Minimal requirements for the epigenetic inheritance of engineered silent chromatin domains

Significance

Heterochromatin is characterized by histone modifications that close regions of the genome to DNA transactions like transcription. Models for how histone modifications silence gene expression have been proposed, but many remain speculative. Furthermore, how heterochromatic domains are epigenetically inherited—a process central to organismal development and evolution—is not fully understood. Here, we transplant a heterochromatin-associated histone modification across vast evolutionary time by incorporating histone H3 lysine 9 methylation (H3K9me) into chromatin of Saccharomyces cerevisiae cells, which naturally lack H3K9me. Using H3K9me as a point of departure, we engineer reduced-complexity silent chromatin domains, the assembly and epigenetic inheritance of which require only two rationally designed chimeric proteins constituting an interdependent positive feedback loop that can preserve memory of silent information.

Abstract

Mechanisms enabling genetically identical cells to differentially regulate gene expression are complex and central to organismal development and evolution. While gene silencing pathways involving DNA sequence–specific recruitment of histone-modifying enzymes are prevalent in nature, examples of sequence-independent heritable gene silencing are scarce. Studies of the fission yeast Schizosaccharomyces pombe indicate that sequence-independent propagation of heterochromatin can occur but requires numerous multisubunit protein complexes and their diverse activities. Such complexity has so far precluded a coherent articulation of the minimal requirements for heritable gene silencing by conventional in vitro reconstitution approaches. Here, we take an unconventional approach to defining these requirements by engineering sequence-independent silent chromatin inheritance in budding yeast Saccharomyces cerevisiae cells. The mechanism conferring memory upon these cells is remarkably simple and requires only two proteins, one that recognizes histone H3 lysine 9 methylation (H3K9me) and catalyzes the deacetylation of histone H4 lysine 16 (H4K16), and another that recognizes deacetylated H4K16 and catalyzes H3K9me. Together, these bilingual “read–write” proteins form an interdependent positive feedback loop that is sufficient for the transmission of DNA sequence–independent silent information over multiple generations.

Heterochromatin preserves genome stability and downregulates gene expression. Histone H3 lysine 9 methylation (H3K9me) is a hallmark of heterochromatin in mammals, plants, and fungi including the fission yeast Schizosaccharomyces pombe. Heterochromatin-mediated gene silencing in S. pombe cells is heritable and couples RNA interference (RNAi) with the recognition and local propagation of H3K9me at centromeres and silent mating type cassettes (1–3). Locus specificity is, in part, conferred by short interfering RNA (siRNA) sequence complementarity to nascent transcripts generated at target loci (3–5). In addition to RNAi, S. pombe cells also employ DNA sequence–dependent mechanisms for gene silencing, whereby transcription factors and the origin recognition complex (ORC) bind specific DNA sequences and recruit heterochromatin-associated factors to chromatin via protein–protein interactions (6–8).

The budding yeast Saccharomyces cerevisiae, which naturally lacks H3K9me and RNAi, assembles heterochromatin at silent mating type loci via composite DNA sequences known as silencers (9). Silencers feature binding sites for essential transcription factors and ORC, which recruit proteins of the silent information regulator (SIR) complex to regions flanking genes specifying mating type (10, 11). The SIR complex is composed of Sir2, the founding member of the sirtuin family of nicotinamide adenine dinucleotide (NAD⁺)–dependent deacetylases, Sir3, and Sir4. Sir2 deacetylates histone H4 lysine 16 (H4K16), and Sir3 recognizes deacetylated H4K16-containing nucleosomes (12–14). As Sir3 dimers associate with Sir2 via Sir4 dimers (15–17), the SIR complex binds cooperatively to pairs of deacetylated nucleosomes and propagates hypoacetylated chromatin states in cis by physically spreading from silencers over mating type-specifying genes (18–20).

Studies of heterochromatin in S. pombe cells revealed that silent chromatin states can propagate over multiple generations independently of underlying DNA sequence (21, 22). Establishment and maintenance of these states require numerous factors, highly diverse enzymatic activities, and notably, the engineered absence of a crucial silencing antagonist (21–24). In contrast, SIR complex–mediated gene silencing in S. cerevisiae cells requires the presence of silencer DNA at all times (25, 26).

We reasoned that incorporating H3K9me into S. cerevisiae chromatin would enable us to test principles of heterochromatin-mediated gene silencing in a controllable environment ~300 million years removed from the complexities of natural H3K9me-dependent processes. This bottom-up approach enabled us to build reduced-complexity domains of silent chromatin that—in addition to recapitulating essential properties of heterochromatic domains found in nature—can be epigenetically inherited. Heritable silencing in this system requires only two engineered chimeric proteins, one that recognizes H3K9me and deacetylates histones and a second that recognizes deacetylated H4K16 and methylates H3K9, thus forming histone modification-based interdependent positive feedback loops.

Results

Introducing H3K9me to S. cerevisiae Cells.

We sought to deposit H3K9me at a specific locus in S. cerevisiae cells by adopting an experimental scheme that has been used to study inducible ectopic heterochromatin domains in S. pombe cells (21–23). Specifically, we replaced the normally silent mating type locus HMR—including silencers E and I—with ten tandem tet operators (tetO_10X) positioned upstream of an ADE2 reporter gene in ade2-1Δ cells (Fig. 1A). Cells in which ADE2 is silent accumulate pigment and form red colonies on medium containing limiting adenine, whereas cells in which ADE2 is active do not accumulate pigment and form white colonies. We then designed a chimeric protein consisting of the bacterial tetracycline repressor TetR fused to the catalytic SET (Suppressor of variegation, Enhancer of Zeste, Trithorax) domain of human Suppressor of Variegation 3-9 Homolog 2 (SUV39H2), an H3K9 methyltransferase (Fig. 1A). We refer to this protein as TetR-SET, which contains a nuclear localization signal (NLS) and a V5 epitope tag between TetR and the SET domain (SI Appendix, Fig. S1A). Production of TetR-SET in reporter-containing cells resulted in local di- and tri-methylation of H3K9 at tetO_10X (Fig. 1B and SI Appendix, Fig. S1B). We took advantage of the reversible binding properties of TetR by treating cells with anhydrotetracycline (aTc) and measured the level of H3K9me2 under conditions where TetR-SET is no longer bound adjacent to ADE2 (Fig. 1A). Our results showed that the addition of aTc induced rapid TetR-SET untethering within 30 min (SI Appendix, Fig. S1C) and rendered H3K9me2 undetectable after 6 h (Fig. 1C), likely via the dilution of modified parental histones with unmodified newly synthesized histones during several generations of growth in the presence of aTc.

