Chromatin regulation: How complex does it get?

Abstract

Gene transcription is tightly regulated at different levels to ensure that the transcriptome of the cell is appropriate for developmental stage and cell type. The chromatin state in which a gene is embedded determines its expression level to a large extent. Activation or repression of transcription is typically accomplished by the recruitment of chromatin-associated multisubunit protein complexes that combine several molecular tools, such as histone-binding and chromatin-modifying activities. Recent biochemical purifications of such complexes have revealed a substantial diversity. On the one hand, complexes that were thought to be unique have been revealed to be part of large complex families. On the other hand, protein subunits that were thought to only exist in separate complexes have been shown to coexist in novel assemblies. In this review we discuss our current knowledge of repressor complexes that contain MBT domain proteins and/or the CoREST co-repressor and use them as a paradigm to illustrate the unexpected heterogeneity and tool sharing of chromatin regulating protein complexes. These recent insights also challenge the ways we define and think about protein complexes in general.

Abbreviations

ATP

adenosine triphosphate

BAP

brahma associated protein

BHC80

BRAF-histone deacetylase complex 80

BRG1

brahma Related Gene 1

CHD

chromo domain helicase DNA binding

CoREST REST

corepressor

dL(3)mbt

Drosophila Lethal 3 malignant brain tumor

DNA

deoxyribonucleic acid

DNMT

DNA methyltransferase

DP-1

dimerization partner 1

ELM2

EGL-27 and MTA1 homology 2

ES cell

embryonic stem cells

E2F

E2 transcription Factor

H

histone

hBRM

human Brahma

HDAC

histone deacetylas

HMTase

histone methylase

HP1

heterochromatin protein 1

K

lysine

Lint-1

l(3)mbt interacting 1

LINT

l(3)mbt interacting

LSD1

lysine-specific demethylase 1

L3MBTL

lethal 3 malignant brain tumor-like

l(3)mbt

lethal 3 malignant brain tumor

MBT

malignant brain tumor

MBTS

malignant brain tumor signature

NPA1

nucleosome assembly protein

NRSF

neural-restrictive silencing factor

NuRD

nucleosome remodeling and deacetylase

PBAP

polybromo-associated BAP

Pc

polycomb

PcG

polycomb group

Ph

polyhomeotic

PHD

plant homeo domain

Pho

pleiohomeotic

PhoRC

Pho repressive complex

PRC1

polycomb repressive complex 1

PRE

polycomb responsive element

Psc

posterior sex combs

REST

repressor element 1 silencing transcription factor

RB

retinoblastoma

RNA

ribonucleic acid

Rpd3

reduced potassium dependency 3

SANT

SWI/ADA2/N-CoR/TFIIIB

Sce

sex combs extra

Scm

sex combs on midleg

SCML

sex combs on midleg-like

Sfmbt

Scm-related gene containing 4 mbt domains

SLC

SFMBT1, LSD1, CoREST

SWH

Salvador-Warts-Hippo

SWI/SNF

switching defective/sucrose non-fermenting

TSS

transcription start site

YY1

ying-yang 1

ZNF

zinc finger

Introduction

In order to specifically and reliably regulate gene transcription the cell uses multisubunit protein complexes to modulate or maintain chromatin structure. A remarkable feature of these complexes is that they combine several activities to manipulate chromatin. These activities include ATP-dependent nucleosome remodeling, histone modification, DNA modification, DNA and RNA binding, as well as the recognition of specific histone modifications. Typically, each activity is provided by one specific complex subunit, mostly through defined protein domains of this subunit. Multisubunit chromatin regulators can be viewed as “toolboxes” which contain the right combination of “tools” to get a particular chromatin job done. Recently, it is becoming apparent that these complexes are not fixed entities. Complexes that were once thought of as being unique have been revealed to exist in several versions that differ in the precise combination of tools they contain and form “complex families.” Moreover, subunits that were considered to exist in separate complexes have been found together in novel molecular assemblies suggesting that a significant amount of tool sharing is going on. In this review, we will highlight recent findings that illustrate the dynamic nature of the composition and function of multisubunit chromatin regulators.

