REPORTS

15. D. J. Sieg, C. R. Hauck, D. D. Schlaepfer, J. Cell Sci.
112, 2677 (1999).
16. P. Y. Chan, S. B. Kanner, G. Whitney, A. Aruffo, J. Biol.
Chem. 269, 20567 (1994).
17. X. D. Ren, W. B. Kiosses, M. A. Schwartz, EMBO J. 18,
578 (1999).
18. X. D. Ren et al., J. Cell Sci. 113, 3673 (2000).
19. N. Watanabe, T. Kato, A. Fujita, T. Ishizaki, S. Naru-
miya, Nature Cell Biol. 1, 136 (1999).
20. A. S. Alberts, J. Biol. Chem. 276, 2824 (2001).
21. T. N. Sims, M. L. Dustin, Immunol. Rev. 186, 100 (2002).
22. D. A. Brown, E. London, Biochem. Biophys. Res. Com- (C.H.E.), and NIH grants CA87038 (D.D.S.), GM 44585
mun. 240, 1 (1997). (E.E.M.), and GM 62939 and GM 68695 (G.G.G.).
23. C. Allal et al., J. Biol. Chem. 275, 31001 (2000).
Supporting Online Material
24. A. F. Palazzo et al., Curr. Biol. 11, 1536 (2001).
www.sciencemag.org/cgi/content/full/303/5659/836/
25. M. A. del Pozo, L. S. Price, N. B. Alderson, X. D. Ren,
DC1
M. A. Schwartz, EMBO J. 19, 2008 (2000).
Materials and Methods
26. We thank J. Taek-Yoon for help with spreading as- Figs. S1 to S5
says, and Y. Wen, S. Thomas, J. Lippincott-Schwartz, Table S1
S. Narumiya, and A. Alberts for cells and reagents. References
Supported by predoctoral fellowships from Fonds de
Recherche en Santé du Quebec (A.F.P.), HHMI 17 September 2003; accepted 5 November 2003

Integrins Regulate Rac Targeting domains in vivo are still uncertain (8–10),
and there is evidence for different types (11).

by Internalization of Known collectively as lipid rafts, they repre-

sent cholesterol-rich regions of higher order
and lower buoyant density than bulk plasma
Membrane Domains membrane. Sphingolipids, including ganglio-
sides such as GM1, are proposed structural
Miguel A. del Pozo,1,2,3* Nazilla B. Alderson,1,2 components of lipid rafts (12). These do-
William B. Kiosses,1 Hui-Hsien Chiang,1 mains have been proposed to compartmental-
Richard G. W. Anderson,4 Martin A. Schwartz1,5 ize and organize signal transduction at the
plasma membrane (5, 6, 13).
Translocation of the small GTP-binding protein Rac1 to the cell plasma membrane is GTP-Rac1 binds more effectively to
essential for activating downstream effectors and requires integrin-mediated adhesion membranes from adherent than from sus-
of cells to extracellular matrix. We report that active Rac1 binds preferentially to pended fibroblasts, indicating that integrins
low-density, cholesterol-rich membranes, and specificity is determined at least in part regulate Rac1 membrane binding sites at
by membrane lipids. Cell detachment triggered internalization of plasma membrane the cell surface (3). RhoA and Rac1 are
cholesterol and lipid raft markers. Preventing internalization maintained Rac1 mem- also thought to be concentrated in lipid
brane targeting and effector activation in nonadherent cells. Regulation of lipid rafts by rafts and caveolae (14, 15). We therefore
integrin signals may regulate the location of membrane domains such as lipid rafts and investigated the involvement of such mem-
thereby control domain-specific signaling events in anchorage-dependent cells. brane domains in Rac1 membrane targeting
and elucidated how they are regulated by