Fig. 1.

Establishment of H3K9me-dependent gene repression in S. cerevisiae cells. (A) Cartoon schematic of how HMR was replaced by a tetO_10X-ADE2 reporter via homologous recombination and how DNA sequence–specific recruitment of TetR-SET—consisting of bacterial TetR fused to the catalytic SET domain of human SUV39H2—results in continuous local H3K9me (red spheres). TetR-SET is depicted as a dimer bound to a single tet operator. H3K9me ceases in the presence of anhydrotetracycline (aTc), which untethers TetR-SET from the tet operator array. (B) ChIP-seq profile of H3K9me2 at the tetO_10X-ADE2 reporter in cells producing TetR-SET. Data ranges (in counts per million) are indicated in brackets. The region visualized encompasses ~50 kb of the right arm of chromosome III. The H3K9me2 profile shown in red encompasses 1,150 bp upstream and 1,825 bp downstream of tetO_10X. The input DNA profile is shown in gray. The position of tetO_10X is marked by a white square, and DNA features including select genes are indicated by arrows. (C) ChIP-qPCR analysis of H3K9me2 levels at the tetO_10X-ADE2 reporter in cells producing TetR or TetR-SET. Primers (half-headed arrows) bind to unique DNA sequences (black rectangle) adjacent to tetO_10X (white square). TetR-SET-producing cells were cultured in the presence or absence of aTc for 6 h. Values represent the averages and SD of three biological replicates. The level of H3K9me2 at the reporter locus in aTc-treated cells approached the limit of detection. (D) Cartoon depiction of Sir2^Sp-HP1—consisting of full-length S. pombe Sir2 fused to the chromodomain (CD), hinge region, and chromo shadow domain (CSD) of S. pombe Chp2—bound as a dimer to a dinucleosome containing two methylated H3K9 residues (red spheres). Arrows are drawn to indicate Sir2^Sp-mediated deacetylation of H4K16 and other lysines. (E) RT-qPCR analysis of ADE2 mRNA levels in cells producing TetR-SET and either HP1, Sir2^Sp-HP1, or a Sir2^Sp-HP1 variant lacking histone deacetylase activity (Sir2^Sp N247A-HP1). Values represent ADE2 levels (normalized to ACT1) in the indicated cells relative to ADE2 levels in reporter gene-containing cells producing TetR alone. Averages and SD of three biological replicates are shown. Significance was assessed by a two-tailed Student’s t-test (P ≤ 0.01). (F) Phenotypes of tetO_10X-ADE2 reporter-containing cells producing TetR-SET and either HP1, Sir2^Sp-HP1, Sir2^Sp N247A-HP1, a Sir2^Sp-HP1 variant containing a mutant chromodomain incapable of methyllysine recognition (Sir2^Sp-HP1 W199A), or a monomeric Sir2^Sp-HP1 variant lacking its CSD (Sir2^Sp -HP1ΔCSD). HP1, Sir2^Sp-HP1, and Sir2^Sp-HP1 variants were produced from pRS315, a low-copy plasmid encoding LEU2. Cells were spotted on leucine-lacking medium containing limiting adenine (10 μg/mL; Ade¹⁰). White dashed lines indicate sites of cropping between spots of cells grown and photographed on the same plate.

H3K9me recruits heterochromatin protein 1 (HP1) family proteins to silent chromatin. HP1 proteins feature an H3K9me-recognizing CD linked to a CSD (27). The CSD mediates HP1 protein homodimerization, which is required for its efficient localization to H3K9me domains, and the direct or indirect recruitment of myriad effector proteins, including histone deacetylases (HDACs) (28–31). As histone deacetylation is a conserved feature of heterochromatin-mediated gene silencing (12, 28, 29, 32, 33) and corepressor-mediated gene repression (34, 35), we hypothesized that producing a single subunit HDAC fused to an HP1 protein would suffice to downregulate ADE2 expression in TetR-SET-containing S. cerevisiae cells.

To test this hypothesis, we designed a chimeric protein consisting of S. pombe Sir2 (Sir2^Sp) fused to an S. pombe HP1 protein (residues 170–380 of Chp2, encompassing the natural Chp2 chromodomain, hinge region, and CSD) (Fig. 1D). We refer to this protein as Sir2^Sp-HP1, which contains an NLS and a Myc_3X epitope tag between Sir2^Sp and HP1. Like its S. cerevisiae homolog Sir2^Sc, Sir2^Sp preferentially deacetylates H4K16 (36). As Sir3 recognizes unmodified H4K16-containing nucleosomes, all experiments described in this work were, unless otherwise noted, performed in sir3Δ cells to rule out contributions of the native S. cerevisiae SIR complex to reporter gene silencing.

Whereas production of HP1 in TetR-SET-containing cells had no effect on ADE2 expression, production of Sir2^Sp-HP1 resulted in decreased ADE2 expression and increased colony pigmentation (Fig. 1 E and F). Downregulation of ADE2 expression required the catalytic activity of Sir2^Sp, a functional HP1 chromodomain, and HP1 dimerization (Fig. 1 E and F). A Sir2^Sp -HP1 variant featuring the dimerization domain of bacteriophage λ CI in place of the Chp2 CSD behaved similarly to Sir2^Sp-HP1 (SI Appendix, Fig. S1D). Reduced colony pigmentation was not attributable to differences in fusion protein levels (SI Appendix, Fig. S1E). We conclude that Sir2^Sp HDAC activity is sufficient to downregulate ADE2 expression, likely via the direct binding of Sir2^Sp-HP1 to H3K9me-modified nucleosomes.

