Skip to content

Bivalent chromatin

A bounded chromatin state with co-enriched H3K4me2 or H3K4me3 and repressive H3K27me3 at the same locus in a specified cell or allele context.

Version
v1 · 2026-10-17 · History
Domain-specific #
14073
Domain group
Natural Sciences
Origin domain
Biology & Ecology
Subdomains
Chromatin Biology, Gene Regulation → Biology & Ecology
Aliases
Bivalent chromatin domain

Core Idea

Bivalent chromatin is a state of a specified chromatin region in a specified biological context: source-supported H3K4me2 or H3K4me3 and the repressive-associated H3K27me3 mark are enriched in that same region.[1][2] The region, cell or tissue context, which of these two K4 grades is measured, and measurement resolution all matter. An H3K4me1-only state is outside the evidence for this entry, without claiming that biology universally excludes it. Bernstein and colleagues identified H3K4me3/H3K27me3 domains at selected developmental loci in mouse embryonic stem (ES) cells.[1] Sanz and colleagues identified an allele-specific H3K4me2/H3K27me3 domain at the mouse Grb10 CGI2 region.[2] Those are instances of the bounded co-marked state without requiring the K4 grade or allelic arrangement to be identical.

The two marks have different usual regulatory associations, but their coexistence is the state being named, not a promise of a particular transcriptional result. Bernstein and colleagues proposed that some ES-cell domains keep developmental genes repressed yet poised for activation.[1] Their selected-locus observations and later differentiation comparisons motivate that model; they do not establish rapid activation, maintenance of pluripotency, or resolution of every bivalent domain as necessary properties.[1] A later normal-colon study found several different tumor-associated chromatin outcomes rather than one obligatory fate.[3]

Structural Signature

The state has five constitutive roles and a separate evidentiary test:

  • Specified chromatin locus. A promoter or other bounded genomic region is the unit classified. K4 signal at one locus and K27 signal at another do not form one bivalent region.[1][2]
  • Specified biological context. The relevant cell or tissue condition must be fixed; where the finding is allele-specific, the parental allele must also be fixed. Combining signals from different cells or parental alleles could manufacture an apparent positive.[2][3]
  • Source-supported H3K4me2 or H3K4me3 signal. One of these two K4 grades must occur at the named region; an H3K4me1-only signal is not a positive under the reviewed evidence. H3K4me3 characterizes the Bernstein and Hahn cases; the direct paternal Grb10 CGI2 assay in Sanz is H3K4me2, not an interchangeable me3 finding.[1][2][3]
  • H3K27me3 signal. The repressive-associated mark must occur in the same bounded context. A K4-only promoter lacks the second side of the configuration.[1][2]
  • Same-region co-enrichment. The two signals must refer to one relevant domain at an explicitly stated grain. Sequential chromatin immunoprecipitation (sequential ChIP or ReChIP) at selected loci supports their presence on immunoprecipitated chromatin fragments; it does not, by itself, place both marks on one histone molecule or prove the state in every cell.[1][2][3]

A discriminating assay and comparator is central to recognizing an instance, though measurement is not an ontic component of the state. One-mark controls, parental-allele contrasts, or selected-locus ReChIP help test whether a reported positive is more than overlapping bulk signals.[1][2][3] A state could exist without having been measured, but a particular claim of its presence needs evidence that can address the relevant mixture alternatives.

What It Is Not

It is not simply a low-expressed gene or a pair of separately measured histone marks, and the inspected positive cases do not license an H3K4me1-only example. A low-expressed promoter may carry only one mark, while pooled measurements can mix different cells, alleles, or loci. Bernstein's H3K4-only Tcf4 and H3K27-only HoxA3 controls show the boundary in the same experimental setting.[1]

It is not an all-instance developmental switch. Several Bernstein loci changed mark balance in neural precursors, but weakly induced Dlx1 retained both marks.[1] Sanz found a bivalent paternal Grb10 domain in differentiated non-brain tissues.[2] Thus neither permanent persistence nor mandatory disappearance after differentiation defines the state.

