Circadian Rhythm of Nitrogenase Gene Expression in the Diazotrophic Filamentous Nonheterocystous Cyanobacterium Trichodesmium sp. Strain IMS 101

Abstract

Recent studies suggested that the daily cycle of nitrogen fixation activity in the marine filamentous nonheterocystous cyanobacterium Trichodesmium sp. is controlled by a circadian rhythm. In this study, we evaluated the rhythm of nitrogen fixation in Trichodesmium sp. strain IMS 101 by using the three criteria for an endogenous rhythm. Nitrogenase transcript abundance oscillated with a period of approximately 24 h, and the cycle was maintained even under constant light conditions. The cyclic pattern of transcript abundance was maintained when the culture was grown at 24 and 28.5°C, although the period was slightly longer (26 h) at the higher temperature. The cycle of gene expression could be entrained with light-dark cues. Results of inhibitor experiments indicated that transcript abundance was regulated primarily by transcription initiation, rather than by degradation. The circadian rhythm, the first conclusively demonstrated endogenous rhythm in a filamentous cyanobacterium, was also reflected in nitrogenase MoFe protein abundance and patterns of Fe protein posttranslational modification-demodification.

A circadian rhythm is an endogenous biological oscillation that persists in constant environmental conditions with a period of about 24 h. In addition, circadian rhythms are temperature compensated and can be entrained to the solar day (24 h) by environmental cues (20, 35, 39). It had long been thought that circadian rhythms were restricted to eukaryotes (11). However, the dogma was seriously challenged a decade ago when several reports provided strong evidence of the existence of circadian rhythms in a few cyanobacterial genera (15, 20, 25, 36). More recently, a circadian rhythm has been convincingly demonstrated for a unicellular cyanobacterium (Synechococcus sp. strain AMC149) which was transformed with bacterial luciferase reporter genes (14, 22, 23). Circadian rhythms have not yet been conclusively demonstrated for filamentous cyanobacteria.

Trichodesmium spp. are marine filamentous nonheterocystous cyanobacteria found worldwide in tropical and subtropical oceans (5). The significance of Trichodesmium is twofold. From an ecological point of view, Trichodesmium plays an important role in global nitrogen fixation. From a biological perspective, the mechanisms allowing photosynthesis and oxygen-sensitive nitrogen fixation, two seemingly incompatible processes, to proceed simultaneously without obvious spatial and temporal separation are intriguing (4, 13, 41). Recently, it has been suggested that nitrogen fixation and photosynthesis may be separated spatially in Trichodesmium (2, 12); however, more direct evidence is needed to substantiate this hypothesis. Saino and Hattori (32) first suggested that nitrogen fixation in Trichodesmium may be due to an endogenous rhythm. Evidence of rhythmic respiration and nifHDK transcript abundance has also been reported in field populations of Trichodesmium spp. (31, 40). Our previous study provided strong evidence that nitrogen fixation in Trichodesmium sp. strain IMS 101 was at least partially under the control of a circadian rhythm (6).

In the past, because of the difficulties in both culture and RNA analysis techniques with Trichodesmium, few studies have studied regulation of nitrogen fixation in this organism at the gene transcription level. In this study, we have demonstrated that the circadian rhythm of nitrogenase gene expression in Trichodesmium sp. strain IMS 101 is at transcriptional and translational as well as enzymatic activity levels. To our knowledge, this is the first evaluation of the three criteria that proves the existence of a circadian rhythm in a filamentous cyanobacterium.

MATERIALS AND METHODS

Culture and growth conditions.

Trichodesmium sp. strain IMS 101 was originally isolated from western Atlantic Ocean waters near North Carolina (29). The cultures were grown in YBCII artificial seawater medium as previously described (6). In comparison to many cyanobacteria, Trichodesmium is difficult to cultivate, grows slowly, and attains only low concentrations even in the stationary phase of growth in laboratory conditions. During the course of this study, stock and control cultures, grown in 800 ml of the medium in 2.8-liter Fernbach flasks, were maintained at 26.5°C with a 12-h light (L)–12-h dark (D) cycle. The L phase was from 1000 to 2200, and the D phase was from 2200 to 1000, all local time. Irradiance was provided by cool-white fluorescent lamps at about 100 μmol of photons m⁻² s ⁻¹. Biomass of the cultures was estimated by measuring chlorophyll a (Chl a) (38). Batch cultures of Trichodesmium sp. strain IMS 101 at mid- to late-logarithmic growth stage, normally 2 to 3 weeks after the inoculation (typical Chl a concentration around 0.2 mg/liter) were used for all experiments.

