Cyanosite webserver: http://bilbo.bio.purdue.edu/www-cyanosite/

C. Cyanothece gene expression and regulation

We analyzed growth in short light-dark periods, along with a comprehensive analysis of growth under different light-dark conditions, to develop a fuller understanding of circadian vs. diurnal regulation of gene transcription in Cyanothece 51142. The results indicated that the main metabolic genes involved in photosynthesis, respiration, nitrogen fixation and central carbohydrate metabolism have strong (or total) circadian-regulated components. The transcription of these gene sets was carefully maintained at the appropriate light or dark period, even during 6 h LD growth, so that the cell could generate sufficient energy and an appropriate anoxic environment, when required. The photosynthate from fixed carbon was placed in glycogen granules and the degradation of this glycogen was used as a substrate for the massive respiratory burst that occurred prior to and concomitant with N2 fixation. Coordinating these metabolic properties in a circadian manner appears to be one of the key regulatory functions of the cell. It is interesting that many PSI and PSII genes are transcribed in a diurnal fashion, whereas the reaction center genes and a few genes encoding assembly-type proteins are transcribed in a circadian manner. These findings might have important ramifications as to the assembly and specific functions of the two photosystems. We have also identified some specific regulatory molecules that may be involved in the intricate timing of gene transcription from light to dark transitions and visa versa. We will complete a more detailed computational analysis of these data to determine if we can predict specific interrelationships between regulatory and structural proteins that can then be tested via genetics and physiology.

It was evident that Cyanothece undergoes several adaptation processes in order to develop the appropriate environment for N2 fixation. Nitrogenase activity and nif gene expression were in good correspondence. Our results indicated that one 6 h light period is not sufficient for induction of the nitrogenase-related genes, a feature that also was demonstrated in Gloeobacter This is consistent with our hypothesis that N2 fixation is dependent upon the energy status of the cells Toepel et al (2008) Toepel et al. (2009). Our results demonstrated that breakdown of stored glycogen and high respiration rates were connected to the nitrogenase activity and that both processes are necessary to provide sufficient energy equivalents for N2 fixation. Furthermore, our data suggested that regulation of photosynthesis is independent of incident light during the first 18 h and the decline of photosynthesis is coordinated with the increase in respiration and the concomitant breakdown of glycogen. Our results support the hypothesis of diurnal control of the major processes in Cyanothece sp., but also suggest a strong feedback regulation regarding the energy status of the cells. Additionally, we demonstrated that respiratory enzyme activity and gene expression patterns did not always correspond, suggesting post-transcriptional, energy dependent activation of respiration and a tight co-regulation with N2 fixation.

Comparison of experiments in which cells were grown in LD + LL Toepel et al. (2009) and in McDermott et al (2011) demonstrated that gene expression for glycogen and cyanophycin production/breakdown was LD-dependent. In both experiments, glycogen breakdown was clearly dark dependent and glycogen phosphorylase appeared to be the initializing enzyme and one that most likely required a dark signal for expression. Comparison between the LD+ LL experiments and the 6 h LD experiment, allows us to better differentiate between LD-dependent and LD-independent expression and identify circadian controlled genes.

Two gene sets, encoding proteins for hydrogenases and photosynthesis, were of particular interest to us. Toepel et al. (2008) showed that the uptake hydrogenase (hup) and the bidirectional hydrogenase (hox), found in Cyanothece sp., were co-transcribed with the N2 fixation genes (nif-genes) and probably consumed the produced hydrogen. The expression pattern of the uptake hydrogenase-encoding hup genes in Cyanothece sp. was in good accordance with previous studies, such that the expression of these genes was closely related to the nif genes. In contrast, the hydrogenase-encoding hox-genes were mostly up-regulated during the first dark period, and probably acted as an electron valve during photosynthesis. Since the fermentative-related genes for acetate and lactate production and genes for the PPP cycle (especially gnd) showed elevated expression during the first dark period, it is likely that Cyanothece sp. could produce reducing power and energy equivalents from fermentation and the PPP cycle during this dark period.

The photosynthesis genes demonstrated rather important behaviors. The majority of the PSI and PSII structural genes were diurnally expressed and had peaks in both light periods when grown in 6 h LD periods. At the same time, the most highly expressed psbA gene, psbA1/ psbA5, also demonstrated a diurnal pattern, although the intensity was lower in the 2nd DP. The expression pattern of the psbA1 gene was in good agreement with previous results. At the same time, psaAB demonstrated a circadian behavior and the psaAB transcript level peak was always near the end of the light period. Additionally, two members of the psbA family, psbA3 and psbA4, also demonstrated a circadian transcriptional pattern, but with peaks at different times. It is possible that the encoded D1 proteins that are produced at specific times specifically to change the properties of PSII, similar to the situation in Synechococcus elongatus.