Evolution of Polymorphic Complexes

SWI/SNF (switching defective/sucrose non-fermenting) complexes are evolutionarily conserved chromatin remodeling complexes that use the energy derived from ATP hydrolysis to modulate chromatin structure.^1-3 They provide a good example to illustrate the combinatorial diversity in subunit composition that is characteristic of many chromatin regulating protein complexes of higher organisms. The unicellular yeast Saccharomyces cerevisiae contains a single SWI/SNF complex harboring the SWI2/SNF2 nucleosome remodeling ATPase. This complex has a defined subunit composition.⁴ Brahma, the homolog of SWI2/SNF2 in the fruit fly Drosophila melanogaster, forms a similar complex.⁵ However, 2 of its subunits (OSA and Polybromo) associate with Brahma in a mutually exclusive manner. This gives rise to 2 variants of the Brahma complex, BAP and PBAP.⁶ Humans have 2 highly related but distinct SWI2/SNF2-type ATPases, BRG1 and hBRM, which reside in distinct complexes.⁷ As in flies, BRG1 shows mutually exclusive interaction with certain subunits resulting in the coexistence of distinct mammalian SWI/SNF complexes.⁸ The differences in composition and complexity of SWI2/SNF2 assemblies that we observe between unicellular and multicellular organisms likely reflect the diversification of an ancestral monomorphic SWI/SNF complex during evolution (Fig. 1). Such diversification equips the cell with several distinct but related toolboxes that can carry out more specialized functions. Moreover, the polymorphic mammalian SWI/SNF complexes switch subunits as cells undergo differentiation from ES cells to neural progenitor cells and to fully differentiated neurons ⁹ (Fig. 2, left panel). In analogy to “protein families” these assemblies can be viewed as “protein complex families.”

Figure 1.

Evolution of a protein complex family. Schematic representation of a monomorphic complex (Complex A) that evolves into a family of polymorphic complexes (Complexes A1, A2 and A3).

Figure 2.

Complex families: developmentally regulated cell type-specific expression versus coexistence in the same cell. Scheme illustrating different modes of expression of protein complex family members. Left: Not all members of a protein complex family coexist. Expression of certain subunits is developmentally regulated. As a consequence certain complex family members are only present at particular developmental stages (e.g., SWI/SNF complexes; see text for details). Right: Complex family members coexisting in the same cell type (e.g., PRC1 complexes, see text for details).

Diversity of Function Within Protein Complex families

Another prominent example for a heterogeneous protein complex family is provided by Polycomb group (PcG) proteins. These epigenetic repressors play a critical role during various developmental processes, such as maintenance of cell identity, stem cell differentiation, and cancer development.^10,11 To exert their repressive function on chromatin, they assemble into multisubunit protein complexes. One of these complexes is polycomb repressive complex (PRC) 1 that was originally described as a single complex in Drosophila.¹² However, in mammals every Drosophila PRC1 subunit is represented by several homologs, providing potential for heterogeneity. Indeed, a combinatorial proteomic and genomic analysis has recently identified 6 major PRC1 complexes (PRC1.1 to PRC1.6) constituting a novel complex family.¹³ These complexes were all purified from the same cell line demonstrating that they can coexist (Fig. 2, right panel). Why do cells need so many different PRC1 complexes? The answer appears to be that they bring different abilities to the table. While they all share common subunits, each complex also contains a specific set of associated proteins. These additional subunits equip each PRC1 complex with different tools, such as histone demethylation or deacetylation activities and different types of histone modification binding modules.

Merging Protein Complex Families

Sorting the myriad of chromatin regulating complexes neatly into different complex families satisfies the human need for order. However, recent findings suggest that the lines separating protein complex families are becoming increasingly blurred (Fig. 3). This raises fundamental questions about the way we think about and define protein complexes and protein complex families.

