Heat Shock Transcriptional Responses in an MC-Producing Cyanobacterium (Planktothrix agardhii) and Its MC-Deficient Mutant under High Light Conditions

Microcystins (MCs) are the most commonly-reported hepatotoxins produced by various cyanobacterial taxa in fresh waters to constitute a potential threat to human and animal health. The biological role of MCs in the producer organisms is not known, and it would be very useful to understand the driving force behind the toxin production. Recent studies have suggested that MCs may have a protective function in cells facing environmental stress. Following this starting premise, we speculate that under adverse conditions the expression of stress-related genes coding for Heat Shock Proteins (Hsp) might be different in an MC-producing strain and its MC-deficient mutant. We therefore used RT-qPCR to compare the expression of 13 hsp genes of an MC-producing strain of Planktothrix agardhii (CYA126/8) and its MC-deficient ΔmcyD mutant over different periods of exposure to high light stress (HL). Three reference genes (RGs) were selected from six candidates to normalize the RT-qPCR data. Of these three RGs (rsh, rpoD, and gltA), gltA is used here for the first time as an RG in prokaryotes. Under HL stress, five genes were found to be strongly up-regulated in both strains (htpG, dnaK, hspA, groES, and groEL). Unexpectedly, we found that the MC-producing wild type strain accumulated higher levels of htpG and dnaK transcripts in response to HL stress than the MC-deficient mutant. In addition, a significant increase in the mcyE transcript was detected in the mutant, suggesting that MCs are required under HL conditions. We discuss several possible roles of MCs in the response to HL stress through their possible involvement in the protective mechanisms of the cells.


Introduction
Cyanobacterial bloom-forming species are a worldwide problem, because of the toxins they produce. The situation has become a cause of increased concern in recent decades as the frequency and intensity of bloom occurrence have increased, due in part to anthropic activities [1]. One of the diverse groups of cyanotoxins, that of the hepatotoxin microcystins (MCs) which includes up to 90 variants [2], is widespread and frequently reported. Microcystins [3] are becoming a real threat to human and animal health due to the contamination of freshwater [4]. However, the ecological significance and biological function of the MCs for the producer cells, which could elucidate the forces underlying toxin production, remain elusive. MCs are typically intracellular components, which are usually released into the environment after the cells die [5]. It has been shown that they are not essential for growth, but are probably involved in intracellular processes [6]. Various different hypotheses for the possible role of cyanotoxins have been proposed, these include a role as: grazer deterrent compounds [7] iron scavenging molecules [8], allelopathic compounds [9], growth regulators permitting successful adaptation [10], light harvesting and chromatic adaptation [11], and infochemicals [11][12][13]. However, the findings of the various studies are contradictory in many respects, and difficult to understand [14].
More recently, emerging investigations have suggested that MCs may have a protective role in the response to unfavorable conditions. Significantly greater growth rates have been observed in an MC-producing strain Planktothrix agardhii, than an MC-free strain; both strains had previously been collected from environmental samples (i.e. had different genomes) [15]. It has been also reported that transcription of the genes responsible for MCs production in Microcystis increased in response to strong illumination or iron starvation suggesting that MCs may play a protective role under various stressful conditions, including oxidative stress [11,16]. To date, most of the data about the physiological roles of MCs has been reported by Neilan and colleagues in Microcystis [17], where they show that MCs production has a complex and deep effect on the proteome. Furthermore, MCs may be involved in the carbon-nitrogen metabolism, in redox control, in the perception of redox changes, and in providing protection against oxidative stress [17][18][19]. However, cyanobacterial responses to abiotic stresses are complex, and several mechanisms usually act in concert and synergistically to prevent cell damage and to reestablish cellular homeostasis [20][21]. The activities of heat shock proteins (hsps) constitute an important component of the cell's response to stress. Depending on their size, Hsps are divided into 5 main classes: Hsp100, Hsp90, Hsp70, Hsp60 and small Hsp; representatives of each class are found in cyanobacteria [21]. Under normal conditions, Hsps play an important role in the folding, assembly, and trafficking of newly-synthesized polypeptides and in the degradation of denatured or aggregated proteins. Under stressful conditions, as a result of increased levels of aberrant proteins, the importance of Hsps increases, and this is usually reflected in their up-regulation [20]. However, much remains to be learned about the ''in vivo'' function of the Hsps proteins in cyanobacteria, as their biological role may extend to stress responses in general (known as HSRs for Heat shock responses) [22][23], and to multiple molecular cell defenses. Indeed, the protective effects of Hsps can be attributed to the network of the chaperone machinery, in which many Hsps play complementary and sometimes overlapping roles [20]. These major studies are initiated at the transcriptional level, using RT-qPCR (real-time quantitative polymerase chain reaction) analysis, which is one of the most powerful tools available for investigating quantitative differences in gene expression responses under experimental conditions [24]. However it is also a demanding tool, and calls for preliminary evaluation of the stability of a panel of reference genes (RGs) to provide accurate normalization of the gene expression analysis.
