TIMI 1303 No.

of Pages 10

Special Issue: Microbial Endurance

Review
Why Be Temperate: Lessons
from Bacteriophage l
Sylvain Gandon1,*
Many pathogens have evolved the ability to induce latent infections of their
Trends
hosts. The bacteriophage l is a classical model for exploring the regulation and
Bacteriophage l is the archetypal tem-
the evolution of latency. Here, I review recent experimental studies on phage perate phage. The study of the genetic
l that identify specific conditions promoting the evolution of lysogenic life switch between lysis and lysogeny in l
led to a deep understanding of the
cycles. In addition, I present specific adaptations of phage l that allow this molecular mechanisms underlying
virus to react plastically to variations in the environment and to reactivate its lytic gene regulation in other organisms.
life cycle. All of these different examples are discussed in the light of evolution-
An understanding of the regulation of
ary epidemiology theory to disentangle the different evolutionary forces acting viral latency in l offers unique oppor-
on temperate phages. Understanding phage l adaptations yield important tunities to explore experimentally and
theoretically the adaptive nature of
insights into the evolution of latency in other microbes, including several life-
fixed and plastic latency in different
threatening human pathogens. environments.

Understanding the evolution of lyso-

Phage Life Cycles geny in l can help to build up a com-
Some pathogens exploit their host for a very limited amount of time. In humans, for instance, prehensive framework to study the
evolution of latency in other infectious
most influenza virus infections last for less than 10 days [1]. The duration of the infection is
diseases, including many life-threaten-
generally limited by host immunity but it can also be reduced by the death of the infected host. In ing human pathogens.
contrast, other pathogens can remain in their host for a very long time in a latent state. Many
human viruses have adopted this alternative life history strategy. For instance, infections by
herpes simplex viruses lead to short acute and usually localized infections of the skin, yielding
durable infection of nearby sensory neurons [2]. These latent infections can later reactivate to
cause recurrent disease and transmission. Hence, latency can guarantee lifetime persistence of
the infection and constitutes a major therapeutic challenge in several human pathogens [3–6].
Yet, the adoption of a latent life cycle is a dramatic feature that requires specific pathogen
adaptations. When should pathogens adopt such latent life history strategies? When the
pathogens are in a latent state, when should they reactivate and attempt to reach another host?

These fundamental questions are relevant for a broad range of pathogens but are often difficult
to tackle experimentally. Bacteriophages, however, provide a powerful model system to explore
these questions. Bacteriophages are viruses that infect bacteria and they are one of the most
abundant organisms in the biosphere [7–9]. They can adopt very different exploitation strategies
of their host, ranging from lytic to lysogenic life cycles (Figure 1, Key Figure). Lytic life cycles are
characterized by bacterial adsorption, replication of the phage genome, synthesis of new viral 1
CEFE UMR 5175, CNRS – Université
particles, and bacterial lysis. Lysogeny, however, is characterized by the integration of the virus de Montpellier, Université Paul-Valéry
Montpellier, EPHE, 1919, route de
as a prophage in the bacterial genome and the indefinite persistence of the infection. Lysogeny Mende, 34293 Montpellier Cedex 5,
was discovered in bacteriophages by Bail and Bordet 90 years ago [10–12]. The dichotomy France
between these two alternative lifestyles provides a fantastic opportunity to study the evolution of
latency in microbes. Bacteriophage l [13] is a temperate coliphage that has been involved in
*Correspondence:
major molecular biology advances [14]. In particular, it has been used as a classical model Sylvain.GANDON@cefe.cnrs.fr
system to unveil the regulation of lysogeny [15]. (S. Gandon).

Trends in Microbiology, Month Year, Vol. xx, No. yy http://dx.doi.org/10.1016/j.tim.2016.02.008 1

© 2016 Elsevier Ltd. All rights reserved.
 TIMI 1303 No. of Pages 10

Key Figure
Schematic Representation of (A) a Lytic and (B) a Lysogenic Bacterioph-
age Life Cycle

(A) Free viruses

Lyc life cycle

(B) Free viruses

Lysogenic life cycle

1–φ Lysis

α
φ
Lysogeny

Prophage

Figure 1. In both cases the infection starts with the adsorption of a free virus particle to[2_TD$IF] a susceptible [3_TD$IF]bacterium.. In virulent
phages, such as T-even coliphages (lytic life cycle), the infection leads rapidly to lysis and to the release of free virus particles.
In temperate phages, such as bacteriophage l (lysogenic life cycle), the infection can either lead to the integration of the
prophage in the bacterial genome (with probability f) or to lysis (with probability 1– f). The prophage can either be
transmitted vertically with the replication of the lysogenic bacteria or reactivate the lysis (at a rate /) and release free virus
particles.

