Transcription factor Pit-1 affects transcriptional timing in the dual-promoter human prolactin gene

Gene transcription occurs in short bursts interspersed with silent periods, and these kinetics can be altered by promoter structure. The effect of alternate promoter architecture on transcription bursting is not known. We studied the human prolactin (hPRL) gene that contains two promoters, a pituitary-specific promoter that requires the transcription factor Pit-1, and displays dramatic transcriptional bursting activity, and an alternate upstream promoter that is active in non-pituitary tissues. We studied large hPRL genomic fragments with luciferase reporters, and used bacterial artificial chromosome (BAC) recombineering to manipulate critical promoter regions. Stochastic switch mathematical modelling of single-cell time-lapse luminescence image data revealed that the Pit-1-dependent promoter showed longer, higher-amplitude transcriptional bursts. Knockdown studies confirmed that the presence of Pit-1 stabilised and prolonged periods of active transcription. Pit-1 therefore plays an active role in establishing the timing of transcription cycles, in addition to its cell-specific functions. at CC at GG reagent. Media transfection and supplemented with  g/ml G418. Media + antibiotic Colonies 2-3 after culturing in selective were re-cloned into individual wells of a 48 well plate. The stable transfectant clones which were for large culture vessels as necessary. These data show that the hPRL alternative and proximal promoters both generate cyclical patterns of transcription, but that transcription in the absence of Pit-1 binding occurs at a reduced rate, with significantly shorter periods of active transcription in any given cycle. the transcription factor Pit-1 to cis-acting regulatory elements plays a key role


Introduction
Prolactin is a multifunctional mammalian hormone with a major role in lactation as well as other biological functions, including reproduction, immunomodulation and behaviour [1,2]. This polypeptide hormone mainly originates in the lactotroph cells of the anterior pituitary gland, but is also expressed at extra-pituitary sites, such as brain, decidualised endometrium, myometrium and circulating lymphocytes, and hence requires tissue-specific transcriptional control mechanisms to regulate its functional versatility [1][2][3].
The human PRL (hPRL) gene is located on the short arm of chromosome 6 and consists of five coding exons within a region of 10 kb. The gene has two distinct, alternative promoter elements that are 5.8 kb apart and show selective and tissue-specific activation [2,4]. The proximal promoter is located immediately upstream of exon 1b and regulates hPRL transcription in the pituitary gland. This is often referred to as the pituitary-specific promoter and contains multiple binding sites for the pituitary-specific transcription factor Pit-1 [5,6]. In lymphocytes and human endometrial cells, hPRL expression has been shown to be driven by an alternative promoter, upstream of the pituitary transcriptional start site [7]. This evolved from a long terminal repeat-like transposon, resulting in transcription of the additional exon 1a, giving rise to mRNA that is 150 bp longer than the pituitary mRNA [2,8]. The different sized transcripts do not give rise to different protein isoforms, but are believed to contribute an additional level of regulation of PRL transcription in different cellular and functional contexts (pituitary versus extra-pituitary) by influencing the stability or translational efficiency of alternatively transcribed messages [2,8]. Expression of the longer alternative promoter transcript has classically been viewed as a product of extra-pituitary sites, and considered to be Pit-1 independent [4,9,10].
Pit-1 (also known as POU1F1) is a member of the POU family of transcription factors that are characterised by the presence of a bipartite DNA binding domain, known as the POU domain [11]. It is essential for the differentiation of lactotroph, somatotroph and thyrotroph cells; Snell and Jackson dwarf mouse models carry mutations in the Pit-1 gene and show no development of these pituitary cell types [6], and Pit-1 mutations commonly result in hypopituitarism [12]. Additionally, Pit-1 binding has been shown to play a crucial role in both basal and hormonally induced activity of the hPRL promoter [13][14][15].
The hPRL genomic locus has many conserved regions far upstream of the transcriptional start site.
Outside of the genomic locus, the PRL gene is surrounded by over 1Mb of non-coding DNA, a gene desert which could have a functional impact on hPRL regulation [16]. It has therefore been important to develop strategies for studying the transcription control function of the complete genomic region.
A c c e p t e d M a n u s c r i p t BAC based reporter systems permit the inclusion of far-distant regulatory elements and can prove particularly beneficial in analysing function of promoters that comprise a complex array of cis-acting regulatory elements [7,17,18]. We previously generated a BAC reporter construct which spanned 163Kb of the hPRL genomic locus including 115kb upstream and 38kb down-stream of the PRL gene and expressed firefly luciferase (Luc) under the control of the entire hPRL gene (referred to as hPRL WT BAC) [18].
In our earlier work, single cell imaging of the hPRL WT BAC in pituitary cell lines and tissues, showed that the hPRL gene displays dramatic pulses in transcriptional activity [19,20]. This activity has been observed for many other genes, including the hGH (human growth hormone) gene [21], and appears to be a general phenomenon that is intrinsic to gene regulation [21][22][23][24][25][26][27]. The characteristics of these transcriptional pulses may be susceptible to modulation, as part of normal physiological control, and this might be expected to impact on overall levels of gene expression. In the present study, to study the role of Pit-1, we have compared the hPRL WT BAC with a construct in which the entire 5kb pituitary promoter was deleted, leaving intact exon 1a and the upstream promoter (referred to as hPRL PitProKO BAC). We compared the transcriptional behaviour of the two constructs when they were stably transfected into GH3 pituitary cell lines and found that the alternate promoters directed distinct patterns of transcriptional bursting in single cells. We examined the role of Pit-1 binding sites and the effect of modulation of Pit-1 levels in pituitary cells, and found that rather than just being necessary for transcription, the binding of Pit-1 regulates the timing of hPRL transcription. These data suggest a new and unsuspected function of this well studied transcription factor in regulating the dynamics of pituitary-specific gene expression in an important tissue model for the control of endocrine function.

