Epigenetic modifications, heritable chemical additions to DNA or histones which are associated with gene expression but do not alter the primary DNA sequence, are well known as connected with many important processes (Law and Jacobsen, 2010). DNA methylation is one of the most important epigenetic modifications and catalyzed by DNA-specific methyltransferase which transfers methyl group from general substrate, S-adenosylmethionine, to specific DNA sequences. This epigenetic modification is usually related to regulation of gene expression, genomic imprinting, X chromosome inactivation, chromosome stability and so on in eukaryotes (Feng et al., 2010).

DNA methylation mainly refers to 5-methylcytosine (thereafter m⁵C) in eukaryotes. In mammals most m⁵C exist within CG context, and methyltransferases DNMT3 and DNMT1 are responsible for establishment and maintenance of the DNA methylation, respectively. In higher plants, m⁵C could exist in CG, CHG, and CHH context. Establishment of the methylation is conducted by methyltransferase DRM2, and two methyltransferases, MET1 and CMT3, are involved in maintenance of the methylation in higher plants (Chan et al., 2005; Law and Jacobsen, 2010). Bisulfite can convert cytosine into uracil but cannot alter m⁵C (Patterson et al., 2011). The whole genome bisulfite sequencing (thereafter WGBS) technology has been widely used in exploring the global pattern of m⁵C in eukaryotes (i.e., Lister et al., 2009; Carvalho et al., 2012). In addition, m⁴C can also show partially resistance to the bisulfite-mediated deamination, therefore WGBS indeed could be used to identify both m⁵C and m⁴C (Vilkaitis and Klimasauskas, 1999).

Besides methylation by DNA methyltransferases, demethylation could also influence the global DNA methylation pattern in mammals and plants. Deaminases (Aid/Apobec), glycosylase (Mbd4), thymine DNA glycosylase (Tdg), and unidentified apyrimidinic lyase are needed for active demethylation in mammals, while bifunctional 5-methylcytosine glycosylases (DME/ROS1) are needed for active demethylation in higher plants (Ikeda and Kinoshita, 2009).

In bacteria, besides m⁵C, N6-methyladenine (m⁶A) and N4-methylcytosine (m⁴C) also exist. Unlike in mammals and plants, DNA methylation in bacteria is believed to be part of the restriction/modification system, which prevents bacteria from being self-digested (Wilson and Murray, 1991). Taking Escherichia coli genome as an example, three DNA methyltransferase genes exist, one of which is clustered with its cognate restriction enzyme gene and the other two encode solitary methyltransferases. In bacteria, DNA methyltransferases may also have other physiological significances. The methyltransferase, Dam, involves in distinguishing newly synthesized DNA from the old DNA (Palmer and Marinus, 1994) and takes part in DNA repair (Pukkila et al., 1983). In Caulobacter crescentus, the methyltransferase CcrM acts as a regulator for cell cycle (Reisenauer et al., 1999). Single molecule real-time (SMRT) sequencing technology could be used to directly detect m⁶A and m⁴C with high degree of accuracy and sensitivity (Flusberg et al., 2010). With this technology, many bacterial and archaeal species were investigated for their global m⁶A and m⁴C distribution patterns (Clark et al., 2012; Fang et al., 2012; Murray et al., 2012; Lluch-Senar et al., 2013; Blow et al., 2016). Besides SMRT sequencing, WGBS has also been applied into bacterial such as E. coli to determine the global m⁵C distribution (Kahramanoglou et al., 2012).