A few studies have identified the intracellular function of MCs in cyanobacteria against abiotic stressors at the transcriptomic level, and they involved the use of two strains sharing identical genomes, but with one gene engineered to inhibit MCs production. Unicellular Microcystis strains were investigated [19], but no study has yet been carried out of two identical clones of filamentous Planktothrix agardhii (differing solely by a single insertional mutation in one mcy gene), despite its hazardous impact on ecosystem functioning and the current increase of blooms in waterbodies. We selected P. agardhii for our study, due to its ecological preference for low light intensities [25], as HL (high light) could be expected to induce more explicit responses. We attempted to i) validate that the RGs were stable under our experimental conditions (i.e. control vs stressed conditions) in order to optimize RT-qPCR accuracy; ii) determine the gene expression profiles of 13 hsps and one mcy gene (mcyE, which is involved in the synthesis of Adda, and the incorporation of D-glu during MCs production); iii) determine the fold changes in the expression levels of the hsps and mcyE in the MC-producing strain (CYA126/8) and its MC-free mutant (DmcyD) when exposed to HL stress, using RT-qPCR analysis. To the best of our knowledge, this is the first study to report the transcriptional shift of a panel of 13 stress related genes (hsps) in MC-producing cells during a short period of exposure to stress (5-24 h).

Strains and Culture Conditions
Two strains of P. agardhii were used in this study. The MCproducing strain CYA126/8 (i.e. the wild type: WT) and its MCdeficient mutant (DmcyD) were kindly provided by Dr Kurmayer (University of Vienna). Both strains were monoclonal and not axenic and are maintained in the PMC (Paris Museum Collection, Paris). They shared an identical genome except that in the DmcyD mutant; a chloramphenicol cartridge was inserted into the mcyD gene to inhibit MC biosynthesis. We used LC/MS to confirm that CYA126/8 WT was an MC-producer, and that the mutant was free of MCs (Combes, unpublished data). Both cultures were maintained in Z8 liquid medium [26]

Experimental RT-qPCR Conditions
The cultures under optimal conditions (i.e. control) were obtained during the exponential growth phase under a continuous light intensity of 22 mmol m 22 s 21 . For the HL treatment, the cultures in the exponential phase were shifted from control conditions to an intensity of 600 mmol m 22 s 21 during 24 h. For both conditions, the temperature was maintained at 20uC using a Binder phytotron. Experiments were performed using equivalent culture densities (OD 750 nm = 0.3) of both strains. Samples were taken at 0 h (control), and 1 h, 2 h, 5 h and 24 h after transferring to HL conditions, and were used for subsequent analysis. Two independent replicates were performed.

RNA Extraction and cDNA Synthesis
For each sample, 40 ml of the culture suspension (OD 750 nm = 0.3) was centrifuged at 4uC, for 15 min, at 4000 rpm. Total RNA extraction was carried out using Trizol reagent (Invitrogen, USA) followed by purification using Pure-Link TM RNA Mini Kit (Invitrogen), according to the Manufacturer's instructions. The pellet was mixed with 3 ml of Trizol, and then immediately frozen in liquid nitrogen and conserved at 280uC until extraction. Phase separation was obtained by adding 600 ml of chloroform to the cell lysate and shaking vigorously by hand for 15 seconds (8 times), storing at room temperature for 5 minutes, and then centrifuging at 12 000 g for 15 minutes at 4uC. RNA purification was performed using PureLink TM RNA Mini Kit (Invitrogen). Purified RNA, previously treated with a DNAfree Kit (Ambion), was quantified using a NanoDrop 2000 Spectrophotometer (Thermo Scientific); and its integrity was checked on 1.5% agarose gel (data not shown). Genomic DNA contamination was checked by PCR on a total RNA template using primers targeting the citrate synthetase sequence (data not shown).
