Blackwell Science, LtdOxford, UKMMIMolecular Microbiology0950-382X© 2005 The Authors; Journal compilation © 2005 Blackwell Publishing Ltd?

2005592367375Review ArticleMulticellular behaviour in cyanobacteriaC.-C. Zhang

et al.

Molecular Microbiology (2006) 59(2), 367–375 doi:10.1111/j.1365-2958.2005.04979.x

First published online 17 November 2005

MicroReview

Heterocyst differentiation and pattern formation in

cyanobacteria: a chorus of signals

Cheng-Cai Zhang,1* Sophie Laurent,1 Samer Sakr,1 neighbouring cells. HetR, a protease showing DNA-
Ling Peng2 and Sylvie Bédu1 binding activity, is crucial to heterocyst differentiation
1
Laboratoire de Chimie Bactérienne, UPR9043-CNRS, and appears to be the central processor of various
Institut de Biologie Structurale et Microbiologie, 31, early signals involved in the developmental process.
chemin Joseph Aiguier, 13402 Marseille cedex 20, How the various signalling pathways are integrated
France. and used to control heterocyst differentiation pro-
2
Département de Chimie, UMR6114-CNRS, 163 avenue cesses is a challenging question that still remains to
de Luminy, 13288 Marseille cedex, France. be elucidated.

Summary Introduction

Heterocyst differentiation in filamentous cyanobacte- The generation of cell-type diversity is a complex mecha-
ria provides an excellent prokaryotic model for study- nism that enables an organism to adapt to environmental
ing multicellular behaviour and pattern formation. In changes and to acquire specialized functions that would
Anabaena sp. strain PCC 7120, for example, 5–10% of otherwise be impossible. Some filamentous cyanobacte-
the cells along each filament are induced, when ria, such as Anabaena/Nostoc sp. strain PCC 7120, pro-
deprived of combined nitrogen, to differentiate into vide an excellent model for studying two important
heterocysts. Heterocysts are specialized in the questions in developmental biology: cell differentiation
fixation of N2 under oxic conditions and are semi- and the formation of a multicellular pattern (Fig. 1). Het-
regularly spaced among vegetative cells. This devel- erocysts differentiated from vegetative cells specialize
opmental programme leads to spatial separation of in nitrogen fixation catalysed by the oxygen-sensitive
oxygen-sensitive nitrogen fixation (by heterocysts) enzyme complex nitrogenase (Wolk et al., 1994; Wolk,
and oxygen-producing photosynthesis (by vegetative 2000; Meeks and Elhai, 2002; Golden and Yoon, 2003).
cells). The interdependence between these two cell The metabolic and morphological changes associated
types ensures filament growth under conditions of with heterocyst development underlie the need for a
combined-nitrogen limitation. Multiple signals have micro-oxic environment in which nitrogenase can function.
recently been identified as necessary for the initiation This need is particularly acute in the case of cyanobacte-
of heterocyst differentiation, the formation of the het- ria, because they produce oxygen. Heterocysts are envel-
erocyst pattern and pattern maintenance. The Krebs oped by two superimposed layers, the one consisting of
cycle metabolite 2-oxoglutarate (2-OG) serves as a polysaccharides and the other of glycolipids. The gly-
signal of nitrogen deprivation. Accumulation of a non- colipid layer reduces the permeation of oxygen, and the
metabolizable analogue of 2-OG triggers the complex polysaccharide layer protects the fragile glycolipid layer.
developmental process of heterocyst differentiation. Heterocysts also display an increased respiration rate to
Once heterocyst development has been initiated, consume the oxygen molecules that manage to go
interactions among the various components involved through or around the cell-envelope barriers. In addition,
in heterocyst differentiation determine the develop- the Photosystem II responsible for oxygen production is
mental fate of each cell. The free calcium concentra- no longer present in heterocysts (Wolk et al., 1994; Wolk,
tion is crucial to heterocyst differentiation. Lateral 2000; Meeks and Elhai, 2002). Heterocysts rely on vege-
diffusion of the PatS peptide or a derivative of it from tative cells as sources of carbon and reductant; in return,
a developing cell may inhibit the differentiation of they supply the surrounding vegetative cells with fixed
nitrogen. Cooperation between the two functionally dis-
Accepted 25 October, 2005. *For correspondence. E-mail cczhang@ tinct cell types is thus essential to filament growth in the
ibsm.cnrs-mrs.fr; Tel. (+33) 491 164 096; Fax (+33) 491 718 914. absence of combined nitrogen.