Thus, we have used a short day-night cycle to help determine the circadian vs. diurnal regulation of genes encoding important nitrogen fixation and energy-producing functions. We have then used these data, in conjunction with experiments using other light-dark patterns, to develop a comprehensive model of circadian and diurnal associations. Importantly, both approaches resulted in very similar results, thus demonstrating the value of the CLR analysis. Our results clearly indicated that the nitrogenase gene cluster is under very tight circadian regulation, but that other energy-producing functions must be modified to ensure that the cell can provide the appropriate energetic and oxygen environment to permit N2 fixation under such anoxic conditions. This is a key property of Cyanothece sp. ATCC 51142 and one that may provide directions for the synchrony of other important metabolic processes, including photosynthetically-driven hydrogen production.

In the more recent study, we report the development of a predictive model of cyclic behavior in Cyanothece 51142 using a previously published method, the Inferelator (McDermott et al 2011). The model is based on a set of transcriptional experiments that are focused on investigating diurnal and circadian processes in this organism. We report that the model can accurately predict the behavior of the system when validated on independent data. We found that topology derived from co-expression networks was correlated with gene conservation and that including topological bottlenecks as potential regulators improves the performance of the predictive model. Functional modules, i.e. targets of the inference process, were defined using an iterative process of modeling. We found that the behavior of portions of the metabolic network representing important metabolic processes, e.g. nitrogenase and ribulose-1,5-bisphosphate carboxylase oxygenase (RuBisCO), could be accurately predicted using our models. Finally, we show that the model trained on cyclic time course data is capable of predicting expression dynamics in an acyclic validation time course experiment following Cyanothece 51142 in low oxygen conditions. The models we describe represent an important step forward in the systems biology of photosynthetic cyanobacteria and provide a large number of insights into important biological processes under cyclic regulation.

In this study, we presented a predictive model of cyclic transcriptional processes in Cyanothece 51142, and show that this model can accurately predict the behavior of co-expressed clusters and important functional complexes under conditions not included in the training data. Additionally, we have extended our network analyses of the transcription of Cyanothece 51142 to highlight the importance of topological bottlenecks to the overall functioning of the system. Importantly, we show that topological bottlenecks are as good at predicting the behavior of the system as traditional regulators defined by gene annotation. Our results represent the first global predictive model of transcriptional behavior in a cyanobacterium. The model we present can be queried in different ways to provide hypotheses pertinent to the functioning of the system as a whole, which can be validated experimentally. One kind of hypothesis is presented in this study, the predicted regulatory connections between functional components (co-expressed clusters and other regulators) and can be validated by experiments that eliminate the activity of the predicted regulator and examine its effect on the expression of the predicted target. Prediction of the behavior of the system under novel conditions (as demonstrated in this study) is also possible, and the expression of a small number of regulators can predict the global behavior of the system. We are currently pursuing these avenues in Cyanothece 51142 and in the closely related and genetically tractable, Cyanothece sp. ATCC 7822.

We have shown that topological analysis of association networks is promising for identification of true bottlenecks that mediate transitions between system states, identifying genes that are apparently more important to the system. These predictions summarize a large amount of information in the system, and thus represent a starting point for further investigation. They are based on analysis of high-throughput data, and therefore are unlikely by themselves to provide mechanistic insight into function. Examining the functions of connected genes in the network, temporally upstream and downstream, will shed light on the general function of the bottleneck in the system. However, experimental investigation is needed to validate and further investigate these predictions. The results we present in this study show that our modeling approach is very useful for understanding the regulation and dynamics of the transcriptomics of functional processes in a highly cyclic system. In some cases, the expression of genes directly reflects their function (for example the nitrogenase complex), however, this will not be true for all (or even most) cases. Therefore, integration of other data types, high-throughput proteomics and metabolomics for example, should provide the basis for a more complete model that can accurately predict more functional processes in the system.

Modules defined from the Cyanothece diurnal cycling transcriptomics data represent system states in which genes important for particular functions have peaks. The system requires regulatory and metabolic transitions to activate the appropriate system states in response to appropriate environmental signals. This allows Cyanothece to be flexible in response to variations in photocycles and availability of nutrients. These transitions are mediated by transcriptional regulators, environmental sensors and proteins with other functions; e.g., ion channels. The essential components of the system can then be thought of as the set of functional modules that actually do the work, and the mediators that join them together and regulate their activity. These 'mediators' of system transitions act as effectors that must be active under both the condition of origination and the 'target' condition. Mediators represent decision points where the system may choose to take a number of different courses based on the input signals, either environmental signals (light or dark) or inputs from the originating module. Our predictive model captures many of these elements, allowing accurate and robust prediction of transitions between system states.

Back to Sherman Lab Homepage