The A 260 /A 280 ratio of the RNA samples was 2.08960.017 (mean 6 SD), indicating the absence of protein and the purity of all the total RNA samples required for an accurate qRT-PCR analysis. First-strand cDNA was synthesized from 0.8 mg total RNA using SuperScript III First-Strand Synthesis SuperMix (Invitrogen, Carlsbad, USA), with 1 ml of random hexamers in a 20-ml reaction mixture, according to the Manufacturer's instructions. cDNA samples were stored at 220uC.

Genes Investigated in the Study
Two series of genes were used in this study ( Table 1). As the relevance of RT-qPCR analysis greatly depends on transcript normalization with stably-expressed reference genes (RGs), 6 candidates were selected from different functional classes. Four conventional candidate RGs were tested: 16S rRNA [27], rpoD [28], GAPDH [27], rsh [27], plus two RGs genes that had never so far been tested in prokaryotes: gltA and rpsL. The sequence of 16S rRNA is available in a database (GeneBank FJ184435.1), but the other 5 RGs candidates were all isolated in this study.
The second series were the target genes. They included nine hsp genes: hspA, hslO, hsp40, grpE, dnaK, hsp70 (1), hsp70(2), hsp70(3), and hsp70(4), which had previously been sequenced, and were kindly provided by Dr Quiblier (MNHN, Paris) plus four other genes: clpC (hsp100), htpG (hsp90), groEL (hsp60) and groES (hsp10)), which were isolated in this study. The full-length sequences of the genes obtained here are available in the EMBL database under the following accession numbers: KF275115 to KF275124, and KF294782 to KF294790. We also included mcyE, which is involved in Adda synthesis and the incorporation of D-Glu in the biosynthesis of MCs [29] as a target gene.
The specificity of the primer sets were tested by Real-time PCR using cDNA of P. agardhii CYA 128/6 and confirmed by melting curve (Fig. S1) and gel electrophoresis (Fig. S2). The identities of all PCR products were further confirmed by TA cloning using pGEM-T Easy vector (Promega, WI, USA) and subsequent sequencing.
The melting curve analysis indicated that all the primer pairs produced a single peak (Fig.S1), and only one band of the expected size was obtained on 2.5% agarose gel electrophoresis, which confirmed the specificity of all the primer pairs ( Fig. S2).
Real time PCR was performed on a LightCycler 2.0 (Roche) using 32-capillary carousel combined with the LightCycler FastStart DNA Master SYBR Green I (Roche). Each 20-ml capillary contained a total volume reaction of 10 ml including: 1 ml of ready-to-use hot start PCR reaction mix; 1 ml of primer mix (Table S1); 4 ml of 1:64 diluted cDNA sample; and 4 ml of MgCl 2 . Each run included a non-template control (NTC). Real-time amplification reactions for each gene of the WT and mutant strains from one biological replicate were performed in a technical duplicate and in the same PCR run. The cycling conditions were: 1 cycle at 95uC for 10 min, followed by 40 cycles at 95uC for 10 s, 62-65uC (depending on the target - Table S1) for 4 s, and 72uC for 10 s. In this study, the C T was automatically identified using the ''Second Derivative Maximum Method'' [33]. At the end of the amplification, the melting temperature of the product was also determined using the melting curve program: 65-95uC, with a heating rate of 0.1uC per s and continuous fluorescence measurement.

Expression Stability of the Candidate Reference gene and Data Analysis
In order to determine the true gene-specific variation, at least one stably-expressed RG is required to normalize the expression level of the target genes. The expression levels of the six candidate RGs were determined by RT-qPCR under the same experimental conditions as for the target genes (i.e. control condition+HL stress).
