrane surface. The net charge on onesqualamine molecule is +2e, the net charge on one Rac1 proteinis +4e, and a charge ratio of 1.5 with respect to the membranecharge was imposed. Accordingly, 60 squalamine molecules and30 Rac1 proteins were placed in the simulation box. In addition,80 positive and 120 negative monovalent counterions werepresent in the system to maintain charge neutrality. No addi-tional salt was added. An orthorhombic simulation box was usedwith a linear size of 172 Å (corresponding to 20 lipid bead di-ameters) and periodic boundary conditions in the x and y di-rections. The height of the cell was 344 Å (40 lipid beaddiameters), with repulsive LJ walls at the top and the bottom ofthe simulation box. These walls interacted with the same purelyrepulsive-shifted and truncated LJ potential as the beads. Themolecular dynamics time step was set to 0.01τ, and a typicalsimulation run took 10 million steps, corresponding to 1 × 105τ.

X-Ray Diffraction Measurements (Laboratory of G.C.L.W.). 1,2-Dio-leoyl-sn-Glycero-3-[Phospho-L-Serine] (sodium salt), 1,2-Dioleoyl-sn-Glycero-3-Phosphocholine, and 1,2-Dioleoyl-sn-Glycero-3-Phos-phoethanolamine, all purchased from Avanti Polar Lipids, wereused without additional preparation. Lipid solutions in chloro-form were dried under N2 and desiccated under vacuum over-night. Dried lipids were rehydrated with buffer containing 50mM Tris (pH 7.4) and 100 mM NaCl to a final concentration of20 mg/mL, incubated at 37 °C overnight, sonicated, and extrudedthrough a 0.2-μm Nucleopore (Whatman) filter to form smallunilamellar vesicles. WT human Rac1 protein, purchased fromCytoskeleton, was reconstituted to 5 mg/mL in the followingbuffer: 50 mM Tris (pH 7.5), 0.5 mM MgCl2, 50 mM NaCl, 0.5%(wt/vol) sucrose, 0.1% (wt/vol) dextran. Squalamine was dis-solved in buffer containing 50 mM Tris (pH 7.4) and 100 mMNaCl to 1 mg/mL. Protein solutions were incubated with smallunilamellar vesicles at specific protein to lipid molar ratios andsealed in quartz capillaries. X-ray diffraction was performed atthe small-angle X-ray scattering spectrometer at the CaliforniaNanosystems Institute, University of California at Los Angeles,using CuKα radiation (λ = 1.54 Å) X-rays. The scattered in-

Zasloff et al. www.pnas.org/cgi/content/short/1108558108 1 of 5

tensity was collected using a MAR Research Image Plate De-tector (pixel size = 150 μm). Additional experimental details canbe found in the work by Yang et al. (6).

Dengue Virus (Laboratory of I.M.).Viral infection of a line of humandermal microvascular endothelial cells that were grown on un-coated glass coverslips followed a published protocol (7). Den V2was obtained from the Institute of Diagnostic and Epidemio-logical Reference in Mexico City. The strain was isolated froma patient who developed dengue fever. Monolayers of humandermal microvascular endothelial cells were pretreated withsqualamine (10–100 μg/mL) for 2 h at 37 °C in MCDB-131Medium (GIBCO); fresh medium without squalamine was addedbefore viral exposure. Virus (at a multiplicity of infection of 3.0)remained in contact with cells for 30 min at 4 °C followed by 90min at 37 °C. The medium was then replaced with fresh mediumlacking virus, and cells were maintained at 37 °C for 48 h. Cellswere then fixed and processed for immunohistochemical de-tection of viral E protein or for F-actin labeling with rhodamine–phalloidin. Viral E protein expression was used to monitor thepercentage of infection (7). In the absence of drug, about 38% ofcells were infected. Ten randomly chosen fields were counted incontrol coverslips and those coverslips treated with squalamine.