Two observations suggest that Sir2^Sp acts independently of the native S. cerevisiae SIR complex: i) The N-terminal domain of Sir2^Sc – including its Sir4 interaction determinants—and the corresponding region of Sir2^Sp lack sequence conservation and are structurally dissimilar (SI Appendix, Fig. S2A), and ii) unlike Sir2^Sc-HP1, production of Sir2^Sp-HP1 failed to rescue silencing in sir2Δ cells (SI Appendix, Fig. S2B). We conclude—especially given that experiments were performed in sir3Δ cells—that Sir2^Sp-HP1 functions independently of the native S. cerevisiae SIR complex.

We asked whether we could distinguish heterochromatin-mediated gene silencing, which is regional in nature, from gene repression, which can occur via the deacetylation of promoter-proximal nucleosomes (34, 35). To do so, we modified our reporter by inserting another gene, URA3 of Kluyveromyces lactis (URA3^Kl), positioned upstream of tetO_10X, with the URA3^Kl promoter situated distal to the operator array (Fig. 2A). As all cells carry a non-functional ura3-1 allele, those in which URA3^Kl is silent are resistant to 5-fluoroorotic acid (FOA) and those in which URA3^Kl is active are not. If TetR-SET and Sir2^Sp-HP1 repress ADE2 expression primarily by modifying nucleosomes adjacent to tetO_10X, then cells producing these two proteins alone are predicted to be sensitive to FOA, which was indeed the case (Fig. 2A). We reasoned that URA3^Kl silencing would require the propagation of H3K9me in cis over the URA3^Kl gene body and promoter.

Fig. 2.

An H3K9me read–write protein allows for distinction between local gene repression and regional gene silencing in S. cerevisiae cells. (A) Schematic of the modified reporter and phenotypes of reporter-containing cells producing either TetR-SET and Sir2^Sp-HP1 alone (row one), or TetR-SET and Sir2^Sp -HP1 in combination with Suv39^χ, a Suv39^χ variant containing a mutant chromodomain incapable of methyllysine recognition (Suv39^χ W31G), or a Suv39^χ variant lacking its chromodomain (Suv39^cΔCD) (rows two through four). Suv39^χ and Suv39^χ variants were encoded on pRS315. Cells were spotted on leucine-lacking medium containing the indicated compounds. White dashed lines indicate sites of cropping between spots of cells grown and photographed on the same plate. (B) Cartoon depiction of how Suv39^χ—consisting of the CD and hinge region of S. pombe Clr4 fused to the catalytic SET domain of human SUV39H2—enables H3K9me spreading. Two Suv39^χ monomers that “read” a nucleosome containing two methylated H3K9 residues (red spheres) and “write” H3K9me on adjacent nucleosomes are illustrated. (C) ChIP-qPCR analysis of H3K9me2 levels at the URA^Kl-tetO_10X-ADE2 reporter and sites located ~3 kb away from tetO_10X. Primer binding sites are indicated by half-headed arrows. Cells produce TetR-SET, Sir2^Sp -HP1, and either Suv39^χ (pink) or Suv39^χ W31G (white). Suv39^χ-producing cells were cultured in the presence (gray) or absence (pink) of aTc for 6 h. Values represent the averages and SD of three biological replicates.

SUV39H family proteins are uniquely poised to support H3K9me propagation due to their ability to recognize (or read) and catalyze (or write) the same histone modification (37). This read–write capability is the basis for a positive feedback loop proposed to enable spreading of H3K9me at target loci while ensuring the faithful transfer of silent information from parental to newly synthesized histones following DNA replication (21, 22, 38). We therefore sought to test whether an enzyme capable of reading and writing H3K9me would permit H3K9me spreading and maintain reporter gene silencing without continuous DNA sequence–dependent recruitment of H3K9 methyltransferase activity.

To do so, we designed a chimeric H3K9me read–write protein consisting of the chromodomain and hinge region of the S. pombe H3K9 methyltransferase Clr4 (residues 1–191 of Clr4) fused to the SET domain of human SUV39H2 (Fig. 2B) (we note that the natural SET domain of Clr4 exhibited reduced function relative to that of SUV39H2 in S. cerevisiae cells). We refer to this protein as Suv39 chi(mera) (Suv39^χ), which contains an NLS and a FLAG_3X epitope tag at its N-terminus. Cells producing TetR-SET, Sir2^Sp–HP1, and Suv39^χ containing a functional chromodomain formed colonies exhibiting increased pigmentation and were resistant to FOA, suggesting that Suv39^χ had the capacity to read and write H3K9me (Fig. 2A). In contrast, cells producing Suv39^χ lacking its chromodomain or a Suv39^χ variant harboring a missense substitution abolishing methyllysine recognition (residue W31G of Clr4) exhibited no change in colony pigmentation and were sensitive to FOA (Fig. 2A). Reduced colony pigmentation and FOA-sensitivity were not attributable to differences in fusion protein levels (SI Appendix, Fig. S3A), and all requirements for Sir2^Sp -HP1 function established previously for cells lacking Suv39^χ also applied to cells containing Suv39^χ (SI Appendix, Fig. S3B). However, cells grown in the presence of aTc formed unpigmented colonies and were sensitive to FOA (Fig. 2A). Analysis of chromatin immunoprecipitates showed that H3K9me2 was present at URA3^Kl and ADE2 promoters and gene bodies (SI Appendix, Fig. S3C). Furthermore, the extent and local density of H3K9me2 was dependent on the read function of Suv39^χ (Fig. 2C). Yet Suv39^χ could not maintain local H3K9me2 levels in the presence of aTc (Fig. 2C).

Sequence-Independent Heritable Gene Silencing in S. cerevisiae Cells.

We reasoned that the inability of Suv39^χ to maintain reporter gene silencing could be explained by the following logic: One chromodomain of Suv39^χ necessarily competes with two chromodomains of the Sir2^Sp -HP1 dimer for association with H3K9me-modified nucleosomes. FOA-resistant cells must therefore produce Suv39^χ at sufficiently high levels to support local spreading. High levels of Suv39^χ, however, can result in off-target H3K9me at sites other than the reporter locus (SI Appendix, Fig. S3D). Off-target H3K9me, in turn, may redistribute Suv39^χ and Sir2^Sp-HP1 genome-wide, effectively diluting both proteins and thereby rendering locus-specific reporter gene silencing unsustainable without continuous sequence-specific recruitment of TetR-SET.