To identify the genes most stably expressed during HL treatment, the C T values of these six candidate RGs were analyzed using three different mathematical algorithms: geNorm [34], Normfinder [35] and BestKeeper [36]. In brief, GeNorm is the one most commonly used in the literature and it relies on the transformation of raw C T values (using the delta-C T method). The gene expression stability measure (M) for a candidate RG is computed by averaging pairwise variations of that gene versus all the other candidates tested. A decrease in the M value reflects an increase in expression stability. NormFinder is a model-based algorithm used to identify the optimum RGs from a group of candidates. This algorithm required the transformation of C T values to linear scale expression quantities. The genes with the lowest stability values have the most stable expression. BestKeeper uses the raw C T as the input for calculation. A Pearson correlation coefficient was calculated for each candidate pair as well as the probability that the correlation was significant. An index value was calculated as the geometric mean of the C T values of all highlycorrelated candidate RGs. Stable RGs show a strong correlation with the BestKeeper index.
The gene-specific PCR efficiency was determined for each pair of primers using a 5-fold serial dilution of cDNA as template. The standard curve was obtained by plotting C T values against a logarithm of serial dilutions of the target nucleic acid. The efficiency of the reaction (E) was calculated from the slope value of a standard curve, as follows: E = 10 (21/sl u pe) -1.
The relative quantity of each gene (Q), which was used in geNorm and NormFinder, was calculated as: Q = E (min C T 2 sample C T ) where Q = sample quantity relative to the sample with the highest expression; E = amplification efficiency; min C T = lowest C T value = C T value of the sample with the highest expression.
To calculate the normalized relative gene expression levels, data were analyzed using Relative Expression Software Tool (REST) (http://gene-quantification.com/rest.html) [37].
Data of transcript expression levels were analyzed using oneway analysis of variance (ANOVA) at a confidence level of p,0.05; followed by Tukey's test on GraphPad Prism 5.0 software.

Selection and Validation of Reference genes for RT-qPCR Normalization
In order to compare the expression levels of hsp target genes in the WT and DmcyD mutant of P. agardhii CYA 126/8 under HL conditions, we normalized all the samples using the same RGs. The C T values were obtained for each candidate gene across all the samples (Fig. 1) and revealed the differences in transcript levels. The 16S rRNA gene gave the lowest C T (8.81), corresponding to the highest expression level, whereas gltA and rsh showed the lowest expression levels, with mean C T values of 25.66 and 25.99, respectively (Fig. 1). The expression stability of the candidate RGs were analyzed using geNorm, NormFinder and BestKeeper, which provided complementary measures of the cDNA samples. Both geNorm and NormFinder classified 16S rRNA as the least stable gene (Table 3), regardless of the data series used (combined or single WT and mutant).The transcript level of this candidate was also much higher than that of the others (Fig. 1). Because it is crucial to use RGs with ranges of expression similar to those of the target genes in the samples for analysis [38], we excluded the 16S rRNA gene from the BestKeeper analysis. The three most stable genes identified by all three programs were similar, especially for the first two in all the data sets (Table 3). Rsh, rpoD and gltA were identified as the best performing genes, whereas rpsL and 16S rRNA were always identified as the least stable genes. The optimum number of RGs required for an accurate normalization was provided by the pairwise variation (V n/n+1 ) calculation using the GeNorm program. The closest value to 0.15 (the cut-off value for validation - [35]) in our analyses, was found for V 2/3 (0.12-0.14 according to data series); the value for V 3/4 was even lower (0.09), indicating that three reference genes were the optimal number for accurately normalized gene expression. Consequently, we validated the 3 most stable reference genes (rsh, rpoD and gltA) for the normalization of all RT-qPCR data.
Under control conditions, the groES gene was the one most highly expressed in both the WT and DmcyD (23.7263.45 and 25.4760.7, respectively). Except for hspA, no significant difference was found under control conditions between the WT and DmcyD (p.0.05). The hspA expression level was significantly higher (p,0.001) in DmcyD than in WT (24.561.26 and 9.2760.52, respectively).