It is not evidence, on its own, that H3K4 methylation prevents irreversible silencing, that every domain preserves pluripotency, or that a chromatin remodeler always evicts a Polycomb complex. Those are separate causal claims. The sources here support a co-marked state, selected expression and chromatin comparisons, and bounded functional interpretations; they do not make those mechanisms universal.[1][3]

Scope of Application

The classic experimental habitat is a selected mouse ES-cell developmental locus. Bernstein and colleagues examined 56 large regions rich in highly conserved noncoding elements, plus controls, and associated bivalent marks with low-expressed developmental transcription-factor genes.[1] That selected design is not a genome-wide census or a requirement that every bivalent locus be an ES-cell transcription-factor promoter.

An unlike habitat is mouse genomic imprinting at the paternal Grb10 CGI2 region in differentiated non-brain tissue. Allele-specific assays and sequential ChIP there concern H3K4me2/H3K27me3.[2] The maternal allele is differently marked; paternal H3K27me3 enrichment is lost in brain.[2] This does not establish the same bivalent mechanism for human GRB10 or a second maternal-brain positive.

Hahn and colleagues studied H3K4me3/H3K27me3 promoter calls in normal human colonic mucosa and their alteration in four matched stage-II colorectal tumor pairs.[3] Peak overlap defined broad promoter calls, while ReChIP tested selected normal-mucosa loci more directly.[3] The study adds an observational tissue-and-disease comparison, not a claim that all tumors arise from bivalent promoters or share one chromatin trajectory. Approximately 40–50 percent of the studied normal-colon bivalent promoters were colon-specific relative to the compared ES-cell data.[3]

Clarity

Calling a locus “bivalent” becomes clearer when a report names the region, cell or allele, K4 grade, K27 mark, and assay grain. The phrase “both marks at Grb10” would hide Sanz's decisive distinction between paternal H3K4me2/H3K27me3 CGI2 chromatin and a separate regional H3K4me3 database observation.[2] Likewise, two aligned ChIP tracks are an operational overlap call, whereas sequential ChIP at a selected locus gives stronger same-fragment evidence.[2][3]

This language also separates an observed state from its proposed consequence. A low transcript level does not demonstrate co-enrichment. Co-enrichment does not, without a comparison, establish that transcription can be induced rapidly.[1] The distinction lets a reader evaluate the chromatin observation and the functional interpretation on their own evidence.

Manages Complexity

Many loci, mark profiles, cell conditions, alleles, and expression measurements can be organized into a smaller classification question: which bounded region in which context has both relevant marks at what resolution? The same question can compare an ES-cell developmental locus with a paternal imprinted locus while preserving their different K4 grades and allelic structures.[1][2] The state criterion therefore reduces descriptive clutter without pretending that their regulatory mechanisms are identical.

Outcomes remain a second layer. Hahn's normal-to-tumor comparisons include H3K27me3 loss associated with either H3K4me3 gain and activation or DNA methylation and repression, as well as other changes including H3K27me3 gain.[3] Treating all of these as “resolution to activation” would discard the very alternatives the study measured. The compact state label is useful only if later trajectories retain their own case and time qualifiers.

Abstract Reasoning

To assess a claimed instance, first fix the locus and biological context, including allele where relevant. Next ask whether each mark is measured there and whether the assay distinguishes co-enrichment from a mixture of different cells, alleles, or fragments. Only then infer the bounded classification. In Bernstein's Irx2 case, reverse-order sequential ChIP and one-mark controls strengthen the positive; in Sanz's Grb10 case, the paternal allele and non-brain context are indispensable to the claim.[1][2]

A separate inference asks what happens after the state is identified. Some ES-cell loci changed marks during neural-precursor differentiation, while Dlx1 retained bivalency.[1] At Grb10, reduced paternal H3K27me3 in neural precursors accompanied paternal exon 1B1 expression, but EED-deficient embryos did not show ubiquitous paternal expression.[2] A change in one mark is therefore a premise for a context-specific interpretation, not a universal rule that fixes transcription by itself.