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd
368 C.-C. Zhang et al.
Fig. 1. Heterocyst development in Anabaena PCC 7120.
A. Filaments grown in the presence of nitrate serving as a nitrogen
source.
B. Filaments transferred from nitrate-containing to a combined nitro-
gen-free medium, and observed under the microscope 24 h later.
C. Heterocysts (arrows) are dimly fluorescent, unlike vegetative cells.
This picture gives the superimposed images of the same filaments
taken under white light and under fluorescence.
In Anabaena sp. PCC 7120, heterocysts are regularly
intercalated among vegetative cells, and are usually sep-
arated by 10–20 vegetative cells. This linear regular
arrangement of heterocysts among vegetative cells con-
stitutes one of the simplest one-dimensional patterns
imaginable in developmental biology (Fig. 1). The aim of
the present review is to summarize the progress made
during the last few years in the identification of several key
signals involved in heterocyst development and pattern
formation, and to provide a basis for future research.
Several comprehensive reviews are available for readers
interested in a more global view of this subject, including
heterocyst maturation and symbioses, which are not cov-
ered here (Wolk et al., 1994; Wolk, 2000; Meeks and
Elhai, 2002). Most of the information obtained to date on
heterocyst development has been based on the use of
model organisms such as Anabaena PCC 7120, Ana-
baena variabilis ATCC 29413 and Nostoc punctiforme
ATCC 29133, although heterocyst-forming cyanobacteria
are widespread in freshwater and brackish environments,
either in symbiosis or free-living. The basic mechanism of
heterocyst development seems to be conserved among
these cyanobacteria, and we will not distinguish between
different strains unless necessary. It is worth noting here
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 59, 367–375
that A. variabilis ATCC 29413 contains three nitrogena-
ses; two are classical Mo-nitrogenases (Nif1 and Nif2),
and the third is a V-nitrogenase (Thiel et al., 1995). Nif1
and V-nitrogenase are active in heterocysts under oxic
conditions, whereas Nif2 functions in all cells under anoxic
conditions (Thiel et al., 1995). In a nif1 mutant, Nif2 can
provide a sufficient nitrogen supply to support the growth
of the filaments, but fails to prevent the formation of
spaced heterocysts; surprisingly, under the same condi-
tions, exogenously supplied ammonium represses hetero-
cyst formation (Thiel and Pratt, 2001). It is unknown how
endogenously fixed nitrogen by Nif2 and exogenously
supplied ammonium can have opposite effects on hetero-
cyst differentiation. It is possible that whether or not the
filaments differentiate heterocysts may depend on the
availability of external supply of fixed nitrogen (Thiel and
Pratt, 2001).
How filaments sense nitrogen starvation
2-Oxoglutarate (2-OG) as a metabolic signal
The presence of ammonium strongly inhibits the differen-
tiation of heterocysts, whereas nitrate is usually some-
what less inhibitory, and heterocysts are normally present
in the absence of fixed nitrogen (Flores and Herrero,
1994; Wolk et al., 1994; Wolk, 2000; Meeks and Elhai,
2002). It has been proposed that in proteobacteria, the
glutamine level may signal nitrogen status (Ikeda et al.,
1996), but this has not been confirmed in cyanobacteria
(Forchhammer, 2004; Herrero et al., 2004). Both in vitro
and in vivo data tend rather to support the idea that
accumulated levels of 2-OG (also called α-ketoglutarate)
constitutes the signal of nitrogen limitation in cyanobacte-
ria. It has been thought for a long time that the Krebs
cycle in cyanobacteria may be incomplete because of the
lack of 2-OG dehydrogenase activity (Stanier and Cohen-
Bazire, 1977). This idea is consistent with the genomic
analyses carried out on several cyanobacterial strains
(Muro-Pastor et al., 2001; Forchhammer, 2004; Herrero
et al., 2004). The main function of 2-OG is therefore to
serve as the carbon skeleton for ammonium assimilation
via the glutamine synthetase-glutamate synthase (GS-
GOGAT) cycle (Vàzquez-Bermùdez et al., 2000). In this
case, 2-OG may accumulate within cells starved of com-
bined nitrogen. This hypothesis is consistent with data
obtained by measuring the 2-OG levels in both unicellular
and filamentous cyanobacteria (Muro-Pastor et al., 2001;
Laurent et al., 2005). Artificially increased levels of 2-OG
mimic nitrogen starvation to some extent in both the uni-
cellular strain Synechococcus elongatus PCC 7942 and
the heterocyst-forming filamentous strain Anabaena PCC
7120 (Li et al., 2003; Vàzquez-Bermùdez et al., 2003). In
the latter strain, heterocyst differentiation occurs concom-
itantly with an increase in the 2-OG level under moder-
© 2005 The Authors

ately repressive conditions. As none of the cyanobacterial
strains tested to date have a 2-OG transport system, they
cannot efficiently take up 2-OG added to the medium.
This problem was elegantly resolved by expressing kgtP,
an Escherichia coli gene encoding a 2-OG permease, in
S. elongatus PCC 7942 (Vàzquez-Bermùdez et al.,
2000). A similar strategy has been used with Anabaena
PCC 7120 (Li et al., 2003).
The keto group of 2-OG is involved in nitrogen assimi-
lation. To test whether 2-OG constitutes the nitrogen sta-
tus signal in vivo, a dozen analogues of 2-OG were
synthesized by replacing the keto group of 2-OG by bro-
movinyl, gem-fluoromethyl, or chlorovinyl groups, etc.
(Chen, 2005). One of the analogues, 2,2-difluoro-pen-
tanedioic acid (DFPA), was extremely useful in the in vivo
studies on the signalling function of 2-OG (Laurent et al.,
2005). Firstly, DFPA is recognized and taken up by kgtP,
and stimulates the in vitro DNA-binding activity of NtcA, a
candidate 2-OG sensor. Secondly, the presence of fluo-
rine makes it possible to trace the fate of DFPA in vivo by
performing magic angel spinning nuclear magnetic reso-
nance spectroscopy (NMR). The NMR data show that
DFPA accumulates within the cells and is not metabolized.
Thirdly, in comparison with other analogues, DFPA com-
petes very poorly (IC50 = 13.6 mM) with 2-OG as a sub-
strate for 2-OG dehydrogenase (P. Ling, unpubl. data);
these data suggest that DFPA has only limited inhibitory
effects on enzymes using 2-OG as substrate. In vivo,
DFPA has no significant effect on the uptake and assimi-
lation of ammonium. The accumulation DFPA triggers het-
erocyst development in the presence of ammonium.
These studies provide in vivo evidence that accumulated
2-OG serves as the nitrogen limitation signal triggering
Multicellular behaviour in cyanobacteria 369
the key signalling pathways leading to heterocyst differen-
tiation (Laurent et al., 2005; Fig. 2).
The candidate 2-OG sensor, NtcA
The 2-OG levels are known to affect the activity of two
proteins in cyanobacteria, namely PII (which is encoded
by glnB) and NtcA (Forchhammer, 2004; Herrero et al.,
2004). Recent experimental data suggest that NtcA rather
than PII, is the main 2-OG sensor responsible for the
initiation of heterocyst differentiation (Fig. 2). Strains bear-
ing mutant glnB alleles encoding altered PII protein that
is insensitive to changes in nitrogen sources still develop
heterocysts normally, although the heterocysts are func-
tionally impaired (Laurent et al., 2004). Whether NtcA is
the sole sensor of 2-OG in the initiation of heterocyst
development still remains an open question.
NtcA is a transcription factor belonging to the cyclic
AMP receptor protein family, which controls a number of
genes involved in nitrogen and carbon metabolism (Her-
rero et al., 2004). It is a well-conserved protein in cyano-
bacteria, including unicellular strains and other strains that
do not form heterocysts. Two lines of evidence suggest
that NtcA is the main 2-OG sensor involved in the initiation
of heterocyst differentiation. The first evidence was pro-
vided by genetic studies showing that ntcA is necessary
for the initiation of heterocyst differentiation to occur (for
reviews, see Golden and Yoon, 2003; Herrero et al.,
2004). The second was the finding that the DNA-binding
activity of NtcA is stimulated by the presence of 2-OG and
the 2-OG analogue DFPA (Laurent et al., 2005). In addi-
tion, in the unicellular cyanobacterium S. elongatus PCC
7942, NtcA alone is able to bind to promoter regions of its