The regulation of lysogenisation is mainly under the control of the repressor gene cI. The
repressor protein turns off the expression of other genes of the phage but activates its own
transcription. This prevents the expression of the lytic life cycle and allows the lysogenic bacteria
to divide and to transmit the prophage vertically. Other phage genes are also involved in this
regulation. The gene cro is an antagonist of cI, and its expression is associated with the
expression of other phage genes leading to the launch of a productive infection and the lysis
of the host. The balance between the expression of cI and cro is at the heart of the genetic switch
that allows l to control the fate (lysogeny or lysis) of its host [15]. Interestingly, genetically
identical cells infected with the same number of viruses can give rise to distinct outcomes
(Figure 1). Detailed models of within-host dynamics of the accumulation of gene products of cI
and cro have shown that this binomial variability can be explained by spontaneous biochemical
noise during the development of the infection [16,17]. Yet, this fact does not imply that cell fate is
exclusively stochastic. First, several mutations are known to alter the propensity for lysogenisa-
tion. For instance, the cI857 mutant has a cI repressor that decreases the probability of lysogeny,
f, and increases the induction rate, /. Second, several environmental factors are known to affect
lysogenisation and induction.

2 Trends in Microbiology, Month Year, Vol. xx, No. yy

 TIMI 1303 No. of Pages 10

Earlier studies of l were mainly focused on the molecular details of phage infections. Here, I want
to discuss the biology of l with an evolutionary perspective to question the adaptive nature of
specific features of its life cycle. First, I discuss the adaptive nature of fixed latency in different
environmental conditions. I review two recent experimental studies that explore the effects of (i)
epidemiological dynamics, and (ii) spatial structure on the evolution of viral latency. Second, I
discuss the ability of l to adopt plastic exploitation strategies in response to fluctuations of its
environment. I use theoretical models to explore different drivers of the evolution of pathogen
plasticity: (i) variations in the environment inside the host, and (ii) variations in the environment
outside the host. These theoretical results are discussed in the light of plastic transmission
strategies observed in l. Finally, I discuss the relevance of these studies for the evolution of both
fixed and plastic transmission strategies of other bacteriophages and other pathogens.

Latency as a Fixed Strategy

An understanding of the regulation of lysogeny in l provides invaluable information on the
evolution of latency in pathogens. When should pathogens adopt such latent life-history
strategies? To answer this question, Berngruber et al. [18,19] studied the competition between
two l variants with distinct life history strategies. The wild-type temperate l has a high
lysogenisation rate (f
0.5) and a low induction rate (/
104). In contrast, the virulent
lcI857 is a thermosensitive mutant with a low lysogenisation rate (f < 0.1) and a high induction
rate (/ >103). Monitoring the frequency of these variants provides a measure of selection on
those life-history traits. In other words, it provides a way to determine when latency is favored
and when it is counterselected.

Evolution during Epidemics

During the early stage of a phage epidemic the density of susceptible bacteria is very large, and
free viruses have many opportunities for horizontal transmission. This promotes the spread of
phage variants with more aggressive host exploitation strategies that rapidly get access to a
larger number of hosts through the production of free virus particles. In contrast, at the end of an
epidemic, the availability of susceptible cells is reduced and the probability of finding a new host
to infect is very low. In this case, more prudent host exploitation strategies become adaptive
because they secure the durable exploitation of a host and invest in vertical transmission [20,21].
This argument is akin to the effect of host availability on optimal lysis time in lytic phages [22,23].
But the dynamic nature of the density of susceptible hosts requires models that keep track of
epidemiology and selection to understand and predict the competition between different life-
history strategies [24].

Berngruber et al. [20] formalized the effects of epidemiology on l evolution throughout epi-
demics and tested these predictions with experimental epidemics in chemostats. Each chemo-
stat was seeded with a 1:1 ratio of the temperate wild-type l and the virulent lcI857. These
experiments supported three theoretical predictions (Figure 2). First, the frequency of the virulent
mutant increases rapidly during the early phase of the epidemics and decreases later on.
Second, the evolutionary outcome is governed by the epidemiology. When the experiment is
initiated with a small number of infected cells the availability of a large number of susceptible cells
favors the virulent lcI857 in the early stage of the epidemic. In contrast, when the experiment
starts with a larger fraction of infected cells, the initial benefit of the virulent lcI857 is reduced.
Finally, the experiments confirm that the frequency of the virulence mutant is always higher in the
free virus compartment. The validation of these three qualitative predictions demonstrates
the predictive power of models integrating epidemiology and evolution, and accounting
for the specificities of l life cycle.