Knowledge Transfer

Within chromatin biology, the recognition questions transfer: name the locus, condition or allele, exact H3K4 grade, H3K27me3 signal, and resolving assay. They work across ES-cell developmental genes, a differentiated imprinted region, and selected normal-colon promoters. The answers do not transfer automatically. In particular, Sanz's paternal H3K4me2 result cannot be silently restated as Bernstein's H3K4me3 result, and Hahn's tumor associations cannot establish the fate of every ES-cell locus.[1][2][3]

The more general question of whether distinguishable features share a bounded region recurs outside biology; that relation is Co-location, an internal constituent of the bounded chromatin state. A nonbiological instance of Co-location is not a new instance of bivalent chromatin. The named state's literal transfer remains within source-supported chromatin settings.

Examples

Mouse ES-cell Irx2 transcription-start region. Bernstein and colleagues measured H3K4me3 and H3K27me3 in mouse ES-cell chromatin at this selected developmental locus.[1] Sequential ChIP, including reversed order, supported both marks on the assayed chromatin fragments.[1] H3K4-only Tcf4 and H3K27-only HoxA3 supplied discriminating controls.[1] Mapped back: the Irx2 region is the locus; ES cells fix the context; H3K4me3 and H3K27me3 fill the two mark roles; sequential ChIP supports same-region co-enrichment at fragment grain; the one-mark controls test the positive. Neural-precursor differentiation is a later comparison, not the membership condition.

Mouse paternal Grb10 CGI2 in differentiated non-brain tissue. Sanz and colleagues observed paternal H3K4me2 and H3K27me3 at CGI2, with sequential ChIP evidence on chromatin fragments in fibroblasts.[2] The maternal allele and brain condition provide different comparisons: paternal H3K27me3 is lost in brain, where paternal expression uses distinct promoter regions.[2] Mapped back: CGI2 is the locus; the paternal allele and non-brain tissue fix the context; H3K4me2 and H3K27me3 fill the mark roles; Figure 2B sequential ChIP supports co-enrichment; parental and tissue contrasts test the classification. The direct result is me2, not a paternal me3 demonstration.

Normal human colonic-mucosa promoters. Hahn and colleagues called promoters bivalent where H3K4me3 and H3K27me3 peaks overlapped in an expanded promoter interval, then used ReChIP to check selected normal-mucosa loci.[3] Four matched stage-II tumor pairs supplied a comparison of later chromatin and expression patterns.[3] Mapped back: each called promoter is a bounded region; normal mucosa is the baseline context; me3 and H3K27me3 are the two marks; peak overlap is the broad operational grain; selected-locus ReChIP provides stronger fragment evidence; matched tumors and controls frame outcomes. The selected ReChIP results cannot be generalized to every called promoter or converted into a universal cancer mechanism.

Structural Tensions

Permissive versus repressive interpretation. Source-supported H3K4me2 or H3K4me3 and H3K27me3 carry opposing regulatory associations, yet the bivalent state contains both.[1][2] Leaning entirely on H3K27me3 would erase the permissive side and the study's poising hypothesis; leaning entirely on H3K4 would wrongly treat co-marking as active transcription or guaranteed activation. Their co-presence does not maximize repression and activation simultaneously in one deterministic output. Diagnostic: at which locus, grade, and sampled context was expression observed, and what evidence connects that expression to the marks?

Coverage versus physical specificity. Broad ChIP maps can screen many promoters for overlapping peaks, but pooled overlap may combine signals from different cells or fragments. Selected-locus sequential ChIP gives more discriminating same-fragment evidence while sacrificing genome-wide coverage.[1][3] Treating the broad screen as molecular proof inflates certainty; demanding ReChIP everywhere may leave many operational calls unclassified. Diagnostic: what grain supports this particular call, and which mixture alternative has actually been excluded?