Based on this logic, we hypothesized that sustainable locus-specific silencing might be achieved by transplanting the H3K9 methyltransferase activity of Suv39^χ to a naturally dimer-forming protein capable of recognizing a histone modification other than H3K9me. Dimerization of the Sir3 protein has previously been shown to mediate its cooperative binding to deacetylated nucleosomes and increase its specific localization to heterochromatin (39). We therefore designed a chimeric protein composed of full-length Sir3—featuring an unmodified H4K16-recognizing bromo-adjacent homology (BAH) domain and a winged helix (wH) homodimerization domain (13, 39)—fused to the SET domain of human SUV39H2. We refer to this protein as Sir3-SET, which contains an NLS and FLAG_3X epitope tag between Sir3 and the SET domain. In the absence of aTc, cells producing Sir3-SET and Sir2^Sp -HP1 accumulated more pigment and were more resistant to FOA than Suv39^χ-containing cells (Fig. 3A). These phenotypes were attributable to decreased reporter gene expression and reduced RNA polymerase II (Pol II) occupancy at the singular ADE2 reporter gene promoter (SI Appendix, Fig. S4 A and B). In the presence of aTc, however, only cells producing Sir3-SET accumulated pigment and remained resistant to FOA (Fig. 3A). As Sir3 naturally interacts with Sir2-bound Sir4 (40), we deleted SIR2 and SIR4, thereby rendering cells sir2Δ sir3Δ sir4Δ. We refer to cells lacking all three components of the SIR complex as sir⁻. Critically, production of Sir3-SET and Sir2^Sp-HP1 in sir⁻ cells was sufficient to establish and maintain reporter gene silencing (Fig. 3A). Silencing required a functional Sir3 BAH domain and Sir3 dimerization (Fig. 3B). Reduced colony pigmentation and FOA-sensitivity were not attributable to loss of fusion protein production (SI Appendix, Fig. S4C). We conclude that Sir3-SET and Sir2^Sp-HP1 together form an interdependent positive feedback loop capable of supporting DNA sequence–independent heritable gene silencing (Fig. 3C).

Fig. 3.

An interdependent positive feedback loop maintains DNA sequence–independent gene silencing in S. cerevisiae cells. (A) Phenotypes of URA3^Kl-tetO_10X-ADE2 reporter-containing cells producing TetR-SET, Sir2^Sp -HP1, and either Suv39^χ (row one) or Sir3-SET. Sir3-SET was produced in sir3Δ cells (row two) or sir2Δ sir3Δ sir4Δ (sir⁻) cells (row three). Suv39^χ and Sir3-SET were encoded on pRS315. Cells were spotted on leucine-lacking medium containing the indicated compounds. White dashed lines indicate sites of cropping between spots of cells grown and photographed on the same plate. (B) Phenotypes of URA3^Kl-tetO_10X-ADE2 reporter-containing sir⁻ cells producing TetR-SET and Sir2^Sp-HP1 alone, or TetR-SET and Sir2^Sp -HP1 in combination with either Sir3, Sir3-SET, a Sir3-SET variant harboring a mutation that prevents Sir3 acetylation and disrupts bromo adjacent homology domain structure (Sir3 A2G-SET), or a monomeric Sir3-SET variant lacking its wH domain (Sir3ΔwH-SET). Sir3, Sir3-SET, and Sir3-SET variants were encoded on pRS315. Cells were spotted on leucine-lacking medium containing the indicated compounds. White dashed lines indicate sites of cropping between spots of cells grown and photographed on the same plate. (C) Cartoon depiction of the interdependent positive feedback loop formed by a pair of bilingual “read–write” proteins: Sir2^Sp-HP1 and Sir3-SET. Sir3-SET is shown bound as a dimer to a dinucleosome containing two unmodified H4K16 residues. The Sir3 wH domain mediates Sir3-SET dimerization, and the Sir3 bromo adjacent homology (BAH) domain mediates the recognition of unmodified H4K16 by Sir3-SET. Sir3-SET methylates nearby H3K9 residues of the dinucleosome to which it is bound (as illustrated) as well as H3K9 residues of flanking nucleosomes (not illustrated). Two newly methylated H3K9 residues (red spheres), in turn, recruit the Sir2^Sp-HP1 dimer (Fig. 1D). Sir2^Sp-HP1 deacetylates nearby H4K16 residues (and other lysines) of the dinucleosome to which it is bound (as illustrated) in addition to H4K16 (and other lysines) of flanking nucleosomes (not illustrated). Two newly deacetylated H4K16 residues, in turn, recruit the Sir3-SET dimer. Arrows are drawn to indicate either Sir3-SET-mediated methylation of H3K9 or Sir2^Sp-HP1-mediated deacetylation of H4K16 and other lysines. (D) ChIP-qPCR analysis of H3K9me2 levels at the URA^Kl-tetO_10X-ADE2 reporter and sites located ~3 kb away from tetO_10X. Primer binding sites are indicated by half-headed arrows. Cells are sir⁻ and produce TetR-SET, Sir2^Sp-HP1, and Sir3-SET. Cells were cultured for 6 h in the presence (gray) or absence (white) of aTc. Values represent the averages and SD of three biological replicates. (E) ChIP-seq profiles of H3K9me2 and Sir3-SET in URA3^Kl-tetO_10X-ADE2 reporter-containing sir⁻ cells producing TetR-SET and Sir2^Sp-HP1. Sir3-SET contains a FLAG_3X epitope tag. Data ranges (in counts per million) are indicated in brackets. The whole genome is visualized, with telomere junctions for all 16 chromosomes indicated on the hash-marked line. H3K9me2 profiles are shown in red, the Sir3-SET profile is shown in black, and input DNA profiles are shown in gray. The H3K9me2 profile of reporter-containing sir3Δ cells producing TetR-SET, Sir2^Sp -HP1, and Suv39^χ is shown for comparison. Black and gray arrowheads point to the reporter locus and rDNA repeats, respectively. (F) ChIP-seq profiles of H3K9me2 and Sir3-SET in sir⁻ cells containing the URA3^Kl-tetO_10X-ADE2 reporter in place of HFL1. In addition to Sir3-SET, which contains a FLAG_3X epitope tag, cells also produce TetR-SET and Sir2^Sp-HP1. Data ranges (in counts per million) are indicated in brackets. The region visualized encompasses ~30 kb of chromosome XI centered about tetO_10X. The position of tetO_10X is indicated by a white rectangle, and select genes are indicated by arrows. MRS4, which is naturally positioned adjacent to HFL1, is highlighted in pink. (G) RT-qPCR analysis of MRS4 mRNA levels in sir⁻ cells containing the URA3^Kl-tetO_10X-ADE2 reporter in place of either HMR or HFL1. Cells produce TetR-SET, Sir2^Sp-HP1, and Sir3-SET. Values represent MRS4 levels (normalized to ACT1) in cells carrying the reporter in place of the indicated locus relative to MRS4 levels in reporter-containing cells producing TetR alone. Averages and SD of three biological replicates are shown. Significance was assessed by a two-tailed Student’s t-test (P ≤ 0.001).

