Silencing and Un-silencing of Tetracycline-Controlled Genes in Neurons

To identify the underlying reason for the controversial performance of tetracycline (Tet)-controlled regulated gene expression in mammalian neurons, we investigated each of the three components that comprise the Tet inducible systems, namely tetracyclines as inducers, tetracycline-transactivator (tTA) and reverse tTA (rtTA), and tTA-responsive promoters (Ptets). We have discovered that stably integrated Ptet becomes functionally silenced in the majority of neurons when it is inactive during development. Ptet silencing can be avoided when it is either not integrated in the genome or stably-integrated with basal activity. Moreover, long-term, high transactivator levels in neurons can often overcome integration-induced Ptet gene silencing, possibly by inducing promoter accessibility.


INTRODUCTION
Since the first description of the bacterially-derived Tet-controlled inducible systems [1][2][3], gene activation and inactivation has been achieved in different cell types in animals. The tetracyclinecontrolled transactivator-responsive minimal Tet promoters (P tet / P tet bi) [1,4] are activated by the tetracycline transactivator (tTA) [1] (Fig. 1A-left) and inactivated by tetracycline (Tet-off) or its derivatives such as doxycycline (Dox). The reverse tTA (rtTA) is a complementary genetic module uniquely suited for rapid gene activation by addition of Dox (Tet-on) ( Fig. 1A-right) in cultured cells [1] and in vivo [5,6]. Both systems have been successfully employed in transgenic mice for studying various biological functions [7].
As of today, the Tet systems are unique for reversible control of gene expression in higher eukaryotes [1][2][3]. Switching gene expression ''on'' and ''off'' is of extreme importance in understanding the function of genes in phenotypes. This becomes especially important when studying gene function(s) in the adult nervous system. Key studies provided the first strong evidence that the Tet-regulated gene expression can be used to analyze the involvement of genes in cognition in the mouse [8]. The successful employment of the a-subunit of CaMKII promoter for regulating gene expression in the forebrain was first demonstrated with the Tet-off system [9] and later with the Tet-on system [10,11].
However, in some studies in the central nervous system (CNS) employing the Tet-off system, the full reactivation of Tetregulated genes after Dox withdrawal was difficult to achieve once their expression had been suppressed prenatally by Dox [12][13][14]. As a consequence, reactivation was slow, expression patterns changed and expression levels failed to reach their original maximal values [12][13][14]. Other studies observed similar difficulties with the rtTA-dependent gene activation in the mouse brain [15]. Possible reasons include poor Dox penetrance across the blood brain barrier (BBB) [15], weak P tet /P tet bi activation in adult mice [16,17] and P tet /P tet bi susceptibility to silencing [18][19][20].
Here we investigated the regulation of several Tet-responder genes integrated at different sites in the genome [21][22][23] and in an episomal state [24][25][26][27] in the mouse. To regulate the Tet-responder genes, we employed transgenic mice with both forebrain-specific tTA [9] and rtTA expression (present study) and monitored responder gene activity by enzymatic activity and immunohistological analyses of brain slices. We have discovered that stably-integrated P tet bi modules in all responder mice become functionally silenced in the majority of neurons if P tet bi remain inactive during development whereas long-term, high transactivator levels in neurons can overcome P tet bi gene silencing. We also found that basal P tet /P tet bi activity yields permissive conditions for faster activation/re-activation, possibly by providing better accessibility of P tet /P tet bi to tTA/rtTA.