Figure 3.

Merging protein complex families. Simplified model of how boundaries between 2 different classes of protein complex families break down as scientific progress (arrow from top to bottom) leads to the identification of novel assemblies. Initially, complex families are defined by the presence of signature subunits (depicted in green/orange or blue/yellow; e.g., MBT domain proteins and CoREST, see text for details). Complex families expand as more complexes are being identified. Some accessory subunits (red) are being found in complexes from both families but the presence of signature subunits still allows an unambiguous classification of complexes. Eventually, complexes are being identified which combine signature subunits from different families. The complex families have merged.

In this review, we will discuss these broader issues by using 2 complex families as examples: MBT domain proteins and complexes containing CoREST proteins ^14-19 (Table 1). When MBT and CoREST complexes were originally isolated they were viewed as belonging to distinct families that did not share common subunits (Fig. 3, top panel). Both were proposed to induce repressive chromatin structures but to employ different mechanisms to do so. Unexpectedly, complexes in which MBT domain and CoREST proteins coexist were recently identified.^20-23

Table 1.

MBT and CoREST protein complexes. As target genes only major groups of special interest are mentioned. Core subunits are depicted in bold. On the left complexes are labeled according to the presence of MBT and/or CoREST proteins as subunits: MBT protein (blue), CoREST (green) and a combination of MBT and CoREST proteins (red).

MBT and CoREST complexes
Name	Subunits	Purification/Material	Target genes	Hypothesized repression mechanism	Ref.
Drosophila
PRC1	Ph Pc Psc Sce/dRing1 dTAFIIs (42, 62, 85, 110, 250) Zeste Scm (?)	Chromatographic fractionation; affinity purification/Drosophila embryos	Hox genes	H3K27me3 binding (modification set by PRC2); H2AK119 monoubiquitination; chromatin compaction	52
PhoRC	Sfmbt Pho	Tandem affinity purification/Drosophila embryos	Hox genes	PRE binding via Pho; targeting of PRC1 and PRC2	27
MybMuvB	Myb Mip130/120/40 Lin52 E2F2/DP RBF1/2 p55 dL(3)mbt, dRpd3	Chromatographic fractionation; affinity purification/Drosophila embryos	Developmentally-regulated E2F target genes	Recruitment of histone modifying co-repressors (e.g., HDAC; HMTs)	101
dLsd1-dCoREST complex	dLsd1 dCoREST dRpd3	Immunoprecipitation/ SL2 cells	Neuronal genes	Histone demethylation (H3K4) and deacetylation	72
LINT	dL(3)mbt dCoREST dLint-1	Flag-affinity purification/SL2 cells	Germline-specific genes	Promoter occupancy and promoter blocking (H4K20me1-independent)	20
Human (Mammalian)
PRC1	PHs PCGFs CBXs RING1A/B SCMH1	Chromatographic fractionation; affinity purification/ HeLa cells	Hox genes	H2AK119 mono-ubiquitination; chromatin compaction; H3K27me3 binding (modification set by PRC2);	53
PRC1.1	PCGF1SKP1 BcoR/L1 KDM2B	Tandem affinity purification/HEK293T-REx cells
PRC1.2/4RING1B/ARYBP or YAF2	PCGF2/4CBXs PHCs SCMs
PRC1.3/5	PCGF3/5AUTS2 FBRS/L1 CKIIs
E2F6.com*/PRC1.6**	L3MBTL2 E2F6/DP Mga/Max RING1B/A HP1γ YAF2 PCGF6 RYBP**, HDAC1/2**, WDR5 **HMTase1+G9a*	Tandem affinity purification/HeLa cells *Tandem affinity purification/HEK293T-REx cells**	E2F target genes	Recruitment of histone modifying co-repressors (e.g., HDAC; HMTs); chromatin compaction	(*)56 (**)13
L3MBTL1 complex	L3MBTL1 HP1γ pRB H1b core histones	Affinity purification/ HEK293 cells	E2F target genes	Nucleosome compaction (H4K20 methylation-dependent)	15
LSD1-CoREST complex	LSD1 CoREST HDAC1/2 BHC80 BRAF35*/**, CtBP*/ **ZnF217*/**/516*/198 **KIAA0182*, KIAA1343 **HMG20A**	Affinity purification/HEK293 cells *Tandem affinity purification/HeLa cells**	Neuronal genes	Histone demethylation (H3K4) and deacetylation	(*)18 (**)19
SLC complex	SFMBT1 LSD1 CoREST	Tandem affinity purification/HEK293T-REx cells	Histone house keeping genes	Nucleosome compaction (H4K20methylation-dependent)	23