Chromatin immunoprecipitation experiments confirmed that H3K9me2 levels persist at reporter genes in interdependent positive feedback loop-containing sir⁻ cells after 6 h of growth in the presence of aTc (Fig. 3D). However, in addition to reporter genes, Sir3-SET and H3K9me2 also localized to telomeres (Fig. 3E). S. cerevisiae telomeres consist of repetitive DNA sequence featuring binding sites for the essential transcription factor Rap1. Rap1 recruits the SIR complex to telomeres via protein–protein interactions with Sir4 and Sir3 (10, 41). As sir⁻ cells lack Sir4, we suspected that telomere localization resulted from Sir3-SET interacting with telomere-bound Rap1. To assess the potential contribution of DNA sequence—particularly that of the HMR-proximal telomere—to reporter gene silencing, we performed three tests.

First, to rule out contribution of DNA-bound Rap1 to reporter gene silencing, we deleted Rap1-interaction determinants of Sir3 (residues 456–479) from Sir3-SET (10). sir⁻ cells producing Sir3Δ456-479-SET behaved similarly to those producing Sir3-SET (SI Appendix, Fig. S4D), indicating that neither establishment nor maintenance of reporter gene silencing requires an interaction between Sir3-SET and DNA-bound Rap1.

Second, to rule out contribution of other DNA sequences in the vicinity of HMR to reporter gene silencing, we relocated tetO_10X—along with flanking URA3^Kl and ADE2 reporter genes—from chromosome III to the non-essential, centromere-proximal, and normally euchromatic HFL1 locus of chromosome XI. Analyses of colony color phenotypes and chromatin immunoprecipitates showed that cells producing Sir3-SET and Sir2^Sp-HP1 silenced reporter genes—as well as a gene naturally flanking HFL1—and maintained silencing in the presence of aTc. These results indicate that reporter gene silencing does not require a specific chromatin context (SI Appendix, Figs. S4E and Fig. 3 F and G).

Third, to rule out any residual contributions of tetO_10X and TetR-SET to reporter gene silencing in aTc-treated cells, we deleted the gene encoding TetR-SET (SI Appendix, Fig. S5A). We performed this deletion in cells cultured in the presence of FOA. URA^Kl silencing can therefore be maintained under selection in cells devoid of TetR-SET (Fig. 4A). However, unlike cells containing TetR-SET, cells lacking TetR-SET formed colonies exhibiting little to no pigmentation on non-selective medium (Fig. 4A). Based on these results, we conclude that interdependent positive feedback loop-mediated reporter gene silencing does not require specific DNA sequence and is intrinsically unstable without selection (22, 42).

Fig. 4.

Reporter gene silencing is heritable but intrinsically unstable. (A) Phenotypes of sir⁻ cells producing Sir2^Sp -HP1 and Sir3-SET in the presence or absence of TetR-SET-encoding DNA. Three strains lacking TetR-SET (Δ #1–3) isolated on and maintained in FOA-containing medium are shown. Cells were grown in the presence of FOA and spotted on medium containing FOA or Ade¹⁰. (B) Silent information can propagate over multiple generations. Colonies—established by a Δ #1, Δ #2, or Δ #3 FOA-resistant founder cell—grown without selection either contain some fraction of FOA-resistant cells (red letter R) or contain FOA-sensitive cells only (black letter S). FOA-resistant daughter cells can be traced from round to round and emerge only from FOA-resistant mother cells, as evidenced by the Δ #1 lineage. Cells are unlikely to acquire resistance to FOA spontaneously, and once resistance to FOA is lost, it becomes irretrievable, as exemplified by the Δ #3 lineage. (C) For 20 Round 1 colonies grown on adenine-replete medium lacking FOA, the fraction of FOA-resistant (FOA^R), pigment-accumulating cells per colony was determined (pink), and the probability (P) of silent information loss per generation was calculated (gray) (Materials and Methods). The final calculated probability of silent information loss per generation represents the average and SD of P for all 20 colonies.

Two possible scenarios can account for the apparent lack of pigment accumulation in cells devoid of TetR-SET: Either i) colonies contain homogeneous populations of cells that all gradually lose silent information during colony formation, or ii) colonies contain heterogeneous populations of cells, a minority of which retain silent information, and a majority of which lose silent information (43). To distinguish between these possibilities, we asked whether colonies—each established by an FOA-resistant founder cell cultured under selection—still contain FOA-resistant cells after growth on medium lacking FOA, and if so, whether FOA resistance was a heritable trait.