Nitrogenase activity.

Nitrogenase activity was assayed by the acetylene reduction technique (3). Ten-milliliter aliquots were placed in 15-ml serum bottles. The vials were sealed with silicone rubber stoppers, and 0.75 ml of air in the headspace was replaced with purified acetylene at each time point. Samples were then incubated under different experimental conditions. Ethylene production was measured by gas chromatography (Shimadzu GC-14A equipped with a flame ionization detector) at 1- to 2-h intervals for up to 4 h (2 h for most samples). Treatments and cultures were duplicated for all experiments. The ethylene production rate was normalized to Chl a.

RNA extraction.

At each sampling point, 20 to 100 ml from each treatment was rapidly filtered onto 25-mm MAGNA nylon membranes (MSI; pore size, 10 μm) and immediately lysed in 1 ml of RNA extraction buffer (5% Triton X-100, 10% sucrose, 20 mM EDTA, 50 mM Tris, 100 mM dithiothreitol). The lysate was then extracted twice with 0.8 ml of low-pH phenol (pH 4.5) equilibrated with the same volume of chloroform and then extracted once with 0.8 ml of chloroform. Total RNA was ethanol precipitated, washed, vacuum dried, resuspended in RNase-free H₂O, and stored at −20°C until analyzed.

Northern blots.

Northern blot analysis was performed to determine the time course abundance of nifHDK transcripts of Trichodesium sp. strain IMS 101 cultures grown under different conditions. RNA isolations and Northern blot analyses were carried out according to widely used protocols (1, 33) with minor modifications. Total RNA extracted from equal amounts of biomass (approximately 1 μg of Chl a per lane) was fractionated by electrophoresis on a 1% agarose gel with 1 M formaldehyde. The intensities of ethidium bromide-stained rRNA bands were examined visually to ensure that equal amounts of total RNA were loaded for all samples. The RNA was then transferred to a charged nylon membrane (Nytran; Schleicher & Schuell) and was fixed to the membrane by being baked in a vacuum oven at 80°C for 2 h. The blots were hybridized overnight at 42°C with an α-³²P-labeled nifH DNA probe cloned from Trichodesmium sp. strain IMS 101 (9). The blots were then washed twice with 2× SSC (0.3 M NaCl and 0.03 M sodium citrate)–0.1% SSC for 30 min at 42°C, rinsed with 0.2× SSC, and exposed to X-ray film (Sterling X-ray film; Bio World). Preparation of the α-³²P-labeled homologous DNA probe, a 1.1-kb fragment containing nifH cloned from Trichodesmium sp. strain IMS 101(pBD511), is described elsewhere (9).

Protein assays.

Protein analysis was performed according to previously described protocols (42) with minor modifications. At the midpoint of each acetylene reduction measurement period, 10-ml subsamples of the cultures were filtered onto 5-μm-pore-size membrane filters (Millipore). The filters were immediately placed into protein solubilization buffer (10% sodium dodecyl sulfate, 75 mM dithiothreitol, 50 mM Tris-HCl, 10% glycerol, 10 mM EDTA, 0.1% bromphenol blue) and boiled at 95°C for 10 min before being stored at −20°C. Proteins extracted from equal amounts of biomass a (about 1 μg of Chl a per lane) were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to polyvinylidene difluoride membranes (Immobilon-P; 0.45-μm pore size; Millipore), and then challenged with antisera raised against the Fe protein and FeMo protein of nitrogenase (generously provided by P. Ludden, University of Wisconsin, Madison).

Densitometric analysis of Northern and Western blots.

Original Northern autoradiograms and Western blots were scanned into a computer with Adobe PhotoShop software (version 3.0) and a Hewlett-Packard ScanJet 4C scanner (optical resolution, 600 by 600 dpi). NIH Image software (Scientific Computing Resource Center, National Institutes of Health) was used to perform densitometric analysis of labeled bands. To minimize artifacts introduced during the quantitative analysis, we compared samples only on the same gel or same batch of gels and plotted the results on a relative scale (percentage of maximum value).