RESULTS
Forebrain-specific rtTA expression in transgenic mice can activate Tet-induced gene expression in some neurons For controlling transgenic Tet-responder genes, we used activator mice which express tTA [1,9] and rtTA-M2 [2,28] from the promoter for aCaMKII [9], which is active mainly in principle forebrain neurons. We engineered novel tTA and rtTA variants by replacing the potent activation domain, VP16, with three minimal activation domains of about 14 amino acids each, abbreviated as tTA2 and rtTA2-M2 [28]. Next, we generated five rtTA2-M2 CaMK mouse lines. For measuring gene activation, we Figure 1. Functional brain specific rtTA mice and tetracycline-induced, rtTA-dependent gene activation in the forebrain. (A) Schematics depicting the double-positive mice harboring constructs of tTA (left) and rtTA (right) under the control of the aCaMKII promoter and responder genetic modules under control of bidirectional Tet promoter (P tet bi) to express Cre recombinase and the firefly luciferase genes. Notice that the tTA and rtTA activate P tet bi in the absence and presence of Dox, respectively. (B) Luciferase activity in different brain subregions in double-postive transgenic mice (rtTA2-M2 CaMK -26LC-1) treated with Dox by different delivery routes (black bars, w (water), f (food) and ip (intraperitoneal injection). For comparison, luciferase activity in double transgenic mice (tTA CaMK 6LC-1) is also shown (grey bars). Luciferase activity presented on a log scale. (C) Forebrain-specific expression of Cre recombinase in tTA CaMK /LC-1 mice (left). Specific detection of rtTA2-M2 RNA in brain slices of rtTA2-M2 CaMK -2 mouse. Gene activation induced for two weeks of 9TB-Dox treatment and Cre signal is visualized by immunohistochemistry (right, insert shows a magnified image of dentate gyrus). doi:10.1371/journal.pone.0000533.g001 employed LC-1 reporter mice [21], which harbor a bidirectional Tet-promoter cassette (P tet bi) with genes for luciferase (Luc) and Cre recombinase (Cre) (Fig. 1A). We treated five rtTA2-M2 CaMK lines crossed to LC-1 with Dox and gene activation was visualized by both luciferase activity measurements from different brain areas and Cre immunohistochemistry on fixed brain slices. Robust transactivation (luciferase activity) was apparent in the olfactory bulb and low transactivation in the forebrain by three routes of Dox delivery; water, food and intraperitoneal injection (i.p.) ( Fig. 1B and data not shown). In none of our five rtTA2-M2 CaMK lines could we detect the expected robust forebrain specific gene activation observed with the tTA system (tTA CaMK /LC-1 mice) [9,14,21] (Fig. 1B and data not shown), despite the fact that rtTA2 mRNA was clearly detectable throughout the forebrain, with dentate gyrus (DG) granule cells having the highest rtTA2 mRNA levels in all rtTA2-M2 CaMK lines tested (Fig. 1C-middle and Fig.  S1A). One possible explanation for these results was lower effectiveness of Dox for rtTA-dependent gene activation.
To find a better substitute for Dox, we used a more hydrophobic Dox derivative, 9-tert-butyl-doxycycline (9TB-Dox), which when compared to Dox has approximately 10-times higher binding affinity to the tet repressor (compound available from Mark Nelson, Paratek Pharmaceutical, Boston, USA, data not shown) and thus it is also approximately 10-times more effective than Dox in rtTA2-M2 dependent gene activation in HeLa cells (data not shown). We found that granule cells of DG, the subregion with the highest rtTA2 mRNA level (Fig. 1C-middle and Fig. S1A), showed the best induced Cre expression in all rtTA2-M2 CaMK lines tested, but only a few Cre positive nuclei were seen in other forebrain regions ( Fig. S1B; see Dox-induction Protocol 1 in Materials and Methods). The influence of high rtTA2 mRNA levels for P tet bi gene activation in different brain regions is clearly apparent in the striatum of rtTA2-M2 CaMK -4 mice (Fig. S1A, B). Increasing the treatment time with 9TB-Dox facilitated rtTA2-M2 dependent gene activation (rtTA2-M2 CaMK -2 mice) in a larger fraction of DG granule cells when tested with the responder mouse line, LC-1 ( Fig. 1C-far right; see Protocol 2 in Materials and Methods), but poorly in the hippocampal CA1 and CA3 fields and in cortex, raising concerns that the BBB might be impeding Dox availability in the mouse brain.