To do so, we plated TetR-SET-lacking cells cultured in the presence of FOA on adenine-replete medium lacking FOA. Cells grown under adenine-replete conditions do not accumulate pigment and therefore form white colonies regardless of ADE2 levels. We randomly picked colonies and patched them onto medium containing limiting adenine and FOA. All of these colonies, which we refer to as Round 1 colonies, formed pigmented patches on medium containing limiting adenine and FOA, indicating that ADE2 and URA3^Kl were co-silenced under selection (Fig. 4B and SI Appendix, Fig. S5B). Round 1 colonies patched onto selective medium were also restreaked on adenine-replete medium lacking FOA for a second round of colony formation without selection. Unlike Round 1 colonies, only 3 out of 36 Round 2 colonies—each established by a founder cell, the phenotype of which was initially unknown—formed pigmented patches on medium containing limiting adenine and FOA (Fig. 4B and SI Appendix, Fig. S5B). We patched and restreaked cells for three additional rounds and found that reporter gene silencing was detectable without selection through Round 5, encompassing ~15 d and ~100 generations (Fig. 4B). We conclude that DNA sequence–independent silent information propagates in a minority of cells belonging to a heterogeneous population.

To quantify the (in)stability of reporter gene silencing, we calculated the probability that silent information-containing mother cells give rise to silent information-lacking daughter cells in a population undergoing exponential growth. To do so, we estimated the number of pigment-accumulating, FOA-resistant cells in each of 20 Round 1 colonies by resuspending and replating ~250 cells onto selective and non-selective medium (Fig. 4C). We used the fraction of FOA-resistant cells per colony to calculate the probability (P) of silent information loss per generation (Materials and Methods). Assuming that differences in the growth rate and replicative lifespan of cells carrying either silent or active reporter genes are negligible during colony formation on adenine-replete medium lacking FOA, we conclude that the probability of silent information loss per generation is ~12% (Fig. 4C).

A Requirement for Native S. cerevisiae Replisome–Associated Factors.

To ask whether our system is regulated by any native S. cerevisiae processes, we performed a transposon mutagenesis screen in sir⁻ cells containing a tetO_10X-ADE2 reporter along with TetR-SET, Sir2^Sp-HP1, and Sir3-SET. In total, we screened ~25,000 transposon-containing mutants for altered colony pigmentation on medium lacking aTc and ~12,500 mutants on medium containing aTc. We failed to isolate a single mutant exhibiting increased colony pigmentation in the presence of aTc—suggesting that S. cerevisiae cells naturally lack non-essential, H3K9me-specific silencing antagonists—but isolated 78 mutants exhibiting decreased colony pigmentation in the absence of aTc. We mapped transposon insertion sites for all 78 mutants, many of which were located at the promoter region of the gene encoding Sir3-SET or the reporter locus itself. Strikingly, however, the majority of remaining transposon insertions mapped to genes—most of which were identified at least twice and up to five times—involved in one of two processes: i) NAD⁺ precursor metabolism, or ii) DNA replication (SI Appendix, Fig. S6A). While we focused on the latter, we note that the availability of histone-modifying enzyme cofactors—in this case, NAD⁺-dependent Sir2^Sp-HP1—can limit heterochromatin-mediated gene silencing (44).

We deleted the coding sequence of five genes that were identified in our screen and encode replisome-associated proteins in either sir3Δ cells containing a URA3^Kl-tetO_10X-ADE2 reporter, TetR-SET, Sir2^Sp-HP1, and Sir3-SET, or SIR⁺ cells containing an ADE2 reporter gene situated between the natural E and I silencers of HMR. Deletion of four genes (DPB4, MRC1, TOF1, and CTF4) in interdependent positive feedback loop-containing cells resulted in reduced colony pigmentation and rendered cells sensitive to FOA, even in the absence of aTc (SI Appendix, Fig. S6B). In contrast, deletion of ELG1, the product of which is thought to play a pivotal role in natural S. cerevisiae gene silencing (45), had more subtle effects. The same gene deletions had comparatively modest effects on colony pigmentation in SIR⁺ cells (SI Appendix, Fig. S6C), suggesting that silencing defects were unlikely to be the result of genome-wide nucleosome depletion. Further characterization of cells lacking either DPB4, encoding a subunit of DNA polymerase ε (Pol ε) required for symmetric parental histone transfer to nascent DNA (46), or MRC1, encoding an S phase checkpoint protein bridging Pol ε and the replisome helicase (47), showed that mutant phenotypes were indicative of increased reporter gene expression (SI Appendix, Fig. S6D) and were not the result of reduced fusion protein levels (SI Appendix, Fig. S6E). These results suggest that interdependent positive feedback loop-containing cells are sensitive to asymmetric parental histone transfer and require MRC1 to assemble silent chromatin.

At first glance, it may be surprising that the deletion of genes encoding replisome-associated factors impaired reporter gene silencing in the absence of aTc, a condition under which DNA sequence–dependent recruitment of H3K9 methyltransferase activity occurs continuously. However, the establishment of heterochromatic domains requires time to reach a critical density threshold of histone modifications at target loci (48). In fact, the de novo establishment of naturally silent chromatin in S. cerevisiae cells requires at least three cell divisions to reach maturation (49, 50). Silent chromatin establishment in our engineered system is therefore expected to require maintenance factors, including replisome components involved in the retention of parental histones.

Discussion

The prime objective of many branches of molecular biology is to reduce the complexity of otherwise staggeringly complex natural phenomena, a process traditionally falling under the jurisdiction of biochemistry and in vitro reconstitution. Our study demonstrates how reductionism can be applied in vivo to capture the essential properties of complex heterochromatic domains found in natural settings. These properties include i) the capacity to assemble regional domains of silent chromatin via read–write mechanisms, ii) the—to our knowledge—universal requirement for histone deacetylation, and iii) the potential to transfer silent information between histone proteins over multiple generations in a manner that is independent of DNA sequence but dependent on replication-coupled parental histone transfer.

Taken together, our findings distill the natural complexities of locus-specific heritable gene silencing to essential components: a pair of bilingual read–write proteins, each of which i) binds cooperatively to nucleosomes, ii) reads one of two histone modifications, and iii) writes the modification distinct from what it reads, with one modification serving as a specificity-determinant as well as a contributor to chromatin silencing. The interdependent positive feedback loop described here relies on CSD- and wH domain-mediated dimerization of CD and bromo adjacent homology (BAH) domain reader modules, respectively, thereby recapitulating the requirement for cooperative binding to nucleosomes (17, 51). Furthermore, by requiring two chimeric proteins (Sir2^Sp-HP1 and Sir3-SET) along with two nucleosomal specificity-determinants (deacetylated H4K16 and H3K9me), it safeguards against unlicensed heterochromatinization and allows for silent chromatin propagation at specific loci. Finally, its mechanistic simplicity stems from the requirement for a dual-function histone modification (deacetylated H4K16) that exemplifies how “erasure” of histone modifications and recognition of unmodified histone lysines can contribute to the establishment and maintenance of silent chromatin domains.