Nitrogenase gene expression in cultures grown under different L-D regimens.

To assess nitrogenase gene expression in cultures grown under different L-D regimens, Trichodesmium sp. strain IMS 101 cultures were grown for 2 to 3 weeks under regular culture conditions except that the incubating temperature was 28.5°C. At the beginning of the experiments, three subcultures were established at the onset of the L period. The L-D subcultures were continuously grown under regular conditions with a 12-h L–12-h D regimen; the L-L subcultures were transferred to a continuously illuminated incubator with the same L intensity used for L-D subcultures. The D-D subcultures were placed in flasks wrapped with two layers of black electrical tape (Scotch; 3M) and then maintained in a D incubator. D-D samples were removed from the D incubator and processed immediately in dim L. The incubation temperature for all subcultures was 28.5°C. The experiment lasted for more than 60 h, during which samples were taken for RNA, protein, and nitrogenase activity assays.

Temperature compensation of rhythm of nitrogenase gene expression.

To examine whether the nitrogenase gene expression rhythm was temperature compensated, Trichodesmium sp. strain IMS 101 cultures were incubated at 26.5°C for 2 weeks after inoculation. The culture was separated into two subculture replicates and incubated at 24°C, starting 1 week before the experiment. The temporal variation of nitrogenase activity, nifHDK transcript abundance, and the MoFe protein abundance were determined, and patterns were compared with those of cultures grown at 28.5°C.

Entrainment of rhythmic nitrogenase gene expression to solar L-D cycle by L-D cues.

To determine if rhythmic nitrogenase gene expression can be entrained to solar L-D cycles by environmental cues, we initially grew the Trichodesmium sp. strain IMS 101 culture for 3 weeks under regular conditions (12-h L–12-h D, 26.5°C). The culture was then divided into two subcultures and placed under continuous L for at least three complete diel cycles (72 h). Both subcultures were given a 12-h D–12-h L–12-h D sequence as an environmental cue to reset the phase, followed by 30 h of continuous illumination, during which various measurements were made to verify whether the rhythmic nitrogenase gene expression was entrained by L-D signals. A phase difference of 90° (6 h) between the two subcultures was imposed during the entraining stage.

Turnover of nifH transcripts and the MoFe protein.

In order to determine the pattern of nifH transcripts and the MoFe protein turnover during the day, rifampin and chloramphenicol (final concentration, 50 μg/ml) were added separately at different times of day to actively growing Trichodesmium sp. strain IMS 101 cultures that were maintained under regular L-D conditions. The concentrations of the inhibitors used in the experiments were comparable to those used in studies of Trichodesmium spp. and other cyanobacteria (24, 27, 42). Samples were taken at short intervals for RNA or protein analysis.

Modification of Fe protein of nitrogenase.

To characterize the modification of the Fe protein of nitrogenase in cultures grown under different L-D conditions, an actively growing Trichodesmium sp. strain IMS 101 culture was split into two subcultures, one of which (L-D) remained in regular incubation conditions while the other (L-L) was transferred to continuously illuminated conditions. On the fourth day after subculturing, 24-h time course measurements of nitrogenase activity and protein abundance were carried out for both subcultures.

RESULTS

Nitrogenase gene expression in cultures grown under different L-D regimens.

The results of time course measurements of nitrogenase activity, nifH transcript abundance, and the MoFe protein abundance in Trichodesmium sp. strain IMS 101 cultures grown under different L-D regimens are summarized in Fig. 1 to 3.

FIG. 1.

Circadian rhythm of nitrogenase activity in Trichodesmium sp. strain IMS 101 cultures grown under different L-D regimens at 28.5°C. Measurements were duplicated (open and filled symbols). L-D cultures (open and filled circles) were incubated under a 12-h-L–12-h-D regimen; L-L cultures (open and filled triangles) were incubated under constant illumination beginning at time zero; D-D cultures (open and filled squares) were incubated under constant darkness since the preceding D phase and throughout the experiment. ∗, note that the nitrogenase activities of L-L cultures (triangles) on days 2 and 3 are scaled to the right y axis.

FIG. 3.