Generation of hPRL pituitary promoter knock-out BAC (hPRL PitProKO BAC)
The generation of hPRL WT BAC with Luc reporter has been described previously [18]. The hPRL WT BAC spans 163 kb of the human PRL genomic locus including 115 kb upstream and 38 kb downstream of the PRL gene (10kb) and expresses the luciferase reporter gene under the control of both the hPRL proximal and alternative promoters. In order to study specific prolactin gene activation via the alternative promoter, we sought to remove pituitary-specific promoter elements from the hPRL WT BAC by the targeting strategy based on the seamless recombineering technology [28]. Briefly, the ~5kb hPRL proximal promoter was first replaced with the GalK (E.coli galactokinase k gene) coding sequence. A positive recombinant selected on galactose containing minimal media, was subjected to a second round of recombination to remove the inserted GalK sequence and replace with the short (256bp) immediate hPRL promoter sequence necessary for transcription initiation which also contains the alternative RNA splice acceptor site at position -246 relative to Luc translation start site [18]. Within this region 3 Pit1 binding sites are found, these elements were mutated in vitro (see A c c e p t e d M a n u s c r i p t site directed mutagenesis below) prior to recombination to remove remnants of pituitary-specific regulatory elements. Positive recombinants were identified through glycerol/deoxygalactose screening and BAC size and integrity confirmed by pulse field gel electrophoresis (PFGE). Mutations in Pit-1 binding sites were confirmed by sequencing. The following two primer sets (italics denote homology arm sequence) were used for primary and secondary recombination respectively;

Site-Directed Mutagenesis
The hPRL 5kb Luc plasmid used for generation of the 256bp intermediate hPRL promoter sequence has been described previously [29]. Site mutations were induced in the three Pit-1 binding sites contained within the 256bp region [13,30]  Site 1, 5′-gcctgattcattCGCtAcCtgaagatatcaag-3′; site 2, 5′-tcttcctgaatatgGatCCgaaataaaatacc-3′; site 3, 5′-cttttggcctaatCCatGGaaatccttcctag-3′. Capital letters represent mutated bases which resulted in the introduction of unique restriction sites as Kpn1, BamH1 and Nco1 respectively. These restriction sites were subsequently used for screening the mutated clones. The 256bp region with mutated Pit-1 binding sites was amplified with BAC homology arms appended to primers and recombineered into the hPRL BAC Luc construct with deleted proximal promoter.

