TREE vol. 5, no. IO, October
1990
ParasitbHost Coevolution
Catherine A, Toft and Andrew J. Karter
Parasite-host coevofution can have many
different endpoints, not simply the commensafism of ‘conventional wisdom’. Empirical studies and mathematical
models
are elucidating the conditions under which
parasite-host systems can coevolve to intermediate and high levels of parasite virulence - and when they can coevolve to
commensalism and mutualism.
The conventional
wisdom of parasitology and medicine
is that parasites evolve to do little or no harm to
hosts. While this view is still prevalent in many circles, it has been
repeatedly
challenged
in the last
dozen years’-7. Although
the conventional
wisdom appeals to common sense and is supported
by
examples,
recent efforts’,5,R,9 have
explored
mechanisms
beyond
the
naive group-selectionist
arguments
that have satisfied
earlier authors.
Meanwhile,
the study of coevolution has made us moreaware
of the
costs, benefits
and trade-offs
involved
in all coevolved
relationships, including both parasitism and
mutualisml~9~‘0.
Lately,
ecologists
and evolutionary
biologists
have
considered
a wider
range of orthan
have
traditional
ganisms
parasitologists8,“,‘2,
yielding
additional
insights
about
the many
pathways that coevolution
can take.
In sum, these studies
reveal that
many coevolutionary
outcomes
are
possible,
mainly
because
of constraints
specific
to the organisms
involved
that direct
or stop the
trajectories
of coevolutionlO.
By ‘coevolution’,
we mean that
two species undergogenetic
change
in response
to the interaction
between
them13, and here we will
specifically
consider
only the coevolution
of virulence. Virulence is a
complex
feature
of parasite-host
interactions.
It may appear
to be
solely an attribute
of the parasite,
but, in fact, virulence and resistance
are each the net effect
of the
physiological,
morphological
and
behavioral
interactions
between
parasite and host.
Virulence,
then.
can be conCatherineToft is at the Dept of Zoology,University
of California,Davis,CA 95616, USA;Andrew Karter is
at the Dept of Epidemiology and Preventive Medicine, School of Veterinary Medicine, University of
California, Davis, CA 956 16, USA
326
sidered
the net effect of parasites
on their hosts, and can be measured
in a number of ways. For example, it
can be estimated
by parasite reproduction, by parasite infectiousness
or by the damage a parasite does to
the host, i.e. the disease’“.
In recent
modeling
efforts (see below), virulence often (but not always15l refers to host death caused directly by
the parasite4-5. However, all studies
emphasize
that there are complex
interrelations
between
pathology
leading
to parasite-induced
host
deaths
and benefits
accrued
to
the parasite from its interaction
with
the host - primarily
transmission
and reproduction5,9.
These
interrelations
have been explored
explicitly
in only a few instances;
unfortunately,
we know all too little
about how parasite virulence, transmission and fecundity
interact with
host resistance in real parasite-host
systems’“.
Evolutionary change in virulence
A variety
of examples
demonstrate (or suggest) that virulence can
change
through
coevolution.
Importantly,
examples
exist for an
array of evolutionary
outcomes3~5
(Boxes I and 2). Coevolution
has
been clearly demonstrated
in only
a few empirical
studies,
the best
being
that of the Myxoma
virus
infecting
rabbits
in Australia
and
Britain17. While many other documented cases provide
evidence
of
evolutionary
change3,7,‘b, we often
do not
know
the
physiological
mechanisms
for virulence
or resistance, much less the details of coevolutionary
change.
The discrepancy
between
parasite and host generation
times has
suggested to some that coevolution
is not as important
as the parasite’s adaptations
to the host. The
Myxoma-rabbit
story
contradicts
this reasoning, however, because in
less than 20 years rabbits showed a
clear evolutionary
change in resistance17, albeit not as much as that in
the virus’s virulence.
(Note that real
changes in virulence
and resistance
in the field were estimated
from
reference
test stocks kept in the
laboratory.)