Effects of HL on the Expression Levels of hsp genes in P. agardhii WT and DmcyD Mutant
The transcriptional response of 13 hsp genes of P. agardhii (WT and DmcyD) was compared at different times (0, 1 h, 2 h, 5 h and 24 h) under HL conditions ( Fig. 3 and Table 4). Based on the relative expression levels (shown as the fold change in gene expression versus control at time T 0 ), three different expression profiles were obtained: (i) clpC, hsp70(1), grpE and hsp40 (Fig. 3.A) showed no significant difference in expression level (p.0.05) under control and HL conditions; (ii) hsp70(2), hsp70(3), hsp70(4) and hslO (Fig. 3.B) showed a slight (,4 fold) but significant change (p,0.05) in expression levels under HL and (iii) htpG, dnaK, hspA, groEL and groES (Fig. 3.C) showed a strong (.4 fold) and significant (p,0.05) increase in expression level under HL conditions. The expression of the following genes was significantly higher under HL than under control conditions (T 0 ) (Fig. 3.B): Hsp70(3) and hslO showed a significant increase at 1 h, with a peak after 2 h under HL (<3.5 fold for both strains) (p,0.05). Hsp70(4) exhibited a significant (p,0.01) increase after 24 hours for both the WT and the DmcyD (1.3160.16 and 1.6760.24, respectively). For hsp70(2), a slight but significant increase (p,0.05) was observed in the WT and the mutant after 2 hours under HL (1.3860.12 and 1.6460.001, respectively) (p,0.05).
Five genes were strongly up-regulated under the HL conditions ( Fig. 3.C) and they displayed a similar expression pattern: a significant increase of expression level (p,0.001) at 1 hour, a peak reached after 1 to 2 hours under HL, and then a decrease to the background level. The expression profiles of htpG, dnaK, and hspA were quite similar in both strains. Interestingly, quantitatively significant differences were found between the WT and the DmcyD for these three genes ( Fig. 3.C). In the WT, the expressions of htpG, dnaK, and hspA were up-regulated to a greater extent after 2 h than in DmcyD (Fig. 3.C). The fold change in dnaK expression induced by HL was 13.34 in the WT versus 8.34 in the DmcyD (p,0.01). For htpG, the expression fold change reached its maximum of 10.0661.08 after 2 h of exposure to HL in the WT, which was significantly higher (p,0.001) than that in the mutant (5.9861.39).
The relative expression of hspA after 2 h under HL was 52.3166.9 in the WT strain compared to 17.8264.8 in the DmcyD (p,0.001). However, it is noteworthy that under control conditions, the level of hspA in WT was significantly lower than in the DmcyD strain (9.2760.52 and 24.561.26, respectively) (Fig. 2). Therefore, even though the maximum level of change in WT reached 52.3166.9 fold versus 17.8264.8 fold in the mutant, there was no significant difference between the amounts of hspA mRNA under HL in the two strains (p.0.05).
For groEL and groES genes, in both strains, the mRNA reached the greatest level 2 hours after being exposed to HL stress and then decreased ( Fig. 3.C). However, the pattern of change differs in timing with a significant increase (p,0.001) right after 1 h in the mutant, in contrast to the unchanged value in the WT (Fig. 3.C). The maximum level was reached after 2 h in both strains. For groEL and groES expression in HL, no significant differences were observed between the maximum levels of the WT and DmcyD.

Effects of HL on Expression Levels of the mcyE gene in P. agardhii WT and DmcyD Mutant
The transcription level of mcyE of the WT strain was not affected by HL during the test period (0-24 h) (Fig. 3.B). The main finding was that the insertion of the Cm R gene into mcyD did not disrupt the expression of another mcy gene (mcyE) in P. agardhii. Under HL, a significant increase in the abundance of mcyE mRNA was observed in the mutant (2.7 fold 5 h after the shift from control conditions to HL) (p,0.001). This mRNA level remained high till 24 h.