Dox delivery is not the only limiting factor for lack of rtTA-dependent gene activation To investigate whether the BBB was the reason for inefficient Tetcontrolled gene activation in hippocampus and cortex of our rtTA2-M2 CaMK -2 mice, we used the transgenic line G3 which harbors under P tet bi control the LacZ and GFP genes [22], injected into the different brain regions 10 mg of Dox in 1 ml (0.7% NaCl) and tested for reporter gene activation. Seven days after Dox injection, GFP expression was induced mainly in DG granule cells on the injected side of the brain ( Fig. 2A), but only in a few cortical neurons. These results demonstrate that P tet bi activation in G3 responder mice failed even when Dox availability was not a limiting factor.
Our result also finds support from Dox-treatment of organotypic hippocampal slices derived from rtTA2-M2 CaMK -2/LC-1 mice. We again used the P tet bi-directed Cre expression of the responder LC-1 to exclude line specific effects. Dox added to the culture medium activated P tet bi expression only in DG granule cells but again not in CA1/CA3 neurons (Fig. 2B). P tet bi activation was not detected without Dox. These results suggest that the responder genes controlled by P tet bi may become silenced in neurons during development.
For P tet bi re-activation, incomplete suppression of tTA-dependent gene expression is necessary and sufficient We described previously reversible P tet bi regulation in neurons in the brains of tTA CaMK /LC-1 mice [21]. Such a reversible P tet bi regulation should not be possible if transcriptionally inactive P tet bi become silenced. Therefore, we analyzed in the brains of tTA CaMK /LC-1 [21] mice the transcriptional activity of P tet bi after switch-off by Dox treatment. For a detailed examination we treated adult mice (P60) with different Dox concentrations (2 to 2000 mg/ml in drinking water) and measured down-regulation of P tet bi-controlled luciferase expression by bioluminescence imaging of brain explants and biochemical measurements of whole brain extracts (Fig. 3A). Increasing Dox concentrations down-regulated Luc activity in a graded manner. High Dox concentration (2 mg/ ml) reduced Luc activity by approximately 95% but not to background levels observed in LC-1 mice not carrying the aCaMKII promoter tTA transgene (Fig. 3A). To bypass the BBB, we treated throughout gestation newborn tTA CaMK /LC-1 mice with 2 mg/ml of Dox in the drinking water ( Fig. 3B) when BBB does not exist and Dox availability into brain tissue is not a limiting factor and found complete suppression of gene activity in the brain of newborn pups. To rule out line specific effects, we also employed a different aCaMKII promoter tTA line, CN12 [29], and again found incomplete gene inactivation by Dox (2 mg/ml, 3 weeks) (Fig. 3B). These results indicate that a Dox concentration needed for complete P tet bi suppression cannot be reached in the brain by including 2 mg/ml of Dox in the drinking water. Since complete suppression of tTA-dependent gene activity in cultured mammalian cells [1] requires 10 ng/ml of Dox, it seems plausible that the effective Dox concentration in brain tissue in vivo will not exceed 10 ng/ml. Our results suggest that persistent basal P tet bi activity prevents the P tet bi locus from being silenced. Transcriptional activity might keep the chromatin locus ''open'', thus providing permissive conditions for gene re-activation upon removal of Dox.

Episomal P tet bi responders are activatable in rtTA2-M2 CaMK -2 mice
To provide direct proof that chromosomally-inserted P tet bi become transcriptionally inaccessible for functional tTA/rtTA dependent gene expression, we delivered by stereotactic injection recombinant adeno-associated virus (rAAV) carrying the P tet bi responder genes Cre and GFP into the hippocampus and cortex of rtTA2-M2 CaMK -2 mice. As AAV remains in an episomal state [24,25], we asked whether that non-integrated state might provide a permissive condition for P tet bi, possibly by increasing promoter accessibility [30]. Indeed, we detected in the cortex and hippocampus of rtTA2-M2 CaMK -2 mouse brains after viral injection with rAAV-P tet bi-GFP/Cre robust GFP expression by 9TB-Dox (i.p; 1.5 mg/injection, every 48 hours, 10 days). A large number of cortical neurons were visible by live GFP fluorescence and in fixed brain slices after 10 days of 9TB-Dox treatment (Fig. 4A, see Protocol 2 in Materials and Methods). Mice not treated with 9TB-Dox did not show P tet bi activation (data not shown). Because non-integrated rAAV-P tet bi-GFP/Cre, without tTA, have basal promoter activity, we used low virus titers to reduce background signal (data not shown). We estimate that 9TB-Dox induced, rtTA2-M2 dependent gene activation of rAAV-P tet bi-GFP/Cre can be as high as 20-fold, as determined by twophoton imaging of GFP fluorescence in fixed brain slices (data not shown). Further, we find that gene activation occurred rapidly within hours after a single pulse of 9TB-Dox by i.p. injection (Fig. 4B).
