REVIEWS

Adaptive significance of circadian

programs in cyanobacteria
Carl Hirschie Johnson, Susan S. Golden and Takao Kondo

C
ircadian rhythms are en- Prokaryotic cyanobacteria express robust modeled assuming only two es-
dogenous biological pro- circadian (daily) rhythms, even when sential components: (1) a circa-
grams with the follow- growing with doubling times that are dian rhythm of psbAI promoter
ing three characteristics: (1) in considerably faster than once every 24 h. activity in every cell and (2) an
constant conditions, they ‘free- This biological clock orchestrates cellular exponential increase in the num-
run’ with a period that is close events to occur in an optimal temporal ber of cells in the culture5. After
to, but not exactly, 24 h in du- program. Competition experiments the growth rate of the culture
ration, (2) their free-running demonstrate that fitness is enhanced when slows, the luminescence pattern
rhythm is temperature compen- the circadian period resonates with the stabilizes into a circadian pat-
sated and (3) the rhythm will period of the environmental cycle. tern of consistent amplitude
entrain to an appropriate envi- (‘sustained phase’). It is in this
ronmental cycle1. Before 1985, C.H. Johnson* is in the Dept of Biology, Vanderbilt growth phase that the circadian
it was believed that circadian University, Nashville, TN 37235, USA; S.S. Golden oscillator can most easily be as-
is in the Dept of Biology, Texas A&M University,
programs were exclusively a College Station, TX 77843, USA; T. Kondo is in the sayed. As the culture ages, the
property of the eukaryotic do- Divn of Biological Science, Graduate School of rhythm slowly damps (‘senesc-
main2. The few dissenting re- Science, Nagoya University, Nagoya 464-01, Japan. ent phase’). The damping ob-
ports suggesting that Escherichia *tel: 11 615 322 2961, served in the senescent phase
fax: 11 615 343 0336,
coli and Klebsiella have circa- e-mail: [email protected] probably results from nutrient
dian clocks were not convinc- depletion, including diminution
ing2. The conclusion that only eukaryotes have circadian of light intensity by high cell densities, which is very
oscillators seemed reasonable, because it was assumed important to the photosynthetic cyanobacteria.
that an endogenous timekeeper with a period close to When we maintained cultures in the actively growing
24 h would not be useful to organisms that divide more state indefinitely by continuously diluting liquid cul-
rapidly than once every 24 h, as do many prokaryotes. tures to keep the average cell density constant, we ob-
Based on data from eukaryotic unicellular organisms, served an interesting pattern in the cell growth profile
this assumption became known as the ‘circadian/ (Fig. 1b). In light/dark (LD) cycles, the cells divided in the
infradian rule’: organisms that divide more rapidly light and did not divide in the dark, as expected, but
than once every 24 h do not express circadian programs3. under constant illumination, a circadian pattern of divi-
sion continued, even though the doubling time of the
Cell division patterns culture was once every 10.5 h (Fig. 1b). In other words,
We and others subsequently found that at least one even though cells are dividing on average more than
group of prokaryotes, the cyanobacteria, has evolved twice every 24 h, when they divide is still ‘gated’ by a cir-
circadian programming of gene expression2,4. We there- cadian oscillator3. The phase of the ‘slowdown’ of divi-
fore set out to test whether the circadian/infradian sion rate coincides with the early subjective night. Using
rule holds for the rapidly growing unicellular cyano- flow cytometry, we have determined that the DNA
bacterium Synechococcus sp. strain PCC 7942. We replication rate and the growth in size of each individual
have transformed Synechococcus with a luciferase cell are not rhythmic over the circadian cycle, but it is the
reporter construct that allows the assay of the pro- timing of division that is controlled by a circadian pro-
moter activity of the psbAI gene, which encodes the gram3. Therefore, in Synechococcus, a circadian clock
D1 protein of the photosystem II reaction center4. We regulates cell division and gene expression even when
found that, in contrast to eukaryotic unicells, cultures cells are dividing more rapidly than once a day.
of Synechococcus growing with doubling times as Based on the earlier work with eukaryotic unicells,
rapid as one division every 10 h continue to exhibit selection for circadian organization had seemed rea-
circadian rhythms of psbAI gene expression3,5. sonable only if the generation time of the cell is as
The amplitude of the luminescence patterns re- long or longer than a day1. To us, however, it seems
flecting psbAI promoter rhythms is a function of the equally plausible that the advantages of anticipating,
stage of the growth cycle for colonies on agar or for for example, the onset of solar illumination, with its
batch liquid cultures. For example, in early log phase, concurrent benefits and deficits, would accrue to all
the luminescence patterns display a clear circadian light-sensitive organisms, even if they divide more
rhythm that grows in amplitude exponentially (Fig. 1a). rapidly than the daily cycle. It is important to recognize
This ‘exponential phase’ pattern can be precisely that unicells that divide asexually are not ‘organisms’
Copyright © 1998 Elsevier Science Ltd. All rights reserved. 0966 842X/98/$19.00 PII: S0966-842X(98)01356-0