Discussion
There is a general consensus that the selection of suitable reference genes for normalization of the target genes is a prerequisite for RT-qPCR [24]. In this study, six candidate genes were chosen from independent pathways to avoid possible effects due to co-regulation. Four genes had previously been used as conventional RGs in prokaryotes (16SrRNA, rpoD, GAPDH, rsh) [27][28]. To the best of our knowledge, the other two candidate genes (gltA, rpsL) had never been tested in prokaryotes. A combination of three computational programs (GeNorm, Norm-Finder and BestKeeper) was used to provide accurate validation of the most stable genes for normalization [39]. The same three genes (rsh, rpoD and gltA) were identified as being the most stable genes, despite some slight differences in ranking order (Table 3). Since these three algorithms rely on different mathematical approaches to calculate stability (see Mat & Meth), these minor differences between their outputs were not unexpected [36,40]. Such discrepancies have been reported in several studies as minor changes in gene stability rankings [41]. According to the geNorm analysis, V2/3 was below 0.15, indicating that the minimum number of RGs required for reliable normalization in this study would be two. However, it should be pointed out that using additional genes is usually an option [40][41], and using the three best RGs is a valid normalization strategy in most cases [35]. Moreover, in our study, including the third most stable candidate gene gave a significantly lower V-value. We therefore decided to use the three most stable genes as RGs for normalization of the RT-qPCR data. These included gltA, which will be listed as a newly-recommended RG for further RT-qPCR data normalization, because validation of the suitability of RGs for normalization in prokaryotes is still lacking [42].
The most unstable genes were also the same in all the sample sets (Table 3). We showed here that 16SrRNA and GAPDH, two RGs used in previous studies [28,43] were not appropriate for normalization under our experimental conditions for the two Planktothrix strains we used. The suitability of 16SrRNA gene for use as an RG is currently disputed [44] due to its high abundance compared to target gene transcripts (resulting in too large a difference in C T values), which can bias interpretations [17,35]. This once more showed that appropriate RGs can be very different in different organisms as well as under different   Under control conditions, all the hsp genes were found to be constitutively expressed. Among them, groES showed the highest expression level, which was at least double that of the others (Fig. 2). This was not surprising, as most Hsps are constantly produced to assist the proper folding of nascent proteins, and to prevent protein aggregation throughout the lifetime of cells [21,45]. Furthermore, groES is a part of the groESL operon, a major chaperone system in bacteria, which plays an important role in the conformational homeostasis of cell proteins [38]. However, under control conditions, a higher expression level (about 2.5-fold greater) of a small hsp (hspA) was observed in the mutant strain than in the WT strain (Fig. 2), something that we cannot explain.
Under HL stress, among the 13 genes under investigation, only five hsps were strongly up-regulated in both strains, with a 6 to 52fold change relative to control conditions, four genes were slightly up-regulated, and the other four were unchanged (Fig. 3, Table 4). It has been observed that the Hsps stress response varies considerably in some organisms depending on the species, Hsp family, developmental stage, and stressor [46][47]. Diverse Hsp isotypes in different species may have different roles and modes of action. Therefore, the true significance and role of Hsps in difference species must be confirmed using methods appropriate for each species.
However, it had previously been reported that HL stress may induce changes in gene transcription within as little as 15 minutes, and the expression level may also return to the basal level very quickly [48]. In this study, over our experimental time course (1-24 h) the five genes were all strongly up-regulated and displayed a similar kinetic pattern. The expression level increased 1 h after the shift to high light; it then reached its maximum level after 1 to 2 h, and thereafter declined (Fig. 3). This pattern of change is typical of how gene expression usually responds to stress [48]. These changes are usually transient and, even with persistent stress, gene expression fairly soon reaches a new homeostasis, in which the physiology of the cell has adjusted to new conditions [48][49][50].
The pronounced inductions of the five hsp genes (htpG, dnaK, hspA, groEL, groES) are consistent with those reported for Synechocystis PCC6803, where htpG, dnaK2, groESL, and hspA were also conspicuously up-regulated in response to a shift from low light to HL [48,51]. This may suggest that these genes play a significant physiological role in protecting cells against this specific abiotic stress, regardless of the cell's ability to produce MCs.