These results reveal two key points: first, episomal rAAV vectors are highly permissive for gene activation in neurons in vivo. Second, neuronal rtTA2-M2 levels in rtTA2-M2 CaMK -2 mice are functionally sufficient for rAAV-P tet bi activation. Therefore, it seems most reasonable to conclude that chromosomally-integrated P tet bi modules are functionally silenced in neurons.

Prolonged high tTA expression overcomes P tet bi silencing in neurons
It is well known that the potent transcriptional activator derived from Herpes Simplex virus, VP16, has anti-silencing gene activity [30][31][32]. Since tTA2 [28] is equipped with three minimal activation domains with similar properties, we investigated whether strong tTA expression in neurons of the responder mice would overcome P tet bi silencing. Therefore, we delivered either tTA2 or rtTA2-M2 under control of the human synapsin promoter (hSyn) by injecting rAAV-hSyn-tTA2 into cortex and hippocampus of Tet-responder line MTH-Cg2-7 that showed robust P tet bi promoter activity in the genetic background of tTA CaMK mice (Fig. S2A). Two weeks after rAAV infection of MTH-Cg2-7, tTA2 expression was apparent in cortical neurons but P tet bi activation was observed in only a very few neurons as revealed by GFP immunostaining (Fig. 5A-upper right panel and Fig. 5C). Similar results were found for eight different ''tight'' Tetresponder mouse lines. In total we analyzed approximately 100,000 neurons on serial sections (6 slices per mouse and 2 mice/line) immunostained for GFP and tTA2. For eight different ''tight'' Tet-responders, we found that after two weeks of virus infection, GFP gene activation was apparent in only a small fraction of neurons in CA1, DG and cortex, whereas after three weeks and beyond GFP activation was nearly complete in CA1 and DG, but remained incomplete in cortical neurons ( Fig. 5C and data not shown). Next, we wanted to determine whether longterm expression of tTA2 would overcome P tet bi silencing. Consistent with the role of the transcriptional activation domain of VP16 in gene un-silencing [30][31][32], we found that high tTA2 levels in neurons gradually increase P tet bi activation over time and thus P tet bi un-silencing (Fig. 5A-lower panel and C). However, in mouse line MTH-Cg2-19 with basal P tet bi activity [23] we observed widespread GFP expression in CA1, DG and cortical subregions already after two weeks of viral tTA2 expression (Fig. 5B-upper panel), and 9TB-Dox induced gene activation in MTH-Cg2-19 mice was also apparent to similar extend when rtTA2-M2 was delivered by rAAV ( Fig. 5B-lower panel). These results indicate a more permissive nature of P tet bi in the MTH-Cg2-19 line (Fig. 5B). Again, GFP expression was incomplete in a majority of cortical neurons (,40%) with lowest response observed in cortical layer 4 (Fig. 5C). It is conceivable that different neuronal populations differ in their requirements for gene re-activation of silenced P tet -controlled genes.
Although viral-mediated tTA/rtTA delivery into neurons of MTH-Cg2-19 mice identifies this line as ''more permissive'' than the other eight responder lines tested so far, 9-TB-Dox induced gene activation in MTH-Cg2-19/rtTA2-M2 CaMK -2 double-positive mice remained restricted to DG granule cells (Fig. S3A). It is possible that basal P tet bi gene activity in neurons of MTH-Cg2-19 mice is still insufficient and rtTA-M2 levels provided by rtTA2-M2 CaMK -2 mice are not high enough for un-silencing. As a control, we can show that MTH-Cg2-19 mice can be activated in a forebrain-specific manner in crosses with tTA CaMK mice (Fig. S3B, also see Fig. S2B).