Integrin-mediated cell adhesion not only ini- flotillins, src-family kinases, and glyco- integrins. Membrane domains such as lipid
tiates signals directly but also modulates trans- sylphosphatidylinositol (GPI)-linked proteins rafts can be disrupted by depleting mem-
mission of signals downstream of growth factor (5–7). The size and composition of these brane cholesterol with methyl-␤-cyclodex-
receptors (1). Among these signals are the Rho
family of small GTP-binding proteins that reg-
Fig. 1. Cholesterol de-
ulate cell polarization and migration, membrane pletion mimics loss of
trafficking, cell cycle progression, gene expres- adhesion. (A) Adherent,
sion, and oncogenic transformation (2). Inte- serum-starved 3T3 cells
grins control the activation of Rho proteins and were incubated with or
separately regulate the translocation of activat- without 10 mM methyl-
ed (GTP-bound) Rac1 and Cdc42 to the plasma ␤-cyclodextrin (CD) for
1 hour at 37°C. Some
membrane (3, 4). Consequently, GTP-Rac1 in CD-treated cells were
nonadherent cells remains in the cytoplasm then incubated with
bound to Rho guanine nucleotide dissociation cholesterol (16 ␮g/ml)–
inhibitor (RhoGDI) and thus is uncoupled from 0.15 mM CD [⫹chol in
downstream signaling. This regulatory mecha- (A), (B), (C), (D), and (F)]
nism may account for a variety of effects of for 1 hour at 37°C to
replenish cholesterol.
integrins in anchorage-dependent cells. Cells were stimulated
The plasma membrane is thought to con- with 10% serum for 10
tain domains enriched in cholesterol, sphin- min, and Rac1 activity
golipids, and proteins including caveolins, was assayed (n ⫽ 5 ex-
periments). (B) Cells
treated as in (A) were
1
Department of Cell Biology, 2Department of Immu- separated into particu-
nology, The Scripps Research Institute, 10550 North late and cytosolic frac-
Torrey Pines Road, La Jolla, CA 92037, USA. 3Centro tions. Rac1 association
Nacional de Investigaciones Cardiovasculares, Madrid, with membranes was determined by Western blot (17). Values are means ⫾ SEM (n ⫽ 3). (C) GFP-
Spain. 4Department of Cell Biology, University of V12Rac1–expressing 3T3 cells treated as in (A). Arrows indicate V12Rac1 at the plasma membrane. (D) Pixel
Texas Southwestern Medical Center, Dallas, TX intensity for GFP-V12Rac1 was assessed starting at the cell edge (18). GFP-V12Rac1 is uniformly distributed
75235, USA. 5Departments of Microbiology and Bio- across CD-treated cells but is concentrated at the edges in control and cholesterol-replenished (⫹chol) cells.
medical Engineering, Mellon Prostate Cancer Research Values are means ⫾ SEM (n ⫽ 3). (E) Isolated plasma membranes from adherent (Att) 3T3 cells were treated
Institute and Cardiovascular Research Center, Univer- with or without CD; plasma membranes from suspended (Sus) cells were analyzed as a control. Binding of
sity of Virginia, Charlottesville, VA 22908, USA. cytosolic myc-tagged V12Rac1 was measured by Western blot. ␤1 integrin blot shows levels of membrane
*To whom correspondence should be addressed. E- protein for normalization. (F) PAK protein was immunoprecipitated and kinase activity assayed in cells treated
mail: mdelpozo@scripps.edu as in (A) (n ⫽ 3). PAK protein is indicated by Western blot.