As an ancient group of photosynthetic bacteria, cyanobacteria plays very crucial role in the earth as the primary producer in the ecosystem. They could be found in many diverse ecological habitats, and survive under adverse environments such as deserts, polar region, and hot springs (Whitton and Potts, 2000; Stomp et al., 2007). Some cyanobacteria species may form water bloom, threatening the health of human (Catherine et al., 2013). Based on sequence homology, some methyltransferase encoding genes have been identified in cyanobacteria. For example, there are nine genes possibly encoding DNA methyltransferase in Anabaena sp. PCC 7120 (Matveyev et al., 2001). Type II restriction/modification system and solitary methyltransferase genes were also predicted in N. punctiforme genome (Meeks et al., 2001). In Synechocystis sp. PCC 6803 (thereafter Synechocystis), a model cyanobacterium, three genes encoding solitary methyltransferases, mbpA, sll0729, and synMI, exist (Kaneko et al., 1995, 1996; Scharnagl et al., 1998). Some limited DNA methylation studies have also been conducted in cyanobacteria. Restriction analysis and quantitative estimation of methylated bases in filamentous and unicellular cyanobacterial DNA shows the presence of methyladenine in the GATC sequence (Padhy et al., 1988). Enzyme digestion experiments in two Anabaena species indicates that DNA from both species is methylated, but that no gross change in methylation occurs during heterocyst formation (Adams, 1988). Recently, global DNA methylation pattern study has been conducted in Synechocystis and more methyltransferases were identified (Hagemann et al., 2018).

Many studies have reported external environment stress could influence the epigenetic processes in eukaryotes (Gao et al., 2010; Lira-Medeiros et al., 2010; Kou et al., 2011; Menzel et al., 2011; Navarro-Martin et al., 2011; Yu et al., 2013). Furthermore, rapid accumulating evidence shows that environmental stress could generate particular defense epigenetic changes across generations in eukaryotes (Hauser et al., 2011; Heard and Martienssen, 2014; Skinner, 2014). This transgenerational epigenetic inheritance has gained interest over the last years and addressed the importance of transgenerational inheritance for adaptation to ever changing environment and for practical applications. For instance, in Oryza sativa L., it has been reported chronic nitrogen deficiency could induce global change in DNA methylation in leaf-tissue of the stressed plant, and part of the change could be stably inherited to its non-stress suffered progenies and improve the resistance to nitrogen deficiency in progenies (Kou et al., 2011). In contrast, in bacteria few studies concerned about the influence of external environment on global DNA methylation pattern, much less the defense transgenerational epigenetic inheritance.

While nutrients are important to microbes, seasonal depletion and repletion of nutrients in the environment are common in nature. In some habitats, nitrogen, one of the most important essential nutrients, is seriously scarce for cyanobacteria to survive (Elser et al., 1990; Vitousek and Howarth, 1991; Conley et al., 2009; Thomas et al., 2015). However, cyanobacteria have evolved sophisticated mechanisms to cope with the nitrogen deficiency. For example, most cyanobacteria could utilize various alternative forms of nitrogen source under nitrogen deficiency (Flores and Herrero, 2005). Moreover, some cyanobacteria could biologically fix nitrogen from air (Wolk, 1996). Considering that DNA methyltransferase is widespread in this group (Kaneko et al., 1995, 1996; Scharnagl et al., 1998; Matveyev et al., 2001; Meeks et al., 2001), it is possible for cyanobacteria to deal with the stress through defense alteration in DNA methylation as in plant (Kou et al., 2011). However, knowledge about the whole-genome DNA methylation in cyanobacteria is still poor, and defense epigenetic modifications such as DNA methylation in this group, especially under any environment stress, remain a mystery.

In this study, DNA from a model single-celled model cyanobacterium Synechocystis cultured under normal nitrogen (sample as NC), nitrogen starved for 72 h (N72), and nitrogen recovered (NR) which was derived from N72 but thereafter NR for 12 days was bisulfite converted for next generation sequencing, and the DNA methylation patterns were compared among the three samples. Questions proposed to answer are: (i) whether widely spread mC exists in Synechocystis? (ii) whether the DNA methylation could be influenced by nutrient depletion or not? and (iii) if influenced whether the altered pattern is inheritable and how? Results indicate that nitrogen deficiency stress can induce DNA methylation alternations, and gives the evidence of transgenerational epigenetic process, inheritance of DNA methylation pattern from N72 to NR, in cyanobacteria. Our current data will contribute to a better understanding of the biological function, epigenetic adaptation to ever changing environment, and even the evolution of DNA methylation in prokaryotes.