One key element of l biology is the ability of lysogenic bacteria to prevent the exploitation of
the host by superinfecting viruses. In other words, superinfection inhibition ensures that the

Trends in Microbiology, Month Year, Vol. xx, No. yy 3

 TIMI 1303 No. of Pages 10

(A) Provirus (B) Free virus

100

1000
Virulent/nonvirulent rao

Virulent/nonvirulent rao

100
10

10
1

0 10 20 30 0 10 20 30
Time (hours) Time (hours)

Figure 2. Virulence Evolution During an Epidemic in Bacteriophage l. (A) Change in the virulent/avirulent ratio in the
provirus stage. (B) Change in the virulent/avirulent ratio in the free virus stage. The initial value of the virulent/avirulent (lcI857/
l) ratio in the provirus was 1:1, and competition was started from two[4_TD$IF] different initial prevalence values: 1% (red) and 100%
(blue). The lines indicate the mean over four replicate chemostats, and the envelopes show the 95% confidence intervals of
the log-transformed data. Modified from Berngruber et al. [18].

prophage controls the fate of its cell. Note, however, that lambdoid phages have evolved the
ability to escape superinfection inhibition, and several immunity groups are circulating. A
phage from a different immunity group is able to escape the repression induced by the
resident phage. This dramatically alters the prediction regarding the evolution of lysogeny
because the coexistence between different immunity groups selects for more virulent
strategies [25].

Evolution in Spatially Structured Environments

The previous section focused on the transient evolutionary dynamics taking place in a well-
mixed environment during an epidemic. Epidemics, however, often spread in spatially struc-
tured environments. Theory predicts that spatial structure generally selects for lower pathogen
virulence [26–28]. Berngruber et al. [19] extended previous models to account for localized
migration of the bacteria and localized transmission of the phage. In addition, this model allows
us to track the contribution of horizontal and vertical transmission of the two phage genotypes.
This helped to identify conditions promoting the evolution of the temperate phage. To test
these theoretical predictions, Berngruber et al. [19] manipulated the amount of mixing among
bacteria and phages and recovered two main theoretical predictions. First, low mixing
promotes the spread of the temperate phage. Indeed, in a spatially structured environment
the virulent phage rapidly runs out of susceptible cells (Figure 3). Second, the relative
contributions of horizontal and vertical transmissions are modulated by spatial structure.
When mixing is high, most of the fitness of the virulent mutant depends on horizontal
transmission. When mixing is low, vertical transmission is key because it allows perpetuation
of the infection. In other words, spatial structure [8_TD$IF]changes the availability of susceptible cells[9_TD$IF]
and empty sites in the [10_TD$IF]local environment of an infection[1_TD$IF]. The magnitude of this effect varies with
the genotype of the parasite and this is what selects for latent (i.e., more prudent) exploitation
strategies[12_TD$IF] [19].

4 Trends in Microbiology, Month Year, Vol. xx, No. yy

 TIMI 1303 No. of Pages 10

0.500
Fitness of virulent λ

0.100

0.050

0.010

0.005

No Low Intermediate High

Mixing

Figure 3. Mixing Selects for Virulence in Bacteriophage l. A 1:1 mixture of temperate l and its virulent mutant lcI857
was used to seed spatial epidemics. Spatial structure was disturbed to different degrees (no mixing, low mixing,
intermediate mixing, and high mixing) and transferred onto a fresh biofilm of susceptible hosts for 3 consecutive days.
Disturbance of the epidemic structure has a strong effect on the competitive fitness of the virulent lcI857. The undisturbed
environment strongly selects against virulence whereas virulence is beneficial in a well-mixed environment. As expected,
intermediate levels of mixing lead to intermediate fitness of lcI857. Modified from Berngruber et al. [19].

Latency as a Plastic Strategy

In the above section, latency is assumed to be a fixed strategy. Bacteriophage l, however, is
also known to exhibit plastic life-history strategies. For instance, the efficacy of lysogenisation
and the induction rate of the prophage can vary with the environment. These conditional
strategies are well studied on a molecular level but the adaptive nature of this plasticity is often
overlooked.

