|Home | About | Journals | Submit | Contact Us | Français|
E.M. constructed all strains and performed and analyzed all experiments except for the FACS-based mating assay. A.S. constructed strains for and performed the FACS-based mating assay. H.E-S. and E.M. wrote custom MATLAB software and conducted data analyses for the FACS assays. E.M., H. E-S. and H.D.M. wrote the manuscript.
Users may view, print, copy, download and text and data- mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use: http://www.nature.com/authors/editorial_policies/license.html#terms
In the Saccharomyces cerevisiae pheromone-response pathway, the transcription factor Ste12 is inhibited by two MAP kinase-responsive regulators, Dig1 and Dig2. These two related proteins bind to distinct regions of Ste12 but are redundant in their inhibition of Ste12-dependent gene expression. Here we describe three unexpected functions for Dig1 that are non-redundant with those of Dig2. First, the removal of Dig1 results in a specific increase in intrinsic and extrinsic noise in the transcriptional outputs of the mating pathway. Second, in dig1Δ cells, Ste12 relocalizes from the nucleoplasmic distribution seen in wild-type cells into discrete subnuclear foci. Third, genome-wide iChIP studies revealed that Ste12-dependent genes display increased interchromosomal interactions in dig1Δ cells. These findings suggest that the regulation of gene expression through long-range gene interactions, a widely-observed phenomenon, comes at the cost of increased noise. Consequently, cells may have evolved mechanisms to suppress noise by controlling these interactions.
Cells respond to environmental fluctuations by transducing signals to networks of DNA-binding proteins. Numerous transcriptional regulators, including p531, E2Fs2 and Smads3,4, are subject to overlapping inhibitory mechanisms, yet the logic underlying these potential circuit redundancies remains poorly understood. A well-defined example of such regulatory architecture occurs in the S. cerevisiae mating pathway in which the transcription factor Ste12 is inhibited by two MAP kinase-responsive regulators, Dig1 and Dig2. These related proteins are redundant in their suppression of Ste12 activity since the removal from cells of both proteins is required to de-repress pathway activity5,6. Despite this redundancy Dig1 and Dig2 bind to distinct regions of Ste12; Dig1 to the activation domain and Dig2 to the DNA-binding domain7,8.
Ste12 lies at the terminus of a signal transduction pathway that is initiated by the binding of extracellular pheromones to a G-protein coupled receptor. This ligand-sensing event triggers the activation of a MAP kinase (MAPK) cascade, which initiates a cytoplasmic response and transmits the mating signal to the nucleus to activate the transcription factor Ste12 (Fig. 1a). Ste12 regulates the expression of a network of genes whose products are required for the process of mating. Unstimulated cells display a basal level of signalling that increases upon stimulation with pheromone. This system has been utilized recently as a model to measure variability, or noise, in a signal transduction cascade and to ascertain whether such noise is controlled9,10. Interestingly, it was found that removal of either of the MAPKs, Fus3 or Kss1, did not affect total output variability, suggesting that this natural system may have evolved overlapping mechanisms that buffer against noise9. Since the regulation of gene expression noise has been suggested to be important for appropriate input-output responses11-13, we reasoned that the investigation of noise in the output of the mating pathway might reveal mechanisms that underlie the redundant regulatory architecture controlling Ste12 activity.
We constructed two Ste12-dependent reporter genes, pAGA1-YFP and pFUS1-YFP. Their output distributions in wild-type and dig1Δ cells overlapped less than 5% with the background autofluorescence of yeast (Supplementary Information, Fig. S1). The mean output of dig1Δ strains increased 1.4-fold over wild-type, while mean fluorescence levels in dig2Δ did not change measurably (Fig. 1b), confirming that Dig1 and Dig2 appear redundant in their inhibition of average Ste12-dependent transcription5,6 when assayed in this manner. As expected, deleting DIG1 and DIG2 resulted in a 19-fold and 9-fold increase in mean expression for pAGA1-YFP and pFUS1-YFP, respectively (Fig. 1b). The mean output of a Ste12-independent reporter, pPMP1-GFP, was unaffected by deletion of DIG1 or DIG2 (Fig. 1b).