Circadian rhythm of MoFe protein abundance in Trichodesmium sp. strain IMS 101 cultures grown under different L-D regimens at 28.5°C. Solid boxes indicate the subjective D phase under constant illumination. (A) Western immunoblot analyses were performed with antiserum raised against the MoFe protein of Rhodospirillum rubrum. Numbers at the top indicate time. (B) Densitometric analysis of the Western blot results in panel A.

In L-D cultures, nitrogen fixation activity was confined to the L period with little change in rate for over 3 days. In L-L cultures, while active nitrogen fixation was essentially confined to the subjective L phase (the period corresponding to the original L phase of regular L-D cultures), there was an approximately 2-h shift of phase (postponed) for each successive diel cycle. The overall rates of nitrogen fixation in the L-L cultures decreased rapidly over time, dropping by 85% or more each day (Fig. 1; note that the nitrogenase activities of L-L cultures on day 2 and day 3 are scaled to the right y axis due to damping of activity). No nitrogen fixation was detected in D-D cultures. The results of Northern blot analysis indicated that, in L-D cultures, nifH transcripts appeared as early as 5 h before the onset of the L period, reached the maximum level well before the midpoint of the L period, and decreased thereafter, disappearing at the beginning of the D period (Fig. 2A). In L-L cultures, while the pattern of cyclic abundance of nifH transcripts resembled that of L-D cultures and net accumulation of the transcripts occurred mainly during the subjective L phase, there appeared to be a daily shift in the phase of the peak, suggestive of a free-running period that was longer than 24 h (Fig. 2). In addition, though the overall level of the transcripts dropped more than 80% after the first diel cycle, the transcripts continued to be present during the subjective D phases in L-L cultures (Fig. 2). Although D-D cultures were completely shielded from L since the prior D period, the pattern of nifH transcript abundance in them was the same as that of the regular L-D culture in the first diel cycle; however, no transcripts were detected during the rest of the experiment (Fig. 2). Western immunoblotting with antisera raised against the MoFe protein of nitrogenase revealed that the MoFe protein abundance pattern reflected the pattern of nifH transcripts (Fig. 3). In L-D cultures, the level of MoFe protein increased at the beginning of the L phase, reached a maximum level 2 to 4 h before the D phase, decreased rapidly over 4 h, and then remained at a minimum level for the rest of the D phase (Fig. 3). A similar pattern was observed for the MoFe protein in L-L cultures; however, the daily shift in the phase of the peak was not as consistent as that of nifH transcripts (Fig. 3). In D-D cultures, the pattern of the MoFe protein abundance was generally similar to that of the MoFe protein abundance in L-D cultures in most of the first diel cycle (Fig. 3). However, as for nifH transcripts, the MoFe protein was not detected after the first day. In addition, unlike nifH transcripts that were similar in abundance in L-D and L-L cultures in the first subjective L phase, the MoFe protein appeared to be much less abundant in D-D than in L-D and L-L cultures (Fig. 3A).

FIG. 2.

Circadian rhythm of nifHDK transcript abundance in Trichodesmium sp. strains IMS 101 cultures grown under different L-D regimens at 28.5°C. (A) Three bands hybridized to the probe; they correspond to 1.1-kb (nifH), 2.8-kb (nifHD), and 4.5-kb (nifHDK) transcripts. Numbers at the top indicate time. (B) Densitometric analysis of the Northern blot results of nifH transcript abundance from panel A.

Temperature compensation of rhythm of nitrogenase gene expression.

Natural populations of Trichodesmium live in waters that range in temperature from the lower to the upper 20°C’s. The rhythmic patterns of nitrogenase activity, nifH transcripts, and the MoFe protein displayed in Trichodesmium sp. strain IMS 101 cultures grown at 28.5°C were examined to determine if they were maintained when the cultures were grown at a different temperature within the bacterium’s physiological limit. Time course measurements of nitrogenase activity and abundance of nifHDK transcripts and the MoFe protein of nitrogenase from the cultures grown at 24°C are summarized in Fig. 4 and 5.

FIG. 4.

Circadian rhythm of nitrogenase activity, as well as densitometric analysis of nifH transcripts and MoFe protein in Trichodesmium sp. strain IMS 101 cultures grown under constant illumination at 24°C. The boldly outlined box indicates the subjective D phase. Measurements were duplicated for nitrogenase activity.

FIG. 5.

