REST (NRSF)
REST, also called NRSF, is a single zinc-finger protein that recognizes RE1 (the neuron- restrictive silencer element), a 21-base-pair site that is one of the longest and most information- rich transcription-factor motifs in the human genome. Eight of its nine zinc fingers read two half-sites separated by a two-base spacer, which is why the element is long and unusually specific. REST recruits SIN3-HDAC and CoREST-LSD1 complexes to silence neuronal genes in non-neuronal cells, and is removed or degraded as neurons differentiate. Human and mouse genomes contain roughly one to two thousand canonical RE1s, many spread by transposable elements, plus a larger set of weaker non-canonical variants. Unusually for a PWM, a strong RE1 match is good evidence of real REST binding.
The factors
This family has one member. REST (RE1-silencing transcription factor) was identified twice in 1995: as NRSF, a coordinate repressor of multiple neuron-specific genes 7871435, and as REST, a silencer protein that restricts type II sodium channel expression to neurons 7697725. It contains nine C2H2 zinc fingers, eight of which form the DNA-binding array, plus a separate C-terminal finger 25092546. REST is not itself a catalytic repressor; it carries two repression domains that dock different corepressor machinery, an N-terminal region that binds SIN3A with HDAC1 and HDAC2, and a C-terminal region that binds CoREST/RCOR1, which in turn brings the demethylase LSD1/KDM1A and associated activities such as G9a, CTDSP1 and MeCP2 25092546. REST is expressed broadly in non-neuronal tissue and in neural stem cells, and its protein level falls sharply at neuronal differentiation 25092546. A truncated splice variant, REST4, lacks most of the finger array. There is no paralogue in these databases, so the family and the protein coincide.
The motif
The canonical RE1/NRSE is written TTCAGCACCACGGACAGCGCC, 21 base pairs 7871435,7697725. It is best read as two half-sites, a left half around TTCAGCACC and a right half around GGACAGCGCC, separated by a two-nucleotide spacer that carries almost no information. Separate groups of zinc fingers engage the two halves, which is the structural reason a single monomer can specify 19 informative positions 33849071; by contrast most vertebrate factors specify six to ten. The spacer explains a discrepancy readers will notice: the canonical string has AC at those two positions, while the JASPAR and HOCOMOCO matrices show a weak preference for at, and the data consensus for this family is TCAGCACCatGGACAGCgcC. These are the same element, differing only where the model is uninformative. Non-canonical RE1s are common and functionally important. They include half-site RE1s, in which only the left or right half is present; compound elements with expanded spacers; and single-substitution variants throughout 17581960. Binding-affinity work shows that canonical motifs give strong occupancy at targets shared by all cell types, whereas atypical motifs give weak occupancy at cell- and tissue-specific targets, and, unexpectedly, that the weak sites are the most evolutionarily constrained 19401398.
Where it acts
RE1 elements sit in promoters, first introns and distal regions of neuronal genes: ion channels, synaptic vesicle proteins, neurotransmitter receptors, neuronal microRNAs and splicing factors. Position-specific scoring searches of the human genome recover on the order of 1,300 canonical RE1s (about 1,000 in mouse) 16899447, and a comparative-genomics model built from conserved occurrences produced 842 high-confidence human sites at 733 genes 16963704; larger surveys that admit weaker matches reach a few thousand 15240883,17581960. The number is small for a vertebrate motif, which is the practical consequence of the element's length. RE1s are present in all sequenced vertebrate genomes and absent from invertebrates and protochordates, as is REST itself. A substantial fraction lie in repeats, and many RE1s fall into closely related families associated with transposable elements, implying that transposon- mediated duplication carried the element to new target genes during vertebrate evolution 16899447. The REST cistrome is not conserved in detail: it is substantially larger in human than in mouse embryonic stem cells 25990720.
Biology and disease
REST silences neuronal genes in non-neuronal cells and keeps them repressed in progenitors, releasing them as neurons mature. Its regulation is therefore mostly post-transcriptional, through splicing to REST4 and through ubiquitin-dependent degradation at differentiation. The first mammalian ChIP-sequencing experiment was performed on NRSF 17540862, and the resulting maps established the tight coupling between the motif and occupancy that makes this factor unusual.
Several diseases turn on REST dosage. Wild-type huntingtin sequesters REST in the cytoplasm; mutant huntingtin fails to, so REST enters the nucleus, occupies RE1s and represses target genes including BDNF, contributing to Huntington disease pathology 12881722. In the ageing human cortex, nuclear REST rises and is required for stress resistance, and is lost in Alzheimer disease 24670762. Germline truncating REST mutations predispose to Wilms tumor 26551668, and REST loss or degradation has been reported in epithelial tumors, so it behaves as a tumor suppressor in some contexts while its gain of function drives pathology in others.