Cell culture and generation of stably transfected BAC cell lines
Pituitary GH3 cells were maintained in phenol red-free DMEM with pyruvate/glutamine and 10% fetal bovine serum (FBS). Jurkat cells were cultured in RPMI (Life Technologies Inc., Gaithersburg, MD) supplemented with glutamine and 10% FCS. Serum starving conditions were in media containing 0.25% BSA in place of FBS.
For stable transfection of cell lines, BAC DNA (hPRL WT BAC or PitProKO BAC) was prepared by maxiprep (BAC100 Nucleobond kit, Macherey-Nagel, Germany) and 3g was used to transfect 10 6 A c c e p t e d M a n u s c r i p t GH3 cells in a 10cm dish or 2 × 10 6 Jurkat cells in 25cm 2 flask using ExGen500 transfection reagent.
Media was changed 48h post transfection and supplemented with 500g/ml G418. Media + antibiotic were refreshed every 3-4 days. Colonies formed 2-3 weeks after culturing in selective media were recloned into individual wells of a 48 well plate. The stable transfectant clones which were found positive for Luc expression were sequentially scaled up to large culture vessels as necessary.

Endpoint luminometry assays
GH3 cells (1×10 6 ) were washed once with PBS then lysed using 200l of lysis buffer (25mM Tris/PO4, 10 mM MgCl 2 , 5 mm EDTA, 15% glycerol, 0.1% Triton X-100, and 0.1 mg/ml BSA). Cell lysis was aided by agitation at room temperature for 15 minutes, ATP added to a final concentration of 1mM and luciferase activity of samples measured using a FLUOstar Omega (BMG Labtech). Each experiment was performed in duplicate in three independent studies.

Live-cell luminometry assays
Pituitary GH3 cells (1.5×10 4 per well) were seeded into 96-well microplates (white opaque, PerkinElmer) in serum-free media containing 1mM luciferin (Biosynth). After 24 hours (h) cells were stimulated as indicated and microplates sealed with Breathe-Easy sealing film (Sigma). Luciferase activity from each well was measured using FLUOstar Omega (BMG Labtech) over a period of 24h while maintaining the cells at 37C in the presence of 5% CO 2 . Photon counts from each well were integrated over 5 seconds (s) after every 15 minutes (min). Results are shown as mean fold induction relative to an untreated control and represent triplicates of three independent experiments.

RT-PCR
Total mRNA was extracted from GH3 and Jurkat cells stably transfected with PRL-Luc BAC or 5kb Null BAC using an RNeasy kit (Qiagen, Valencia, CA). Genomic DNA was removed using gDNA eliminating columns provided with the kit. First strand cDNA was synthesized using random octamers in VILO Superscript II (Life technologies). RT-PCR for the detection of pituitary or extra-pituitary A c c e p t e d M a n u s c r i p t derived Luciferase mRNA was performed using forward primer 98 [7] in PRL exon-1a and reverse primer 11 [18] in the Luciferase gene.