Fig. 2. The regulatory circuit possibly involved in the early stages of heterocyst development. Processes or components promoting heterocyst
differentiation are marked in red, and those suppressing heterocyst formation are in black. NtcA and HetR together constitute the main regulatory
loop controlling the initiation of heterocyst development and are therefore shadowed. The expression of hetR and ntcA is autoregulatory (Black
et al., 1993; Ramasubramanian et al., 1996; Muro-Pastor et al., 2002) as well as being mutually dependent (Muro-Pastor et al., 2002). The action
of CcbP is possibly subject to nitrogen control (Zhao et al., 2005), but it has not yet been established whether a factor such as NtcA may regulate
the action of CcbP directly or via some other factor (labelled with a question mark). It has been established that excessive levels of expression
of ccbP suppresses the increase in the free Ca2+ levels, which is necessary for the proper upregulation of hetR expression (Zhao et al., 2005).
Where and how free Ca2+ acts still remains unknown, and any of the steps involved in the auto- and mutual regulation of hetR-ntcA could be
potential targets of this regulation process. The arrow depicting the action of free Ca2+ is therefore tentatively positioned above the shadowed
regulatory loop composed of NtcA and HetR. HetR is necessary for the expression of patS to occur in developing cells, and PatS can in turn
inhibit the DNA-binding activity of HetR (Huang et al., 2004). The interactions occurring between HetR and PatS may contribute importantly to
cell-type determination (Yoon and Golden, 1998; 2001; Huang et al., 2004; Khudyakov and Golden, 2004; Wu et al., 2004). The expression of
hetC is controlled by NtcA (Muro-Pastor et al., 1999). It has been shown that the hetC mutant was able to initiate heterocyst differentiation but
stopped at an early stage of the process, and that the developing cells fail during a transition step to reach a non-dividing state (Khudyakov and
Wolk, 1997; Xu and Wolk, 2001).

© 2005 The Authors

Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 59, 367–375
370 C.-C. Zhang et al.
target genes (Tanigawa et al., 2002; Vàzquez-Bermùdez
et al., 2002), but the transcription will not start unless 2-
OG is present (Tanigawa et al., 2002).
2-OG threshold concentrations and cell behaviour
Ammonium is the preferred nitrogen source for cyanobac-
teria, as it can be directly assimilated via the GS-GOGAT
cycle, whereas nitrate has to be reduced by nitrate reduc-
tase and nitrite reductase to ammonium (Flores and Her-
rero, 1994). Ammonium exerts a global control over the
use of alternative nitrogen sources, and this control is
mediated by NtcA (Herrero et al., 2004). When nitrate
replaces ammonium, NtcA activates genes involved in
nitrate uptake and reduction. There exist good reasons to
believe that the level of 2-OG is also involved in the regula-
tion of nitrate uptake and utilization, at least by a unicellular
cyanobacterial strain (Vàzquez-Bermùdez et al., 2003). If
the same elements, NtcA and 2-OG, control both nitrate
utilization and heterocyst development, how could nitrate
still repress heterocyst development? One fairly plausible
explanation is based on the idea that 2-OG reaches differ-
ent cellular concentrations that serve as thresholds above
which nitrate utilization or heterocyst formation is acti-
vated. Different intracellular levels of 2-OG were measured
experimentally in the unicellular cyanobacterium Syn-
echocystis sp. PCC 6803 (Muro-Pastor et al., 2001) and
Anabaena PCC 7120, although the changes in the levels
of 2-OG in the latter organism occurred in a narrower
range than that observed in the case of unicellular strains
(Laurent et al., 2005; S. Laurent, unpubl. results). In the
presence of ammonium, 2-OG remains at a low level
(Muro-Pastor et al., 2001; Laurent et al., 2005). When
nitrate is the only nitrogen source available, 2-OG reaches
a level that may suffice to allow NtcA to activate the genes
involved in nitrate uptake and assimilation, but not to acti-
vate the regulatory cascade resulting in heterocyst devel-
opment. When filaments are starved of combined nitrogen,
2-OG accumulates up to the highest level (Muro-Pastor
et al., 2001; Laurent et al., 2005) and the autoregulatory
effect of ntcA gene expression may further amplify the
nitrogen starvation signal and lead to the expression of
hetR (Muro-Pastor et al., 2002), a gene required for het-
erocyst differentiation (Buikema and Haselkorn, 1991).
When deprived of combined nitrogen, genes involved in
the use of alternative nitrogen sources such as nitrate are
also highly activated (Cai and Wolk, 1997). If an alternative
nitrogen source is available, the heterocyst differentiation
process may be reversed (Fig. 2).
Relationship between 2-OG and free calcium ions
Nitrogen starvation not only leads to increased 2-OG lev-
els, but also to increased concentrations of free calcium
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 59, 367–375
ion, which was recently found to be necessary for hetero-
cyst development (Torrecilla et al., 2004; Zhao et al.,
2005). Using the Ca2+-binding luminescent photoprotein
aequorin as a reporter, a transient, approximately three-
fold increase in the Ca2+ level was detected about 1 h after
combined-nitrogen step-down (Torrecilla et al., 2004). The
results of studies in which several agents affecting Ca2+
homeostasis were used suggest that the intracellular Ca2+
concentrations must be finely tuned for heterocyst differ-
entiation to be possible. It has also been suggested that
the Ca2+ signal may act earlier than hetR, as the transient
increase in the Ca2+ concentration continued to occur in a
hetR mutant. Interestingly, the transient increase in the
Ca2+ level observed in response to combined-nitrogen
removal was found to mainly result from the mobilization
of internal Ca2+ storage (Torrecilla et al., 2004). This find-
ing has been confirmed by recent studies on CcbP, a
calcium-sequestering protein (Zhao et al., 2005). The
inactivation of ccbP leads to the formation of multiple-
contiguous heterocysts (Mch phenotype), whereas over-
expression of ccbP results in a low level of free Ca2+, which
is correlated with lower rates of hetR induction and a lack
of heterocyst differentiation. A different Ca2+ reporter, obe-
lin, has been used to monitor the calcium levels in individ-
ual cells along the filaments (Zhao et al., 2005). The Ca2+
level was found to be about 10-fold higher in heterocysts
than in vegetative cells, which correlates with the levels of
expression of ccbP (which are high in vegetative cells and
undetectable in heterocysts). CcbP itself might not be
directly involved in heterocyst development but rather
exert its effects by sequestering Ca2+ under conditions of
deprivation of combined nitrogen. This assumption is sup-
ported by the finding that the heterologous calcium-bind-
ing protein calmodulin, when produced in Anabaena PCC
7120, also prevents heterocyst development in a similar
way to CcbP. These results suggest that under nitrogen
sufficiency conditions, CcbP may bind to Ca2+ and main-
tain the free Ca2+ at a low level, and that the CcbP–Ca2+
complex may serve as a Ca2+ storage device (Fig. 2).
Under conditions of limitation of combined nitrogen, CcbP
may be preferentially inactivated by an unknown mecha-
nism in developing cells, which may increase the Ca2+
levels, resulting in heterocyst differentiation.
The fact that both free Ca2+ ions and 2-OG are early
signals to heterocyst development raises several ques-
tions. They could act sequentially, independently or syn-
ergistically. There is too little evidence available at the
moment to be able to speculate what the relationship
between these two signals may be. The 2-OG level
increases immediately in response to nitrogen limitation
(Laurent et al., 2005); Ca2+ ions start to accumulate
45 min (Torrecilla et al., 2004), or 4 h (Zhao et al., 2005)
after the withdrawal of combined nitrogen, depending on
the reporter used to measure the Ca2+ ion levels. In addi-
© 2005 The Authors