Reading a hit
This is the family where a PWM hit means the most. With about 19 informative positions, a full-length canonical RE1 match essentially does not occur by chance in a three-gigabase genome, and the great majority of strong matches are occupied wherever REST is expressed. There is no paralogue problem: the motif belongs to one protein. Two caveats remain. First, half-site and other non-canonical RE1s score poorly and will be missed by a threshold tuned to the canonical element 36777181,17581960, yet they carry the cell-type-specific part of the regulon, so a negative result at a neuronal gene is weak evidence. Second, occupancy is not the same as repression: REST is bound at many sites in cells where the target is expressed anyway, and the functional output depends on which corepressor complex is recruited and on REST protein level, which falls in neurons. Because the element is long and distinctive, ChIP-derived REST models are relatively clean, and an RE1-like motif appearing in another factor's ChIP data usually means that factor is part of a REST corepressor complex.
How this was worked out
The same protein is reported twice in one year, as NRSF, a repressor acting on a shared element in many neuron-specific genes, and as REST, which restricts type II sodium channel expression to neurons; both define the 21-base RE1/NRSE.
Still open then: How many RE1s exist in the genome, and whether the element was always this long, were unknown.
A combined computational and biochemical survey catalogs 1,892 human and 1,894 mouse RE1s and shows an in vivo hierarchy: some sites are occupied at endogenous REST levels and others only when REST is raised.
Still open then: Which of the predicted sites are occupied in any given cell type.
- 2006mediumPSSM genome search plus BLAST clustering; comparative-genomics model trained on conserved sites16899447 16963704
Two independent catalogs converge on roughly 1,000-1,300 canonical RE1s per mammalian genome, far fewer than for a typical vertebrate motif, and show that many RE1s belong to sequence families carried by transposable elements.
Still open then: Whether transposon-borne RE1s are functional silencers or passengers.
Serial analysis of chromatin occupancy in a kidney cell line recovers far more REST-occupied sites than the canonical catalogs, and defines non-canonical motifs: half-site RE1s and versions with expanded spacers.
Still open then: Whether non-canonical sites are bound with enough affinity to repress, or are low-occupancy background.
NRSF is the subject of the first mammalian ChIP-sequencing experiment, producing genome-scale occupancy maps whose peaks align tightly with RE1 matches, which is why this motif became the standard demonstration that sequence can predict binding.
- 2009mediumQuantitative ChIP across cell types with affinity ranking, comparative sequence analysis19401398
Measured in vivo binding affinities split the regulon: canonical RE1s are strongly occupied at targets shared across cell types, while atypical low-affinity sites give cell-type-restricted occupancy, and those weak sites are the most evolutionarily constrained.
Still open then: Why selection would preserve weak sites more strongly than strong ones.
The REST cistrome in human embryonic stem cells is substantially larger than in mouse, so the regulon defined by occupancy is not conserved in detail even where the motif is.
- 2023mediumModified massively parallel reporter assay (MPRAduo) across RE1s and 1,500 variants36777181
A massively parallel reporter assay tests RE1s directly for silencing and finds a minimum REST binding strength below which cofactors do not assemble, and that non-canonical motifs silence only when their half sites are aligned at particular spacings, qualifying the view that any half-site RE1 is a functional element.
Still open then: The assay measures episomal silencing, so which of these elements silence in their native chromatin is still untested.
Still unresolved
- How much of the non-canonical RE1 repertoire is functional is contested: affinity-based work argues that weak, atypical sites carry the cell-type-specific part of the regulon and are the most conserved, whereas reporter-based testing finds that non-canonical motifs silence only when their half sites sit at particular spacings.
- Why the weakest RE1s are the most evolutionarily constrained has no accepted explanation; one possibility is that weak occupancy is itself the regulated quantity, another that these sites are constrained for reasons unrelated to REST.
- Whether the much larger human than mouse REST cistrome reflects ongoing transposon-mediated expansion of the RE1 repertoire, or differences in cell state and assay sensitivity between the compared datasets, is unresolved.
- What determines whether an occupied RE1 is actually repressed is unsettled, since REST is bound at many sites in cells where the target gene is expressed, and the corepressor complex recruited is not predictable from the site.
This is the best case for trusting a PWM hit: a full canonical 21-bp RE1 has roughly 19 informative positions, has no paralogue competing for the label, and is rare enough that chance matches are negligible. The failure mode is the opposite of usual: thresholds set for the canonical element miss half-site and other non-canonical RE1s, which are numerous, weakly bound and carry the tissue-specific part of the regulon. A hit also does not imply repression, since REST occupies sites in cells where the target is expressed and its protein level and corepressor choice determine the output. Conversely, an RE1-like motif recovered from another factor's ChIP data usually means that factor sits in a REST corepressor complex.