The MBT Protein Family

The founding member of the MBT protein family is Drosophila Lethal 3 malignant brain tumor (dL(3)mbt).²⁴ A recessive-lethal mutation of the l(3)mbt gene results in malignant transformation of the larval brain.²⁵ Analysis of the dL(3)mbt polypetide sequence revealed 3 tandem repeats of a novel MBT domain.²⁴ In addition to dL(3)mbt, Drosophila encodes 2 more MBT domain proteins with 2 and 4 MBT repeats, respectively: Sex combs on midleg (Scm) and Scm-related gene containing 4 mbt domains (Sfmbt). Both are PcG proteins and thus are required for stable silencing of Hox genes during development.^26,27 During evolution the MBT domain protein family has expanded to include 9 members in humans.²⁸ In accordance with the l(3)mbt phenotype in Drosophila several human MBT domain proteins (L3MBTL1, L3MBTL2, L3MBTL3, L3MBTL4, and SCML2) have been implicated in tumorigenesis.^28-33

In addition to MBT modules, these proteins often contain zinc fingers and Scm, Ph and MBT homology (SPM) domains.³⁴ The SPM domain mediates oligomerization of MBT domain proteins.^35-37

In reporter gene assays, MBT domain proteins function as potent transcriptional repressors when tethered to DNA and the MBT domains appear to play a critical role here.^27,38-41 Deletion studies of L3MBTL1 revealed that the 3 MBT domains are essential for full repressive activity.³⁹ Furthermore, the 4 MBT repeats of SFMBT1 are sufficient to mediate transcriptional repression.⁴¹

In vitro studies using isolated MBT domains and histone peptides have demonstrated a general preference for binding to mono- or di-methylated, over un- or tri-methylated peptides.^27,42-49 An exception to this general rule is provided by the Caenorhabditis elegans protein LIN-61, which displays specificity toward di- and tri-methylated H3K9 histone peptides.⁵⁰ A recent study investigating the MBT domains of all 9 human family members has found that some MBT domains (L3MBTL1, L3MBTL3) recognize mono- and/or di-methyl-lysine in a promiscuous, non-sequence-specific fashion, whereas others (SCML2, L3MBTL4, MBTD1, L3MBTL2) specifically bind to only a few selected histone sequences.⁴⁹

Studies focusing on L3MBTL1 and L3MBTL2 have revealed a possible molecular mechanism by which MBT domains could contribute to transcriptional repression. Incubation of the MBT domains of L3MBTL1 or of full-length L3MBTL2 with oligo-nucleosomal arrays results in chromatin compaction in vitro.^14,15 Moreover, expression of recombinant full-length SFMBT1 or its MBT domains in cell lines reduces chromatin accessibility to nuclease digestion.²³ As we will detail below, MBT domain proteins are found in several multisubunit protein complexes (see Table 1 for a list of MBT and CoREST complexes discussed in this review). They provide these complexes with a tool (the MBT domains) for binding to mono- and di-methylated nucleosomes and for establishing repressive chromatin structures, potentially by way of nucleosome compaction.