tion, although the accumulated levels of 2-OG analogue
DFPA suffice to trigger heterocyst development, even
when ammonium is available, Ca2+ ions exert their effects
only under conditions of combined-nitrogen deprivation.
In the ccbP mutant, heterocyst differentiation did not occur
when a combined-nitrogen source was present, even
when the Ca2+ ion concentrations were high. The Mch
phenotype was observed in this mutant only when the
combined-nitrogen source was removed (Zhao et al.,
2005). These data suggest that the effects of CcbP are
subject to nitrogen control, which is possibly mediated by
2-OG and NtcA (Fig. 2).
HetR, the central signal processor involved in
heterocyst development
HetR appears to be the main positive regulatory factor
involved in heterocyst development and pattern formation
(Buikema and Haselkorn, 1991; 2001; Black et al., 1993;
Khudyakov and Golden, 2004). Two activities have been
defined for HetR, a self-degrading protease activity, and
a DNA-binding activity requiring homodimer formation
(Zhou et al., 1998; Huang et al., 2004). Although NtcA
may be the first to perceive the 2-OG signal, further
enhancement of ntcA expression requires hetR. Con-
versely, hetR expression depends on ntcA (Muro-Pastor
et al., 2002). Both hetR and ntcA are also positively auto-
regulated (Black et al., 1993; Ramasubramanian et al.,
1996; Muro-Pastor et al., 2002). The regulatory loop com-
posed of NtcA and HetR is central to heterocyst develop-
ment (Fig. 2). NtcA is involved in the regulation of activities
other than heterocyst development, whereas HetR seems
to be required mostly for cell development. HetR expres-
sion is necessary and maybe sufficient for the formation
of heterocysts to occur (Buikema and Haselkorn, 1991;
2001). High-level expression of hetR leads to heterocyst
development, even under nitrogen-replete conditions
(Buikema and Haselkorn, 2001). Ectopic expression of a
gain-of-function allele of a hetR mutant (hetRR223W) was
able to circumvent all the main inhibitory signals, resulting
in near-complete heterocyst differentiation under condi-
tions of combined-nitrogen limitation (Khudyakov and
Golden, 2004). HetR may be able to directly or indirectly
balance the effects of most of the activating and inhibitory
signals involved in heterocyst differentiation.
In addition to ntcA, two other genes, patA and hetF are
also required for the upregulation of hetR expression dur-
ing the early phase of heterocyst development. How hetF
acts is not yet known (Wong and Meeks, 2001). In a patA–
mutant, only terminal heterocysts were formed, even
when hetR was overexpressed by using a heterologous
promoter (Liang et al., 1992; Buikema and Haselkorn,
2001). This finding indicates that PatA may affect HetR
activity at a level after transcription. PatA is a CheY-type
© 2005 The Authors
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 59, 367–375
Multicellular behaviour in cyanobacteria 371
response regulator that probably exerts its effects via pro-
tein–protein interactions. A mutant allele of hetR has been
isolated; when located on a plasmid in either a hetR-
deletion mutant or a hetR–, patA– double mutant, this
mutant allele of hetR shows a similar terminal-heterocyst
phenotype to that of the patA– single mutant (D. Risser
and S.M. Callahan, unpubl. results). In cells destined to
become heterocysts, PatA might interact directly with
HetR, either preventing HetR autoproteolysis or relieving
the PatS-mediated inhibition of the DNA-binding activity
of HetR (see also below). Another gene, hanA, encoding
a protein similar to the histone-like DNA-binding protein
HU, is also necessary for heterocyst development to be
initiated (Khudyakov and Wolk, 1996).
Heterocysts are terminally differentiated cells. The main
cell division protein, FtsZ, is undetectable in mature het-
erocysts (Kuhn et al., 2000). The hetC gene, under the
control of NctA (Muro-Pastor et al., 1999), is involved in
the regulation of early steps in heterocyst differentiation
(Khudyakov and Wolk, 1997). In a hetC mutant, partially
differentiated cells have been detected, and these cells
keep on dividing, resulting in chains of small cells (Xu and
Wolk, 2001). These partially differentiated cells are thus
unable to stop the process of cell division or to proceed
with a further step prior to envelope morphogenesis,
which is necessary for proheterocyst formation to occur
(Fig. 2). No direct interactions between HetC and FtsZ
could be detected using a yeast two-hybrid system
(unpublished results from our laboratory).
Pattern formation and maintenance
Competition via lateral inhibition exerted by the
diffusible PatS signal
It was observed long ago that a group of cells often begin
to differentiate and that competition involving an inhibitory
signal determines which one becomes a heterocyst, while
the others revert to the vegetative state (Wilcox et al.,
1973a,b). This finding has been strongly confirmed by the
identification of an inhibitory signal, PatS (reviewed by
Golden and Yoon, 2003). The patS gene from Anabaena
PCC 7120 was found to encode a 13- or 17-amino-acid
polypeptide, depending on which translation start codon
was used in vivo (Yoon and Golden, 1998; 2001). The
putative orthologue of patS from N. punctiforme ATCC
29133 contains only 13 codons (Meeks et al., 2001; Wu
et al., 2004). patS is strongly expressed in developing
cells and mature heterocysts (Yoon and Golden, 1998;
2001). A strain without patS develops multiple contiguous
heterocysts amounting to 30% of all the cells, as com-
pared with 10% in the wild-type strain, and overexpression
of patS suppressed heterocyst differentiation. Adding a
synthetic peptide corresponding to the last five residues
372 C.-C. Zhang et al.