Thus we assume
for
other examples
that coevolutionary
change may well have occurred and
that virulence
and resistance
are
not sharply separable. Nevertheless,
from here on, we will for simplicity
emphasize
virulence
(taking
the
parasite’s
point
of
view),
but
readers should keep these caveats
in mind.
We now review
case studies
under
the three
qualitative
outcomes of evolution
of virulence:
virulence
decreases
to ‘intermediate’ levels; virulence
increases from
some starting point; and ‘virulence’
decreases
to zero (commensalism)
or
even
becomes
positive
lmutualism
I.
Virulence settles to ‘iNtermediate’ levels
The best-documented
cases so
far fall under this evolutionary
outcome Imyxomatosis;
avian malaria,
Box I ; but see Box 2). We put quotes
around
‘intermediate’
because we
will be emphasizing
relative
rather
than absolute
measures
of virulence. ‘Intermediate’
is appropriate
to describe
these
examples
because
virulence
decreased
over
time but did not decrease to zero
lthe commensalism
of conventional
wisdom I.
The Myxoma-rabbit
system, familiar by now to most ecologists
and
evolutionary
biologists,
settled into
intermediate
levels
of virulence
within a year or so and has remained
roughly
steady
for two decades.
However,
an equilibrium
may not
yet have been reached, and some
of the fluctuations
in prevalence
of various
strains
may represent
evolution
of increased resistance in
rabbits,
or introduction
of new
vector+“.
Less well-studied
but nonetheless
documented
examples
in
humans are diseases, such as smallpox and measles,
that are much
more
virulent
in populations
to
which they have been recently
inwhere
troduced than in populations
there
has been
long exposureb.
Similarly,
‘zoonoses’
(diseases
introduced
to humans from wild and
domestic
animals) are often severe
or lethal in humans and milder
in
the ‘reservoir’
(animal)
host, although much of the evidence
comprises anecdotes
from the medical
literature
(e.g. rabies, Lassa fever,
bubonic
plague, and perhaps
HIV;
see also Box 21. These diseases generally are not completely
benign in
the animal,
i.e. more coevolved,
hosts. For none of these examples
besides
the Myxoma-rabbit
case is
it known what constrains
the viru-
rREE vol. 5, no. 10, October
1990
lence in the low direction,
i.e. preventing
it from reaching commensalism. However,
the dependency
Df transmission
on factors leading to
host pathology
and death is a likely
explanation,
and one that has rezeived most attention.
Virulence settles to ‘fiig& levels
Once again, ‘high’ is a relative
neasure;
we refer here to cases
Nhere
virulence
increased
over
:ime. Ewald3 argues that in vector3orne parasites - including
malaria
disease
Plasmodium),
Chagas’
Trypanosoma I, Leishmania,
relapsng fever,
Rickettsia
and yellow
,‘ever (but see Box I I - the cost of
:3athogenicity
causing host immolbility is removed.
He argues that
*:here is no penalty
(and perhaps
tzven an advantage)
for these para&sites in increasing their exploitation
of the host until it is immobilized,
because a vector can still find the
host, unlike in directly
transmitted
parasites. However, some examples
I including
the Myxoma-rabbit
sysI em15; see also Box 2 I do not corroborate this idea, and we might expect
each vector-primary
host-parasite
system
to be slightly
different,
depending
on biological
details.
Nevertheless,
in some host-paraGte systems, the upper constraint
IO virulence
might be removed, and
in those instances, virulence
could
,:ontinue to increase.
‘lirulence goes to zero: tommeflsalistic
,rnd mutualistic endpoints
Alternatively,
the
lower
conr-,traint to virulence can be removed,
‘:lnd those
parasite-host
systems
evolving
lower virulence
or greater
resistance
have no barrier
along
the
parasitism-commensalismmutualism
continuum’.
As Thomp:.on’O has pointed
out, mutualisms
itIre not the happy,
‘peaceful’
arrangements
they are conventionally
thought
to be. Many are so full of
manipulation,
costs and even antagonism that they are suspected
to
have
arisen
evolutionarily
from
parasitism.