References
- Schoenherr CJ, Anderson DJ. The neuron-restrictive silencer factor (NRSF): a coordinate repressor of multiple neuron-specific genes. Science 1995;267:1360-3. PMID 7871435
- Chong JA, Tapia-Ramírez J, Kim S, et al. REST: a mammalian silencer protein that restricts sodium channel gene expression to neurons. Cell 1995;80:949-57. PMID 7697725
- Johnson R, Gamblin RJ, Ooi L, et al. Identification of the REST regulon reveals extensive transposable element-mediated binding site duplication. Nucleic Acids Res 2006;34:3862-77. PMID 16899447
- Mortazavi A, Leeper Thompson EC, Garcia ST, et al. Comparative genomics modeling of the NRSF/REST repressor network: from single conserved sites to genome-wide repertoire. Genome Res 2006;16:1208-21. PMID 16963704
- Bruce AW, López-Contreras AJ, Flicek P, et al. Functional diversity for REST (NRSF) is defined by in vivo binding affinity hierarchies at the DNA sequence level. Genome Res 2009;19:994-1005. PMID 19401398
- Bruce AW, Donaldson IJ, Wood IC, et al. Genome-wide analysis of repressor element 1 silencing transcription factor/neuron-restrictive silencing factor (REST/NRSF) target genes. Proc Natl Acad Sci U S A 2004;101:10458-63. PMID 15240883
- Johnson DS, Mortazavi A, Myers RM, et al. Genome-wide mapping of in vivo protein-DNA interactions. Science 2007;316:1497-502. PMID 17540862
- Rockowitz S, Zheng D. Significant expansion of the REST/NRSF cistrome in human versus mouse embryonic stem cells: potential implications for neural development. Nucleic Acids Res 2015;43:5730-43. PMID 25990720
- Zuccato C, Tartari M, Crotti A, et al. Huntingtin interacts with REST/NRSF to modulate the transcription of NRSE-controlled neuronal genes. Nat Genet 2003;35:76-83. PMID 12881722
- Lu T, Aron L, Zullo J, et al. REST and stress resistance in ageing and Alzheimer's disease. Nature 2014;507:448-54. PMID 24670762
- Mahamdallie SS, Hanks S, Karlin KL, et al. Mutations in the transcriptional repressor REST predispose to Wilms tumor. Nat Genet 2015;47:1471-4. PMID 26551668
- Thiel G, Ekici M, Rössler OG. RE-1 silencing transcription factor (REST): a regulator of neuronal development and neuronal/endocrine function. Cell Tissue Res 2015;359:99-109. PMID 25092546
- Otto SJ, McCorkle SR, Hover J, et al. A new binding motif for the transcriptional repressor REST uncovers large gene networks devoted to neuronal functions. J Neurosci 2007;27:6729-39. PMID 17581960
- Tang Y, Jia Z, Xu H, et al. Mechanism of REST/NRSF regulation of clustered protocadherin α genes. Nucleic Acids Res 2021;49:4506-4521. PMID 33849071
- Mouri K, Dewey HB, Castro R, et al. Whole-genome functional characterization of RE1 silencers using a modified massively parallel reporter assay. Cell Genom 2023;3:100234. PMID 36777181
4 motifs in this family — JASPAR 1 · HOCOMOCO 1 · CIS-BP 1 · Vierstra 1
| motif | database | consensus | bits |
|---|---|---|---|
| REST (MA0138.3) | JASPAR | TCAGCACCatGGACAGCgcC | 22.8 |
| AC0137:REST:C2H2_ZF | Vierstra | CTGTCCatGGTGCTG | 19.3 |
| REST.H14CORE.0.P.BPB | HOCOMOCO | tCAGcACCatGGACAGCgCC | 18.6 |
| M07846_2.00 | CIS-BP | TCCrtGGTGCTGAA | 18.1 |
HOCOMOCO tags read as assay + quality grade: P ChIP-seq, S HT-SELEX, M methyl-HT-SELEX, G GHT-SELEX, I SMiLE-seq, B PBM; grade A means two or more assay types agreed. A P-only matrix is a record of where the protein was crosslinked, which need not be a sequence it binds itself.
The factors, elsewhere
These entries are written from the primary literature and every citation is checked against PubMed, but this one can still be out of date, too confident, or simply wrong — especially where a motif’s published site and its database matrices disagree. Corrections and additions are welcome, and they are the fastest way to make this page better for everyone.