(RGSGR) of PatS (PatS-5) to the growth medium inhibited
heterocyst development, suggesting that PatS-5 may be
diffusible and may correspond to the mature product of
PatS, but no definite evidence is available to date to sup-
port this assumption.
The fact that the ectopic expression of patS driven by
heterologous promoters at work in either early developing
cells, vegetative cells, or all cells abolishes the process of
heterocyst differentiation suggests that PatS does not act
cell-autonomously (Yoon and Golden, 1998; Wu et al.,
2004). However, as no lateral inhibition occurs when a
patS minigene corresponding to PatS-5 is expressed in a
developing cell, PatS-5 may fail to diffuse from the
producing cell to its neighbours and may act only cell-
autonomously (Wu et al., 2004). This situation contrasts
strongly with the reported effects of exogenous PatS-5
(Yoon and Golden, 1998). It may therefore be possible for
the putative mature form of PatS (PatS-5) to be imported
into cells, but not exported (Wu et al., 2004). These results
suggest that full-length PatS is exported and matures
during or after its exportation, and that the mature form is
then imported by neighbouring cells, and inhibits their
development. This mechanism is similar to the signalling
mechanism involving peptide pheromone and compe-
tence factor previously described in other bacteria. In
Streptococcus pneumoniae, for example, the competence
pheromone originates from a larger precursor peptide
encoded by comX, and matures via proteolysis while
being exported by the ABC-type exporter ComA (Michiels
et al., 2001). The competence pheromone is then sensed
by a two-component signalling system composed of
ComD and ComE, which regulates the expression of
genes involved in quorum sensing.
PatS receptor
Studies by J. W. Golden’s group have shown that proteins
bearing the PatS-5 sequence RGSGR located in the cyto-
plasm can fully exert their inhibitory effects, which sug-
gests that the receptor of the PatS signal is cytoplasmic
(Wu et al., 2004). Recent genetic and biochemical data
suggest that one PatS receptor might be HetR itself. The
HetR homodimer displays a DNA-binding activity required
for the expression of patS and hetR to be upregulated
(Huang et al., 2004). This DNA-binding activity is inhibited
by PatS-5 in vitro in a dose-dependent fashion, which
suggests that the PatS to HetR ratio plays an important
role in the outcome of cell development. A high PatS to
HetR ratio in vegetative cells might prevent them from
developing (Huang et al., 2004; Fig. 2). Interactions
between HetR and PatS have also been found to occur in
molecular and genetic studies (Huang et al., 2004; Khudy-
akov and Golden, 2004). PatS-5 added to the growth
medium abolished the upregulation of hetR expression
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 59, 367–375
(Huang et al., 2004). A suppressor mutant, which
becomes resistant to the overexpression of patS, carries
a mutation in the hetR gene (Khudyakov and Golden,
2004). The strain bearing this mutant allele of HetR,
HetRR223W, developed an abnormal pattern of hetero-
cysts and Mch phenotypes.
Immunity against PatS inhibition in developing cells
In lactic acid bacteria, the bacteriocin operon responsible
for the production of the antibacterial peptide also
includes a cognate immunity gene, the product of which
specifically protects the producer (Michiels et al., 2001).
In the case of heterocyst development, how developing
cells, within which PatS is synthesized, remain immune to
PatS inhibition is still a puzzling question. One might eas-
ily imagine that the full-length PatS is not active when it
is first produced in developing cells, and becomes active
only when it (putatively) matures into the PatS-5 form
before or when it is exported into neighbouring cells via
cell junctions or the open periplasmic space along the
filament. But this may only be partly true. The full-length
PatS, when produced ectopically in vegetative cells, com-
pletely inhibits heterocyst differentiation (Wu et al., 2004).
The PatS-5 motif embedded within different polypeptides
are able to suppress heterocyst differentiation, at least
when these polypeptides are expressed at high levels,
although it has not yet been established whether or not
the PatS-5 motif present in these polypeptides is directly
responsible for the observed phenotype. These data sug-
gest that the maturation of the full-length PatS, if it actually
occurs, might not be an absolute prerequisite for PatS to
inhibit heterocyst differentiation within the cells in which it
is produced. This leaves us with at least two other possible
mechanisms to account for the immunity of developing
cells against PatS-mediated inhibition of HetR. The first
possible explanation is based on the timing of patS
expression, which is dependent on and thus lags behind
hetR expression (Huang et al., 2004). Consequently, at
the beginning, these developing cells may contain high
HetR levels, but low PatS levels. The second explanation
is that while it is activating patS, HetR may also activate
the expression of a gene, such as patA (Liang et al., 1992;
Buikema and Haselkorn, 2001), the product of which may
have protective effects on HetR from PatS via protein–
protein interactions (see also the above discussion).
Unpublished results from S. Callahan’s lab suggest that
PatA may contribute to attenuating the effects of PatS on
the suppression of heterocyst differentiation. Within the
first 24 h of combined-nitrogen deprivation, a patS–, patA–
double mutant gives rise to the Mch phenotype similar to
that of a patS– single mutant. This phenotype is consistent
with the possibility that the PatS signal may be attenuated
by PatA and shows that the terminal heterocyst phenotype
© 2005 The Authors