MBT Complexes

In Drosophila the MBT domain proteins Scm and Sfmbt are associated with different multisubunit PcG complexes. Drosophila PRC1 contains the core subunits Polycomb (Pc), Polyhomeotic (Ph), Posterior sex combs (Psc) and Sex combs extra (Sce)/Ring1 ^12,51,52 (Table 1). Several observations suggest that Scm also associates with PRC1 albeit not as a stoichiometric subunit: Biochemical purifications of PRC1 contain substoichiometric amounts of Scm,^52,53 Scm interacts with the PRC1 subunit Ph in vitro ^37,54 and can stably assemble into a reconstituted PRC1 complex.⁵⁵ However, co-immunoprecipitations failed to confirm a robust interaction between Scm and PRC1 components and the bulk of Scm in embryo extract separates from Ph in gel filtration analysis.^27,55 Nevertheless, Scm repressive activity is sensitive to Ph dosage in reporter gene assays.³⁸ The exact nature of the PRC1-Scm interaction in vivo remains somewhat enigmatic. It is conceivable that this interaction is dynamically regulated, only takes place in certain scenarios, or stably occurs only in the chromatin-bound state.

In mammalian genomes each Drosophila PRC1 subunit is encoded by several homologous genes. This has made it difficult to decipher the composition and molecular function of mammalian PRC1. A comprehensive study has recently identified 6 distinct classes that constitute a multivalent family of human PRC1 ¹³ (Table 1). Each class consists of one of the 6 Pc homologs PCGF1 to PCGF6 (PRC1.1-1.6), H2AK119-specific monoubiquitin-E3-ligases RING1B/A and a specific set of additional factors. Among these complexes PRC1.2 and PRC1.4 contain PCGF1 and PCGF4, respectively, in addition to CBXs (Pc homologs), PHCs (Ph homologs) and the human Scm homologs SCMH1, SCML1 and SCML2 and therefore resemble the canonical Drosophila PRC1 complex, this time with MBT proteins as bona fide subunits. PRC1.6 contains a different MBT protein: L3MBTL2. Additional subunits of PRC1.6 are RING1A/B, the transcription factors E2F6/DP-1 and Mga/Max as well as HP1γ. E2F6/DP-1 and Mga/Max are believed to contribute to targeting the complex to specific sites. An independent study has isolated a very similar assembly incorporating the histone methyltransferases G9a and EuHMTase/GLP as additional subunits.⁵⁶ PRC1.6 complexes have both been suggested to repress E2F- and Myc-responsive genes in G0 quiescent cells and to repress transcription in actively dividing cells.^14,56 Thus, 3 members of the PRC1 complex family are equipped with MBT domains. L3MBTL1 is a component of another complex containing core histones, the linker histone H1B, HP1γ, and the retinoblastoma protein (pRB).¹⁵ In line with the pRB association, L3MBTL1 is bound to proximal promoter regions of E2F target genes where its presence contributes to their repression.^15,57

In contrast to Drosophila PRC1, canonical PhoRC comprises only 2 subunits, Sfmbt and Pleiohomeotic (Pho) or Pho-like.²⁷ The latter 2 are the only known PcG proteins that contain sequence-specific DNA binding activity and recognize polycomb response elements (PREs).^58,59 The closest vertebrate ortholog of Pho is the transcription factor Yin Yang-1 (YY1).⁶⁰ However, although YY1 and SFMBT2 co-immunoprecipitate upon overexpression in mammalian cells,⁶¹ they do not appear to form a stable complex.