Hepatitis Viruses (Laboratory of J.M.T.). Monolayer cultures ofprimary human hepatocytes were infected with human hepatitis Bvirus (HBV) or human hepatitis δ-virus (HDV) using a modifi-cation of procedures as described in the work by Taylor and Han(8). Cells were exposed to HBV in the presence of 5% poly-ethylene glycol in serum-free hepatocyte growth medium (HGM;Cellzdirect) for 16 h and then replaced with virus-free medium.In one series of experiments, squalamine was present 1 h beforeaddition of virus and remained in the medium for 16 h. Ina second series, squalamine was added at 24 h after viral in-fection and remained in the medium for 16 h. Medium-con-taining virus and squalamine was replaced with fresh medium(drug- and virus-free), and both treated and untreated cells weremaintained for 12 d. Total RNA was extracted, with concen-trations as indicated, and assayed per unit of RNA by real-timePCR for HBV sequences as described (8). The experimentalpoints were performed in at least triplicate.Infections with HDV were performed and assayed in a manner

very similar to HBV, except that HDV exposure was limited to 3 hand was performed in the absence of polyethylene glycol (8). Theperiod of infection was reduced to 7 d.We could not directly evaluate the virucidal activity of squal-

amine against HBV because of the dilute concentration of virusavailable to us, which would require additional dilution of thepretreated sample before assay. BecauseHDV uses the same viralenvelope as HBV, we evaluated the virucidal activity againstHDV by incubating the HDV inoculum with squalamine (20 μg/mL) in HGM or an equivalent volume of drug-free HGM for 3 hat 37 °C. The samples (run in triplicate) were diluted with HGMcontaining 5% PEG, and primary hepatocytes were exposedovernight followed by replacement with fresh medium. Viralreplication was assayed as above at 7 d.

Yellow Fever Virus (Laboratory of J.J.). Animals. Female GoldenSyrian hamsters (Mesocricetus auratus) with an average weight of110 g were used after a quarantine period of greater than 48 h.Experiments were conducted in the biosafety level 3 animal suiteat the Utah State University Laboratory Animal Research Cen-ter. All personnel continue to receive special training on blood-borne pathogen handling by this university’s EnvironmentalHealth and Safety Office. Standard operating procedures forbiosafety level 3 were used.Virus. Jimenez, a hamster-adapted yellow fever virus strain, wasobtained as a gift fromRobert B. Tesh (Galveston, TX). The virus

was inoculated into five adult female hamsters. The livers of theinfected hamsters were removed 3 d postinjection and homog-enized in a 2× volume of sterile PBS. This liver homogenate hada titer of 106.0 50% cell cultures infectious doses/mL (CCID50)and served as the stock for subsequent studies.Virus inoculations, treatment, and postexposure monitoring. Hamsterswere randomly assigned to groups, with 10 included in each groupand 20 placebo-treated controls. A 10−4 dilution (102.0 CCID50/mL) of the virus was prepared in minimal essential media.Hamsters were injected i.p. with 0.1 mL diluted virus (10CCID50/animal). Squalamine was administrated s.c. at a totaldaily dose of either 15 or 30 mg/kg as indicated one time per day.To evaluate protection, dosing began at 24 h before introductionof virus and ended on day 6; in the case of the treatment ex-periment, dosing began on either day 1 or 2 after infection andcontinued until days 8 or 9, respectively. Ribavirin was admin-istered i.p. at total daily doses of 3.2, 10, or 32 mg/kg per d ad-ministered two times daily beginning 24 h before introduction ofvirus and ending on day 6 (protection experiment) or at a totaldaily dose of 75 mg/kg administered i.p. two times daily (37.5 mg/kg). Hamsters were observed daily for mortality for a total of21 d, and body weight was recorded on 0, 3, and 6 d postinfec-tion. Serum alanine amino transferase was measured as pre-viously described (9).Statistical analysis. Survival data were analyzed using the Wilcoxonlog-rank survival analysis. and all other statistical analyses weredone using one-way ANOVA followed by Bonferroni’s multiplecomparison test (Prism 5; GraphPad Software, Inc). Values ofP ≤ 0.05 were considered significant.