In contrast, examination of the single-cell output distributions of the Ste12-dependent reporters revealed a non-redundant role for Dig1 that is distinct from Dig2. Deletion of DIG1, but not DIG2, significantly increased the variability as measured quantitatively by the coefficient of variation or CV (Fig. 1c), and qualitatively by the spread of the pAGA1-YFP and pFUS1-YFP distributions (Fig. 1d). The CVs of the pFUS1-YFP dig1Δ and pAGA1-YFP dig1Δ output distributions were 29.6% (P = 0.0003) and 12.5% (P = 0.0014) higher, respectively, than those of wild-type and dig2Δ (Fig. 1c,d). Cell sorting experiments indicated that a cell population isolated from the middle of the dig1Δ output distribution could regenerate the entire distribution within 1-2 cell cycles (Fig. 2). Thus, while the steady-state fraction of cells experiencing the high-expression state at any given time point in the dig1Δ mutant is modest, the entire population of dig1Δ cells is likely to dynamically experience inappropriately high-expression states over time. The larger CV of dig1Δ output distributions was unexpected, and all the more significant, because the slight increase in mean output in dig1Δ cells might be predicted to generate a decrease, rather than an increase, in noise14. Furthermore, the increase in noise in dig1Δ cell populations was independent of forward scatter and side scatter, flow cytometric surrogate measures of cell size and shape (Fig. 1e, see Methods). As expected from the rise in mean expression, dig1Δdig2Δ double mutants displayed less variability than wild-type in mating pathway outputs (Fig. 1c). The effect of deleting DIG1 on noise is specific to outputs of the mating pathway, as the deletion of DIG1 or DIG2 did not affect the variability in three Ste12-independent reporters, pPMP1-GFP, pYEF3-GFP and pAGP1-GFP (Fig. 1c,d, Supplementary Information, Fig. S2). Furthermore, the changes in noise cannot merely be due to changes in the mean expression or growth rate since the analysis of several additional mutants illustrate that increased mean output and decreased growth rate do not result in increased noise (Supplementary Information, Fig. S3).
Gene expression noise can be decomposed into intrinsic and extrinsic components using a two-colour reporter gene system in which distinct fluorescent proteins are expressed from identical promoters in the same cell15. Intrinsic noise is defined as the uncorrelated cell-to-cell variation in levels of these two fluorescent proteins and is thought to reflect stochastic fluctuations in gene expression itself16-19. Extrinsic noise is defined as the correlated variation in the levels of the two proteins. Although extrinsic noise is thought to be impacted by cell-to-cell variability in the global cellular state, its origins and effectors are considerably less well-understood9,11,14.
Using a two-colour assay with strains containing GFP and mCherry driven by pAGA1 (Fig. 3a), we observed that both intrinsic and extrinsic noise increased in dig1Δ cell populations as compared to wild-type and dig2Δ cell populations. This result can be seen qualitatively by the reduced density of cells in the centre of the scatter plot of the data for the dig1Δ mutant relative to wild-type and dig2Δ (Fig. 3b), indicating an increased spread in fluorescence values. Quantitative calculations also reveal increases in the CV measurements (Fig. 3c, Supplementary Information, Fig. S4). The extrinsic noise (ηext) was 22.8% (P = 0.035) greater in magnitude in dig1Δ cells as compared to wild-type, while the intrinsic noise (ηint) was 14.9% (P = 0.009) higher (Fig. 3c). These patterns of increased intrinsic and extrinsic noise in dig1Δ populations were independent of cell size and shape and were specific to Ste12-dependent outputs (Fig. 3d-f, Supplementary Information, Fig. S4d,e).