Altogether, we have demonstrated here that prolonged high tTA2 expression in responder mice can induce P tet bi re-activation, possibly due to the counter-silencing effects of the transcriptional activator [30][31][32].

DISCUSSION
Over the last 10 years, the reported performance of the tTA system for reversible control of gene expression in neurons has been consistently remarkable. For example, an activator line driving tTA expression in principle neurons under the control of the aCaMKII promoter, tTA CaMK [9], has been elegantly used to reversibly regulate the expression of various responder transgenes for studying synaptic plasticity and learning and memory [9,12]. Targeted expression of the tTA protein specifically in CA1 principle neurons has further revealed a direct link between selective neuronal circuit function(s) and memory consolidation [33]. On the other hand, in spite of a few reported successes [10,11,34], others report difficulties with the Dox-induced, rtTAdependent gene expression in neurons [15,16,18]. The controversial performance of the rtTA system in neurons prompted us to systematically examine individual components of the Tet system, namely Dox availability across the BBB, rtTA levels and P tet /P tet bi activation.
We have clearly demonstrated here that lack of gene activation by rtTA in neurons in adult mice is mainly due to the silencing of stablyintegrated P tet /P tet bi modules. By combining rtTA2-M2 and an improved Dox derivative, 9TB-Dox, we have provided strong evidence for fast and robust gene activation in neurons when episomal P tet bi modules had been introduced into the brain of our rtTA2-M2 CaMK -2 mice (live GFP fluorescence, Fig. 4A). Consistently, we could not achieve gene activation in eight different ''tight'' Tet-responder lines tested with tTA2 delivered via rAAV in vivo into brain areas (Fig.5A-upper panel, Fig. 5C and data not shown). This indicates that the poor activation of P tet bi responder genes in the mouse brain is neither due to the poor penetration of Dox across the BBB nor to low rtTA levels in neurons, but is most likely due to silencing of the stably-integrated P tet bi in the genome.
We asked whether there are specific environmental conditions which would prevent the P tet bi locus from becoming silenced. Towards this question, we found that unlike ''tight'' responders where P tet bi modules integrated in a genomic site with no intrinsic activity, P tet bi integrated in a genomic site with basal P tet bi activity will keep the locus permissive. We were able to identify one such mouse line, MTH-Cg2-19 [23], which we label as ''permissive line'' because tTA-dependent P tet bi activation in these mice occur earlier compared to ''tight'' lines (compared Fig. 5A-upper panel with Fig. 5B-upper and lower panels, also see Fig. 5C). These observations strongly support the hypothesis that basal P tet bi activity can keep the chromatin locus accessible for tTA/rtTA dependent gene activation.
Consistent with this idea, we have also discovered that basal P tet bi activity is essential for achieving multiple cycles of gene inactivation by Dox and re-activation upon Dox withdrawal in tTA-activator/P tet bi-responder mice [9,21,35]. Previously, we showed that tTA-dependent gene expression in tTA CaMK /LC-1 mice begins prenatally, at embryonic day 12.5 (E12.5) [14]. It is possible that early tTA expression in neurons during development activates P tet bi and antagonizes silencing. Moreover, even when tTA CaMK /LC-1 and tTA CN12 /LC-1 mice are kept on Dox (2 mg/ ml) for weeks, luciferase activity is not completely suppressed to levels observed in single-positive LC-1 mice (Fig. 3A, B), rendering the P tet bi locus permissive in neurons and enabling gene reactivation upon Dox removal [9,21].