These include
the fig
wasp pollination
system,
lichens,
2nd
mycorrhizal
fungi.
Similarly,
c.ommensalisms
may not represent
‘peaceful’
coexistence
either,
but
rather a stand-off between
parasite
ii ttack and host counterattack;
in few
instances,
however,
have costs to
Posts and parasites been estimated.
Evidence from the new molecular
and other experimental
techniques
unequivocally
demonstrates
that
mitochondria,
chloroplasts
and bacterial
plasmids
are associations
between
organisms
of different
speciesls. Mitochondria
and chloroplasts may have arisen from originally
parasitic
relationships18,
although tests of this postulate
are
difficult to imagine. Studies on bacteria and their phages and plasmids, however,
have been quite
fruitful, because of their short generation
times
and
evolutionary
plasticity’9-2’.
In bacteria and their
viruses,
evidence
suggests
that
evolution
has occurred
independently in both directions
a number
of times’9-2’.
Thus, perhaps ironically,
the conventional
wisdom now under attack
may be correct
about
the evolution of the association
between
such microorganisms,
which
was
responsible
for the evolution
of
eukaryote@
and indeed life as we
know it on this planet. Because the
bacterial
phage-plasmid
pathway
can take place in a few days in the
laboratory,
it may represent
the
best experimental
model to determine what conditions
can overrule
the ‘lower’ constraint
to parasitehost coevolution
i9-21 (inasmuch
as
these prokaryote
systems can be
generalized
to multicellular,
eukaryote
systems).
Experiments
in
progress suggest that the key lies
in the nature
of transmission
(R.
Lenski,
pers. commun.1,
just as
models assert.
Models of parasite-host coevolution
Models
of parasite-host
coevolution have provided
significant
insights
into
the constraints
that
direct and impede the coevolutionary trajectories
of virulence
and
resistance,
and
their
results
in
turn guide further empirical
study.
These early efforts corroborate
the
examples we have provided
above:
multiple
coevolutionary
endpoints
are possible.
In model
systems,
any of the
following
conditions
can (theoretically)
result
in the reduction
of
virulence
to commensalism.
First,
if parasite-induced
host
mortality,
transmission,
host recovery and perhaps
other factors are
not interrelated,
then there is no
theoretical
lower constraint
on virulence. Because the basic parasite
reproductive
rate, R,, is inversely
related
to parasite-induced
host
mortality
(virulence),
R, is maximized by reducing virulence to zero
- assuming that there are no interactions
between
virulence
and
other parameters4,5.
Second, if resistance
has no cost
to the host, it can counter the parasite’s
attack
(virulence)
with
no
limit, and thus the parasite’s
net
effect on the host is negligible9,22.
Examples
of situations
matching
these conditions
may include some
gut symbionts.
Third, in certain modes of transmission, such as vertical transmission (i.e. from parent to offspring),
large benefits accrue to the parasite
by increased host reproduction, and
this will select for lower virulence in
model systems7.
Fourth, if group, or perhaps more
selection
accurately
interdemic,
should predominate
over individual selection, then the lower constraint on virulence is removed or
lessened; i.e. parasites can be prudent” , as the conventional wisdom
holds them to be. lnterdemic selection is not improbable in parasites.
Structured habitats with limited immigration can lead to group-type
selection in model systems2?, which
could theoretically result in lowered
virulence.
However,
others have
postulated the opposite outcome patterns
of increased
virulence
with periodic disease outbreaks from structured habitatsI (see also
Box 21.
If any of the above conditions do
327
TREE vol. 5, no. 10, October
1990
no t a p p ly, virule nc e
will b e c o nstra ine d fro m re a c hing ze ro . Exa c t
p re d ic tio ns on whe re virule nc e will
se ttle a re p o ssib le fro m se ve ra l o f
the a b o ve mo d e ls. Two c o nstra ints
o n d e c re a sing virule nc e a re mo st
c o mmo nly
inve stig a te d
in mo d e l
syste ms, a nd ind e e d a re like ly to b e
c o mmo n in re a l syste ms.