observed in the patA mutant requires a functional patS
gene (D. Risser and S.M. Callahan, unpubl. results).
HetN-mediated heterocyst suppression
PatS is not the only inhibitory signal involved in heterocyst
development. HetN, a protein similar to ketoacyl reduc-
tase, also suppresses heterocyst development (Black and
Wolk, 1994; Callahan and Buikema, 2001; Borthakur
et al., 2005). In filaments grown in the presence of a
source of combined nitrogen, the patS gene was
expressed at a low basal level and HetN was detected in
higher levels than those occurring during the first few
hours of nitrogen step-down (Callahan and Buikema,
2001; Yoon and Golden, 2001; Li et al., 2002). Studies on
single and double mutants of patS and hetN have shown
that these two genes are involved in inhibiting cell differ-
entiation when a combined-nitrogen source is available
(Black and Wolk, 1994; Yoon and Golden, 1998; 2001;
Callahan and Buikema, 2001; Borthakur et al., 2005;
Fig. 2). The initial heterocyst pattern is normal when hetN
is not expressed, but becomes aberrant later, resulting in
the formation of Mch. This finding indicates that hetN is
required to maintain the heterocyst pattern (Callahan and
Buikema, 2001).
The two heterocyst-suppressing pathways mediated by
HetN and PatS, respectively, are independent of each
other (Borthakur et al., 2005). However, when both patS
and hetN are inactive, combined-nitrogen deprivation
induces the formation of increasingly large numbers of
heterocysts, and with time, almost all the cells along a
filament become heterocysts. Even the presence of nitrate
does not prevent this outcome, although ammonium still
represses heterocyst development. hetN- and patS-medi-
ated pathways are therefore the main mechanisms pre-
venting vegetative cells from developing into heterocysts
under nitrogen-limiting conditions.
The cells of the double mutant with inactivated patS and
hetN and those of the mutant that overexpresses
hetRR223W do not all become heterocysts at once: the
process of differentiation develops asynchronously and it
is only in the course of time that most of the cells become
heterocysts (Khudyakov and Golden, 2004; Borthakur
et al., 2005). This raises the possibility previously sug-
gested by other findings (for a review, see Meeks and
Elhai, 2002) that at a given time, all cells are not equally
competent to develop into heterocysts. The factors deter-
mining this developmental competence are not yet known.
Conclusion
The signalling network regulating the early steps in het-
erocyst differentiation and pattern formation is proposed
in Fig. 2. In short, heterocyst differentiation is not induced
© 2005 The Authors
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 59, 367–375
Multicellular behaviour in cyanobacteria 373
under conditions of combined-nitrogen sufficiency
because the 2-OG levels are low (Laurent et al., 2005),
free Ca2+ is mostly sequestered by CcbP (Zhao et al.,
2005), and the presence of PatS (although in low levels)
and HetN keeps the expression of the hetR gene under
check (Yoon and Golden, 2001; Li et al., 2002; Huang
et al., 2004; Khudyakov and Golden, 2004; Bothakur
et al., 2005). When filaments are starved of combined
nitrogen, 2-OG accumulates and enhances the DNA-bind-
ing activity of NtcA (Laurent et al., 2005), which in turn
leads to the expression of hetR (Muro-Pastor et al., 2002).
HetR and NtcA together provide the basic control mech-
anism for signal amplification and initiate heterocyst dif-
ferentiation via their auto- and mutual regulation (Black
et al., 1993; Ramasubramanian et al., 1996; Muro-Pastor
et al., 2002). HetR activates patS expression and PatS, or
one derivative of PatS, may diffuse along the filaments
and thus inhibit the differentiation of neighbouring cells
(Yoon and Golden, 1998; 2001; Huang et al., 2004;
Khudyakov and Golden, 2004). This cell-to-cell competi-
tion may be at least partly responsible for the formation of
heterocyst patterns. Early on, NtcA activates genes
involved in the use of alternative nitrogen sources (Cai
and Wolk, 1997), and if any such sources are available,
their activation by NtcA might lead to the reversal of the
heterocyst differentiation process before it becomes
irreversible.
Although several key signals have by now been defined,
the corresponding signalling pathways are still poorly
understood. It is not yet clear, for example, how the vari-
ous signalling pathways interact with each other, or how
HetR may counteract the effects of various signals so that
the developmental programme is initiated or not. We still
know very little about how signals and metabolites of
various kinds are exchanged between the cells located
along the filaments. Nevertheless, in view of the large
amount of information obtained to date, heterocyst devel-
opment provides a most exciting model for studies on
multicellular behaviour, a field involving a whole range of
processes that have been greatly overlooked to date in
prokaryotes.
Acknowledgements
The research conducted at our laboratories was supported
by the CNRS, an ATIP-Microbiologie programme, and
AFSSE (the Environnement et Santé programme). We thank
D. Risser and S. M. Callahan (University of Hawaii) for com-
municating unpublished results. We thank Dr Y. D. Tan, Dr L.
Wang (Huazhong Agricultural University, China), Dr J. W.
Golden (Texas A&M University) and Dr X. Xu (Institute of
Hydrobiology, China) for their helpful comments. We apolo-
gize to any authors whose publications have not been cited
because of the limited number of references allowed by the
Journal.
374 C.-C. Zhang et al.
References
–
Black, T.A., and Wolk, C.P. (1994) Analysis of a Het muta-
tion in Anabaena sp. strain PCC 7120 implicates a second-
ary metabolite in the regulation of heterocyst spacing. J
Bacteriol 176: 2282–2292.
Black, T.A., Cai, Y., and Wolk, C.P. (1993) Spatial expression