How would one create conditions to unlock the silencing in ''tight'' Tet-responder lines? We found that even high tTA2 expression can not activate P tet bi in neurons two weeks after rAAV infection but long-term (three weeks and beyond) high tTA2 levels in neurons can overcome P tet bi gene silencing (Fig. 5A-lower  panel, C), perhaps involving chromatin remodeling by transcriptional activation domain in tTA/rtTA modules [28,36]. We have also found that tTA2-dependent P tet bi un-silencing is nearly complete in neurons of hippocampus, particularly in the DG subregion but incomplete in the cortex (Fig. 5C). Why about 40% of cortical neurons remain resistant to gene activation even in the presence of high tTA2 levels is not clear. It is conceivable that diverse neuronal types might have different requirements for gene activation, silencing and un-silencing (or re-activation). In our studies, gene activation appears permissive in olfactory receptor neurons and DG granule cells, consistent with previous observations [37]. These cell types might be generally more resistant to gene silencing, possibly because they regenerate throughout life.
To reveal molecular changes which might orchestrate P tet /P tet bi silencing, we performed DNA methylation analyses on P tet bi, but were unable to correlate the observed fraction of DNA methylation of genome-integrated P tet bi with the fraction of P tet bi silencing in neurons. We also failed to find a significant difference in DNA methylation in brain and liver tissues (data not shown). Individual DNA sequencing of clones of P tet bi from brain tissue revealed that only 4% of the clones show strong methylation over the entire sequence of P tet bi (Fig. S4, also see ''Notes on P tet bi methylation'', Fig. S5). This result does not account for about 100% of cortical neurons (representing approx. 60% of total DNA pool) for which we observe strong functional P tet bi silencing (Fig. 1C-far right and Fig. S1B, Fig. 5A-upper panel and 5C). The simple fact that we have identified a strongly methylated P tet bi in cortical genomic DNA hints at epigenetic control mechanisms in certain cell types in the brain. Therefore, we conclude that for P tet bi silencing other mechanisms, besides DNA methylation, are likely to be more relevant, such as chromatin-controlled promoter accessibility to transcriptional factors [30].
We thus favor the idea that different cell types might have different degrees of accessibility to P tet /P tet bi [30]. What determines whether a gene is either silenced permanently or is allowed to un-silence by an environmental signal? Of special interest is the ability of immediate-early-genes (IEGs), including cfos and arc, to become transcriptionally activated by a specific stimulus [38]. It is well known that IEGs have different levels of basal transcriptional activity in different cell types in the ''uninduced state'' [38]. It is still unclear whether basal transcriptional activity of IEGs provide a permissive condition for gene activation in certain neuron types. It is tempting to speculate that basal IEG activity might be needed for activityinduced rapid gene activation in neurons.
Our studies have delineated the necessary conditions for reversible control of gene expression in neurons. First, we have shown that the stably-integrated P tet bi becomes silenced, provided it is inactive during development, and that high tTA/rtTA levels in neurons allow for P tet bi un-silencing over time. Second, we have introduced an alternative approach for reversible control of gene expression in neurons by using episomal responder delivery via rAAV. This combinatorial genetic approach of combining viruses with genetically altered mice is eminently suitable for targeting cell type specific gene expression in select brain subregions. For example, in combination with in vivo 2-photon imaging [23,39], it is likely to facilitate direct visualization of functional neuronal circuits, by employing genetically-encoded fluorescent biosensors as reporters of cell physiology such as for calcium [40,41] and synaptic transmission [42].
The defined transcriptional elements of the Tet systems described here should be useful to probe mechanistic issues in regard to chromatin-remodeling in the mammalian neurons and help in identifying factors that modulate promoter silencing and un-silencing in neurons.

In situ RNA hybridization
In situ hybridization studies using 35 S-labeled oligonucleotide probes were performed according to the method described previously [43]. Saggital 10 mm thick frozen brain slices were prepared by cryostat sectioning from individual animals derived from five different independent rtTA mouse lines. Three oligonucleotide sequences from different coding regions of synthetic tTA/rtTA were hybridized to different brain slices in parallel. All probes gave similar results. The oligonucleotide sequences used are indicated below. Data shown in Fig. 1C- To assess for specificity and rule out non-specific labeling of brain sections, each radiolabeled oligonucleotide was hybridized with a 100-fold excess of unlabelled oligonucleotide. Images were processed after 4 week exposure using Biomax MR (Kodak) X-ray films.