Interestingly, a recent purification of Drosophila Sfmbt identified several associated polypeptides in addition to Pho: NAP1, HP1b, Rpd3 and Mga.⁶² The latter 3 are Drosophila homologs of PRC1.6 subunits.^13,56 This indicates that this larger PhoRC assembly (PhoRC-L) resembles human PRC1.6 complexes, but differs in the type of transcription factors targeting it.⁶²

Collectively, MBT complexes appear to function in Polycomb and E2F/Rb repression systems. Interestingly, MBT often coexist with heterochromatin protein 1 γ (HP1γ), core and linker histones. This reinforces the notion that MBT proteins contribute to nucleosome compaction and the formation of heterochromatin.

The CoREST Protein

The human CoREST family comprises 3 related proteins that are encoded by separate genes (CoREST1, CoREST2, and CoREST3). CoREST1 was first identified as an interaction partner and co-repressor protein of the RE1 silencing transcription factor/neural-restrictive silencing factor (REST/NRSF).⁶³ REST is crucial for the restriction of neuronal gene expression by silencing genes via binding to RE-1 elements in non-neuronal tissues.^64,65 Unlike MBT proteins, CoREST proteins do not contain a defining signature motif. Instead, they possess an ELM2 (EGL-27 and MTA1 homology 2)⁶⁶ and 2 SANT (SWI3/ADA2/N-CoR/TFIII-B)⁶⁷ domains that are shared with several other regulatory chromatin factors. SANT domains frequently act to stimulate histone-modifying enzymes by mediating interactions with their nucleosome substrates.⁶⁸ ELM2 domains were shown to interact with histone deacetylase in several studies.^69-71

CoREST proteins are integral subunits of chromatin modifying complexes that contribute to the generation of repressive chromatin. These conserved complexes combine histone demethylase and deacetylase activities and have been identified in Drosophila,⁷² C. elegans^73,74 and mammals.^16-19,75 CoREST1 complexes have been studied most extensively (Table 1). Their core consists of CoREST1, lysine-specific demethylase 1 (LSD1), histone deacetylases 1/2 (HDAC1/2) and the PHD finger protein BHC80. Thus, CoREST1 complexes combine 2 repressive histone modifying activities (LSD1 and HDAC1/2). Additional factors associated with the core that have not been identified in all studies include CtBP, BRAF35 and Zn-finger proteins such as ZNF217 (Table 1). ZNF217, a frequently amplified oncogene, is involved in the negative regulation of the p15ink4b tumor suppressor gene.⁷⁶ This is accomplished in concert with DNA methyltransferase 3A (DNMT3A), which prevents active DNA demethylation of the locus. This example illustrates that CoREST complexes coordinate several enzymatic activities (histone demethylase, histone deacetylase, DNA methyltransferase) to generate stable repressive chromatin structures.

A recent study revealed that even though CoREST family members CoREST1, CoREST2 and CoREST3 co-purify LSD1 equally well, enzymatic activities of these complexes differ significantly.^77,78 Thus, association of LSD1 with CoREST3 leads to diminished catalytic activity, that can antagonize the histone demethylation activity of CoREST1 complexes.^77,79 Therefore, it is plausible that the balance between homologous CoREST complexes and the active exchange of CoREST subunits could be important in different scenarios of cellular differentiation.^78,79

Structural studies have provided insight into how some of the CoREST complex subunits interact with each other and their chromatin substrate.^80-83 Importantly, the LSD1 enzyme alone demethylates H3K4 within peptides or bulk histones in vitro. However, only LSD1 in complex with CoREST is capable of catalyzing H3K4 demethylation within nucleosomes.^18,19 Deletion studies revealed that the major stimulatory activity for nucleosome demethylation comprises the second CoREST SANT domain (SANT2) and the linker connecting the 2 SANT domains.¹⁹

There are interesting parallels and differences between the functions that MBT domain and CoREST proteins play in their respective complexes. Both proteins are non-enzymatic components of repressive histone-modifying complexes. They both function as tools that directly contact the nucleosome substrates. In the case of MBT proteins, this can have a direct consequence on chromatin structure by promoting nucleosome compaction. In the case of CoREST proteins, the effect is more indirect and stimulates the nucleosome-modifying activities of enzymatic complex subunits.