and autoregulation of hetR, a gene involved in the control
of heterocyst development in Anabaena. Mol Microbiol 9:
77–84.
Borthakur, P.B., Orozco, C.C., Young-Robbins, S.S.,
Haselkorn, R., and Callahan, S.M. (2005) Inactivation of
patS and hetN causes lethal levels of heterocyst differen-
tiation in the filamentous cyanobacterium Anabaena sp.
PCC 7120. Mol Microbiol 57: 111–123.
Buikema, W.J., and Haselkorn, R. (1991) Characterization of
a gene controlling heterocyst differentiation in the cyano-
bacterium Anabaena 7120. Genes Dev 5: 321–330.
Buikema, W.J., and Haselkorn, R. (2001) Expression of the
Anabaena hetR gene from a copper-regulated promoter
leads to heterocyst differentiation under repressing condi-
tions. Proc Natl Acad Sci USA 98: 2729–2734.
Cai, Y., and Wolk, C.P. (1997) Nitrogen deprivation of Ana-
baena sp. strain PCC 7120 elicits rapid activation of a gene
cluster that is essential for uptake and utilization of nitrate.
J Bacteriol 179: 258–266.
Callahan, S.M., and Buikema, W.J. (2001) The role of HetN
in maintenance of the heterocyst pattern in Anabaena sp.
PCC 7120. Mol Microbiol 40: 941–950.
Chen, H. (2005) Synthesis and characterization of non-
metabolizable analogues of 2-ketoglutaric acid. PhD
Thesis, Wuhan University, China.
Flores, E., and Herrero, A. (1994) Assimilatory nitrogen
metabolism and its regulation. In The Molecular Biology of
Cyanobacteria. Bryant, D.A. (ed.). Dordrecht: Kluwer, pp.
487–517.
Forchhammer, K. (2004) Global carbon/nitrogen control by
PII signal transduction in cyanobacteria: from signals to
targets. FEMS Microbiol Rev 28: 319–333.
Golden, J.W., and Yoon, H.S. (2003) Heterocyst development
in Anabaena. Curr Opin Microbiol 6: 557–563.
Herrero, A., Muro-Pastor, A.M., Valladares, A., and Flores,
E. (2004) Cellular differentiation and the NtcA transcription
factor in filamentous cyanobacteria. FEMS Microbiol Rev
28: 469–487.
Huang, X., Dong, Y., and Zhao, J. (2004) HetR homodimer
is a DNA-binding protein required for heterocyst differenti-
ation, and the DNA-binding activity is inhibited by PatS.
Proc Natl Acad Sci USA 101: 4848–4853.
Ikeda, T.P., Shauger, A.E., and Kustu, S. (1996) Salmonella
typhimurium apparently perceives external nitrogen limita-
tion as internal glutamine limitation. J Mol Biol 259: 589–
607.
Khudyakov, I.Y., and Golden, J.W. (2004) Different functions
of HetR, a master regulator of heterocyst differentiation in
Anabaena sp. PCC 7120, can be separated by mutation.
Proc Natl Acad Sci USA 101: 16040–16045.
Khudyakov, I., and Wolk, C.P. (1996) Evidence that the hanA
gene coding for HU protein is essential for heterocyst dif-
ferentiation in, and cyanophage A-4(L) sensitivity of, Ana-
baena sp. strain PCC 7120. J Bacteriol 178: 3572–3577.
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 59, 367–375
Khudyakov, I., and Wolk, C.P. (1997) hetC, a gene coding
for a protein similar to bacterial ABC protein exporters, is
involved in early regulation of heterocyst differentiation in
Anabaena sp. strain PCC 7120. J Bacteriol 179: 6971–
6978.
Kuhn, I., Peng, L., Bédu, S., and Zhang, C.-C. (2000) Devel-
opmental regulation of the cell division protein FtsZ in
Anabaena sp. strain PCC 7120, a cyanobacterium capable
of terminal differentiation. J Bacteriol 182: 4640–4643.
Laurent, S., Forchhammer, K., Gonzalez, L., Heulin, T.,
Zhang, C.-C., and Bédu, S. (2004) Cell-type specific mod-
ification of PII is involved in the regulation of nitrogen
metabolism in the cyanobacterium Anabaena PCC 7120.
FEBS Lett 576: 261–265.
Laurent, S., Chen, H., Bédu, S., Ziarelli, F., Peng, L., and
Zhang, C.-C. (2005) Nonmetabolizable analogue of 2-oxo-
glutarate elicits heterocyst differentiation under repressive
conditions in Anabaena sp. strain PCC 7120. Proc Natl
Acad Sci USA 102: 9907–9912.
Li, B., Huang, X., and Zhao, J. (2002) Expression of hetN
during heterocyst differentiation and its inhibition of hetR
up-regulation in the cyanobacterium Anabaena sp. PCC
7120. FEBS Lett 517: 87–91.
Li, J.-H., Laurent, S., Konde, V., Bédu, S., and Zhang, C.-C.
(2003) An increase in the level of 2-oxoglutarate promotes
heterocyst development in the cyanobacterium Anabaena
sp. PCC 7120. Microbiology 149: 3257–3263.
Liang, J., Scappino, L., and Haselkorn, R. (1992) The patA
gene product, which contains a region similar to CheY of
Escherichia coli, controls heterocyst pattern formation in
the cyanobacterium Anabaena 7120. Proc Natl Acad Sci
USA 89: 5655–5659.
Meeks, J.C., and Elhai, J. (2002) Regulation of cellular dif-
ferentiation in filamentous cyanobacteria in free-living and
plant-associated symbiotic growth states. Microbiol Mol
Biol Rev 66: 94–121.
Meeks, J.C., Elhai, J., Thiel, T., Potts, M., Larimer, F., Lamer-
din, J., et al. (2001) An overview of the genome of Nostoc
punctiforme, a multicellular, symbiotic cyanobacterium.
Photosynth Res 70: 85–106.
Michiels, J., Dirix, G., Vanderleyden, J., and Xi, C. (2001)
Processing and export of peptide pheromones and bacte-
riocin in Gram-negative bacteria. Trends Microbiol 9: 164–
168.
Muro-Pastor, A.M., Valladares, A., Flores. E., and Herrero,
A. (1999) The hetC gene is a direct target of the NtcA
transcriptional regulator in cyanobacterial heterocyst
development. J Bacteriol 181: 6664–6669.
Muro-Pastor, M.I., Reyes, J.C., and Florencio, F.J. (2001)
Cyanobacteria perceive nitrogen status by sensing intrac-
ellular α-ketoglutarate levels. J Biol Chem 276: 38320–
38328.
Muro-Pastor, A.M., Valladares, A., Flores, E., and Herrero,
A. (2002) Mutual dependence of the expression of the cell
differentiation regulatory protein HetR and the global nitro-
gen regulator NtcA during heterocyst development. Mol
Microbiol 44: 1377–1385.
Ramasubramanian, T.S., Wei, T.-F., Oldham, A.K., and
Golden, J.W. (1996) Transcription of the Anabaena sp.
strain PCC 7120 ntcA gene: multiple transcripts and NtcA
binding. J Bacteriol 178: 922–926.
© 2005 The Authors