Circadian rhythm of nifHDK transcripts (A) and MoFe protein abundance (B) in Trichodesmium sp. strain IMS 101 cultures grown under constant illumination at 24°C. The boldly outlined boxes indicate the subjective D phase. Numbers indicate time.

Rhythmic nitrogenase activity was maintained at 24°C with a period close to 24 h (Fig. 4). Active nitrogen fixation was confined to the subjective L phase, and the overall rates of nitrogen fixation in the second and third subjective L phases were about 40% of that of the first subjective L phase. Results of Northern blots revealed a rhythmic pattern of nifH transcript abundance which mimicked that in L-L cultures grown at a higher temperature (28.5°C) (Fig. 5A). The length of the period was substantially different from 24 h, partly because of the low resolution resulting from a relatively long sampling interval (4 h) (Fig. 4). The cyclic pattern of the MoFe protein abundance was well maintained at 24°C (Fig. 5B). The pattern of the MoFe protein abundance had a period of about 24 h and was clearly coupled with nitrogenase activity (Fig. 4).

Entrainment of rhythmic nitrogenase gene expression to solar L-D cycle by L-D cues.

Experiments were carried out to examine whether L-D signals can reset the phase of the Trichodesmium circadian rhythm, i.e., if the rhythmic nitrogenase gene expression can be entrained by environmental cues such as L-D signals. The results are summarized in Fig. 6 and 7.

FIG. 6.

Entrainment of nitrogenase activity rhythm by 12-h D–12-h L–12-h D pulses in Trichodesmium sp. strain IMS 101. A 6-h phase difference was imposed during the entrainment. Subculture A: subjective L phase, 1600 to 0400; subjective D phase, 0400 to 1600. Subculture B: subjective L phase, 2200 to 1000; subjective D phase, 1000 to 2200. Measurements were duplicated for nitrogenase activity.

FIG. 7.

Entrainment of nifH transcription (A) and MoFe protein abundance (B) rhythm by 12-h D–12-h L–-12-h D pulses in Trichodesmium sp. strain IMS 101. A 6-h phase difference was imposed during the entrainment. Subculture A: subjective L phase, 1600 to 0400; subjective D phase, 0400 to 1600. Subculture B: subjective L phase, 2200 to 1000; subjective D phase, 1000 to 2200. The boldly outlined boxes indicate the subjective D phase.

Rhythmic nitrogen fixation with a period of about 24 h was observed in two subcultures after the subcultures were entrained with separate L-D cues that were out of phase by 6 h (Fig. 6). Nitrogenase activity reached a maximum and virtually disappeared 3 to 5 and 12 h after the onset of the L phase, respectively. A cyclic abundance of nifH transcripts and the MoFe protein was demonstrated in both subcultures (Fig. 7). For each measured parameter, the 6-h phase difference was generally maintained between the two subcultures.

Turnover of nifH transcripts and the MoFe protein.

In order to characterize the regulation of nitrogenase gene expression at both transcriptional and translational levels during the diel cycle, the half-lives of nifH transcripts and the MoFe protein were estimated. Results of Northern blot analysis are summarized in Fig. 8. While the estimated half-life of nifH transcripts varied depending on sampling time in the rifampin-treated cultures, it was always much shorter in chloramphenicol-treated cultures than in rifampin-treated cultures (Table 1). There was very little change in the abundance of the MoFe protein in cultures treated with chloramphenicol during the early L phase; however, a significant decrease in abundance occurred in the culture treated with chloramphenicol after the midpoint of the L phase, with an estimated half-life of 81 min (Fig. 9).

FIG. 8.

Northern blot assay of nifH transcript turnover in Trichodesmium sp. strain IMS 101 during L period (1000 to 2200). (A) Treatments started at 1000, the onset of L period. (B) Treatments started at 1600 (above) and 1800 (below). Numbers above blots indicate time.

TABLE 1.

Half-lives of nifH transcripts in the presence of rifampin or chloramphenicol during the L period (1000 to 2200) in Trichodesmium sp. strain IMS 101 culture^a

Time when inhibitor added	Half-life (min) with:
	Rifampin	Chloramphenicol
1000	60	35
1600	74	52
1800	45	NAb

^aThe final concentration for both rifampin and chloramphenicol was 50 μg/ml. See Fig. 8 for corresponding Northern blot analysis of nifH transcripts. 