Real time luminescence imaging
GH3 BAC transfectant cells and collagenase dispersed primary pituitary cells were cultured in 35mm glass coverslip-based dishes (Greiner Bio-One, UK) in the presence of 10% FBS. Luciferin (1mM) was added at least 10h before the start of the imaging. Cells were transferred to a Zeiss Axiovert 200 microscope in a dark room, equipped with an incubator maintained at 37C, 5% CO 2 in humid atmosphere. A bright-field image was taken before and after the luminescence imaging to track the localisation of cells. Luminescence images were obtained using a Fluar 10x, 0.75 NA objective.
Images were captured using a photon-counting charge coupled device camera (Orca II ER, Hamamatsu Photonics, UK). Sequential images, each integrated over 10 min, were acquired using Kinetic Imaging software AQM6 (Andor, Belfast, UK). The same software was used for the analysis of the imaging data. Regions of interest were drawn around each single cell, and total photon counts for individual cell areas were obtained from each image. Mean luminescence intensity data were collected after the average instrument dark count (corrected for number of pixels being used) was subtracted from the luminescence signal. Each imaging experiment was performed at least 3 times, with a representative experiment presented in figures.

Real-time luminescence imaging of primary bone marrow cells from hPRL-Luc transgenic rats
The generation of hPRL-Luciferase rats has been described previously [18]. Bone marrow (BM) cells were harvested from male hPRL-Luc rats, and erythrocytes were removed with ammonium chloride lysis solution. BM cells were cultured in RPMI 1640 containing 10% (V/V) FBS in 35mm glass coverslip based dishes (Greiner Bio-One, UK) and left to adhere. Luciferin (1mM (Biosynth)) was added at least 6h before the start of the experiment. Cells were imaged with a Zeiss Axiovert 200, equipped with an XL incubator (maintained at 37C in a 5% CO2 in humidified conditions) in a dark room. Luminescence images were collected using a Fluar 10×, 0.5-NA objective (Zeiss), and captured using a photon-counting charge coupled device camera (Orca II ER; Hamamatsu Photonics).
Sequential images were taken with a 30 minute integration period, then analysed using Kinetic imaging software AQM6 (Andor). Lipopolysaccharide (LPS) was added directly to the dish (0.05% final concentration. Imaging experiments were performed 3 times, with a representative experiment shown in Figure 1b.

Pit-1 siRNA knockdown
For siRNA transfections 1.5x10 5 hPRL WT GH3 cells were seeded into 12 well plates with 15pmol of Pit-1 Stealth RNAi TM (ThermoFisher Assay ID HSS108266) in the presence of Lipofectamine RNAimax reagent (ThermoFisher Scientific) according to the manufacturers' protocol. Cells were A c c e p t e d M a n u s c r i p t also transfected in the same manner with Stealth RNAi TM siRNA negative control, medium GC duplex to control for sequence independent effects. Seventy two hours after transfection cells were used for live cell luminescence imaging experiments or lysed to assess the efficiency of Pit-1 knockdown.

Cell lysis and immunoblot analysis
After transfection with either Pit-1 siRNA, siRNA negative control or lipofectamine RNAimax reagent only, hPRL-WT GH3 cells were lysed with Laemlli sample buffer (2% SDS, 60mM Tris-Cl (pH 6.8) 0.02% bromophenol blue, 0.1M DTT). Proteins were resolved by SDS-PAGE, transferred to nitrocellulose and blocked in 5% non-fat milk. Blots were incubated in primary antibodies over night at 4 o C (Abcam), washed and then incubated with horse radish peroxidase conjugated secondary antibodies and developed using enhanced chemiluminescence. Efficiency of Pit-1 knockdown was assessed in at least 3 experiments.

Stochastic switch model analysis
To quantify the duration of transcriptional phases raw single-cell luminescence data from 3 independent experiments for each imaging condition (hPRL WT, PitProKO, hPRL WT Pit-1 knockdown and hPRL WT scrambled siRNA experimental conditions) was analysed using a previously developed stochastic switch model (SSM) [33]. This model utilized a reversible jump Markov chain Monte Carlo algorithm using back-calculated transcription rates from an observed luciferase signal to identify transcriptional switches, using known luciferase protein and mRNA degradation rates previously reported in GH3 cell lines, as used in the present study [33]. This analysis was performed using Matlab 2014a software (MathWorks), including the Bioinformatics and Statistical toolboxes.