Escaping a Stressful Environment

One of the best studied examples of plasticity is perhaps the effect of stress on the induction
rate, /, of bacteriophage l. Whenever a lysogen undergoes DNA damage due to UV irradiation
or other chemical stress, the protein RecA is activated. RecA is involved in the inactivation
(cleavage) of the LexA repressor and in the launch of the SOS response, activating different
mechanisms of DNA repair. Interestingly, RecA is also involved in the inactivation (cleavage) of
the cI repressor and thus in the launch of the lytic cycle of l. Is this conditional induction rate
adaptive? One way to answer this question is to show that such plasticity may be adaptive under
some theoretical scenario. In Box 1 I developed a simple model where the infected host can be in
two different states. The host can be in a ‘normal’ state in which host cells replicate and have low
mortality, but with a rate s host cells may enter a ‘stressful’ state where both fecundity and
survival are reduced. The model shows that the virus infecting a stressed cell should shift
towards a more intensive exploitation strategy. In other words, this simple model provides an
adaptive explanation for the conditional induction rate of l because the activation of RecA
signals that the host is in a critical state.

Lysogenisation or Not in a Variable Environment

The probability to lysogenize, f, is another life history trait of l modulated by variations in the
environment. In particular, this probability increases when the number of virus particles

Trends in Microbiology, Month Year, Vol. xx, No. yy 5

 TIMI 1303 No. of Pages 10

Box 1. Leaving a Sinking Ship

Let us assume a simple epidemiological model where the host can fall into two different states: (i) a ‘normal’ host with high
fecundity, l, and low mortality, d, and (ii) a ‘stressed’ host with low fecundity, l*, and high mortality, d*. We can model the
dynamics of phage l under these assumptions which yields the following epidemiological model[5_TD$IF] (the dot notation refers
to differentiation with respect to time):
L_ ¼ lL þ faSVðd þ a þ s ÞL
L_  ¼ l L þ sLðd þ a ÞL
V_ ¼ BðaL þ a L Þ þ Bð1fÞaSVmV
where L and L* are the densities of normal and stressed hosts respectively, V is the density of free viruses, and S is the
density of susceptible hosts. The parameters l, f, d, a, m, a, s and B refer to the fecundity of lysogens, the lysogenisation
rate, the natural death rate of lysogens, the adsorbtion rate of the virus, the death rate of free virus, efficiency of the
induction rate of lysogens, the rate at which a bacterium enters the stressed state, and the burst size of a productive
infection ([6_TD$IF]see Figure 1[7_TD$IF] in main text). This simple model can be used to explore the effect of host stress on the evolution of
the parasite exploitation strategy. To study the evolution of conditional lysis strategies, we focus on the ability of a mutant
pathogen (with a strategy am and am  ) to invade a resident pathogen population (with a strategy a and a*) at equilibrium. In
other words, we assume that the density S ^ of susceptible hosts is fixed by the resident strategy of the pathogen. The
dynamics of the mutant can be fully described by the matrix Fm which accounts for how many mutants are created in the
two different types of host and in the environment (as free-living viruses) and the matrix Vm which refers to transition
between these three compartments:
0 1
l 0 aSf ^
F m ¼ @ 0 l 0 A
^
0 0 0m aBSð1fÞ 1
dþa þs 0 0
V ¼
m @ s d þ a 0 A
m

am B am B m

The per-generation growth rate of the mutant is given by the dominant eigenvalue Rm of the next generation matrix [49]:
0 m 1
tL ! L t mL ! L tm
V !L
F m :V m1 ¼ @ t m
L ! L tm
L ! L 0 A
tmL!V tm
L ! V tm
V !V

where tmX ! Y is the per-generation transition between compartment X and Y. For the sake of simplicity, we focus on the
special case where l* = 0. We derive the gradient of selection on conditional lysis:
@Rm ^ ðdðd þ a Þ þ d s Þ  mðd þ a Þl
/ aBSf
@am |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl}
benefit of lysis cost of lysis
@Rm ^  >0
/ saBSfd
@am


In other words, lysis can evolve towards intermediate rates in normal hosts to balance the benefit associated with access
to new hosts and the cost of losing the current host. In stressed hosts, however, lysis is maximized because there is no
benefit associated with delaying lysis in a sterilized and moribund host.