^bNA, not analyzed. 

FIG. 9.

Western blot immunoassay of half-lives of MoFe protein in Trichodesmium sp. strain IMS 101. During the L period (1000 to 2200), chloramphenicol (50 μg/ml) was added at 1000 (A) and 1600 (B). Numbers indicate time.

Modification of Fe protein of nitrogenase.

As for other diazotrophic cyanobacteria, the nitrogenase Fe protein in Trichodesmium changes in apparent molecular mass under certain conditions, resulting in a modified form (the upper band in the Western blot) and a demodified form (the lower band in the Western blot) (42). Diel abundance and modification of the nitrogenase Fe protein have been documented in field populations and laboratory cultures of Trichodesmium spp. (6, 26, 42). We examined whether modification of the Fe protein was under the control of a circadian rhythm. While the duration of the active nitrogen-fixing period in L-L subcultures did not differ significantly from that in L-D subcultures, it was about 8 h out of phase and the overall rates were less than 15% of those in L-D subcultures (Fig. 10A). A Western blot showed that the appearance of the lower-apparent-molecular-mass form of the Fe protein as well as the amount is related to the period of active nitrogen fixation in both L-D and L-L subcultures (Fig. 10B).

FIG. 10.

Nitrogenase activity (B) and Western blot immunoassay of Fe protein (A) in Trichodesmium sp. strain IMS 101 cultures grown in different L-D regimens. L-D culture was incubated under 12-h L–12-h D; L-L cultures were incubated under continuous L for 72 h prior to zero time point (1000 local time). Numbers at top of panel A indicate time.

DISCUSSION

Previous results provided strong evidence that nitrogen fixation in Trichodesmium sp. strain IMS 101 was likely to be under the control of a circadian rhythm (6). In this study, we investigated the periodicity of nitrogen fixation at the levels of gene transcription and translation, as well as enzymatic activity. The cyclic patterns of each parameter were examined to confirm that a circadian rhythm controls nitrogen fixation and to determine the levels at which the rhythm is manifested.

Oscillations of nitrogenase gene expression persisted under constant L for at least 60 h (Fig. 1, 2A, and 3A). The estimated average periods of the rhythms were around 24 and 26 h at 24 and 28.5°C, respectively, suggesting that the rhythm was temperature compensated (Fig. 1, 2A, and 4). The periods are within the range of periods for circadian rhythms reported for other organisms (14, 19, 39). All rhythms were entrained to the daily L-D cycle with a period of about 24 h following a 12-h D–12-h L–12-h D cue (Fig. 6 and 7). Furthermore, the imposed phase difference between the two subcultures in the entrainment experiment was maintained. The results clearly indicated that the nitrogenase gene expression rhythm of Trichodesmium sp. strain IMS 101 satisfies all three criteria for a circadian rhythm, i.e., persistence under constant conditions, temperature compensation of the period, and entrainment to the daily L-D cycle by environmental cues (35, 39).

Transcriptional analysis of the nifHDK operon of Trichodesmium sp. strain IMS 101 has been reported in detail previously (9, 41). When blots were hybridized with a homologous nifH probe (cloned from Trichodesmium sp. strain IMS 101), Northern blot analysis of total RNA from Trichodesmium sp. strain IMS 101 cultures during the L phase showed three distinct bands of approximately 1.1, 2.8, and 4.5 kb (Fig. 2A). These bands correspond to nifH, nifHD, and nifHDK transcripts, respectively, which is consistent with studies of other cyanobacteria (8, 16). In our Northern blots, not only did the 1.1-kb nifH bands have the densest signal while the 4.5-kb nifHDK band had the least dense signal, but all three bands increased or decreased proportionally to one another. Therefore, the abundance pattern of nifH transcripts was representative of nifHDK operon transcription.