The increased extrinsic noise in dig1Δ cell populations could result from the breakdown of a mechanism in which Dig1 limits fluctuations in the levels of the transcription factor Ste12 through an autoregulatory feedback loop at the Ste12 promoter20-22. However, this was not the case since replacing the Ste12-dependent Ste12 promoter had no effect on noise (Fig. 4). This posed the possibility that the mechanism by which Dig1 acts on Ste12-dependent genes to limit extrinsic noise is beyond correlations in upstream factors. Extrinsic noise is typically measured by quantifying the correlated variability in the expression from two identical promoters, in this case pAGA1. However, more generally, correlated or extrinsic noise in pAGA1 output would be expected to increase in dig1Δ cells if Dig1 limited the correlated expression of all Ste12 outputs in the cell. One way for this to occur would be if Ste12 target genes co-localized in space in the absence of Dig1. If this were the case, the spatial proximity of these genes could increase the dependence of the expression of one Ste12 target gene on the expression of another, perhaps due to increased local concentration of activators. For example, if Ste12 target genes co-localized in space, the induction of one gene could stimulate the induction of a neighbouring Ste12 target gene. Thus, it would be expected that the expression of such co-localized genes would be more correlated, in turn resulting in an increase in extrinsic noise. Given that Ste12 has many known interacting partners and exhibits self-cooperativity5,6,23-27, Dig1 may function to shield protein-protein interaction domains on Ste12 that would otherwise cause Ste12 to homo-dimerize or bind to other proteins. Therefore, the loss of Dig1 might allow DNA-bound Ste12 proteins to enable long-range interchromosomal interactions between Ste12 target genes.
Consistent with this possibility, Ste12-GFP molecules localized to subnuclear foci in dig1Δ cells (Fig. 5a, white arrowheads), while Ste12-GFP displayed granular nucleoplasmic staining in both wild-type and dig2Δ cells (Fig. 5a). Approximately 65% of dig1Δ cells showed one or more Ste12-GFP foci (Fig. 5b). These foci did not co-localize with the nucleolus (Supplementary Information, Fig. S5a) and focus formation could not explained by changes in total Ste12 protein levels since these levels were unaltered in dig1Δ and dig2Δ cells, as measured by quantitative immunoblotting (Supplementary Information, Fig. S5b). dig1Δdig2Δ double mutants also exhibited Ste12-GFP foci, but a slightly higher nucleoplasmic accumulation of Ste12-GFP protein precluded accurate assessment and quantification (Supplementary Information, Fig. S5c). Focus formation in dig1Δ cells was specific to Ste12 as the transcription factor Reb1-GFP displayed nucleoplasmic staining in wild-type, dig1Δ and dig2Δ cells (Fig. 5a).
In wild-type cells, stimulation with pheromone does not induce formation of Ste12-GFP foci (Fig. 5c), indicating that an increase in signalling and transcriptional output is not sufficient to induce their formation. While it has been suggested that mating signalling inactivates Dig15,6, we found that this protein remains physically associated with target genes (presumably via Ste12) in cells treated with pheromone (Fig. 5d). Thus, consistent with our finding that Ste12-GFP foci do not form in wild-type cells upon pheromone stimulation, not all activities of Dig1 are eliminated by signalling.