www.sciencemag.org SCIENCE VOL 303 6 FEBRUARY 2004 839

 REPORTS
trin (CD) (5, 6 ). Depletion of cholesterol These and other findings (14–16) suggest sphingomyelin (Sph) that is in a liquid-
from adherent 3T3 cells did not alter acti- a role for cholesterol-rich domains in the ordered state that may be similar to that of
vation of endogenous Rac1 but prevented translocation of Rac1 to membranes. To fur- endogenous lipid rafts (5, 6, 18) supported
its translocation to the membrane (Fig. 1, A ther investigate Rac1 targeting, we used a greater GTP-dependent Rac1 binding than
and B, and fig. S1A), consistent with a recombinant, isoprenylated human Rac1- other lipid mixtures (Fig. 2, B and C).
recent study on A431 epidermoid carcino- RhoGDI complex purified from yeast (17) Phosphorus assays showed equal recovery
ma cells (16 ). A constitutively active mu- (fig. S2A). Rac1 showed selective and GTP- of liposomes of all compositions (17 ).
tant, V12Rac1, behaved similarly (Fig. 1, C dependent binding to membranes isolated RhoGDI did not bind under any circum-
and D, and fig. S1B), further indicating that from adherent 3T3 cells, whereas RhoGDI stance and remained in the supernatant
changes in Rac1 GTP loading were not showed no membrane association (fig. S2B). (19). In contrast, the phospholipid-binding
involved. In vitro binding of V12Rac1 to Plasma membranes from adherent fibroblasts bee venom peptide melittin (17 ) bound bet-
isolated membranes was similarly inhibited were then fractionated with a detergent-free den- ter to liposomes with greater PC content
by cholesterol depletion (Fig. 1E). Treat- sity gradient centrifugation method (17) (fig. (Fig. 2, B and C). The physical state of the
ment of adherent cells with CD reversibly S2C). Rac1 showed greater GTP-dependent lipids may therefore contribute to the bind-
inhibited serum stimulation of the Rac1 binding to the low-density, cholesterol- and ing of Rac1 to specific membrane domains.
effector p21-activated kinase (PAK) (Fig. caveolin-1– enriched fraction relative to Although lipid rafts are thought to be
1F and fig. S1A), consistent with the re- whole plasma membranes and no detectable heterogeneous, ganglioside GM1 is a widely
quirement for membrane translocation in GTP-dependent binding to the high-density, used marker for these domains (12, 20).
effector activation (4 ). Depletion of mem- cholesterol-depleted fraction (Fig. 2A). Therefore, localization of the GM1-binding
brane cholesterol therefore mimics the ef- When liposomes were prepared from cholera toxin subunit B (CTxB) was com-
fect of cell detachment on Rac1 targeting purified lipids, an equimolar mixture of pared with that of Rac1. Following treatment
and effector activation (3, 4 ). phosphatidylcholine (PC), cholesterol, and of cells with serum to activate Rac1, Rac1
and CTxB colocalized extensively, primarily
at cell edges, whereas little colocalization
was observed in unstimulated cells (Fig. 2D).
Rac1 and GM1 showed negligible colocaliza-
tion in nonadherent cells, and staining of
adherent cells with wheat germ agglutinin to
label the entire cell surface showed no con-
centration of stain at cell edges (19). Further-
more, Rac1 colocalized with GM1 that was
clustered with CTxB-coated latex beads (5
␮m). Rac1 was not observed around control
beads coated with antibody to transferrin re-
ceptor (TfR) (Fig. 2E). Activation of Rac1 by
CTxB beads was not detected (fig. S5B),
confirming an effect on Rac1 targeting rather
than on GTP loading. Thus, Rac1 preferen-
tially associates in vivo with regions of the
membrane enriched in GM1.
If low-density, cholesterol-rich domains
provide the membrane binding sites for Rac1,
then these domains may be the targets for
integrin regulation of Rac1 recruitment to the
membrane. We therefore assayed the effects
of cell adhesion on GM1 localization. 3T3
cells, either attached to glass cover slips or
incubated in suspension for various times,
were chilled on ice and incubated with fluo-
rescently labeled CTxB [fluorescein isothio-
cyanate (FITC)–CTxB] to label surface GM1.