With SMRT sequencing, many methylomes were explored in bacterial. At first, Clark et al. (2012) introduced plasmids containing cloned distinct DNA methyltransferases into E. coli mutant lacking endogenous DNA methyltransferases, and examined the methylation status of the plasmids. Then they investigated the methylation status of six bacterial genomes with the same technology, including Geobacter metallireducens GS-15, Chromohalobacter salexigens, Vibrio breoganii 1C-10, Bacillus cereus ATCC 10987, Campylobacter jejuni subsp. jejuni 81-176, and C. jejuni NCTC 11168 (Murray et al., 2012). Finally, Blow et al. (2016) investigated DNA modifications among 230 diverse bacterial and archaeal species and found lots of reproducibly methylated target motifs. All these studies have been largely limited to m⁶A and m⁴C due to the weak and somewhat diffuse SMRT signal generated by m⁵C (Flusberg et al., 2010). The first genome-wide study of m⁵C at base resolution was carried out by applying WGBS into E. coli and found cytosine methylation may take part in regulation of gene expression in stationary phase as a regulator (Kahramanoglou et al., 2012). With respect to cyanobacteria, SMRT sequencing showed the existence of m⁶A in Microcystis (Zhao et al., 2017), and Hagemann et al. (2018) profiled the global DNA methylation (including m⁶A, m⁴C, and m⁵C) in Synechocystis by using combination of SMRT sequencing and WGBS, which indicated 5-methylcytosine and m⁴C in Synechocystis were in the context of m⁵CGATCG and GGm⁴CC, respectively. All these studies were conducted to determine the specificity of methyltransferases.

In our study, the CG methylation indeed corresponds to the m⁵CGATCG, and CHG and CHH methylations comprise the GGm⁴CC. Considering the methylation level change of CHG obviously differed from that of CHH in TEs (Supplementary Figure S4), we proposed the differential influence of bases downstream the methyltransferase target motif (GGm⁴CC), and therefore distinguished CHG from CHH. A similar situation for the differential influence of base outside of the target motif was found in E. coli K12 of mid-exponential phase, where the methylation level for CmCCWGG was lower than [ATG]mCCWGG (mCCWGG was the methyltransferase target motif) (Kahramanoglou et al., 2012). Due to the partially resistance of m⁴C to the bisulfite-mediated deamination, the methylation level of GGm⁴CC context is much lower than that of m⁵CGATCG (Hagemann et al., 2018). Consistent with this, our study also showed the methylation level of mCG methylation was higher than those of mCHG and mCHH.

Some features in Synechocystis were also identified in this study. Firstly, we found about 3.0% CG, 0.9% CHG, and 1.0% CHH were methylated in Synechocystis, which is different from chordata (31.1–80.3% CG, 0.17–1.22% CHG, and 0.12–0.91% CHH were methylated) and magnoliophyta (22.3–59.4% CG, 5.92–20.9% CHG, and 1.51–3.25% CHH were methylated) (Cokus et al., 2008; Feng et al., 2010), but similar with the eukaryotic microalga Chlamydomonas reinhardtii (5.4% CG, 2.6% CHG, and 2.5% CHH were methylated) (Feng et al., 2010). Then, we found no obvious differences in methylation level distribution among functional regions, which maybe implies that there is no preference of hyper/hypo-methylation to a certain functional region in Synechocystis. In contrast, in eukaryotic microalga Chlorella variabilis, CG within gene body is universally methylated while CG within promoter region is seldom methylated (Zemach et al., 2010). These differences are maybe due to the lack of nucleosomal structure and different protein factors in Synechocystis. It is also reported that hyper-methylated CG in nucleosome linkers is involved in nucleosome positioning and chromatin compaction (Huff and Zilberman, 2014).

In eukaryotes (Lindahl, 1981; Bird, 2002) and some bacteria such as E. coli (Roberts et al., 1985; Kahramanoglou et al., 2012; Militello et al., 2014) and Helicobacter pylori (Kumar et al., 2012, 2018) there are evidence for the regulation role of DNA methylation on gene expression. However, when the correlation between our DNA methylation data of N72 and the published transcriptome data at 96 h after nitrogen starvation (Krasikov et al., 2012) was investigated, neither preference of hyper-methylated region to down-regulated expression nor preference of hypo-methylated region to up-regulated expression were found, we therefore speculate that there was probably no correlation between methylation status and gene expression in Synechocystis. Also, we found the expression of some photosynthesis-related genes seem to be independent of their methylation status in Synechocystis.