At both 24 and 28.5°C, the patterns of abundance of nifH transcripts, the MoFe protein, and nitrogenase activity were similar. The nifH transcript rhythm was offset by 2 to 4 h from protein abundance and nitrogenase activity (Fig. 1, 2B, 3B, and 4). This suggests that nitrogenase protein expression is primarily regulated at the transcriptional level. The expression of the nifHDK operon of Trichodesmium sp. strain IMS 101 appeared to be tightly controlled at the transcriptional level, with the net transcript accumulation limited to about 6 h in L-D and D-D cultures and about 8 h in L-L cultures (Fig. 2B and 4). As in a Cyanothece sp. (8), the similarity of transcriptional patterns between the cultures in different L-D regimens suggests that the accumulation of nitrogenase gene transcripts is not directly regulated by L. The estimated half-lives of nifH transcripts did not vary dramatically during the L period (Fig. 8; Table 1). The half-lives of nifH transcripts are much longer than those of other cyanobacterial gene transcripts documented so far (24, 30). The low turnover rate of nifH mRNA may be reasonable given that active gene transcription is initiated several hours before the onset of the L period. The concentration of rifampin used in this study is similar to that used in other studies, and a higher concentration of rifampin (65 μg/ml) did not affect the results significantly (data not shown). Therefore, it appears that transcription initiation and rate, rather than transcript degradation, are the major mechanisms involved in nifHDK regulation in Trichodesmium sp. strain IMS 101. It should be pointed out that the interpretation of the turnover experiment might be complicated by the fact that the inhibitors could have indirect effects on degradative processes or even the clock proteins themselves.

Previous studies indicated that the nitrogenase Fe protein of Trichodesmium underwent diel modification and demodification and that the appearance and amount of the lower-apparent-mass form (demodified form) were exclusively related to the time of day when nitrogen fixation is active (6, 27, 42). The modification-demodification of the Fe protein may play an important role in the regulation of nitrogenase activity in Trichodesmium (26, 27, 42). Western blots showed that modification-demodification of the Fe protein was also under the control of the circadian rhythm (Fig. 10). The cycle of modification-demodification of the Fe protein is maintained under constant L conditions, and the Fe protein is not modified when cultures are shifted to L during a D phase (data not shown). The modification-demodification might be controlled by circadian expression of an enzyme involved in demodifying (activating) the Fe protein. Western blot analysis also revealed a similar pattern in the MoFe protein abundance from cultures grown under different L-D regimens (Fig. 3). Specifically, the MoFe protein in a D-D culture displayed the regular pattern of increase-maximum-decrease during its first subjective L phase. In contrast, in Cyanothece sp., the pattern of the MoFe protein abundance was different under different L-D regimens (8). These results strongly suggest that L might not be directly involved in translational regulation of nitrogenase gene expression in Trichodesmium sp. strain IMS 101. However, L may be involved in the degradation of nitrogenase, as the half-life of the MoFe protein decreased to 81 min in the late L period compared to little degradation in the early L period (Fig. 9). Enhanced nitrogenase degradation in the late L period was also found in field populations of Trichodesmium spp. (42). The switch-off of nitrogenase activity in the late L period, as reported for both field populations (42) and cultures of Trichodesmium spp. (6, 26), may also be controlled by factors other than L. Unlike the situation for Cyanothece (8), Gloeothece (10), and Synechococcus (7) spp., nitrogenase proteins of Trichodesmium sp. strains IMS 101 were present at significant levels throughout the non-nitrogen-fixing period despite apparent proteolytic activity (Fig. 3A, 5B, and 10A).

There were substantial discrepancies among the abundances of nifH transcripts and the MoFe protein and nitrogenase activity in Trichodesmium sp. strain IMS 101 cultures grown in L-L conditions. At 28.5°C, while the overall nitrogenase activity of L-D cultures changed little over a 3-day time course, it decreased rapidly in L-L cultures, by 85% or more each day (Fig. 1). During the same time, however, the relative abundance of nifH transcripts decreased to a lesser degree than did nitrogenase activity in L-L cultures, remaining just below 20% of that of day 1 on days 2 and 3 (Fig. 2B). The decrease of the MoFe protein in L-L cultures was even less over the same period of time (Fig. 3B). The discrepancies were also found in L-L cultures grown at 24°C, though they were smaller than those in L-L cultures grown at 28.5°C (Fig. 4). These results suggest that other mechanisms, such as proteolysis and posttranslational modification of the nitrogenase Fe protein, may play an important role in regulation of nitrogen fixation in Trichodesmium sp. strain IMS 101. It also suggests that the signals that trigger and control these mechanisms may not directly affect nifHDK gene transcription.