It is interesting to note that CoREST complexes lower the methylation state of nucleosomes and that MBT proteins preferentially bind to undermethylated (mono- or di-methylated) histone tails. It is tempting to speculate that the action of CoREST complexes increases the affinity of chromatin binding of MBT complexes and that both types of complexes act subsequently. However, as we will discuss in the following section, the link between these complex families can be much more direct.

MBT and CoREST Proteins in Common Complexes

As described above, MBT and CoREST proteins are subunits of distinct protein complex families. However, recent studies in human and Drosophila have identified novel assemblies that combine MBT and CoREST proteins ^20-23 (Table 1). Human SFMBT1 was isolated in a complex with LSD1 and CoREST (SLC complex) and the remaining subunits of known CoREST complexes.²⁴ SLC complex subunits are stably associated in both human cells, as well as mouse testes, where SFMBT1 expression levels are highest compared to other tissues.²³ The recently identified Drosophila LINT complex also combines MBT and CoREST proteins. LINT has 3 core subunits: dL(3)mbt, dLint-1 (Drosophila L(3)mbt interacting protein-1), and dCoREST and is expressed in cell lines, embryos and larval brain.²⁰ During chromatographic fractionation LINT can be separated from Drosophila dLsd1-containing assemblies. Therefore, LINT, unlike SLC, contains dCoREST disconnected from its usual interaction partner dLsd1. The direct interaction between LSD1 and CoREST and its function in boosting the LSD1 demethylase activity has been well established.⁸² It remains to be seen what the role of dCoREST within the LINT complex is. It seems likely, that CoREST regardless of the presence or absence of LSD1 facilitates nucleosome binding and may promote a stronger interaction of MBT domains with modified nucleosomes in vivo.

Both SLC and LINT bind preferentially in the vicinity of transcription start sites (TSSs), supporting a major role of both complexes in transcriptional regulation.^20,23 Roughly one third of SLC bound targets, are enriched for genes with a function in chromatin and nucleosome assembly, among them replication-dependent histone gene clusters.²³ The latter constitute multiple gene copies encoding canonical histones that need to be transcribed efficiently during S-phase.^83,84 Interestingly, the occupancy of SLC complex at histone genes appeared to be cell cycle dependent and an enrichment of RNA polymerase II at the promoters during S-phase was accompanied by a loss of all 3 SLC complex components.²³ In addition, chromatin binding of SLC to target genes was also developmentally regulated during spermatogenesis.

In flies, 2 studies identified genes that are misexpressed in brain tumors that are caused by l(3)mbt^ts mutation and contribute to tumorigenic overgrowth.^85,86 These genes have been termed malignant brain tumor signature (MBTS) genes.⁸⁵ A significant proportion of MBTS genes encode proteins with a function in the germline. In addition, a group of target genes downstream of the Salvador-Warts-Hippo (SWH) signaling pathway are upregulated in l(3)mbt mutant brain tumors.⁸⁶ Chromatin immunoprecipitations reveal a direct binding of LINT subunits to the promoter regions of a majority of MBTS germline and SWH target genes that contribute to tumorigenesis in l(3)mbt mutant larval brains.^20,86

Collectively, these studies suggest that complexes combining MBT and CoREST proteins associate with and repress developmentally and cell cycle-regulated genes.

The identification of SLC and LINT weakens the boundaries between MBT-containing PcG complexes and CoREST complexes (Fig. 3). Intriguingly, LSD1 seems to reach out for yet another protein complex family traditionally viewed as separate from both CoREST and PcG complexes: Nucleosome remodeling and histone deacetylase (NuRD) complexes combine the nucleosome remodeling activity of CHD3/CHD4 ATPases with the histone deacetylase activities of HDAC1 and HDAC2. They play important roles in DNA damage repair ^87-89 and in generating repressive chromatin environments during development and differentiation.^83,86,90,91 NuRD complexes have been the subject of several excellent reviews recently.^92-96 LSD1 was found to associate with NuRD subunits in a number of recent studies. This novel assembly functions in the regulation of breast cancer metastasis and in the decommissioning of developmentally regulated enhancers during stem cell differentiation.^97-99 These manifold connections between protein complex families once viewed as distinct from one another illustrate the dynamics of versatile protein assemblies that regulate fundamental cellular processes.