FITC-CTxB efficiently labeled the cell sur-
face before or immediately after detachment,
Fig. 2. GTP-dependent binding of Rac1 to membranes. (A) Purified recombinant isoprenylated Rac1-RhoGDI but surface staining rapidly decreased with
complex was loaded with GTP␥S or guanosine 5⬘-diphosphate (GDP), then incubated with whole plasma time in suspension (Fig. 3A). CTxB staining
membranes (PM), light membrane fractions (LMF), or heavy membrane fractions (HMF) from human
fibroblasts. Membranes were sedimented and pellets analyzed by Western blot with anti-His to detect Rac1,
decreased ⬃10-fold after detachment, where-
or with antibodies to TfR, RACK-1, and caveolin-1. (B) Liposomes were prepared in vitro from purified as surface staining of CD44 increased, and ␣5
phosphatidylcholine (PC), cholesterol (Chol), and sphingomyelin (Sph) at the molar ratios indicated, then integrin was unchanged (Fig. 3B). Total GM1
binding of Rac1-RhoGDI complex or melittin was determined by Western or dot blot (n ⫽ 5). (C) levels were unchanged (19) but showed ac-
Quantification of liposome-binding assays. Values are means ⫾ SD (n ⫽ 5). (D) Colocalization of GM1 with cumulation inside the cells (fig. S3).
endogenous Rac1. Serum-starved cells or cells treated with 10% serum for 10 min were stained with To follow the fate of surface GM1, we first
FITC-CTxB to label GM1, or with anti-Rac1. Arrows indicate regions of colocalization. Magnified areas of cells
are shown (n ⫽ 4). (E) Cells in 10% serum were incubated with 5-␮m latex beads (arrows) coated with CTxB
labeled adherent cells with FITC-CTxB and
or with antibodies to the transferrin receptor (TfR), then stained for endogenous Rac (n ⫽ 5). Images are detached them from the substratum. CTxB
single confocal sections. CTxB beads showed colocalization, and transferrin beads were negative in every focal localized sharply at the cell surface immedi-
plane (19). Bar, 10 ␮m. ately after detachment but was subsequently

840 6 FEBRUARY 2004 VOL 303 SCIENCE www.sciencemag.org

 REPORTS
internalized, and within 30 to 60 min accu-
mulated in a central region of the cell (Fig.
3C). Adherent cells showed slower and less
complete internalization during this time (fig.
S4), consistent with published results (21).
Replating cells on fibronectin (FN) or anti–
␤1 integrin, but not on anti-CD44, reversed
these effects (Fig. 3C). A second lipid raft
marker, aerolysin, which labels GPI-
anchored proteins, showed a similar shift in
localization (Fig. 3D).
To determine whether movement of these
lipid raft markers correlated with changes in
cholesterol distribution, we stained cells with
the fluorescent cholesterol-binding antibiotic
filipin (22). As in other cell types (22), ad-
herent 3T3 cells showed perinuclear and ve-
sicular staining, and less intense but positive
plasma membrane staining. Cells labeled im-
mediately after detachment also showed both
plasma membrane and internal staining (Fig.
3E). Further incubation in suspension de-
creased surface staining and increased ac-
cumulation of cholesterol in a central com-
partment, similar to the effects on GM1 and
GPI-linked proteins. Replating cells on FN
restored surface cholesterol (Fig. 3, E and F).
Caveolin-1 also moved from the cell surface
to an internal compartment upon loss of ad-
hesion (19). Therefore, loss of integrin-
mediated adhesion induced internalization of
components thought to be in lipid rafts. When
cold Triton-X detergent extracts of cells were
separated on sucrose gradients, GM1 shifted
from the light to the heavy fractions (percent
in the light fraction: 50 ⫾ 6% in adherent
versus 27 ⫾ 3% in suspended cells), suggest-
ing that membrane domain structure is altered
when cells become nonadherent.