The first is whe n virule nc e is a
func tio n o f e ithe r tra nsmissio n
or
re c o ve ry ra te . Virule nc e a nd re c o ve ry ra te we re e stima te d
in the
Myxo ma -ra b b it
syste m a nd fo und
to b e no nline a rly re la te d” . The e stima te s we re the n p ut into a simp le
mo d e l to d e te rmine
the virule nc e
le ve l tha t the o re tic a lly
ma ximize s
p a ra site
re p ro d uc tive
ra te . This
le ve l wa s re ma rka b ly c lo se to the
p re va le nt virule nc e le ve l o f the myxo ma virus in Austra lia a nd Brita in4,
p ro vid ing stro ng e vid e nc e fo r suc h a
c o nstra int.
Se c o nd , the costs o f virule nc e a nd
re sista nc e o fte n le a d to inte rme d ia te le ve ls o f virule nc e in mo d e l syste ms9,22: the c o sts p la c e a n up p e r
c o nstra int o n virule nc e , a nd a n up p e r c o nstra int o n re sista nc e (whic h
tra nsla te s into a lo we r ‘ c o nstra int’
o n virule nc e ). To te st the se id e a s,
c o sts c o uld b e e stima te d fro m la b o ra to ry stud ie s a nd use d to c a lc ula te
o p timum
inve stme nts
in
virule nc e
a nd in re sista nc e , fro m
the p a ra site ’ s a nd ho st’ s p o ints o f
vie w re sp e c tive ly. Suc h c a lc ula tio ns
mig ht re ve a l tra d e -o ffs tha t c o uld
re sult in intermediate
levels of virusyslence in so me p a ra site -ho st
te ms, suc h a s in the Myxo ma -ra b b it
e xa mp le .
Ma ny o f the mo d e ls me ntio ne d
a b o ve c o uld be used to estimate
the value that virulence
will reach
whe n p a ra me te rs e stima te d
fro m
re a l syste ms
a re p ut into
the
mo d e ls. Whe the r virule nc e is ‘ lo w’ ,
‘ inte rme d ia te ’
o r ‘ hig h’ wo uld d e - p e nd o n the p a rtic ula rs.
Othe r mo d e ls a sk whe the r g e ne tic p o lymo rp hisms
in virule nc e
a nd
re sista nc e
c an be
ma inta ine d 4,5,9,24, a s o p p o se d to a sking
wha t the a ve ra g e le ve l o f virule nc e
mig ht b e fo r a g ive n p o p ula tio n the se
a re d iffe re nt
b ut re la te d
q ue stio ns. Whe n g e ne tic p o lymo rp hism in virule nc e / re sista nc e
oc c urs, the n susc e p tib ility
is no t fixe d
(virule nc e ma ximum), no r is re sista nc e fixe d
(virule nc e
minimum I.
Po p ula tio ns
o f p a ra site s ma y b e
mixture s o f c lo ne s o f d iffe re nt viru328
‘FREE vol. 5, no. 70, Ocrober
1990
ience levels,
as in the Myxomarabbit case4. If any polymorphism
exists in host resistance,
we may
assume that total resistance doesn’t
pay
evolutionarily,
perhaps
because of costs of resistance.
At
i he same time, the costs of excessive virulence
accrue to parasites,
usually
in reduced
transmission.
Hence,
polymorphisms
are more
likely when both virulence
and re!;istance have some kind of cost9,22.
modeling
Thus,
efforts
have
revealed
a variety
of conditions
directing
the trajectories
of coevolution between
parasite
and host.
.rheir predictions
are easily testable
(at least in principle),
and
c;hould help us to understand
the
evolutionary
dynamics
of many
parasite-host
systems. Importantly,
commensalism
is not the only out,;:ome predicted
by models.