Eastern Equine Encephalitis Virus (Laboratory of A.P.A. and S.C.W.).Animals. Sixteen 6-wk-old Golden Syrian hamsters were purchasedfrom Charles River Laboratories and housed in the animal bio-safety level 3 facilities at the University of Texas Medical Branch(UTMB). All experimental animal protocols were reviewed andapproved by the Institutional Animal Care and Use Committeeat UTMB.Virus.TheNorth American eastern equine encephalitis virus strainFL93-939 was provided by the UTMB World Reference Centerfor Emerging Viruses and Arboviruses and was originally isolatedfrom a 1993 Florida pool of Culiseta melanura mosquitoes. Aninfectious cDNA clone of this strain was subsequently developedby RT-PCR amplification and cloning as previously described(10). Virus was subsequently rescued from the infectious cDNAclone, and virus stocks were titered by plaque assay using Vero(monkey kidney) cells as previously described (11).Virus inoculations, treatment, and postexposure monitoring. Sixteenanimals received an s.c. inoculation of 3.0 log10 pfu strain FL93-393 in a 0.1-mL volume; 24 h before infection and on days 0–6after infection, one cohort of 10 animals was treated withsqualamine (10 mg/kg s.c.), and another cohort of 6 animals wastreated s.c. with D5W placebo. After infection, all animals weremonitored daily for clinical signs of disease, including anorexia,lethargy, abnormal neurologic signs, or mortality. At −1 to 14 dpostinfection, body weight was recorded, and at 1–4 d post-infection, blood samples were collected from the retroorbitus forvirus titration. Virus titrations were performed on the sera byplaque assay using Vero cells as previously described (11). Thelimit of detection for the assay was 100 pfu/mL.Statistical analysis. Survival data were analyzed using the log-ranktest, and for viremia data, a statistical comparison was performedusing a one-way ANOVA followed by Bonferroni’s multiplecomparison test (GraphPad). Values of P ≤ 0.05 were consid-ered significant.

Murine Cytomegalovirus (Laboratory of A.C.). Animals.Male BALB/cmice (H-2d haplotype) were purchased from Harlan Laborato-ries and housed in the animal facility at East Virginia Medical

Zasloff et al. www.pnas.org/cgi/content/short/1108558108 2 of 5

School in sterile microisolator cages with sterile food, water, andbedding. All animal procedures were approved by the AnimalCare and Use Committee of Eastern Virginia Medical Schooland adhered to the guidelines established by the US AnimalWelfare Act.Virus. The Smith strain of murine cytomegalovirus (VR-194) wasused and prepared for inoculation through passage in 3-wk-oldmale BALB/c weanling mice as previously described (12).Virus inoculations, treatment, and postexposure monitoring. In allexperiments, adult mice (6 wk of age) were infected by i.p. in-jection with 1 × 103 pfu virulent salivary gland passaged murinecytomegalovirus. Squalamine was administered either i.p. or s.c.at a dose of 10 mg/kg beginning 1 d before virus inoculation.Controls received an equivalent volume of vehicle (D5W)through the i.p. route. Each cohort was represented by 18 ani-mals. Groups of mice (n = 6) were killed at days 3, 7, and 14 dpostinfection, and livers, spleens, lungs, and salivary glands wereharvested. These tissues were then processed for viral titers asdescribed (12). The limit of detection was 10 pfu/mL.Statistical analysis. For viral titers, statistical comparisons wereperformed using a one-way ANOVA (GraphPad). Values of P ≤0.05 were considered significant.

Videos of Rac1 and Squalamine near a Model Membrane. Descriptionof video. rac1_only.mp4. This video (Movie S1) shows 68 Rac1proteins (accompanied by 180 positive and 272 negative mono-valent counterions) in thermal equilibrium. The purple Rac1tails associate with the blue membrane beads representingcharged lipids. For Rac1 in solution, the tails are frequently seento latch onto nearby Rac1 proteins; this finding is because ofnegatively charged domains on those proteins.The run time of this video is 679τ (see description of the time

unit τ below), with one frame per time unit.squalamine_only.mp4. This video (Movie S2) shows 135

squalamine molecules (accompanied by 180 positive and 270negative monovalent counterions) in thermal equilibrium. Thered (positive) sides of squalamine near the membrane associatewith blue (negatively charged) lipids.The run time of this video is 679τ, with one frame per time unit.squalamine_rac1_mixture.mp4. This video (Movie S3) shows 68

squalamine molecules and 135 Rac1 proteins (accompanied by180 positive and 542 negative monovalent counterions) in thermalequilibrium. The Rac1 proteins can now be observed to be con-

siderably more dispersed through the bulk and less associated withthe membrane than in the absence of squalamine (Movie S1).The run time of this video is 679τ, with one frame per time unit.rac1_displacement.mp4. The components in this system (Movie

S4) are identical to the components in Movie S3, but now, theinitial state is created from an equilibrium configuration of 68Rac1 proteins associated with the membrane (Movie S1), towhich 135 squalamine molecules have been added in the topone-half of the simulation cell. Movie S4 displays how the systemreaches thermal equilibrium (i.e., the state depicted in MovieS3). During this process, squalamine diffuses to the membraneand displaces part of the Rac1 proteins. This process is partic-ularly noticeable in the last one-third of Movie S4, where thenumber of Rac1 proteins near the membrane has decreasedsignificantly.The run time of this video is 2,716τ (four times longer than