Quantifying gene activity and expression patterns
Mouse brains were saggitally cut into two halves. One half was fixed in 4% paraformaldehyde, the other half was used for the measurement of luciferase activity in different brain subregions. Fixed brain slices were cut to a thickness of 75-100 mm using vibratome (VT 1000S; Leica Instruments, Wetzlar, Germany). Luciferase activity was measured from brain extracts as described previously [6,21] and immunohistochemistry for Cre recombinase and b-galactosidase was performed also as described previously [14,21]. Green fluorescence protein (GFP) [44] was visualized in fixed slices either by live fluorescence imaging or by immunohistochemistry using GFP-specific polyclonal rabbit antibodies (Clontech, Mountain View, California, USA) (Krestel et al. 2001) and the DAB peroxidase system (Vectastain ABC Kit; Vector Laboratories, Burlingame, California, USA) or by direct observation of fluorescence with an upright microscope (Zeiss, Oberkochen, Germany) equipped with GFP filters. Dual labeling of neurons for both tTA and GFP was not possible because we needed to employ antigen-specific polyclonal rabbit antibodies for high sensitivity detection of both tTA and GFP.

Sodium bisulphite conversion, PCR amplification and sequencing
Genomic DNA was prepared according to standard procedures. Sodium bisulphite treatment of genomic DNA was performed as described previously [49] with minor modifications. Since methylated cytosines have the same base-pairing characteristics as unmethylated cytosines, DNA is chemically modified to distinguish between the two species. Purified genomic DNA was treated with sodium bisulphite, resulting in the conversion of unmethylated cytosine to uracil. In a subsequent PCR uracil is replicated as thymine. Methylated cytosines are protected from conversion and remain as cytosines. Thus, detection of a ''C'' in a sequencing reaction reflects methylation at that site whereas detection of a ''T'' indicates no methylation. PCRs were performed on MJ Research thermocyclers (Waltham, Massachusetts, United States) in a final volume of 25 ml containing 16 PCR Buffer, 1 U / Taq/ DNA polymerase (Qiagen), 200 mM dNTPs, 12.5 pmol each of forward and reverse primers, and 7 ng of bisulphite-treated genomic DNA. The amplification conditions were 95uC for 15 min and 40 cycles of 95uC for 1 min, 55uC for 45 sec and 72uC for 1 min and a final extension step of 10 min at 72uC. PCR products were purified using ExoSAP-IT (USB Corp., Cleveland, Ohio, United States) and sequenced employing the PCR primers and the ABI Big Dye Terminator v1.1 cycle sequencing chemistry (Applied Biosystems, Foster City, California, United States) followed by capillary electrophoresis on an ABI 3100 genetic analyzer. AB1-files were interpreted using ESME, which normalizes sequence traces, corrects for incomplete bisulphite conversion and allows for quantification of methylation signals. The following primers were used both for PCR amplification of bisulfite converted genomic DNA and sequence reactions: Amp806 fw: 59-TATAGTTTTATGTAGTTGTTT-TTTAG-39 and Amp806 rev: 59-AATAAATTAAACACCTTCCTC-39.
PCR products were cloned using the TOPO TA Cloning Kit (Invitrogen). Sequencing was carried out using the M13 Forward and M13 Reverse primers by 3100 Genetic Analyzer (Applied Biosystem). All bisulphite sequencing experiments were performed in duplicate and sequencing data was analyzed using SeqMan (Lasergene, DNASTAR, USA). The conversion rate was approxinately 95%. We have fully sequenced and characterized 71 clones. clones is plotted with the number of methylated CpG detected (yaxis) against the number of independent clones (x-axis). The actual data set collected is based on methylation detected from sequencing (red) and the calculated data is based on 5% C-to-T conversion loss (blue). Three of seventy-one clones (4%) are strongly methylated.