I:uture directions
Lastly, we briefly
mention
some
obvious
next
steps
to increase
(:)ur understanding
of parasite-host
interactions.
First, we need a family of theoretical
models
that
bridge
the
interface’,
r)arasitism-mutualism
,:Icross which parameters can change.
-*bus, parasite-host
systems may be
,:lble to reach commensalistic
and
rnutualistic
endpoints
even if par,rlmeters are linked,
but linked
in
‘ways different from those leading to
i:ln endpoint
of intermediate
viruience. Work on bacterial
plasmids
and phages may inspire
the first
&tempts
in this
area,
because
I:hese viruses evolve quickly
backand-forth
across
the parasitismmutualism
interface19.
Second,
we need
more
work
on
coevolution
of multicellular
‘nacroparasites’,
such as helminths.
Macroparasites
are invariably
ag6;regated
in the host population;
that is, most hosts have few or no
parasites
and relatively
few hosts
carry high parasite
burdens16. This
aggregated
distribution,
which has
not yet been fully explained,
probably derives from heterogeneity
in
the parasite and host populations,
ii attributes
such as transmission
rates, host immunity
and host susc.eptibility,
as well as in temporal
and spatial patterns. Because variability
in host susceptibility
is at
Ilzast partly genetic, the aggregated
distribution
bears directly
on coc,volution
between
macroparasites
and hosts, and constitutes
a major
unexplored
area.
Third, coevolutionary
models so
far have typically
taken the simplest
form, lacking heterogeneity
in time
(i.e. age structure) and in space. As
we have already mentioned,
spatial
structuring
is crucial
to whether
group selection can occur15, and the
role of group selection
is still an
open question
in parasite-host
coevolution9,‘9. Also, spatially complex
patterns
in virulence,
such as in
Trypanosoma
(Box 21, suggest that
(group selection
or not1 conditions
affecting
coevolution
of virulence
vary spatially
and this variation
is
worth pursuing in modeling and empirical
studies.
Such efforts
have
been initiated
for plant-funga125 and
the Myxom+rabbit15
systems, and
reveal
that temporal
and spatial
structuring
was important
in their
coevolution.
Fourth, interactions
among mortality factors are of potential
importance in parasite-host
systems, and
such interactions
have
received
little attention.
For example,
individuals weakened
by disease might
be easier for predators
to capture2h,
and studies in humans have linked
resistance
to host nutritional
state.
Interactions
among parameters
can
generate
complicated
population
behaviors2’,
and would surely affect
selection
pressures
on parasitehost interactions.
Also, effects are
likely to be different
in intermediate and definitive
hosts (depending
on mode of transmission).
Further
study explicitly
exploring the effects
of interactions
among mortality
factors is warranted.
Finally, far more empirical
work
on parasite-host
associations
is
needed.
In particular,
collaboration
among specialists
in ecology, physiimmunology,
systematics,
ology,
epidemiology
and medicine,
including laboratory,
field and theoretical approaches,
is necessary
to
understand
parasite-host
coevolution in any detail’QR.
Acknowledgements
T. Carlton
helped
extensively
in finding
references
on Trypanosoma.
0. Halvorsen. S.
Levin. R. May, A. Skorping and 1. Theis made
useful comments
on the manuscript.
References
1 Levin.
S. and Pimentel,
D. (1981 I Am. Nat.
Il7,308-315
2 Levin, B.R. eta/
119821 in Population
Biology of Infectious
Diseases IMay, R.M.
and Anderson,
R.M., eds), pp. 213-244,
Springer-Verlag
3 Ewald. P.W. 11983) Annu. Rev. Ecol. Syst.
I4.465-485
4 May, R.M. and Anderson,
R.M. t 1983) Proc.