Stanier, R.Y., and Cohen-Bazire, G. (1977) Photosynthetic
prokaryotes: the cyanobacteria. Annu Rev Microbiol 31:
225–274.
Tanigawa, R., Shirokane, M., Maeda, S.I., Omata, T., Tanaka,
K., and Takahashi, H. (2002) Transcriptional activation of
NtcA-dependent promoters of Synechococcus sp. PCC
7942 by α-ketoglutarate in vitro. Proc Natl Acad Sci USA
99: 4251–4255.
Thiel, T., and Pratte, B. (2001) Effect on heterocyst differen-
tiation of nitrogen fixation in vegetative cells of the cyano-
bacterium Anabaena variabilis ATCC 29413. J Bacteriol
183: 280–286.
Thiel, T., Lyons, E.M., Erker, J.C., and Ernst, A. (1995) A
second nitrogenase in vegetative cells of a heterocystform-
ing cyanobacterium. Proc Natl Acad Sci USA 92: 9358–
9362.
Torrecilla, I., Leganés, F., Bonilla, I., and Fernández-Piñas,
F. (2004) A calcium signal is involved in heterocyst differ-
entiation in the cyanobacterium Anabaena sp. PCC7120.
Microbiology 150: 3731–3739.
Vàzquez-Bermùdez, M.F., Herrero, A., and Flores, E. (2000)
Uptake of α-ketoglutarate in Synechococcus strains trans-
formed with the Escherichia coli kgtP gene. J Bacteriol
182: 211–215.
Vàzquez-Bermùdez, M.F., Herrero, A., and Flores, E. (2002)
α-Ketoglutarate increases the binding affinity of the NtcA
(nitrogen control) transcription factor for the Synechococ-
cus glnA promoter. FEBS Lett 512: 71–74.
Vàzquez-Bermùdez, M.F., Herrero, A., and Flores, E. (2003)
Carbon supply and 2-oxoglutarate effects on expression of
nitrate reductase and nitrogen-regulated genes in Syn-
echococcus sp. strain PCC 7942. FEMS Microbiol Lett
221: 155–159.
Wilcox, M., Mitchison, G.J., and Smith, R.J. (1973a) Pattern
formation in the blue-green alga Anabaena. I. Basic mech-
anisms. J Cell Sci 12: 707–723.
Wilcox, M., Mitchison, G.J., and Smith, R.J. (1973b) Pat-
© 2005 The Authors
Journal compilation © 2005 Blackwell Publishing Ltd, Molecular Microbiology, 59, 367–375
Multicellular behaviour in cyanobacteria 375
tern formation in the blue-green alga Anabaena. II.
Controlled proheterocyst regression. J Cell Sci 13: 637–
649.
Wolk, C.P. (2000) Heterocyst formation in Anabaena. In
Prokaryotic Development. Brun, Y.V., and Shimkets, L.J.
(eds). Washington, DC: American Society for Microbiology,
pp. 83–104.
Wolk, C.P., Ernst, A., and Elhai, J. (1994) Heterocyst
metabolism and development. In The Molecular Biology
of Cyanobacteria. Bryant, D.A. (ed.). Dordrecht, the
Netherlands: Kluwer Academic Publishers, pp. 769–
823.
Wong, F.C., and Meeks, J.C. (2001) The hetF gene product
is essential to heterocyst differentiation and affects HetR
function in the cyanobacterium Nostoc punctiforme. J Bac-
teriol 183: 2654–2661.
Wu, X., Liu, D., Lee, M.H., and Golden, J.W. (2004) patS
minigenes inhibit heterocyst development of Anabaena sp.
strain PCC 7120. J Bacteriol 186: 6422–6429.
Xu, X.D., and Wolk, C.P. (2001) Role for hetC in the transition
to a nondividing state during heterocyst differentiation in
Anabaena sp. J Bacteriol 183: 393–396.
Yoon, H.S., and Golden, J.W. (1998) Heterocyst pattern for-
mation controlled by a diffusible peptide. Science 282:
935–938.
Yoon, H.S., and Golden, J.W. (2001) PatS and products of
nitrogen fixation control heterocyst pattern. J Bacteriol 183:
2605–2613.
Zhao, Y., Shi, Y., Zhao, W., Huang, X., Wang, D., Brown, N.,
et al. (2005) CcbP, a calcium-binding protein from Ana-
baena sp. PCC 7120, provides evidence that calcium ions
regulate heterocyst differentiation. Proc Natl Acad Sci USA
102: 5744–5748.
Zhou, R., Wei, X., Jiang, N., Li, H., Dong, Y., Hsi, K.L., and
Zhao, J. (1998) Evidence that HetR protein is an unusual
serine-type protease. Proc Natl Acad Sci USA 95: 4959–
4563.