Retention versus change. A co-marked state can persist in differentiated non-brain Grb10 tissue or at Dlx1 in a neural-precursor comparison, while other selected loci lose a mark or change expression.[1][2] Assuming permanent retention misses observed change; assuming mandatory resolution excludes real positives after differentiation. Neither pattern can be imposed on all loci without losing the state/context distinction. Diagnostic: which lineage, allele, and sampled interval directly show persistence, mark loss, or altered expression?

Autonomy versus reduction to a shared-region relation. The three primary cases require distinguishable mark-bearing features to occupy one bounded region at the stated grain.[1][2][3] Ignoring that common relation makes the chromatin state look unrelated to a general structural pattern. Yet reducing the whole named state to Co-location would lose the H3K4me2/me3 and H3K27me3 identities, cell or allele context, and assay-dependent boundary. Treating it as wholly autonomous preserves those biological requirements but risks concealing the common same-region test. Neither choice can absorb the other without a loss of explanatory specificity. Diagnostic: do the mark-bearing features share the declared locus at the actual assay grain, with the biological mark identities still separately established? The strict Co-location constituent requires that relation in every actual bivalent state; broad pooled calls remain provisional until mixture alternatives are addressed.

Structural–Framed Character

Bivalent chromatin is a structurally describable, domain-specific state. Its evidential weight comes from a repeatable relation among a locus, two marks, and a bounded context, not from an evaluation that the state is desirable. Calling it “poised” adds a functional interpretation whose warrant varies by case; it cannot replace the measured configuration.[1][2]

The state does not require a human institution or a laboratory to exist. Human practice does determine which locus is sampled, which allele is resolved, how ChIP peaks are called, and what resolution a published positive can support. The term arose in chromatin research and travels through developmental, imprinting, and cancer studies when those same biological roles are shown.[1][2][3] It does not travel literally to any situation with two opposed influences. Importing the word without source-supported H3K4me2 or H3K4me3, H3K27me3, and a chromatin region would be analogy rather than recognition of this named state.

Its character: structural in its testable co-marked configuration, framed by chromatin biology in every literal instance. The generic same-region relation is more portable than the histone-defined state and supplies the strict Co-location constituent; it does not replace the biological identity.

Structural Core vs. Domain Accent

The portable skeleton is distinguishable features present in one bounded place during an overlapping observation interval. The live Co-location Prime names that internal relation. It is a strict composition/part_of constituent here, with parent_in_child direction: removing same-region overlap removes bivalent identity even if both marks exist elsewhere. At a fixed locus and assay grain the relation is symmetric; no unrestricted transitivity across loci, alleles, cells, or changing grain follows. Co-location itself requires no regulatory interaction, and its only live Neighborhood edge is typical, so no strict radius obligation is inherited.

The chromatin domain does more than accent an otherwise generic relation. The specified locus, biological and sometimes allelic context, source-supported H3K4me2 or H3K4me3, H3K27me3, and assay-grain limits determine what can count as this particular state.[1][2] Remove those biological roles and only a general co-presence pattern remains. The named bivalent-chromatin entry therefore does not clear the Prime bar through cross-domain metaphor; any wider reach belongs to an independently justified structural relation, not to these histone marks.

This entry is part of Co-location.

The two distinguishable mark-bearing features must share a bounded region and relevant observation interval.[1][2] The sole typed parent is Co-location, a strict composition/part_of internal constituent, with parent_in_child direction. This placement tracks a necessary spatial relation rather than treating the whole biological state as just a spatial relation. Broad overlapping ChIP peaks are operational calls; the strict relation applies to actual source-bounded co-marked states, with selected sequential-ChIP results supplying stronger evidence at fragment grain.

Epigenetics is a broader biological field and regulatory context, while Histone methylation names a methyl-transfer process. Neither is interchangeable with a particular co-marked locus state; topical proximity alone does not supply a strict parent. A general Co-location relation also leaves the biological marks and their conditional outcomes unspecified.