Using a genome-wide adaptation of the single-locus iChIP technique28, we examined interactions between the Ste12 target locus, pFUS1, and the rest of the genome in wild-type and dig1Δ cells (see Methods). The locus efficiently immunoprecipitated as seen by the large peak centred on the FUS1 promoter on the left arm of Chromosome III (Fig. 6a). No enrichment was observed at the pFUS1 locus in the absence of LacI (Supplementary Information, Fig. S6). The 5% of genes (269 genes) whose promoters displayed the largest differences in ChIP-chip signals between dig1Δ cells and wild-type were analyzed (Supplementary Information, Table S1). Remarkably, of the 203 gene regulators for which genome-wide localization data are available21, only targets of Ste12 and Tec1 displayed a statistically significant increase in interactions with the FUS1 locus in dig1Δ cells (Fig. 6b, similar results obtained for 1%, 3% and 10% cutoffs). Moreover, these physical interactions were dependent on the presence of Ste12 (Fig. 6c, Supplementary Information, Table S1). Tec1 and Ste12 are known to interact and are found at promoters of genes involved in both mating and filamentous growth22,27. Well-studied genes implicated in these processes were prominently featured among those that displayed increased physical interactions with the FUS1 locus in dig1Δ cells (Fig. 6d). We constructed promoter-YFP fusions for 11 of these Ste12-target genes and found that the mean expression increased for seven upon deletion of DIG1 (Supplementary Information, Fig. S7a). Rigorous analysis of the changes in noise for these genes is complicated by the fact that the means increase significantly and the relationships between the means and CVs are unknown. However, we note that the removal of Dig1 induces a broadening of the output distributions that is highly reminiscent of trends seen with the pFUS1-YFP and pAGA1-YFP reporter stains (Supplementary Information, Fig. S7b).
Under basal conditions, the mating pathway must appropriately balance the level of signalling to avoid cell cycle arrest and mating projection formation induced by pathway activation with a requirement for maintaining basal signalling to express key pathway components29. This balance might be expected to be disrupted in dig1Δ cells, with repercussions for growth under basal conditions and mating in the presence of a pheromone signal. Therefore, cell-to-cell variability in outputs of the mating pathway could influence fitness. We found that dig1Δ cells grow more poorly than wild-type or dig2Δ cells and this defect is rescued by the deletion of STE12 (Fig. 7a,b). Additionally, dig1Δ cells display a kinetic defect in cell-cell fusion compared to wild-type and dig2Δ, as measured quantitatively using a fluorescent-based assay in which the accumulation of double-positive fluorescent cells was scored (Fig. 7c-e, Supplementary Information, Fig. S8, see Methods). This defect is unlikely to be due to the slight increase in mean pathway output in dig1Δ cells since previous studies found that even large increases in basal signalling does not reduce mating efficiency30. The defect in fusion between mating partners is mirrored by two quantitative changes in the induction of pheromone-inducible genes in dig1Δ cells (Fig. 8a). First, dig1Δ cells display a larger proportion of cells that do not induce pAGA1-YFP or pFUS1YFP reporter genes in response to pheromone treatment (Fig. 8b). Second, the population of dig1Δ cells that does respond to pheromone displays a reduced dynamic range in the induction of pheromone inducible gene expression (Fig. 8c).
Recent work has shown that DIG1 and DIG2 were derived from a single parental gene that existed prior to the whole-genome duplication (WGD) that occurred in the ancestor of S. cerevisiae 100-200 million years ago31. Their continued presence in the genome suggests that their maintenance has an adaptive role. Indeed, previous work indicates that Dig1 and Dig2 inhibit Ste12 by interacting with distinct domains of the transcription factor, implying biochemical specialization 7,8. However, their genetic redundancy for inhibiting Ste12 was puzzling. Studies presented here revealed three functions of Dig1 that are not redundant with those of Dig2: 1) control of gene expression noise, 2) regulation of the intranuclear distribution of Ste12, and 3) the control of long-range interactions between Ste12-target genes. We discuss below how these three functions may be related and the broader implications of these findings.
Dig1 is a well-studied regulatory protein that functions specifically in the pheromone response pathway and has only one reported biochemical function: to bind to a domain of Ste12 involved in protein-protein interactions5-8,32,33. The loss of Dig1 is, therefore, expected solely to unshield protein-protein interaction domains on the Ste12 transcription factor. Although indirect mechanisms are always difficult to rule out, we propose that this unshielding induces aggregation of Ste12 molecules and target genes, which results in increased cell-to-cell variability in the basal output of the pheromone response pathway. Dig2, which binds the distinct DNA-binding domain of Ste127, 8, does not share these functions. The aggregation of Ste12 molecules into one or two foci may create a domain within the nucleus where the transcription of Ste12-target genes can be activated. Our model suggests that the transcription of Ste12-target genes within the focus is more coordinated such that if one gene in the focus is transcribed, the others are, in turn, more likely to be expressed. Thus, such correlated expression within a single cell would be expected to yield increased correlated cell-to-cell variability in the transcriptional output of the pathway.