attempting to infect a host cell is high [29,30]. Some of the molecular details governing this
regulation have been largely uncovered and modeled [16,17,31]. Yet, the adaptive nature of this
regulation remains unclear. Why should pathogens coinfecting the same host adopt a latent
strategy? Classical virulence evolution theory, in contrast, often predicts that coinfections should
yield more aggressive exploitation strategies [21]. Such conditional lysogenisation strategy,
however, may be adaptive in some environments. Avlund et al. [32] argue that it may evolve as a
response to the stochasticity of the environment. Indeed, conditional dormancy in plants and
other organisms has often been viewed as a bet-hedging strategy that evolved in random
environments [33,34]. Yet, I want to argue here that conditional latency may also be adaptive in
fully deterministic but temporally variable environments. Indeed, the fact that several viruses are
attempting to enter the same cell indicates that very few hosts are available and, in this case, it
may be advantageous to take care of the host. Lysogenisation may thus be a way to stay in the
host and avoid the perils of the free-living stage of the virus life cycle. Box 2 is an attempt to
formalize the evolution of plastic transmission strategies in a variable environment. Theory shows
that plasticity may evolve when it allows transmission to coincide with a transitory increase in the
availability of susceptible cells. But the model does not specify which cues can be used by the
virus to trigger a life-history switch. The difficulty for the virus is to use local information obtained

6 Trends in Microbiology, Month Year, Vol. xx, No. yy

 TIMI 1303 No. of Pages 10

Box 2. Plastic Transmission Strategies in Variable Environments Outstanding Questions

Fluctuation in the quality of the environment is modeled with uðtÞ which describes the periodic influx of susceptible cells in The details of the genetic switch in l
the population. The epidemiological model introduced in Box 1 becomes: have been unveiled. Is it also possible
to disentangle the molecular mecha-
S_ ¼ uðtÞaSVdS
L_ ¼ lL þ faSVðd þ aÞL nisms underlying the regulation of
V_ ¼ BaL þ Bð1fÞaSVmV latency in other pathogens?

Under the assumption that the dynamics of the free virus compartment is much faster than the other processes, we can What is the amount of natural variation in
assume that the density of free viruses is always at this equilibrium: the propensity to develop fixed or plastic
BaL latency among pathogens? [14_TD$IF]Is it possible
^¼
V
mBð1fÞaS to relate this variation to specific features
of their local environment?
For the sake of simplicity, we focus on the case where f = 1. The dynamics of a mutant virus may thus be described by:
 
faBSam How do stochastic and deterministic
L_ m ¼ l þ ðd þ am Þ Lm
m variations in the environment interact
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
rm with the evolution of latency?

If the system oscillates periodically, the mutant will invade if its average growth rate over one period T (noted with a tilde) is Is it possible to demonstrate experi-
positive: mentally the adaptive nature of plastic
Z
1 T ~ am   latency in the laboratory?
~r m ¼ ~ þ faBS~
r m dt ¼ l  dþa ~m Þ þ
faB ~ a
cov S; ~m Þ > 0
T 0 m m
Do different pathogen species share
The covariance in the above condition captures what may select for plasticity [43]. A strategy building up a covariance similar environmental cues that trigger
between the timing of lysis and the high density of susceptible hosts is going to be selected because it synchronizes the relapses?
release of free virus with the availability of susceptible hosts. Multiplicity of infection may be used by the virus to build up
this covariation because it carries valuable information on the epidemiological state of the population. Note that, in Is it possible to develop new drugs that
contrast to Avlund et al. [32], phages do not ‘play dice’ in the above model because the dynamics is fully deterministic. prevent latency and facilitate within-
They do not evolve plastic transmission strategies because of stochasticity but because of the periodic fluctuation of the host eradication?
environment.

within an infected host to infer the epidemiological state of the population (i.e., outside the
infected host). The number of coinfecting viruses carries valuable information on the epide-
miological state of the population, and this is why it may be a good cue to evolve conditional
lysogenisation. But other cues may also be used to modulate lysogenisation. In particular,
host starvation is known to increase the probability of lysogenisation [29]. Starvation is a very
good indicator of the environmental state of the population because it signals that very little
bacterial growth is feasible. In a variable environment, where starvation oscillates, it is adaptive
for the virus to hide in infected cells (lysogenisation) during a famine to avoid the risks
associated with the free-living stage. The above argument does not account for the potential
benefits on host fitness associated with the integration of the phage in the host genome. For
instance, Chen et al. [35] showed that lysogenisation with l may carry benefits for the
Escherichia coli bacteria because it lowers the growth rate of lysogens in energy-poor
environments. The potential symbiotic nature of lysogenisation in some environments clearly
disserves further investigation [36].