Concluding Remarks and Perspectives

In the past decade chromatin regulation in diverse biological processes has been intensively studied. This has unraveled an astounding degree of complexity concerning both the mechanisms and the factors employed. One central theme that has emerged is that chromatin regulators often act in the context of multisubunit protein complexes.

Recent findings suggest that these chromatin regulating complexes - particularly in higher eukaryotes - exist in multiple versions and form complex families instead of single entities. Accordingly, paradigmatic complexes such as PRC1, which were originally viewed as a single entity, have now been revealed to exist in many different versions and to constitute a protein complex family.^13,100 Furthermore, variations in the stoichiometry of subunit composition of several of the newly identified complexes indicates that they can be further sub-divided ^13,18,19,56 (Table 1). Metaphorically speaking, multisubunit protein complexes function as “toolboxes” which in order to accomplish specific tasks on chromatin combine different “tools” or up-grade their standard equipment by adding specific “tools." For example, distinct PRC1 subcomplexes can couple H2A ubiquitination with various other activities, such as the H3K36me3 and H3K4me3 demethylase activity of KDM2B (in PRC1.1), H3K27me3 binding by CBXs (in PRC1.2/1.4) or chromatin compaction through HP1γ and L3MBTL2 (in PRC1.6). The recently purified SLC and LINT assemblies provide another example of combining different tool sets (MBT domains and CoREST/LSD1) enabling them to use novel strategies to silence genes.

The increasing number of multisubunit protein assemblies that regulate chromatin structure that we have discussed above raises more general questions that go beyond the field of chromatin biology. These questions are not new but in the face of an increasingly overwhelming number of related complexes and subcomplexes they acquire new relevance.

What exactly defines a “complex subunit” as opposed to an associated polypeptide? And to which extent is our use of the term “protein complex” an accurate reflection of the dynamic protein assemblies existing within cells? In the classical, biochemical view protein complex subunits form a stable, interlocking assembly that is resistant to the relatively high salt concentrations commonly used during the purification of protein complexes. However, it is clear that many protocols that are used to lyse cells and tissues and isolate proteins from extracts are prone to disrupt bona fide protein interactions that exist in vivo. In addition, in vivo protein complexes will engage in transient interactions that form and disintegrate continuously.

Many of the apparent discrepancies in the current literature concerning subunit composition and identity of protein complexes might result from differences in the preparation of extracts and purification procedures. In addition, the ever-increasing sensitivity of protein identification by mass spectrometry techniques accounts for the identification of novel subunits that were missed in previous approaches.

Currently, we no longer think of single complexes (e.g., PRC1) but rather of complex families (e.g., PRC1.1-1.6). As the size of these complex families increases and as once distinct protein families merge the classic idea of the protein complex as a defined entity carrying out a defined set of functions becomes less and less useful (Fig. 3). It appears more sensible to think of a continuum of protein complexes whose exact composition and function is dynamic and context-dependent. The protein assemblies that survive our purification regimens are those that are particularly stable and abundant. Undoubtedly, as our purification procedures and protein identification methods become more and more refined we will be able to see more and more of the continuum that currently is still largely hidden from our view.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We apologize to all colleagues whose work could not be cited due to shortage of space and the limited scope of our article. Work in the lab of A.B. is supported by the DFG (BR 2102/6). We are grateful to members of the F. Recillas-Targa research group for discussion.

Funding

K.M. is supported by a postdoctoral fellowship from the DGAPA-UNAM.