Movies S1–S3), with one frame per four time units.Color coding. The lipid membrane is represented by 900 blue andyellow beads of uniform size of 8.5 Å. Yellow beads are neutral,and blue beads carry a charge of −1e. Charged beads represent20% (180 of 900) of all lipids.Each Rac1 molecule is represented by two domains: the body

is composed of cyan-colored beads of size 11.6 Å, and the tail iscomposed of purple-colored beads of size 8.5 Å. The body ofRac1 carries a net charge of −1e, and the tail carries a net chargeof +5e.Each squalamine molecule is represented by two pink beads

and one red bead of uniform size of 8.5 Å. The red bead carriesa charge of +3e, and the pink bead on the opposite end carriesa charge of −1e.Ions are represented by green and orange beads, with orange

ions carrying a positive unit charge and green ions carryinga negative unit charge. The counterions are simulated using size8.5 Å, but for visual clarity, they are represented at one-half theirsize in Movies S1–S4.Technical information regarding the simulations. Movies S1–S4 werecreated from coarse-grained molecular dynamics simulations inwhich all excluded volume and electrostatic interactions werefully taken into account (model as described in SI Materials andMethods). The simulation cell has a width and depth of 255 Åand a height of 340 Å, and it is periodically replicated in the xand y directions. All simulations used a molecular dynamics timestep of 0.01τ, with τ defined in Eq. S3.

1. Plimpton S (1995) Fast parallel algorithms for short-range molecular dynamics.J Comput Phys 117:1e19.

2. Arkhipov A, Freddolino PL, Schulten K (2006) Stability and dynamics of virus capsidsdescribed by coarse-grained modeling. Structure 14:1767e1777.

3. Arkhipov A, Yin Y, Schulten K (2008) Four-scale description of membrane sculpting byBAR domains. Biophys J 95:2806e2821.

4. Modha R, et al. (2008) The Rac1 polybasic region is required for interaction with itseffector PRK1. J Biol Chem 283:1492e1500.

5. Stevens MJ, Kremer K (1995) The nature of flexible linear polyelectrolytes in salt freesolution: A molecular dynamics study. J Chem Phys 103:1669e1690.

6. Yang L, et al. (2004) Self-assembled virus-membrane complexes. Nat Mater 3:615e619.

7. Zamudio-Meza H, Castillo-Alvarez A, González-Bonilla C, Meza I (2009) Cross-talkbetween Rac1 and Cdc42 GTPases regulates formation of filopodia required fordengue virus type-2 entry into HMEC-1 cells. J Gen Virol 90:2902e2911.

8. Taylor JM, Han Z (2010) Purinergic receptor functionality is necessary for infectionof human hepatocytes by hepatitis delta virus and hepatitis B virus. PLoS One 5:e15784.

9. Julander JG, Morrey JD, Blatt LM, Shafer K, Sidwell RW (2007) Comparison of theinhibitory effects of interferon alfacon-1 and ribavirin on yellow fever virus infectionin a hamster model. Antiviral Res 73:140e146.

10. Aguilar PV, et al. (2008) Structural and nonstructural protein genome regions ofeastern equine encephalitis virus are determinants of interferon sensitivity andmurine virulence. J Virol 82:4920e4930.

11. Powers AM, Brault AC, Kinney RM, Weaver SC (2000) The use of chimeric Venezuelanequine encephalitis viruses as an approach for the molecular identification of naturalvirulence determinants. J Virol 74:4258e4263.

12. Cavanaugh VJ, Deng Y, Birkenbach MP, Slater JS, Campbell AE (2003) Vigorous innateand virus-specific cytotoxic T-lymphocyte responses to murine cytomegalovirus in thesubmaxillary salivary gland. J Virol 77:1703e1717.

Zasloff et al. www.pnas.org/cgi/content/short/1108558108 3 of 5

Movie S1. Rac1 only. Details of movie in SI Materials and Methods.

Movie S1

Movie S2. Squalamine only. Details of movie in SI Materials and Methods.

Movie S2

Movie S3. Squalamine Rac1 mixture. Details of movie in SI Materials and Methods.

Movie S3

Zasloff et al. www.pnas.org/cgi/content/short/1108558108 4 of 5

Movie S4. Rac1 displacement by squalamine. Details of movie in SI Materials and Methods.

Movie S4

Zasloff et al. www.pnas.org/cgi/content/short/1108558108 5 of 5