TRENDS IN MICROBIOLOGY 407 VOL. 6 NO. 10 OCTOBER 1998

REVIEWS

lem for photosynthetic diazotrophs, such as nitrogen-

(a) fixing cyanobacteria, whose photosynthesis releases
Exponential Sustained Senescent oxygen. Apparently, these cyanobacteria have separated
phase phase phase photosynthesis and nitrogen fixation temporally, as
Luminescence

one of several tactics to accomplish mutually incom-

patible biochemical tasks2,7,9–11. However, not all non-
heterocystous cyanobacteria pursue this strategy. For
example, the filamentous cyanobacterium Tricho-
desmium is able to fix nitrogen in phase with photosyn-
thesis12. The mechanism by which Trichodesmium is able
Days to accomplish this feat is unknown.
(b) Synechococcus sp. strain PCC 7942 does not fix
nitrogen. However, as part of an investigation with a
different purpose, we have uncovered an analogous
Cells per ml

DT = 10.5 h case of temporal separation by the circadian oscillator

in this species. In a global hunt for clock-controlled
genes by promoter trapping, we found some genes
whose promoters are regulated in phases opposite to
that exhibited by our original psbAI reporter con-
struct13. The activity of the psbAI promoter is low at
dawn and increases throughout the day, reaching its
1 2 3 4 5 6 peak at dusk; thus, it is a ‘day-active’ or ‘Class 1’ gene
(Fig. 2a). But a few genes show the opposite pattern
Days (Class 2) in constant conditions. One of these genes is
Fig. 1. Luminescence and cell division patterns in populations of the purF, which encodes the first enzyme of the purine
unicellular cyanobacterium Synechococcus. (a) Circadian patterns of biosynthetic pathway (Fig. 2b). The Synechococcus
gene expression (monitored by luminescence) in three different phases version of purF is homologous to the purF from Bacil-
of the growth cycle: exponential, sustained and senescent. (b) Circadian lus spp., which encodes an oxygen-sensitive enzyme14.
control of cell division in cultures that are continuously diluted with me-
dium to maintain an approximately equal cell density. The figure plots the Although Synechococcus sp. strain PCC 7942 does not
time-dependent cell division rate based on the dilution rate. The culture fix nitrogen, it does have other oxygen-sensitive
was in a light/dark (LD) cycle for the first 1.5 days (black bars along the processes that are regulated so that they occur when
x-axis indicate dark phases; white bars indicate illuminated phases) and intracellular oxygen levels are low, and it therefore
in constant light thereafter (gray bars indicate extrapolated night phases,
i.e. subjective night). Abbreviation: DT, doubling time.
provides another example of temporal separation.
Surprisingly, purL, which encodes a later enzyme
in the same pathway, is not regulated in the same
phase as purF (Ref. 14). So how does the pathway
in the same way that sexually reproducing multicellu- work if there is such a big time delay between purF
lar organisms are. Asexually reproducing unicells can and purL expression? The answer might lie in the en-
be thought of as a single protoplasm that grows larger zymatic steps between the first (PurF) and the fourth
and larger and incidentally subdivides. If the bio- (PurL) steps of the pathway, especially the second
chemical reactions within that protoplasm change as step, which is catalyzed by the enzyme encoded by the
a function of daily alterations in, for example, light purD gene (Fig. 2b). While the intermediate phos-
intensity and temperature, then its ability to adapt opti- phoribosylamine (PRA) is very unstable, phospho-
mally to such daily changes in its environment must ribosylglycinamide (GAR) is stable. Evidence now
be useful, even though it may have subdivided into supports the idea that substrate channeling between
different compartments (i.e. cells) within the 24 h span. the enzymes encoded by purF and purD rapidly con-
verts PRA into GAR. Because GAR is stable, it could
Temporal separation accumulate and be used later in the day when the
How might a daily clock be useful to a cyanobac- other enzymes of the purine biosynthetic pathway,
terium? One strong candidate is optimal temporal pat- such as the enzyme encoded by purL, are turned on14.
terning of biochemical processes. For example, the first If this hypothesis is correct, we predict that the purD
persuasive evidence suggesting that prokaryotes might gene will have the same expression pattern as that of
have circadian oscillators came from three different purF. Research in this area is currently under way.
species of cyanobacteria that display daily rhythms of
nitrogen fixation in LD cycles and in constant light6–8. Competition as a test of fitness
One of these studies showed that the cyanobacteria Although it is logical that circadian programming of
also express a daily rhythm of photosynthesis that is the steps of biochemical reactions could be adaptive,
1808 out of phase with the nitrogen-fixation rhythm; it has not been tested in any organism, either eukary-
photosynthesis peaked at midday, whereas nitrogen otic or prokaryotic. We therefore set out to test the
fixation occurred at night7. The nitrogen-fixing en- value of circadian organization to reproductive fit-
zyme nitrogenase is inactive in the presence of even tiny ness in cyanobacteria using wild-type and mutant
amounts of oxygen, which creates a major design prob- strains that exhibit different free-running periods15.