Concluding Remarks
The discovery and the characterization of bacteriophage l has led to major discoveries in
molecular biology [14]. In particular, the regulation of the lysis–lysogeny decision remains a
very fruitful topic that stimulated a broad range of research studies [26,37]. The molecular
details of the genetic switch of l have been scrutinized for decades and led to a deep
understanding of gene regulation that shed light on many other biological regulatory pro-
cesses. The lysis–lysogeny decision has clear-cut implications on the virus life cycle and offers
unique opportunities to study the ecology and evolution of virus latency. Here, I reviewed a
few recent studies and discussed the adaptive nature of both constitutive and conditional
latency. Competition between different l strains (with different life-history strategies) has
allowed the study of which environmental factors promote the evolution of fixed lysogenisation

Trends in Microbiology, Month Year, Vol. xx, No. yy 7

 TIMI 1303 No. of Pages 10

strategies. To explore the adaptive nature of conditional lysogenisation I developed new

models of life-history evolution in fixed and variable environments. Evolutionary epidemiology
theory provides a way to illustrate how l developed various adaptations to cope with variable
environmental conditions. I do not claim that every feature of the biology of l is well adapted.
As any organism, l evolves under a set of evolutionary constraints. Evolutionary epidemiol-
ogy, however, provides a fruitful framework for trying to make sense of some peculiar features
of the l life cycle. As such, it may also help to understand the evolution of latency in other
phages and other pathogens.

From l to Other Bacteriophages

It is important to remember that while lysogenisation is the classical way to generate latent
infections, alternative strategies are used by lytic phages. For instance, lysis inhibition is the
ability to delay the timing of lysis in T4 phages, and this inhibition is triggered by multiplicity of
infection [38]. It is tempting to think that the same forces have led to the evolution of the
regulation of lysogenisation in l and lysis inhibition in T4. In both cases, multiplicity of
infection carries information about the epidemiological state of the population that may be
used by the virus to better time the induction of a productive infection. Similarly, pseudo-
lysogeny provides an alternative life-history strategy that enables the virus to delay lysis in
starved hosts [39]. Pseudolysogeny is distinct from lysogeny because the phage genome
does not integrate into the host chromosome. The effect of starvation, however, recalls the
increased lysogenisation of l in starved cells. In both cases, conditional latency is likely to be
adaptive because it allows the virus to avoid the risks associated with the free-living stage of
the life cycle.

From Bacteriophages to Other Pathogens

It is also tempting to use our understanding of the evolution of latency in l to study the evolution
of both constitutive and plastic latency in other pathogens. We explored the effect of epidemiol-
ogy and spatial structure on l evolution, but other environmental factors may trigger higher
investment in latency. In malaria, for instance, it is clear that the seasonality of the environment
has dramatic effects on the availability of mosquito vectors and thus on epidemiological
dynamics. Plasmodium vivax is one of the main causative agents of malaria in humans, and
this pathogen has the ability to remain dormant for several months before relapsing and causing
new infections. The rate at which P. vivax relapses is variable among different pathogen isolates
and has been found to depend on latitude [40,41]. In particular, the relapsing rate is much lower
(i.e., latency is more pronounced) at high latitude where seasonality implies that transmission is
hindered by the lack of mosquito vectors in winter. This pattern is consistent with classical
theoretical predictions on the evolution of latency [33,34,42–44].

The analysis of the l life cycle indicates that different types of cue may trigger the relapse of
latent infections. First, relapses may be induced by cues indicating that the environment inside
the host has changed and is unlikely to sustain long-lasting infections (Box 1). Many infections
by herpes simplex viruses and varicella zoster virus have been shown to reactivate when the
host is immune-depressed or physiologically stressed [2]. For instance, asymptomatic reacti-
vation of varicella zoster virus has been observed in astronauts after a mission in space [45]. Of
course, a drop in the efficacy of host immunity may provide a proximal cause for the replication
of the virus leading to a relapse. Yet, the ultimate cause of the relapse may be related to the
evolved ability of the virus to react actively to variations in the physiological state of its host.
Second, relapses may also be induced by cues indicating that the environment outside the host
has changed and is becoming suitable (Box 2). Higher multiplicity of infection carries informa-
tion regarding the epidemiological state of the population. Interestingly, malaria parasites,
similar to bacteriophages, have plastic transmission strategies with regard to the multiplicity of
infection [46]. Again, it is unclear whether this pattern is adaptive or not, but given the

8 Trends in Microbiology, Month Year, Vol. xx, No. yy

 TIMI 1303 No. of Pages 10

tremendous amount of plasticity in Plasmodium parasites [43,47,48] this evolutionary per-

spective is worth investigating.