Contrasting cycles of hPRL transcription directed by the alternative and proximal promoters
Pulsatile cycles of prolactin transcription have previously been shown in single cells from pituitary cell lines and in primary pituitary cells [17-20, 29, 34-37]. hPRL promoter directed reporter gene expression has also been observed in primary cultures of bone marrow myeloid cells (taken from transgenic rats expressing luciferase under the control of the hPRL WT BAC), where it is driven by the alternative promoter. This expression is significantly enhanced through cytokine stimulation, such as LPS treatment [35].
The first aim was to compare the dynamic behaviour of the wild-type hPRL promoter (hPRL WT BAC) in pituitary and non-pituitary cells, using primary pituitary cells and bone marrow myeloid cells from hPRL WT BAC transgenic rats (Figure 1, a and b). Pulsatile cycles of transcriptional activity

Generation of the proximal-promoter knock-out hPRL BAC Luc (hPRL PitProKO BAC) construct
In order to evaluate the influence of promoter architecture on transcriptional bursting the hPRL WT BAC was modified to delete the entire 5kb proximal pituitary promoter, leaving the upstream exon 1a and the alternative promoter intact (Figure 2a). This provided a construct in which luciferase expression was directed from the upstream alternative hPRL promoter, enabling direct comparison with the hPRL WT construct. The hPRL proximal promoter contains an alternative splice acceptor site (ASAS) at position -246 relative to its transcription start site, which is important for the function of the alternative promoter [18] (for diagrammatic analysis of the region and constructs see Figure 3).
In addition, this region contains three Pit-1 binding sites, which could have the potential to drive expression from the pituitary promoter. We therefore performed site-directed mutagenesis to remove the three Pit-1 binding sites in the hPRL Luc 5kb plasmid (described previously [29], Figure 2b).
After mutation, a 256bp DNA sequence containing the ASAS and mutated Pit-1 binding sites was amplified from the hPRL Luc 5kb plasmid and inserted into the hPRL Luc BAC with the 5kb pituitary promoter region deleted (Figure 2c). Chromatin immunoprecipitation (ChIP) confirmed absence of Pit-1 binding to the mutated construct (Figure 2d). The resulting hPRL BAC, termed PitProKO BAC, expresses luciferase under the control of the alternative promoter, and contains the ASAS and mutated Pit-1 sites. The remaining putative Pit-1 binding sites in the alternative promoter have previously been found to be non-functional [7], and were therefore not removed or mutated.

The hPRL alternative promoter is active in GH3 pituitary cells independently of Pit-1
The hPRL alternative promoter is often referred to as the extra-pituitary promoter, based on the notion that its activity is confined to sites outside of the pituitary. Here, to assess the contribution the alternative promoter plays in the transcriptional regulation of the hPRL gene in the pituitary, we generated rat pituitary GH3 stable cell lines expressing hPRL WT BAC-Luc or hPRL PitProKO BAC-Luc. Incorporation of the entire BAC without truncation into the cell lines was confirmed following genomic DNA extraction, with PCR using primer sets spanning the entire BAC construct, including the deleted 5kb region (Supplemental, Figure S1 [38]). The luciferase expression levels of three WT and three PitProKO clonal GH3 cell lines were compared. All cell lines displayed high and fluctuating levels of reporter gene activity (Figure 4a-e). However, WT clonal lines showed markedly higher luciferase activity compared with PitProKO clones (as seen in Figure 4a-b, with quantitative luminometry in Figure 4e). These data confirm that the alternative promoter is active in pituitary cells, but that the activities of both promoters are required for higher-level gene expression.  [17,39].