Fig. 3. Cell adhesion regulates lipid raft marker localization. (A) 3T3 cells that were adherent (Att)
The effects of integrins on Rac1 mem- or suspended (Sus) for the indicated times were chilled and labeled with FITC-CTxB to visualize GM1
brane targeting can be exerted locally (4). on the cell surface. (B) Cells were surface labeled with FITC-CTxB or with antibodies to CD44 or to
Therefore, to determine whether integrins can ␣5 integrin immediately or 2 hours after detachment, then analyzed by flow cytometry. Black line,
affect membrane domains locally, we incu- nonspecific staining; red line, suspended for 2 hours; blue line, suspended for 30 s. (C) 3T3 cells were
bated cells with beads coated with antibodies surface labeled with FITC-CTxB, rinsed, then suspended at 37°C for the indicated times. Some cells
to ␤1 integrin or to CD44 as a control. Bind- that had been suspended for 2 hours were replated for 1 hour on cover slips coated with FN,
anti–␤1 integrin, or anti–mouse CD44. (D) Adherent (Att) 3T3 cells or cells suspended for the
ing was specific because beads coated with indicated times were fixed, permeabilized, and stained with aerolysin to detect GPI-anchored
anti–␤1 integrin and anti-CD44 bound proteins. (E) 3T3 cells were adherent, suspended for the indicated times, or suspended for 2 hours
7.77 ⫾ 2.01 and 9.07 ⫾ 1.56 beads per cell, then replated on FN. Cells were fixed and stained with filipin to visualize cholesterol. (F) Pixel
respectively, whereas nonspecific rat immu- intensity for filipin was assessed starting at the cell edge and moving toward the cell center (17).
noglobulin G–coated beads bound 0.25 ⫾ Values are means ⫾ SEM (n ⫽ 3). Arrows indicate cell surface (red) or intracellular (green) staining.
0.13 beads per cell. Only beads coated with
anti–␤1 integrin triggered local accumulation Fig. 4. Internalization of GM1-containing
of GM1 and green fluorescent protein (GFP)– membrane domains regulates Rac1 sig-
V12Rac1 (fig. S5A). naling. 3T3 cells were incubated with
To determine whether integrin-regulated beads coated with antibodies to CTxB or
internalization of membrane domains medi- to TfR and then placed in suspension for 2
hours. Bound beads per cell were scored
ates loss of Rac1 membrane targeting and
in an identical experiment: CTxB beads:
downstream signaling, we blocked internal- 3.3 ⫾ 0.2; TfR beads: 3.8 ⫾ 0.4. After
ization of at least a portion of the GM1. stimulation with 10% serum for 10 min,
Adherent cells were treated with CTxB-beads cells were fixed and stained for endoge-
before detachment and incubation in suspen- nous Rac1 (A) or assayed for PAK kinase
sion for 2 hours. Beads remained bound to the activity (B). Arrows indicate the area of
cell-bead contact. Bar, 5 ␮m (n ⫽ 3).
cell surface and were not engulfed (19). Cells
were then treated with 10% serum for 10 min
to activate Rac1. Trapping GM1-containing

www.sciencemag.org SCIENCE VOL 303 6 FEBRUARY 2004 841

 REPORTS
domains at the surface of suspended cells
maintained Rac1 localization at the plasma
membrane (Fig. 4A) and PAK activation
(Fig. 4B). Control TfR-beads bound to cells
but had no effect. CTxB-beads had no effect
on Rac1 GTP loading (fig. S5B). Additional-
ly, no integrin ␤1 staining was detected at the
bead surface (fig. S5C). These experiments
demonstrate that loss of GM1-containing do-
mains from the cell surface in suspended cells
is required for the loss of Rac1 targeting and
subsequent effector activation.