Bacteriophages live in complex environments and they have evolved elaborate strategies to
cope with the variability of their habitat. In particular, bacteriophage l has the ability to adopt
constitutive and conditional latency. The study of evolutionary forces leading to these fascinating
adaptations give insight into the diversity of other pathogens’ life-history strategies[13_TD$IF] (see Out-
standing Questions). Because the dormant stages of many human pathogens is challenging the
efficacy of classical therapeutic interventions [3–6], a better understanding of the ultimate causes
of these pathogen life histories may thus inspire new strategies to control and eradicate those
infectious diseases.

Acknowledgments
This work benefited a lot from discussions with Thomas Berngruber, Marc Choisy, Troy Day, Bryan Grenfell, Quentin
Legros, Bruce Levin, Sébastien Lion, Ana Rivero, and Joshua Weitz.

References
1. Carrat, F. et al. (2008) Time lines of infection and disease in human
influenza: a review of volunteer challenge studies. Am. J. Epide-
miol. 167, 775–785
2. Kinchington, P.R. et al. (2012) Herpes simplex virus and varicella
zoster virus, the house guests who never leave. Herpesviridae 3, 5
3. Field, H.J. et al. (2006) Herpesvirus latency and therapy – from a
veterinary perspective. Antiviral Res. 71, 127–133
4. Baird, J.K. et al. (2012) Chapter four – diagnosis and treatment of
Plasmodium vivax malaria. Adv. Parasitol. 80, 203–270
5. Shan, L. and Siliciano, R.F. (2013) From reactivation of latent HIV-1
to elimination of the latent reservoir: the presence of multiple
barriers to viral eradication. Bioessays 35, 544–552
6. Esmail, H. et al. (2014) The ongoing challenge of latent
tuberculosis. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 369,
20130437
7. Edwards, R.A. and Rohwer, F. (2005) Viral metagenomics. Nat.
Rev. Microbiol. 3, 504–510
8. Abedon, S.T. (2008) Bacteriophage Ecology: Population Growth,
Evolution, and Impact of Bacterial Viruses, Cambridge University
Press
9. Clokie, M.R.J. et al. (2011) Phages in nature. Bacteriophage 1, 31–45
10. Bail, O. (1925) Der kolistamm 88 von Gildmeister und Herzberg.
Med. Klin. (Munich) 21, 1271–1273
11. Bordet, J. (1925) Le problème de l’autolyse microbienne trans-
missible ou du bactériophage. Ann. Inst. Pasteur 39, 711–763
12. Lwoff, A. (1953) Lysogeny. Bacteriol. Rev. 17, 269–337
22. Bull, J.J. (2006) Optimality models of phage life history and paral-
lels in disease evolution. J. Theor. Biol. 241, 928–938
23. Heineman, R.H. and Bull, J.J. (2007) Testing optimality with exper-
imental evolution: lysis time in a bacteriophage. Evolution 61,
1695–1709
24. Day, T. and Gandon, S. (2007) Applying population-genetic
models in theoretical evolutionary epidemiology. Ecol. Lett. 10,
876–888
25. Berngruber, T.W. et al. (2010) Inhibition of superinfection and the
evolution of viral latency. J. Virol. 84, 10200–10208
26. Stewart, F.M. and Levin, B.R. (1984) The population biology of
bacterial viruses: Why be temperate? Theor. Pop. Biol. 26, 93–117
27. Lion, S. and Boots, M. (2010) Are parasites “prudent” in space?
Ecol. Lett. 13, 1245–1255
28. Lion, S. and Gandon, S. (2015) Evolution of spatially structured
host-parasite interactions. J. Evol. Biol. 28, 10–28
29. Kourilsky, P. (1973) Lysogenization by bacteriophage lambda. I.
Multiple infection and the lysogenic response. Mol. Gen. Genet.
122, 183–195
30. Zeng, L. et al. (2010) Decision making at a subcellular level deter-
mines the outcome of bacteriophage infection. Cell 141, 682–691
31. Robb, M.L. and Shahrezaei, V. (2014) Stochastic cellular fate
decision making by multiple Infecting lambda phage. PLoS
ONE 9, e103636
32. Avlund, M. et al. (2009) Why do phage play dice? J. Virol. 83,
11416–11420