The hPRL alternative promoter displays bursting gene transcription
We have previously shown the pulsatile nature of hPRL transcription in pituitary cells using both 5kb and BAC constructs in pituitary cell lines and in normal pituitary tissue in transgenic rats (17,20,25,(27)(28)(29)(30)(31)(32). The recombinant BAC constructs allowed us to test whether this bursting transcriptional behaviour was specific to activity directed by the proximal promoter and whether Pit-1 binding elements were required. Real-time luminescence imaging was conducted over 20h periods in living cells to directly compare the dynamics of luciferase expression from single unstimulated hPRL WT and PitProKO cell lines (Figure 4a-d). Cyclical patterns of transcriptional activity were observed in both cell lines, with lower amplitude in the PitProKO cells (normalised data shown in Figure 4c and d). The amplitude difference was comparable to the difference in luciferase activity observed between hPRL WT and PitProKO clonal cell lines seen in luciferase assays (Figure 4e). In addition to a reduction in amplitude, analysis of single cell traces also revealed that the active 'on' periods of gene transcription appeared significantly shorter in PitProKO cells than in WT cells, in other words that the pulses seemed to be shorter and sharper than in the WT cells (Figure 4c and d insets). To quantify the differences in transcription timing between the two promoters, single-cell luminescence data was analysed using a previously described stochastic switch model (33). This modelling analysis confirmed a significant reduction in the 'on' time of the PitProKO promoter compared to WT cells (Figure 4f), confirming the initial impression that the peaks of transcriptional activity were briefer. A c c e p t e d M a n u s c r i p t These data show that the hPRL alternative and proximal promoters both generate cyclical patterns of transcription, but that transcription in the absence of Pit-1 binding occurs at a reduced rate, with significantly shorter periods of active transcription in any given cycle.

Binding of the pituitary-specific transcription factor Pit-1 to cis-acting regulatory elements plays a key role in prolonging the duration of 'on periods' of active gene transcription
One major factor that could explain these differences in transcription timing was the presence or absence of Pit-1 binding to the different hPRL promoters, in different cell contexts. We hypothesised that rather than simply being required for pituitary hPRL gene expression, Pit-1 might also play a role

Asymmetry between activation and deactivation during a transcriptional pulse
In previous work [21] on the transcription of human growth hormone we identified an asymmetry in the number and size of rate-increasing and rate-decreasing switches, resulting in a predominance of an all-or-nothing activation step, followed by a multi-step graded reduction. We therefore investigated if there was such an asymmetry in the current system. To do this, each dataset (WT, PitProKO, WT-Pit-1 knockdown and WT scrambled siRNA) was processed by the SSM, and to each switch in the transcription rate we associated a score S which is the ratio of the change in the rate to the highest of the rates before and after the switch. Thus, a switch with S close to 1 is more binary (complete) while one with a smaller S is partial. We found that for each construct and condition considered, the distributions of S for up and down switches were significantly different with the asymmetry clearly present. The down switches had substantially more partial events than the up switches which were dominated by complete switches (Figure 6 A c c e p t e d M a n u s c r i p t when just the alternative promoter was available. This may be partly explained by the fact that the switches are smaller in magnitude. On the other hand, the distributions for the hPRL WT cells and those with the Pit-1 knockdown were very similar as can be seen from the cumulative distribution functions in Figure 6 m and n. This suggests that the proximal pituitary promoter is responsible for the more graded response seen in the WT, but that this response is not due to interaction of Pit-1 with the promoter. We hypothesise that the graded response is associated with extra transcriptional availability modulated by the proximal pituitary promoter, and that Pit-1 binding facilitates higher transcription rates during the pulses.