TRENDS IN MICROBIOLOGY 408 VOL. 6 NO. 10 OCTOBER 1998

 REVIEWS

(a) (a)

Luminescence
Class 1 SP22
Luminescence

(b)

Luminescence
Class 2

Wild-type
1 2 3 4
Days in constant light
(c)
Luminescence

(b)
purF purD
P-Rib-PP PRA GAR
Class 2 Class 2?
P28
purN
0 24 48 72 96
Hours in constant light
purL
Fig. 3. Circadian phenotypes of (a) SP22 (period Å 23 h),
AMP and GMP FGAM FGAR (b) wild-type (period Å 25 h) and (c) P28 (period Å 30 h)
Class 1
strains. The y-axis represents the luminescence level, which
Fig. 2. Regulation of the purine biosynthetic pathway by the acts as an indicator of psbAI promoter activity.
circadian clock. (a) Rhythms of luminescence of two reporter
strains under constant illumination. The psbAI (Class 1; day-
active) reporter is depicted by open circles, whereas the purF
(Class 2; night-active) reporter is depicted by filled circles. with SP22, the wild-type cells grew better during
The gray areas represent the extrapolated times of night. 27 days of competition in the 30-h cycle, but the short-
(b) The purine biosynthesis pathway. PurF is an oxygen- period SP22 outgrew the wild-type strain in the 22-h
sensitive enzyme. PRA is very unstable, whereas GAR is cycle. When competed against wild-type, the long-
stable. Abbreviations: FGAM, phosphoribosylformylglycinami-
dine; FGAR, phosphoribosylformylglycinamide; GAR, phospho- period P28 took over the culture in the 30-h cycle, but
ribosylglycinamide; PRA, phosphoribosylamine; P-Rib-PP, was overgrown in the 22-h cycle (Fig. 4c). When P28
5-phosphoribosyl 1-pyrophosphate. was competed against SP22, the results were clear –
the strain whose period most closely matched that of
the LD cycle eliminated the competitor16 (Fig. 4d). The
The circadian phenotypes of the strains we used results illustrated in Fig. 4 are from batch liquid cul-
[SP22 (period Å 23 h), wild-type (period Å 25 h) and tures, but the same result was obtained from cultures
P28 (period Å 30 h)] are illustrated in Fig. 3. These maintained at constant cell density in a turbidostat.
strains grow in pure culture at the same rate in con- Estimation of the selective pressure, by taking meas-
stant light and in LD cycles, so there does not appear urements at more time points, suggests that the
to be a significant advantage or disadvantage to selective coefficient is surprisingly strong. To a first
having different circadian periods when the strains approximation, the relative fitness of the less success-
are grown in single-strain cultures16. As depicted in ful strain could be as low as 0.7–0.8 (Ref. 16). Be-
Fig. 4a, the concept of the fitness test is to mix differ- cause the growth rates of the strains in pure culture
ent strains together and to grow them in competition are not different by a factor of 20–30%, this is a case
to determine if the composition of the population of ‘soft selection’, where the poorer fitness of inferior
changes as a function of time. genotypes becomes obvious during competition17.
The results of this competition test were remark- Figure 4 indicates that the strain whose period is nearest
able. When each of the strains was mixed with another to the period of the environmental cycle is the most fit
strain and grown in competition, a pattern emerged in these competition experiments. Because the mu-
that depended upon the nature of the LD cycle. We tants P28 and SP22 can defeat the wild-type strain in
tested two different LD regimens that had equal LD cycles in which the periods are similar to their en-
amounts of light and darkness but in which the fre- dogenous periods, the differential effects we observed
quency of the LD cycle differed: a 22-h cycle (LD are likely to result from differences in the circadian
11:11) and a 30-h cycle (LD 15:15). As illustrated in clock rather than an unrelated mutation that is deleteri-
Fig. 4b, when the wild-type strain was in competition ous for growth. We have confirmed this conclusion