Coupled feedback loops feature prominently in natural heterochromatin-mediated gene silencing systems found throughout Eukarya. For example, RNAi and H3K9me are coupled in fungi and plants (52–54), and DNA methylation can be coupled to H3K9me in fungi, plants, and mammals (55–58). Furthermore, evidence for nucleic acid sequence–independent coupling of histone H3 lysine 27 methylation and histone H2A lysine 119 monoubiquitination can be found in studies of mammals (59), and critically, mounting evidence for coupling of H3K9me and histone H3 lysine 14 monoubiquitination can be found in studies of native S. pombe heterochromatin (60–62).

Unlike these evolutionarily diverse and highly complex natural systems, the requirements for our engineered system (two bilingual read–write proteins, the functions of which we disentangle from native silencing pathways) can be minimally defined. Operating principles of this system promise to inform future studies of more complex silencing mechanisms—particularly those involving interdependent histone methylation and monoubiquitination—as well as other potentially memory-conferring cellular processes.

Materials and Methods

Saccharomyces cerevisiae Cell Growth, Genetic Manipulation, and Plate-Based Phenotypic Analysis.

S. cerevisiae cells derived from strain W303-1a were grown in YPAD (1% yeast extract, 2% peptone, and 2% glucose supplemented with 50 μg/mL tryptophan and 50 μg/mL adenine) or synthetic medium in the presence or absence of 10 μM anhydrotetracycline (aTc) at 30 °C.

Genes were integrated onto chromosomes in two steps. First, a target locus (ade2-1, LYS2, or CAN1) was replaced with a URA3^Kl- and KanMX -encoding cassette conferring sensitivity to 1 mg/mL 5-fluorootic acid (FOA) and resistance to 200 μg/mL G418, respectively (63). The cassette was subsequently replaced with a gene expression module consisting of a natural S. cerevisiae promoter (P_SIR2, P_RAP1, or P_SUP35), a fusion protein-encoding gene, and a terminator (t_TEF, t_ADH1, or t_CYC1). Conventional integration of pRS304-encoding Sir3-SET was also employed (64).

Gene deletions were performed by transforming cells with DNA fragments—containing sequence homologous to target loci at their ends—encoding KanMX (conferring resistance to 200 μg/mL G418), NatMX (conferring resistance to 100 μg/mL nourseothricin sulfate), HphMX (conferring resistance to 300 μg/mL hygromycin B), HIS3^SkMX (conferring histidine prototrophy), or LYS2 (conferring lysine prototrophy). All chromosomal genetic modifications were confirmed by PCR analysis of both sites of homologous recombination. Non-chromosomally encoded proteins were produced from the ARS CEN plasmid pRS315 (64), derivatives of which were maintained in cells grown in medium lacking leucine.

Plate-based phenotypic analyses were performed by normalizing overnight cell cultures to an OD₆₀₀ of 5.0, serially diluting cells in phosphate-buffered saline (PBS), and spotting dilutions onto medium containing limiting adenine (10 μg/mL; Ade¹⁰) and/or 1 mg/mL FOA in the presence or absence of aTc. Plates were incubated for 3 d at 30 °C and stored at 4 °C for 2 d before being photographed.

S. cerevisiae strains and plasmids used in this study are listed in SI Appendix, Table S1.

Calculating the Probability of Silent Information Loss.

Round 1 colonies grown under adenine-replete conditions without selection were thoroughly resuspended in PBS and diluted. ~250 colony forming units (CFUs) were plated on either selective medium containing FOA and Ade¹⁰ or non-selective, adenine-replete medium lacking FOA. The fraction of silent information-retaining cells was determined by dividing the number of pigmented, FOA-resistant colonies by the number of CFUs plated.

We calculated the probability of silent information loss per generation (P) in a population of cells undergoing exponential growth using the following equation: P^N = x, where N equals the number of generations that occurred during the formation of each Round 1 colony tested, and where x equals the fraction of FOA-resistant cells per colony. This calculation assumes that differences in the growth rate and replicative lifespan of cells carrying either silent or active reporter genes are negligible during a round of colony formation.

Chromatin Immunoprecipitation and Quantitative PCR.

First, 50 to 100 mL cultures of S. cerevisiae cells were grown in YPAD medium to an OD₆₀₀ of 1 to 2. Then, cells were fixed in the presence of 1% formaldehyde for 15 min at room temperature with gentle mixing on a moving platform. For Sir3-FLAG_3X-SET immunoprecipitations, cells were fixed in the presence of 1.5 mM ethylene glycol bis(succinimidyl succinate) (EGS) for 30 min at room temperature and then fixed in the presence of 1% formaldehyde for 30 min at room temperature. Crosslinking reactions were quenched in the presence of 125 mM glycine. Crosslinked cells were pelleted, washed twice in cold tris-buffered saline (TBS), once in cold water, and stored at −80 °C.

Frozen samples were thawed and resuspended in cold ChIP lysis buffer (0.1% sodium dodecyl sulfate, 1% Triton X-100, 0.1% sodium deoxycholate, 140 mM NaCl, 1 mM EDTA, and 50 mM HEPES-KOH, pH 7.5) supplemented with commercial protease inhibitor tablets. Cells were disrupted in a cold room by bead beating for a total of 6 min, with 1-min incubations in an ice-cold water bath following every minute of disruption. The insoluble, chromatin-containing fractions of broken cell lysates were collected by centrifugation in a cooling centrifuge set at maximum velocity (4 °C at 21,130 g) for 5 min. The insoluble fractions were washed once with 1 mL of ChIP lysis buffer, pelleted again by centrifugation, and thoroughly resuspended in 600 μL of ChIP lysis buffer. Resuspended samples were sonicated in a cold room using a tip sonicator (Sonics VCX 500; set at 30% amplitude) for a total of 80 s, with 5-min incubations on ice after every 20 s of sonication. The soluble, chromatin-containing fractions of sonicated samples were collected by centrifugation in a cooling centrifuge set at maximum velocity for 15 min.