Relationships to Other Abstractions

Local relationship map for Bivalent chromatinParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Bivalent chromatinDOMAINPrime abstraction: Co-location — is part ofCo-locationPRIME

Current abstraction Bivalent chromatin Domain-specific

Parents (1) — more general patterns this builds on

  • Bivalent chromatin is part of Co-location Prime

    Same-region co-location of the two distinct mark-bearing chromatin features is an identity-bearing constituent of the bivalent chromatin state.

Hierarchy path (1) — routes to 1 parentless root

Neighborhood in Abstraction Space

Bivalent chromatin sits in a sparse region of the domain-specific corpus (93rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Molecular & Cellular Biology Mechanisms (16 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-17

Not to Be Confused With

Epigenetics concerns a much wider family of regulation; an epigenetic finding need not contain both marks at one locus. Histone methylation is a modification process, whereas bivalent chromatin is a state identified by a particular co-marked arrangement. A one-mark promoter, or two marks only in separate cells or alleles, fails the positive boundary.[1][2]

Same-fragment evidence is not the same as a universal same-nucleosome architecture. The sequential-ChIP experiments used for the two complete cases do not establish that every bivalent nucleosome holds both marks, let alone both on one H3 molecule.[1][2] Nor does a tumor's DNA methylation pattern prove that H3K4me3 previously prevented permanent silencing.[3] The frozen member's proposed BAF/PRC1 eviction step is not established here as a universal or constitutive requirement.

Imprinting is not an automatic alias. In the Grb10 study the bivalent CGI2 finding is on the paternal mouse allele in non-brain context.[2] The differently marked maternal allele, brain-specific paternal expression, and human GRB10 literature cannot be folded into that same measured positive.[2]

References

[1] Bernstein, Bradley E., Tarjei S. Mikkelsen, Xiaohui Xie, Michael Kamal, Dana J. Huebert, James Cuff, Ben Fry, Alex Meissner, Marius Wernig, Kathrin Plath, Rudolf Jaenisch, Alexandre Wagschal, Robert Feil, Stuart L. Schreiber, and Eric S. Lander. (2006). “A bivalent chromatin structure marks key developmental genes in embryonic stem cells.” Cell 125(2): 315–326. doi:10.1016/j.cell.2006.02.041. Original article PDF: https://plathlab.org/wp-content/uploads/2006/04/2006-Bernstein-1-s2.0-S0092867406003801-main.pdf . Relevant: selected 56 HCNE-rich loci; Figure 3 sequential ChIP and one-mark controls; Figure 5 neural-precursor comparison. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y ↩z ↩27 ↩28 ↩29 ↩30 ↩31

[2] Sanz, Lionel A., Stormy Chamberlain, Jean-Charles Sabourin, Amandine Henckel, Terry Magnuson, Jean-Philippe Hugnot, Robert Feil, and Philippe Arnaud. (2008). “A mono-allelic bivalent chromatin domain controls tissue-specific imprinting at Grb10.” The EMBO Journal 27(19): 2523–2532. doi:10.1038/emboj.2008.142. Original full text: https://pmc.ncbi.nlm.nih.gov/articles/PMC2567399/ . Relevant: Figures 1–4, especially paternal CGI2 H3K4me2/H3K27me3 sequential ChIP in Figure 2B and conditional EED result. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y ↩z ↩27 ↩28 ↩29 ↩30

[3] Hahn, Maria A., Arthur X. Li, Xiwei Wu, Richard Yang, David A. Drew, Daniel W. Rosenberg, and Gerd P. Pfeifer. (2014). “Loss of the Polycomb mark from bivalent promoters leads to activation of cancer-promoting genes in colorectal tumors.” Cancer Research 74(13): 3617–3629. doi:10.1158/0008-5472.CAN-13-3147. Original author manuscript: https://pmc.ncbi.nlm.nih.gov/articles/PMC4079738/ . Relevant: four matched patient pairs; promoter peak definition; selected normal-mucosa ReChIP; Figure 3 and alternative tumor-associated outcomes. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r