With applying methylation sensitive amplified polymorphism analysis to Oryza sativa L, Kou, H.P. has proved a whole-generation nitrogen deficiency stress (10 or 20 mg L^-1 nitrogen) could induce methylation alteration in the leaf-tissue of the stressed plants (Kou et al., 2011). In addition, with multiple electrophoresis-based polymorphism analysis techniques Yu et al. (2013) reported that nitrogen addition stress could also induced alterations in cytosine methylation patterns in natural populations of Leymus chinensis. However, these electrophoresis technology-based studies cannot provide the details about methylation change at base resolution. To investigate whether DNA methylation in bacterial could be influenced by surrounding environment, and if influenced, whether the altered DNA methylation pattern is inheritable, we compared the global DNA methylation among normal nitrogen (NC), nitrogen starvation (N72), and nitrogen recovery (NR) in the model cyanobacteria, Synechocystis. After 72 h of nitrogen deficiency about 40% mC sites were demethylated, but the methylation level of the remaining mC sites increased, which results in the almost unchanged overall methylation level. Because most Synechocystis cells can become dormant and survive as long as 45 days under nitrogen deficiency (lab observation), the altered DNA methylation pattern in N72 reflects the cellular response to nitrogen deficiency rather than the death or enrichment of some subpopulation. Conclusion can therefore be drawn that in cyanobacteria, environment stress could induce alternations in global DNA methylation pattern through reducing the number of mC sites but increasing the methylation level of the remaining mC sites.

The similar DNA methylation pattern in NR and N72 is interesting. Considering the cell doubling time for Synechocystis is about 23 h, cells in NR sample have already propagated for about 12 generations since being transferred into normal nitrogen, i.e., one cell propagates for 12 times and becomes into 4,096 cells. However, NR regained only 1% of mC sites which demethylated after nitrogen deficiency and preserved about 45% of mC sites induced by nitrogen deficiency. Furthermore, mC sites in NR preserved high levels of methylation as those in N72. All these implied that the altered DNA methylation pattern induced by nitrogen deficiency stress could be partly inheritable even though the environment stress was removed, which demonstrates the environmental memory of epigenetic modifications in cyanobacteria, as in higher plants (Henderson and Jacobsen, 2007).

Environmental stress could generate particular defense epigenetic changes across generations in plants (Holeski et al., 2012), and in Oryza sativa L, alteration in DNA methylation induced by nitrogen deficiency stress could contribute the resistance to the nitrogen stress (20 mg L^-1 nitrogen) for the non-stress suffered progenies which inherited the altered methylation pattern (Kou et al., 2011). The physiological significance of mC in bacterial genomes has raised enigmatic questions during several decades (Sanchez-Romero et al., 2015). Even though slightly higher growth rate seemed in NR cells, no significant growth difference between NC and NR can be found when being recultured in nitrogen limited condition. Considering that nitrogen deficiency stress lasts only 72 h, we speculate DNA methylation alteration in NR is not enough to increase the adaptation to nutrient limitation significantly. In the future examination of growth and genome-wide DNA methylation of Synechocystis having suffered nitrogen deficiency repeatedly or long term of nitrogen deficiency would provide more information. Furthermore, due to WGBS used in this study could not cover N⁶-methyladenine, combination of SMRT sequencing and WGBS will therefore be needed for more comprehensive understanding of epigenetic adaptation to the ever changing environment in Synechocystis.

In summary, we confirmed the widely spread mC methylation and response of DNA methylation to nutrient depletion in Synechocystis. What’s more, we verified the modified DNA methylation pattern could be partly inheritable in cyanobacteria. The results of this study might provide a primer to explore the wealth of information on the epigenetic transgenerational inheritance in prokaryotes.

JW contributed to conception and design of the study. LH, MD, PX, ZC, and YJ performed the experiments. LH, AL, and PX performed the statistical analysis. JW and LH wrote the first draft of the manuscript. JW, LH, and PX wrote the sections of the manuscript. YJ, MD, ZC, HL, and ZH contributed to manuscript revision. All authors read and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.