Rac1 association with the plasma mem-
brane and activation of effectors require mem-
brane binding sites that are controlled by inte-
grins (3, 4). These binding sites are components
of cholesterol-rich membrane domains. Integrin-
mediated adhesion maintains membrane do-
mains at the plasma membrane. When cells are
detached, domains are cleared from the cell
surface through internalization. Preventing in-
ternalization maintains Rac1 plasma membrane
localization and Rac1 signaling in suspended
cells. Although selectivity of Rac1 for mem-
brane domains is unexpectedly determined to
some extent by the state of the lipids them-
selves, it is unlikely that lipids alone completely
account for this effect. This effect may also
provide a means by which adhesion can influ-
ence many growth factor pathways that are
dependent on integrins (1) to confer anchorage
dependence of growth. Local regulation of
membrane domains by integrins may explain
their ability to locally regulate Rac1 targeting
(4) and to recruit many signaling proteins
thought to associate with domains (23). Regu-
lation of Rac1 localization by integrins is likely
to be important for cell migration and polarity
in many systems where precise spatiotemporal
control of guanosine triphosphatase function is
crucial. Although fibroblasts provide a good
model for integrin signaling in anchorage-
dependent cells, Rac1 binding sites in specific
membrane domains may diverge between epi-
thelial, mesenchymal, and hematopoietic cells,
where Rac1 function can also differ (24).
References and Notes
1. M. A. Schwartz, M. H. Ginsberg, Nature Cell Biol. 4,
E65 (2002).
2. E. Sahai, C. J. Marshall, Nature Rev. Cancer 2, 133 (2002).
3. M. A. del Pozo, L. S. Price, N. B. Alderson, X. D. Ren,
M. A. Schwartz, EMBO J. 19, 2008 (2000).
4. M. A. del Pozo et al., Nature Cell Biol. 4, 232 (2002).
5. K. Simons, D. Toomre, Nature Rev. Mol. Cell Biol. 1,
31 (2000).
6. D. A. Brown, E. London, J. Biol. Chem. 275, 17221 (2000).
7. F. R. Maxfield, Curr. Opin. Cell Biol. 14, 483 (2002).
8. I. A. Prior, C. Muncke, R. G. Parton, J. F. Hancock,
J. Cell Biol. 160, 165 (2003).
9. C. Dietrich, B. Yang, T. Fujiwara, A. Kusumi, K. Jacob-
son, Biophys. J. 82, 274 (2002).
10. M. Edidin, Annu. Rev. Biophys. Biomol. Struct. 32, 257
(2003).
11. C. Gómez-Mouton et al., Proc. Natl. Acad. Sci. U.S.A.
98, 9642 (2001).
12. R. G. Parton, J. Histochem. Cytochem. 42, 155 (1994).
13. R. G. Anderson, Proc. Natl. Acad. Sci. U.S.A. 90,
10909 (1993).
842 6 FEBRUARY 2004 VOL 303 SCIENCE www.sciencemag.org
14. P. A. Michaely, C. Mineo, Y. S. Ying, R. G. Anderson, RhoGDI vectors; I. S. Trowbridge for H68 antibody; and P.
J. Biol. Chem. 274, 21430 (1999). Liu, M. Zhu, S. Revak, and E. M. Moreno for advice and
15. H. Kumanogoh, S. Miyata, Y. Sokawa, S. Maekawa, technical assistance. This work was supported by a Lady
Neurosci. Res. 39, 189 (2001). Tata Memorial Trust International Award for Research in
16. S. Grimmer, B. Van Deurs, K. Sandvig, J. Cell Sci. 115, Leukemia, a Special Fellow Award from the Leukemia and
2953 (2002). Lymphoma Society of America, and MCYT (Ministerio de
17. Materials and methods are available as supporting Ciencia y Tecnologı́a) grant SAF2002-02425 to M.A.d.P.;
material on Science Online. U.S. Public Health Service grant RO1 GM47214 to M.A.S.;
18. A. Radhakrishnan, T. G. Anderson, H. M. McConnell, and grants from NIH (HL 20948, GM52016) and the Perot
Proc. Natl. Acad. Sci. U.S.A. 97, 12422 (2000). Family Foundation to R.G.W.A. This is manuscript no.