Discussion
Heterogeneity in gene expression within a cell population has been observed in numerous single-cell imaging studies where it has been shown in a diverse range of organisms that individual genes can be transcribed in short bursts of variable duration and frequency [19, 22-26, 37, 40]. Evidence supports several molecular mechanisms that together control and modulate transcriptional bursting, including nucleosome occupancy [41], chromatin modifications [19,23,42], transcription factor availability [23,26,42,43] and promoter structure [23] (reviewed in [44]. We have used pituitary cells as We previously generated a BAC reporter construct (hPRL WT) spanning 163 Kb of the human PRL genomic locus and engineered it to insert the luciferase reporter gene at the start of the hPRL exon 1b, thus preventing expression of the hPRL coding sequence [18]. We have studied the behaviour of this hPRL WT BAC reporter construct using in vivo and in vitro models [18,20,34,35]. Whole body in vivo imaging of hPRL-luc BAC-transgenic rats using this construct revealed striking evidence of alternative promoter activation after immune challenge, demonstrating that this construct can display transcriptional control by both the exon 1b promoter and the alternative upstream exon 1a promoter [20,35]. In order to determine the relative contributions of the two alternative promoter regions, in the present study we engineered the hPRL-luc BAC (referred to as the WT construct) by deleting the entire 5 kbp region of proximal promoter and inserting a 256 bp DNA fragment which contained alternative splice acceptor site and mutated non-functional Pit-1 binding sites (hPRL PitProKO BAC).
This strategy resulted in a functional upstream exon 1a promoter driving the luciferase reporter gene regions, but no functional Pit-1 responsive elements. The initial assumption in this work was that the two promoters in the hPRL gene locus would display clear differential cell-type-specific activation, namely that the exon 1b promoter with its multiple Pit-1 binding sites would generate 'pituitaryspecific' activation, and that the upstream exon 1a promoter, which lacks Pit-1 binding sites, would be active only in non-pituitary cells [10]. The recombinant BAC approach also allowed us to evaluate possible differences in transcriptional timing in relation to promoter structure and transcription factor binding.
Alternative promoters are a common occurrence in the mammalian genome and can allow diversity and flexibility in gene expression [45,46]. The hPRL proximal and distal promoter regions differ greatly in their architecture with distinct enhancer and regulatory region configuration (Supplemental Figure 1 [38]). A major difference is the presence of multiple Pit-1 binding sites in the proximal promoter region. We have shown here that the alternative 'non-pituitary' promoter is transcriptionally active in pituitary cells, albeit at a greatly reduced level. This suggests that the two hPRL gene promoters display tissue preference rather than absolute tissue specificity in activation. This phenomenon of tissue preference rather than tissue exclusivity has been reported in a variety of human genes [45]. Single-cell analysis of reporter gene expression confirmed that both promoters displayed heterogeneous and bursting transcriptional activity, with the alternative promoter associated with a significantly reduced transcriptional 'on' time and a more binary response. The short 'on' timing appears to be a specific feature of alternative 'non-pituitary' promoter activity as similar timing characteristics were observed in primary bone marrow myeloid cells from hPRL-WT transgenic rats (Figure 1).
The activity of Pit-1 is regulated in response to signal-transduction pathways by its interaction with co-repressor and activator complexes containing nuclear receptor co-repressor 1 (N-CoR) and CREB binding protein (CBP) respectively [47]. In response to physiological stimuli Pit-1, through its interaction with activator complexes uses the histone acetyltransferase function of CBP [47]. Pit-1 has been reported to direct changes in the chromatin structure of the hPRL promoter [48], and changes in chromatin structure have been implicated in the timing of hPRL gene transcription [19]. In this work we examined the effect of modifications of promoter structure by inactivation of Pit-1 binding sites, as well as reduction of transcription factor availability in Pit-1 knockdown experiments. The evidence presented here shows that burst timing is affected by the number of transcription factor binding sites and transcription factor availability. Binding of transcription factors to regulatory regions can influence burst frequency [27] and burst size [42,49] [20,34].
Pit-1 mRNA is first detected in the rat anterior pituitary at embryonic day 15 and the protein is initially expressed at very low levels [50]. Our data suggest that the difference seen in transcriptional bursts during development could be due to the availability of the Pit-1 transcription factor.