13. Lederberg, E. (1951) Lysogenicity in E. coli K-12. Genetics 36, 560
14. Gottesman, M.E. and Weisberg, R.A. (2004) Little lambda, who
made thee? Microbiol. Mol. Rev. 68, 796–813
15. Ptashne, M. (2004) A Genetic Switch: Phage Lambda Revisited.
(3rd edn), Cold Spring Harbor Laboratory Press
16. Arkin, A. et al. (1998) Stochastic kinetic analysis of developmental
pathway bifurcation in phage l-infected Escherichia coli cells.
Genetics 149, 1633–1648
17. Joh, R.I. and Weitz, J.S. (2011) To lyse or not to lyse: transient-
mediated stochastic fate determination in cells infected by bacter-
iophages. PLoS Comput. Biol. 7, e1002006
18. Berngruber, T.W. et al. (2013) Evolution of virulence in emerging
epidemics. PLoS Pathog. 9, e1003209
33. Cohen, D. (1966) Optimizing reproduction in a randomly varying
environment. J. Theor. Biol. 12, 119–129
34. Vitalis, R. et al. (2013) The joint evolution of dispersal and dor-
mancy in a metapopulation with local extinctions and kin competi-
tion. Evolution 67, 1676–1691
35. Chen, Y. et al. (2005) Population fitness and the regulation of
Escherichia coli genes by bacterial viruses. PLoS Biol. 3, e229
36. Paul, J.H. (2008) Prophages in marine bacteria: dangerous molecular
time bombs or the key to survival in the seas? ISME J. 2, 579–589
37. Mittler, J.E. (1996) Evolution of the genetic switch in temperate
bacteriophage. I. Basic theory. J. Theor. Biol. 179, 161–172
38. Bode, W. (1967) Lysis inhibition in Escherichia coli infected with
bacteriophage T4. J. Virol. 1, 948–955

19. Berngruber, T. et al. (2015) Spatial structure, transmission modes
and the evolution of viral exploitation strategies. PLoS Pathog. 11,
e1004810
20. Lenski, R.E. and May, R.M. (1994) The evolution of virulence in
parasites and pathogens: Reconciliation between two competing
hypotheses. J. Theor. Biol. 169, 253–265
39. Miller, R.V. and Ripp, S.A. (2002) Pseudolysogeny: a bacterio-
phage strategy for increasing longevity in situ. In Horizontal Gene
Transfer (Syvanen, M. and Kado, C.I., eds), pp. 79–89, Academic
Press
40. White, N.J. (2011) Determinants of relapse periodicity in Plasmo-
dium vivax malaria. Malar. J. 10, e297

21. Frank, S.A. (1996) Models of parasite virulence. Q. Rev. Biol. 71, 41. Battle, K.E. et al. (2014) Geographical variation in Plasmodium
37–78 vivax relapse. Malar. J. 13, 144

Trends in Microbiology, Month Year, Vol. xx, No. yy 9

 TIMI 1303 No. of Pages 10
42. Cohen, D. (1967) Optimizing reproduction in a randomly varying
environment when a correlation may exist between the conditions
at the time a choice has to be made and the subsequent outcome.
J. Theor. Biol. 16, 1–14
43. Cornet, S. et al. (2014) Evolution of plastic transmission strategies
in avian malaria. PLoS Pathog. 10, e1004308
44. Scott, M.F. and Otto, S.P. (2014) Why wait? Three mechanisms
selecting for environment-dependent developmental delays. J.
Evol. Biol. 27, 2219–2232
45. Cohrs, R.J. et al. (2008) Asymptomatic reactivation and shed of
infectious varicella zoster virus in astronauts. J. Med. Virol. 80,
1116–1122
10 Trends in Microbiology, Month Year, Vol. xx, No. yy
46. Morlais, I. et al. (2015) Plasmodium falciparum mating patterns and
mosquito infectivity of natural isolates of gametocytes. PLoS ONE
10, e0123777
47. Mideo, N. and Reece, S.E. (2012) Plasticity in parasite pheno-
types: evolutionary and ecological implications for disease. Future
Microbiol. 7, 17–24
48. Cameron, A. et al. (2012) Plasticity in transmission strategies of the
malaria parasite, Plasmodium chabaudi: environmental and
genetic effects. Evol. Appl. 6, 365–371
49. Hurford, A.D. et al. (2010) Next-generation tools for evolutionary
invasion analysis. J. R. Soc. Interface 7, 561–571