TRENDS IN MICROBIOLOGY 409 VOL. 6 NO. 10 OCTOBER 1998

REVIEWS

(a) Log-phase cultures (grown in LL) Questions for future research

• What is the mechanism by which a circadian pro-
gram enhances the fitness of cyanobacteria?
• Do prokaryotes other than cyanobacteria have circadian
oscillators?
Mix • Are circadian programs more beneficial to slowly
growing organisms (doubling time .24 h) than to
rapidly growing organisms?
Strain 1 Strain 2
directly by a genetic test, which shows that only the
differences between the circadian periods of these
Grow together strains are responsible for the competitive advantage/
in competition disadvantage observed16.
How do the victorious strains defeat the van-
quished? Other than demonstrating that soft selec-
tion is operating, we do not yet know the precise
Plate aliquots
mechanism of the selection. Several factors could be
responsible, including competition for limiting re-
sources, such as light, nutrients and carbon dioxide,
or the rhythmic secretion of an inhibitory factor by
the cells. However, we do know that the phasing of
psbAI gene expression is disrupted within strains in
Strain 1 phenotype
non-optimal LD cycles16. This is consistent with the
idea that the circadian program orders cellular
Assay luminescence processes to optimally match environmental cycles;
rhythms of single colonies when this order is disturbed, fitness is reduced. Thus,
in conclusion, the circadian pacemaker in cyanobac-
Strain 2 phenotype teria confers a significant competitive advantage
when the period of the clock resonates with the envi-
ronmental cycle, thus achieving optimal phasing of
cellular events. This is the first rigorous demonstra-
(b) 100 SP22 tion in any organism of an advantage conferred by a
Percentage of each strain in mixture

Wild-type circadian system to fitness.

0 Acknowledgements
We are grateful for support from the National Institute of Mental Health
(MH01179 to C.H.J.), the National Science Foundation (MCB-
9633267 to C.H.J. and MCB-9513367 to S.S.G.), the Japanese Ministry
(c) 100 P28 of Education, Science and Culture (10874121 to T.K.) and the Research
Wild-type for the Future Program of JSPS (JSPS-RFTF96L00601 to T.K.).
0
References
1 Pittendrigh, C.S. (1993) Annu. Rev. Physiol. 55, 17–54
2 Johnson, C.H. et al. (1996) Mol. Microbiol. 21, 5–11
(d) 100 SP22 3 Mori, T., Binder, B. and Johnson, C.H. (1996) Proc. Natl. Acad.
Sci. U. S. A. 93, 10183–10188
P28 4 Kondo, T. et al. (1993) Proc. Natl. Acad. Sci. U. S. A. 90,
0 5672–5676
LD 11:11 Initial LD 15:15 5 Kondo, T. et al. (1997) Science 275, 224–227
composition 6 Stahl, L.J. and Krumbein, W.E. (1985) Arch. Microbiol. 143, 67–71
7 Mitsui, A. et al. (1986) Nature 323, 720–722
Fig. 4. Competition of three circadian strains in different light/dark (LD) 8 Grobbelaar, N. et al. (1986) FEMS Microbiol. Lett. 37, 173–177
cycles. (a) Different strains were mixed together in batch cultures and 9 Grobbelaar, N., Lin, H-Y. and Huang, T.C. (1987) Curr.
grown in competition under different LD cycles. Every seven days, the Microbiol. 15, 29–33
cultures were diluted with fresh medium. At various times during the 10 Gallon, J.R. (1992) New Phytol. 122, 571–609
competition, aliquots were plated as single colonies and their lumines-
cence rhythms were monitored to determine the frequency distribution 11 Schneegurt, M.A. et al. (1994) J. Bacteriol. 176, 1586–1597
of the different circadian phenotypes. Results of competition between 12 Capone, D.G. et al. (1997) Science 276, 1221–1229
(b) SP22 (period Å 23 h) and wild-type (period Å 25 h) strains, (c) P28 13 Liu, Y. et al. (1995) Genes Dev. 9, 1469–1478
and wild-type (period Å 30 h) strains and (d) SP22 and P28 strains. 14 Liu, Y. et al. (1996) Mol. Microbiol. 20, 1071–1081
Competition was for 27 days in LD cycles of either LD 11:11 or LD 15 Kondo, T. et al. (1994) Science 266, 1233–1236
15:15. The initial composition of the cultures before competition is 16 Ouyang, Y. et al. (1998) Proc. Natl. Acad. Sci. U. S. A. 95,
shown by the middle histogram in each part. 8660–8664
17 Futuyma, D.J. (1998) Evolutionary Biology (3rd edn), Sinauer

TRENDS IN MICROBIOLOGY 410 VOL. 6 NO. 10 OCTOBER 1998