An important finding of the present work was that manipulation of the promoter structure had clear effects on the timing and structure of transcriptional cycles in living, intact cells. We and others have previously found that the prolactin gene, like other genes studied, displays cycles of transcriptional activity that are likely to involve chromatin remodelling [19,23,37,51]. The question arises as to what elements of promoter structure may control the existence or timing of these cycles. Pit-1 is a well-studied transcription factor thought to be necessary for tissue-specific expression of the prolactin gene in the pituitary. The prolactin gene proximal promoter contains multiple Pit-1 binding sites that are thought to mediate responses to signalling stimuli [4]. Deletion of the Pit-1-regulated promoter elements in the human prolactin gene locus did not prevent transcriptional cycles from occurring, but did markedly reduce the duration of active periods of transcription, modelled here as 'on' periods using a stochastic switch model, and also resulted in a less graded, more binary, response profile. This suggests that Pit-1 binding in the 'pituitary-specific' promoter is important to stabilise transcriptional complexes for longer periods to allow higher rates of transcription to occur, and our data are the first to indicate that the action of this transcription factor has an important effect on timing.
An important challenge is to assess the physiological significance of pulsatile transcription and the role that its modulation may play on normal physiology. The present studies were performed in the pituitary GH3 cell line, which has significant limitations: it is an immortalised clonal cell line, and the cells lack functional dopamine receptors. Nonetheless, they have proved a valuable test-bed in which to explore how pituitary hormones may be regulated, and a system in which complex genetic manipulations may be trialled before conducting studies in living animals. In our previous work using transgenic hPRL-EGFP reporter rats, we have found identical transcriptional pulses in living intact normal pituitary cells [18,20,34,37]. In addition to studying the effects of tissue structure and development on dynamic transcriptional patterns, we showed how modification of pulse characteristics in individual cells changes the overall mRNA production in a larger population of cells (36). Important questions still to be addressed in vivo include studies on the effects of oestrogen and dopamine, and also how hormone production by clonal pituitary tumours might differ from that of lactotrophs in the intermingled cell populations of the normal pituitary. This work will require further use of such animal models, but in the meantime our cell line data indicate that transcriptional timing is an important aspect of overall physiological control of pituitary hormone production, and that transcription factors such as Pit-1 appear to have a key role in stabilising transcriptional pulses to ensure high-level hormone production.
A c c e p t e d M a n u s c r i p t In summary, Pit1 plays an important role in the timing of transcription cycles, rather than simply being necessary to permit tissue-specific gene expression. The proximal promoter displays a binary ('all-or-nothing') activation step, with the presence of Pit-1 and Pit-1 binding sites associated with prolongation of the subsequent 'on-phase', and multi-step graded inactivation. In the absence of Pit-1 or Pit-1 binding, as occurs with the alternative upstream promoter in non-pituitary tissues, both the activation and inactivation steps of the transcriptional cycles have binary characteristics of a smaller amplitude. Pit-1 is localised in nuclear foci and dynamically partitioned, with a key role in interacting with the nuclear matrix [52]. In addition, Pit-1 has been shown to reorganise long range looping [53], nuclear co-repressors [54], nucleosome location and histone acetylation [48]. The current results suggest a dynamic rather than passive role for Pit-1 in transcriptional regulation. These results may be applicable to the mechanism of action of other master cell-lineage specific transcription factors and to the dynamic behaviour of genes which are differentially expressed through alternative promoters.
A c c e p t e d M a n u s c r i p t

Figure 6
Distribution of the switching score S for up and down switching for each of WT, PitProKO, WTPit-1 knockdown and WT scrambled siRNA. a,d,g,j: the distributions of scores S for up switches for each of the four datasets. b,e,h,k: as before (a,d,g,j) but for down switches. c,f,i,l: cumulative distribution functions for each of the four datasets. In each case there is a significant difference between the up and down switches, with up switches having larger values of S. m: a comparison of the cumulative distribution functions of the four datasets.