Incubation temperature appeared to have a significant impact on the manifestation of rhythmic nitrogenase gene expression. The amplitude and periodicity of nitrogenase gene expression of L-L cultures grown at 28.5°C varied much more than those of L-L cultures grown at 24°C. Furthermore, when cultures were grown at 28.5°C, the overall nifHDK transcript abundance decreased substantially, with signs of enhanced transcript degradation (Fig. 2A). We also found that at 31°C (higher than the upper limit of the temperature range of the waters which natural populations of Trichodesmium inhabit) the rhythms were no longer maintained after the first 24 h (no nitrogenase activity, though nifH transcripts and the MoFe protein were still detectable), while the controls (L-D) seemed to function normally (data not shown). In addition, nifHDK transcripts of L-L cultures, unlike those of L-D cultures, did not completely disappear during the subjective D phase (the period corresponding to the original D phase). It has been reported that for Chlamydomonas, the rhythmic degradation of the tufA gene transcript requires a D period (17). We have yet to determine how the L-D regimen and temperature affect the turnover of nifHDK transcripts.

We found that nifHDK transcripts of a Trichodesmium sp. strain IMS 101 culture treated with the protein synthesis inhibitor chloramphenicol were degraded much faster than those of a culture treated with the RNA synthesis inhibitor rifampin (Fig. 8). The decay was even more evident with the larger nifHDK transcripts (4.5 and 2.8 kb) as both bands disappeared completely after only 15 min of treatment with chloramphenicol. The results indicate that protein synthesis or, more specifically, peptide elongation could be important for the stability of nifHDK transcripts. Considering the time scale of this decay as well as the results of the rifampin treatment, we think that it is possible that coverage by ribosomes may protect nifHDK mRNA from both endonucleolytic and processive nucleolytic degradation, as is the case with some bacteria (28).

While unicellular cyanobacteria seem to grow well in continuous L for a relatively long time (18, 34, 36), Trichodesmium sp. strain IMS 101 grew very little, apparently experiencing immense stress in such conditions. The culture began to form spherical colonies after 2 to 3 days in constant illumination while turning more reddish in color, and it rarely survived for more than 4 weeks. We have already shown that nitrogenase gene expression is severely dampened under L-L conditions; however, it is not clear whether decreased nitrogen fixation rates are responsible for the poor growth in continuous L, or whether it is merely a secondary impact resulting from impairment of other important metabolic activities such as photosynthesis and/or respiration under such conditions. The latter scenario may have been the case for the diazotrophic filamentous nonheterocystous cyanobacterium Oscillatoria sp. strain 23, as its photosynthesis ceased under continuous L at a certain L intensity (37).

It has been proposed that, in some cyanobacteria, circadian rhythms or the cell cycle may be the underlying mechanism that temporally separates the two incompatible processes of photosynthesis and nitrogen fixation in the same cells (14, 25, 36). This is not likely to be the case in Trichodesmium, where nitrogen fixation and photosynthesis coincide temporally, either within the same cell (41) or in different cells (2, 12). The role of the circadian rhythm in nitrogen fixation of Trichodesmium spp., therefore, is not to avoid simultaneous nitrogen fixation and photosynthesis, but the rhythm may be critical for other metabolic functions.

Prior to this study, Oscillatoria sp. strain 23 was the only other filamentous nonheterocystous cyanobacterium whose nitrogen fixation was reported to be possibly under the control of a circadian rhythm (36). However, the existence of the rhythm in Oscillatoria sp. was not examined to determine if it met all criteria for a circadian rhythm. In this study, the rhythm displayed in Trichodesmium has been demonstrated to meet all criteria of a circadian rhythm. With the minimum doubling time at about 3 days (6), the rhythm in Trichodesmium does not result from cell cycle, which was suggested to be a possible explanation for the cyclic rhythms observed in two unicellular Synechococcus species (15, 21, 25).

In summary, we have demonstrated that cyclic nitrogenase gene expression persists under constant conditions, that the rhythms can be entrained to daily L-D cycles by L-D cues, and that the rhythms are temperature compensated between 24 and 28.5°C. We conclude that Trichodesmium sp. strain IMS 101 possesses a circadian clock and that its nitrogenase gene expression is controlled by a circadian rhythm. Results of transcriptional analysis of nifHDK genes also indicate that nitrogenase protein expression is primarily regulated at the transcription level, which is likely regulated by transcription initiation rather than by degradation.