19. M. A. del Pozo, data not shown. 15179-VB from The Scripps Research Institute.
20. B. J. Nichols, Curr. Biol. 13, 686 (2003).
Supporting Online Material
21. R. Montesano, J. Roth, A. Robert, L. Orci, Nature 296,
651 (1982). www.sciencemag.org/cgi/content/full/303/5659/839/DC1
22. M. Holtta-Vuori, K. Tanhuanpaa, W. Mobius, P. Som- Materials and Methods
erharju, E. Ikonen, Mol. Biol. Cell 13, 3107 (2002). SOM Text
Figs. S1 to S5
23. S. Miyamoto, S. K. Akiyama, K. M. Yamada, Science
267, 883 (1995). References and Notes
24. K. Simons, E. Ikonen, Science 290, 1721 (2000).
25. We thank P. Read and R. Nakamoto for the Rac1 and 14 October 2003; accepted 23 December 2003
Large Shifts in Pathogen
Virulence Relate to Host
Population Structure
M. Boots,1* P. J. Hudson,2 A. Sasaki3
Theory on the evolution of virulence generally predicts selection for an optimal
level of virulence determined by trade-offs with transmission and/or recovery. Here
we consider the evolution of pathogen virulence in hosts who acquire long-lived
immunity and live in a spatially structured population. We show theoretically that
large shifts in virulence may occur in pathogen populations as a result of a bistability
in evolutionary dynamics caused by the local contact or social population structure
of the host. This model provides an explanation for the rapid emergence of the
highly virulent strains of rabbit hemorrhagic disease virus.
Over the past 30 years, emerging diseases General theory on the evolution of viru-
have caused unexpected and, in some lo- lence (the death rate due to infection) is
calities, significant human mortality (1). focused on the maximization of the epidemi-
This increase in the prevalence of novel ological basic reproductive number of the
diseases has generally been associated with pathogen R0 (3); single infections in com-
anthropogenic changes of the environment, pletely mixed host populations should evolve
such as a change in farming practices or in a manner that maximizes R0. In turn, this
urbanization, as well as the zoonotic trans- suggests that the evolutionarily stable (ES)
fer of pathogens from wildlife to humans transmission rate will be the maximum pos-
(2). However, there is increasing concern sible, the recovery rate the lowest possible,
that some pathogens may emerge as a con- and the ES virulence the minimum possible.
sequence of evolutionary changes in viru- Virulence can, therefore, be seen as a conse-
lence. Here we show that rapid evolution of quence of trade-offs with transmission and
virulence can occur as a consequence of recovery, due to underlying mechanisms as-
bistability in the evolutionary dynamics of sociated with factors such as pathogen repli-
pathogens associated with changes in host cation rates (4). Specifically, a finite ES vir-
social structure. Evidence from molecular ulence will occur when fitness benefits to the
epidemiology studies leads us to suppose parasite in terms of increased transmission or
that this may have occurred in the emer- decreased recovery rates become increasingly
gence of the virulent pathogen rabbit hem- costly in terms of increased virulence. Evo-
orrhagic disease virus (RHDV). lution toward an evolutionarily stable strate-
gy (ESS) with higher virulence (determined
by a trade-off relationship) would typically
1
Department of Animal and Plant Sciences, University of
be gradual, and we would not expect the rapid
Sheffield, Western Bank, Sheffield S10 2TN, UK. 2Depart-
ment of Biology, Mueller Lab, Penn State University, Uni-
emergence of a highly virulent strain.
versity Park, PA 16802, USA. 3Laboratory of Mathematical One important assumption within these
Biology, Kyushu University, Higashi-ku, Fukuoka-shi, Japan. classical virulence models is that the host
*To whom correspondence should be addressed. E- population is free-mixing, whereas in nature
mail: m.boots@sheffield.ac.uk hosts typically live in spatially structured