Transcriptional regulation that involves looping of DNA between distant sites via protein-protein interactions has been observed the lac operon34-38 and λ phage39,40. In the context of the results described here, it is notable that computational models of the lac system suggest that gene regulation by DNA looping can affect fluctuations in transcription41. These models predict that for transcriptional activators, DNA looping should increase noise in transcriptional outputs. Our model for the function of Dig1 is consistent with these theoretical predictions.
Recently, inter- and intrachromosomal interactions have been detected in other systems42-45. In erythroid cells, for example, Klf1-regulated genes, including Hba and Hbb globin genes, display long-range inter- and intrachromosomal interactions42. Although such interactions tend to correlate with transcriptional regulation and sites of active transcription, their precise functions remain a matter of considerable debate. Our observations suggest that while these long-range interactions could be important for gene expression, they may come at the cost of increased variability. This notion is in concordance with an emerging view that, in some cases, such gene interactions can be deleterious and even mutagenic46. It will be interesting to explore whether mechanisms of noise regulation are pervasive among regulatory circuits that involve long-range DNA interactions and the extent to which gene localization is balanced with a need for limiting noise.
While establishing the generality of the effect of aggregate formation on output variability will require further investigation, we note that subcellular protein and DNA aggregation is not uncommon in biology. DNA replication and gene activation can occur in “factories” located at the nuclear periphery47-51. Sites of DNA damage along with proteins that respond to DNA damage form nuclear foci in yeast52,53. Telomeres are also known to cluster in the nucleus54. Cytoplasmic P-bodies are foci of proteins involved in mRNA degradation and translational inhibition55-57. Given our data, these foci may serve, in some cases, to promote simultaneity in cellular transactions. The development of assays that can distinguish between correlated and uncorrelated noise in these systems would allow the testing of such concepts.
We are grateful to J.S. Weissman, E.K. O’Shea, J.E Haber, W.A. Lim, A.D. Johnson and D.J. Sherratt for plasmids and protocols. We thank C.D. Chun and P.D. Hartley for help in conducting and analyzing the ChIP-chip experiments and W.F. Marshall and K. Wemmer for assistance with microscopy. We are especially grateful to A.D. Johnson, S. Komili, W.F. Marshall and S. Shankar for helpful comments on the manuscript. This work was supported by a Genentech Fellowship and an NSF Predoctoral Fellowship to E.M., an NIH Ruth L. Kirschstein National Research Service Award to A.S., as well as funding from the UCSF Program for Breakthrough Biomedical Research and an NIH grant (GSE17583) to H.D.M. and H.E-S.
All yeast strains used are derived from BY4743, of the s288c background, and are described in Supplementary Information, Table S2. Yeast knockouts were generated by conventional lithium acetate and polyethylene glycerol procedures. YFP, eGFP (from pFA6a-EGFP-HIS3MX) and mCherry (from pFA6a-mCherry-HIS3MX or pFA6a-GFPtomCherry-URA3MX from J.S. Weissman) reporters for the mating pathway were constructed using methods as described58, while pPMP1-fluorophore fusions were constructed using plasmids pFA6a-EGFP-HIS3MX6 and pFA6a-GFPtomCherry-URA3MX from J.S. Weissman.