R. Sot. London Ser. B 2 19, 2 18-3I3
5 May, R.M. and Anderson,
R.M ( 19831 in
Coevolution
f Futyuma, D.I. and Slatkin. M..
eds), pp. 186-206, Sinauer Associates
6 Johnson, R. II9861 Med. Hypotheses
19, 341
7 Ewald, P.W. 11987) Ann. N.Y. Acad. Sci.
503, 295-306
8 Gill, D.E. and Mock, B.A. f 19851 in Ecology
and Genetics of Host-Parasite
Interactions
IRollinson,
D. and Anderson,
R.M., edsf, pp.
157-183, Academic
Press
9 Levin, S.A. (19831 in Coevolution
INitecki.
M.H.. ed.1, pp. 21-65. University
of Chicago
Press
IO Thompson,
I.N. II9861 Trends Ecol. Evol.
I, 105-107
I I Price, P.W. ( I9801 Evolutionary
Biology of
Parasites, Princeton
University
Press
I2 Toft. C.A. ( 19861 in Community
Ecology
(Diamond.
1. and Case. T.I.. edsl. pp
445-463, Harper & Row
I3 Thompson,
I.N. II9891 Trends Ecol. Evol.
4, 179-183
I4 Bremermann.
H.f. and Pickering, I. (I9831
1. Theor. Bio/. 100, 4 I l-426
I5 Dwyer, G.. Levin, S.A. and Buttel, L. Ecol.
Monogr. fin press)
I6 Dobson, A. and Merenlender,
A. in
Parasitism:
Coexistence
or Conflict? IToft.
C.A.. Aeschlimann,
A. and Bolis, L.., eds),
Oxford University
Press (in press)
I7 Fenner. F. ( 19831 froc. R. Sot. London
Ser. B 2 18, 259-285
I8 Cavalier-Smith,
T. II9871 Ann. N.Y Acad.
Sci. 503, 17-54
I9 Levin, B.R. and Lenski, R E. II9851 in
Ecology and Genetics of Host-Parasite
Interactions
(Rollinson.
D. and Anderson,
R.M.. edsf, pp. 227-242, Academic Press
20 Bouma, I.E. and Lenski, R.E. II9881
Nature 335, 35 l-352
21 Rosqvist, R., Skurnik. M and Wolf-Watz,
H. (I9881 Nature 334, 522-524
22 Gillespie,
J.H. ( 19751 &co/ogy 56. 493-495
23 Wilson, D.S. (19801 The Natural Selection
of Populations
and Communities,
Benjamin/
Cummings
Publishing
24 Cohen, I.E. and Newman,
Nat. 133, 533-552
C.M. II9891
Am.
25 Burden, 1.1.and Jarosz, A.M. (19881 Philos.
Trans. R. Sot. London Ser. B 23 I, 349-36 I
26 Holmes, I.C. II9821 in Population
Biology
of Infectious
Diseases (Anderson,
R.M. and
May, R.M.. edsl, pp. 37-52, Springer-Verlag
27 May, R.M. and Anderson,
R M. (19791
Nature 280.455-461
28 Holmes, l.c. (1987) Int / Parasitol.
17.
203-208
29 van Riper, Ill, C, van Riper, S.G., Coff.
M.L. and Laird, M t 19861 Ecol. Monogr. 56,
327-344
and
30 Baker, J.R. I 19741 in Trypanosomiasis
Leishmaniasis
with Special Reference
to
Chagas’ Disease (Lumsden,
W.H.R.. ed.1. pp.
29-50. Elsevier
31 Gibson, W.C., de Marshall, T.F. and
Godfrey,
D.G. (I9801 Adv. Parasjtol.
18,
175-246
32 Mehlitz,
D., Zillman, U. and Sachs, R.
I I985 1 Int Sci. Count. Trypanosomiasis
Res.
Control I 13, 128-l 29
33 Lake, LA.. de la Cruz. V.F.. Ferreira. P.C.,
Morel, C. and Simpson, L II9881 Proc. Nat/
Acad. Sci. USA 85, 4779-4783
34 Gardiner.
P.R and Wilson, A.I. II9871
Parasitol. Today 3. 49-52
329