The protein concentration of chromatin samples was measured by the Bradford method against a bovine serum albumin standard curve. First, 600 μg of chromatin was precleared with 30 μL of Protein A- or Protein G-coupled magnetic beads (Invitrogen Dynabeads) for 1 h at 4 °C on a rotating platform. Then, for each immunoprecipitation, 100 to 200 μg of precleared chromatin was incubated at 4 °C for 3 h with magnetic beads that had been prebound to one of the following antibodies: 3 μg of α-V5 antibody (Abcam ab9116), 1.5 μg of α-H3K9me2 antibody (Abcam ab1220), 3 μg of α-H3K9me3 antibody (Abcam ab8898), 4 μg of α-FLAG antibody (Sigma M2), or 4 μg of α-RPB1 (Biolegend 8WG16).

Immunoprecipitations and quantitative PCR (qPCR) analysis of immunoprecipitates were performed essentially as described (65), with the following modifications: samples were washed three times in ChIP lysis buffer, DNA was purified using an E.Z.N.A. Cycle Pure Kit (Omega Bio-tek), and real-time PCR was performed using the QuantStudio 6 Pro system (Applied Biosystems). Primers used for qPCR analysis were designed in Primers3 4.1.0 and are listed in SI Appendix, Table S2.

Chromatin Immunoprecipitate Sequencing.

Chromatin immunoprecipitates and input DNA samples were further sheared by sonication using a Q800R3 sonicator (Qsonica; set at 20% amplitude). Samples were sonicated for a total of 15 min, with 15-s incubations without sonication following every 15 s of sonication. DNA was purified using a NucleoSpin Gel and PCR Clean-up kit (Machery-Nagel). Samples were barcoded and pooled as previously described (66). Pooled samples were sequenced on an Illumina NextSeq 500 system at the Bauer Core Facility (Harvard University). In-line barcode demultiplexing was performed courtesy of the Bauer Core Facility, and reads were mapped to the S288C reference genome R64.3.1, the sequence of which was manually edited to lack the native S288C ADE2 gene and to contain tetO_10X-ADE2 or URA3^Kl- tetO_10X-ADE2 reporters in place of either HMR or HFL1. Read alignments were performed using Bowtie with default settings. Data were normalized to counts per million using the bamCoverage module of deepTools 3.5.2 and visualized in IGV 2.16.0.

RNA Extraction and cDNA Synthesis.

Total RNA was collected from S. cerevisiae cells by hot acid phenol-chloroform extraction. RNA samples were treated with TURBO DNase (Invitrogen) and purified using a Quick-RNA Miniprep Kit (Zymo Research). cDNA was synthesized using the SuperScript IV First-Strand Synthesis System (Invitrogen) with Oligo(dT) primers. cDNA was diluted 1:10 or 1:50, and cDNA levels were quantified by real-time PCR as previously described (65). Primers used for qPCR analysis were designed in Primers3 4.1.0 and are listed in SI Appendix, Table S2.

Western Blot Analysis.

Yeast whole cell extracts were prepared by bead beating and trichloroacetic acid (TCA)-mediated protein precipitation. Samples were solubilized in sodium dodecyl sulfate (SDS) loading buffer containing 50 mM Cleland’s reagent (DTT), boiled for 5 min, and clarified by centrifugation. Gel electrophoresis was performed using NuPAGE 4-12% Bis-Tris precast gels (Invitrogen) and MOPS-SDS running buffer (Boston Bioproducts). Samples were transferred to nitrocellulose membranes, which were stained with Ponceau S Staining Solution (Thermo Scientific), photographed, and blocked in tris-buffered saline containing 0.1% Tween 20 (TBST) and 5% milk. The same antibodies used for chromatin immunoprecipitations were also used to probe membranes for epitope-tagged proteins of interest. Primary antibodies were diluted 1:2,500 and secondary antibodies conjugated to HRP were diluted 1:5,000 in TBST milk. Chemilluminescence was detected by a ChemiDoc XRS+ System (Bio-rad).

Transposon Mutagenesis and Mapping of Transposition Sites.

hmrΔ:: tetO_10X-ADE2 reporter-containing sir⁻ cells producing chromosomally encoded TetR-SET, Sir2^Sp-HP1, and Sir3-SET were transformed with ARS CEN plasmid pSG36, which encodes URA3, a Hermes-NatMX transposon, and a hyperactive Hermes transposase variant produced under the control of a galactose-inducible promoter (67). Six transformants were individually cultured in glucose-containing medium lacking uracil. Cell cultures were diluted to an OD₆₀₀ of 0.05 in 25 mL of galactose-containing medium lacking uracil—to induce transposition—and grown to saturation overnight. The next day, cell cultures were diluted to an OD₆₀₀ of 0.25 in 100 mL of glucose-, uracil-, and FOA-containing medium—to cure cells of pSG36—and grown to saturation overnight. The following day, cell cultures were diluted to an OD₆₀₀ of 0.5 in 100 mL of glucose-, uracil-, FOA-, and nourseothricin sulfate–containing medium—to select for transposon-containing cells lacking pSG36—and grown to saturation overnight. Finally, cells were plated on glucose-, uracil-, FOA-, nourseothricin sulfate-, and 10 μg/mL adenine (Ade¹⁰)-containing medium in the presence or absence of aTc. Plates were incubated for 3 d at 30 °C and stored at 4 °C for 2 d before being visually inspected.

Genomic DNA of mutants exhibiting altered colony pigmentation was purified, and Hermes-NatMX transposon insertion sites were mapped by arbitrary PCR (68) and Sanger sequencing. Primers used for arbitrary PCR are listed in SI Appendix, Table S2.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Raw and processed ChIP-seq data are available at the NCBI Gene Expression Omnibus (accession number: GSE236173 (69)).