Single fluorescent strains used: YM1968, YM2091, YM2100, YM2105, YM2109, YM2112, YM3550, YM3593, YM3594, YM3760, YM3762, YM3763, YM3764, YM3766, YM3767, YM3769, YM3770, YM3771, YM3772, YM3773, YM3776-YM3782 YM3804-YM3814. Dual fluorescent strains used: YM2636, YM2871, YM2876 and YM3132. Cells were grown in 1 mL cultures for 36 hr in 96-well deep pocket plates (Costar). OD600 measurements were taken and cultures were diluted to an OD600 = 0.08 and grown for 10 hr. A Becton Dickinson LSR-II flow cytometer was used, along with an autosampler device (HTS) controlled by custom software, to collect data over a sampling time of 7 sec11. YFP and GFP were excited at 488 nm and fluorescence was collected through a 505-nm long-pass filter and HQ530/30 and HQ515/20 band-pass filters (Chroma Technology), respectively. mCherry was excited at 532 nm and fluorescence collected through 600-nm long-pass filter and 610/20 band-pass filters (Chroma Technology).
All data analysis was done using custom MATLAB software. Raw cytometry data were filtered to eliminate errors due to uneven sampling time and negative fluorescence readings. Bulk calculations were done on these processed data. To control for cell aggregates, as well as cell size and shape, forward and side scatter (FSC and SSC) data were expressed on orthogonal axes and subpopulations of cells were selected using circular gates of increasing radii centred on the median FSC and SSC values11. Nineteen circular bins were created with radii of 6000, 9000, 104, 2*104, 3*104,…,17*104 arbitrary units were used. Results are shown for data in bin 5, with a radius of 3*104. Data were used if at least 5000 cells were in this bin, but on average between 20,000-40,000 cells had FSC and SSC values within this gate. The coefficient of variation (CV) was used as a measure of total noise, while intrinsic/uncorrelated and extrinsic/correlated noise were calculated as described15 using GFP and mCherry dual-colour strains (Supplementary Information, Table S2). T-tests were used to calculate level of significance for increases in noise in the dig1Δ mutant strains.
YM2105 (pAGA1-YFP dig1Δ) were grown to mid-log phase. The fluorescence distribution was determined. A narrow gate cantered on the middle of this distribution was created and cells with expression levels within this gate were sorted using a Becton Dickinson FACSAria cell sorter. Cells were spun down, resuspended in YPAD and grown at 30 C. Aliquots were removed and the fluorescence distributions determined for 30,000 cells using a Becton Dickinson LSR-II flow cytometer. Data was analyzed as described above.
YM2910, YM3102, YM3103, YM3104, YM3722, YM3723, YM3724, YM3774 and YM3775 were grown overnight to saturation in YPAD. Cultures were diluted back to an OD600 of 0.1 in YPAD and grown for 4 hr. Microscopy was performed using a DeltaVision deconvolution microscope, which was outfitted with Olympus Plan Apo 60- and 100-X objectives. Z-stacks were taken with 0.3 μm steps. DeltaVision deconvolution software was used to deconvolve and analyze these images. For Ste12-GFP, a 1 s exposure was used and for Nup188-mCh, a 0.5 s exposure was used. For Reb1-GFP, a 1.0 s exposure was used. For the Ste12-GFP and Nop7-mCherry colocalization experiments, a 0.5 s exposure was used for the FITC channel and a 0.2 s exposure was used for the rhodamine channel.
YM1953, YM2101, YM2315, and YM2102 were grown to log phase in YPAD and 3 OD600 were collected by centrifugation and snap-freezing. Pellets were re-suspended in 100 μl 2X protein loading buffer and 1:100 Sigma phosphatase inhibitor cocktails 1 and 2 and 1:260 Sigma protease inhibitor cocktail. Samples were boiled for 2 min and 50 μl Zirconia/silica beads (Biospec Products) were added. Samples were then vortexed on a platform vortex for 2 min. Samples were again boiled for 2 min and centrifuged at 14,000 x g for 10 min to remove cell debris. The supernatants were pulled off, boiled for 3 min and resolved on a 10% SDS-PAGE gel. Proteins were then transferred to nitrocellulose and immunoblotting was performed as described in the Li-COR Odyssey manual using αSte12 (1:1000, a gift from Ira Herskowitz), αTubulin (1:3000, AbCam), αRabbit-IR800 (1:1000) and αRat-IR680 (1:1000).