The most important finding of our study was detecting some transcriptional differences between the WT and its MC-deficient mutant. Indeed, two hsps genes (htpG and dnaK) were < 1.7-fold more highly expressed under HL stress in WT than in DmcyD (Table 4 and Fig. 3). HtpG had previously been reported to play a role in the ability of cyanobacteria to tolerate various stresses [52][53][54], including providing effective protection against the oxidative stress caused by HL in unicellular cyanobacteria [50]. Some studies have suggested that HtpG is involved in regulating the biosynthesis of tetrapyrrole [55] and that it interacts with the linker polypeptides of phycobilisome in cyanobacteria to prevent their samples. An asterisk indicates a significant difference versus control (T 0 ) *: p,0.05. Circles indicate a significant difference in the expression level between the WT and the mutant strains; uu: p,0.01; uuu: p,0.001. doi:10.1371/journal.pone.0073198.g003 Table 4. Relative fold change of transcript of the genes obtained by qRT-PCR. thermal aggregation [56]. These activities may endow HtpG with an effective photoprotective role in response to HL stress. For the dnaK gene, it has been shown that cyanobacteria contain multiple dnaK homologs, the expressions of which are differently regulated [57]. However, the alignment of our sequence (Accession no.: KF294788) revealed the greatest similarity to dnaK2 of Synechocystis PCC6803 (91% similarity-data not shown). dnaK2 has been reported to be induced by various abiotic stresses [50,[58][59], including HL [50], although its essential function is still elusive [60].
Even if their multiple functions are still unclear, one would expect that higher expression levels of dnaK and htpG may contribute to better protection of macromolecular complexes, such as the photosynthetic apparatus, and thus enable the WT strain to tolerate HL better.
Finally, an unexpected finding was about the expression profile of the mcyE in both strains when they were transferred from control conditions to HL stress (Fig. 3.B, Table 4). In the WT strain, stable expression of the mcyE gene was observed after the transition from control to HL stress conditions for a short period of time (0-24 h), in contrast to some previous findings in Microcystis strains. An up-regulation of mcyB and mcyD was found as a result of HL intensities [11]. Using RT-qPCR we showed that the expression level remained constant, which was corroborated by constant MCs production by the P. agardhii WT cells during the first 24 h (450 ng eq. MC-LR per mL of culture normalized to OD 750 nm = 1, unpublished data). The mcy operon seems to be expressed at a basic level corresponding to the intracellular-MC present in the cell. Up-regulation in the WT seems to be unnecessary during the first 24 h under this level of HL. Unexpectedly, a basal level of the mcyE transcript was observed under control conditions in the MC-deficient mutant that was similar to that in the WT. The disruption of mcyD by Cm R has no effect on the expression of mcyE gene expression in P. agardhii. This absence of any polar effect of the mutation of one mcy gene on the others had previously been reported in Microcystis [61]. In the MCdeficient mutant, HL induced a significant increase of the mcyE transcript that reached its maximum level from 5 to 24 h (Fig. 3.B). The enhancement of the mcyE transcript induced by HL conditions in the DmcyD mutant strain might reflect a requirement for MCs production under such stress. HL is known to cause direct severe damage of the photosynthetic apparatus, and an indirect increase in ROS production (which induces oxidative stress). As a consequence, many different mechanisms and substances may act as cellular defenses in different ways. MCs may be one of them, as suggested by Zilliges and colleagues [19], as MCs bind to cysteineresidues and to specific protein targets involved in photosynthesis processes and against oxidative stress conditions. The possibility cannot be excluded that the depletion or the absence of MCs in deficient cells could increase the damage caused, and may thus contribute to an increase in the susceptibility to environmental stress. This might explain the sporadic changes seen within populations, where MC-producing genotypes can replace non-producing strains in the field under unfavorable conditions [15,62].
In conclusion, our findings support the hypothesis that MCs have an intracellular function in Planktothrix agardhii related to the transcriptional variations of mRNA, and that this could be attributed to the intracellular presence of MCs in the producer cells (related to HL stress). However, further investigations are needed to identify the nature of the interactions between MCs and Hsps modulated-responses (if any), and finally, to define a possible connection between MCs and the primary metabolism of cyanobacteria that produce this ''secondary'' metabolite.