YM1953, YM2101, YM2315, YM2643, YM2248, YM3776, YM3777 and YM3778 were grown to log phase overnight in YPAD. These cultures were then diluted back to an OD600 = 0.2 (YM1953, YM2101, YM2315 and YM2643) or OD600 = 0.05 (YM2248, YM3776, YM3777 and YM3778) at t=0 and OD600 measurements were taken every hour. To avoid cultures reaching saturation and entering stationary phase, cultures were diluted periodically. This dilution was accounted for in the subsequent OD600calculations. OD600 measurements at later time points were normalized to the OD600 at time = 0 min. Best-fit lines were calculated using DeltaGraph 5 graphing software.
A MATa strain (YM2901) containing at the TRP1 locus a construct consisting of the N-terminus (AA 1-158) of eGFP fused to a leucine zipper dimerization domain59 was constructed. MATα strains (YM2903, YM3085, YM3086 and YM3087) containing at the LEU2 locus a construct consisting of the C-terminus (AA 159-240) of eGFP fused to a leucine zipper dimerization domain59 as well as an mCherry marker driven by pTEF2 integrated at the LYS1 locus were also constructed. See supplementary methods for experimental details.
YM1968, YM2091, YM2100 and YM2105 were grown into log phase over night in YPAD. Cultures were diluted back to an OD600 = 0.4 and 50 nM α-Factor was added. 1 mL aliquots were removed at t = 0, 15, 30, 60, 90, 120, 150, 180 and 240 min, washed with water, resuspended in 1 mL TE pH = 8 and fluorescence distributions were measured by flow cytometry. Data were analyzed as described above.
An 11 kb construct consisting of 240 tandem arrays of Lac operators62 and an associated HIS3MX marker was inserted 331bp upstream of the FUS1 ATG in strains containing a mCherry-LacI plasmid (BHM1336 adapted from pJH212, strains: YM3587, YM3588 and YM3687). Cultures were prepared for ChIP-chip by overnight growth to saturation in –Ura medium. Cultures were then diluted to an OD600 of 0.01 and grown for 4 hr in –Ura medium. These cells were again diluted to an OD600 of 0.01 in YPAD and collected 4 hr later. Chromatin immunoprecipitation was performed as described60,61. However, the protein crosslinker ethylene glycolbis (succinimidylsuccinate) (EGS) was added to a final concentration of 1.5 mM for 30 min before the addition of formaldehyde. Additionally, DNA was lightly sonicated in a Diagenode Bioruptor for 2×5 min on the low setting with 1 s on/ 0.5 s off pulses. To immunoprecipitate mCherry-LacI, a polyclonal anti-DsRed antibody from Clontech (catalogue number 632496) was used at a 1:100 dilution. Following ChIP, strand displacement amplification and labelling were performed as described to generate DNA probes with incorporated aminoallyl-dUTP63. Probes representing mCherry-LacI immunoprecipitates and whole cell extracts were differentially labelled with Cy fluorescent dyes and hybridized on Agilent yeast whole-genome tiling microarrays (G4491A). Hybridization and array washing were performed as described by Agilent Technologies (Version 9.2). In addition, after the acetonitrile wash, slides were rinsed in Agilent Stabilization and Drying Solution (5185-5979). Microarrays were scanned at 5 μm resolution on a GenePix 4000B scanner (Molecular Devices) using GenePixPro 6.0 software. Microarray analysis was done using in-house software as described64. See supplementary methods for details of data analysis.
Competing Financial Interests
The authors declare no competing financial interests.
Methods and associated references are available in the online version of the paper at http://www.nature.com/naturecellbiology