SAND domain (AIRE/SP100/GMEB)
The SAND domain is a small DNA-binding fold named after the four proteins in which it was first recognized: Sp100, AIRE, NucP41/75 and DEAF1. Sequence-specific binding depends on a conserved KDWK motif on the domain's surface, and only some SAND proteins have it. The best-defined members are GMEB1 and GMEB2, which bind as a heterodimer, also known as parvovirus initiation factor, to two CpG-containing tetranucleotide half-sites with flexible spacing: ACGY-N(1-9)-RCGY, with an optimum spacer of six bases. Binding is abolished by methylation of the CpG in either half-site. AIRE and the speckled proteins SP100 and SP140 lack the KDWK motif and are recruited through chromatin rather than sequence, so motif models attributed to them should be treated as descriptions of where they were found.
The factors
SAND is a fold, not a functional group. The domain was named from Sp100, AIRE-1, NucP41/75 and DEAF1, and its NMR structure showed a novel alpha-beta fold whose DNA-binding activity maps to a conserved surface motif, KDWK, which gives the family its alternative name 11427895,10523663. Human SAND proteins fall into distinct groups. GMEB1 and GMEB2 (p96 and p79) were identified twice over: as the heterodimer that binds the glucocorticoid modulatory element upstream of the tyrosine aminotransferase gene 7665613, and as parvovirus initiation factor, the host factor required for minute virus of mice origin nicking 10523663. They carry the KDWK motif, bind DNA cooperatively, and are ubiquitously expressed 7665613,10523663,11743720. DEAF1 (NUDR) is a related KDWK protein that binds TTCG-containing repeats 9773984; notably, it has no motif in any of the four databases used here despite being named in this family's definition. The speckled proteins SP100, SP110, SP140 and SP140L are nuclear-body components that combine a SAND domain with PHD fingers and bromodomains and act as chromatin readers in immune cells 32386862. AIRE, the autoimmune regulator, is the most important special case: its SAND domain lacks the KDWK residues required for sequence-specific DNA contact and instead mediates homodimerization, while its two PHD fingers read histone H3, PHD1 binding H3 tails that are unmethylated at lysine 4 18292755. AIRE is essentially restricted to medullary thymic epithelial cells 12376594.
The motif
For GMEB1/GMEB2 the site is well defined and unusual. Degenerate oligonucleotide selection showed that the heterodimer, and each homodimer, binds two tetranucleotide half-sites conforming to ACGY-N(1-9)-RCGY, with an optimum spacer of six bases 11743720; the spacing is genuinely flexible, so the element is a pair of short half-sites at variable distance rather than a fixed-width motif 10523663,11743720. The invariant feature is the CpG in each half-site, and 5-methylation of those cytosines abolishes binding completely 11743720, which makes GMEB an unusual example of a factor whose occupancy is directly methylation-sensitive at its core. Because CpG is suppressed in vertebrate genomes, such sites should be rare, yet over half of a sample of 100 human promoters contained at least one, often within 100 bases of the start site 11743720. The models in these databases capture only part of this grammar. Most GMEB matrices collapse the pair into a single half-site region: the representative model here is TACGTAA, and JASPAR's GMEB1 and GMEB2 matrices are tACGTa and TACGTAA. One CIS-BP model, M05735_2.00, does show the doubled form (TACGTACGGTACGTA), which is the two half-sites resolved. The consequence for scanning is severe: a six- or seven-base ACGT-centered model sits in the middle of the sequence space used by bZIP factors, since TACGTA, TGACGT and CACGTG all contain ACGT 10523663, and the data reflect this directly. Two GMEB1 models have a higher similarity to their best out-of-family neighbour than to any in-family motif, and the family's cohesion is only 0.70 across 15 motifs. The remaining models are not ACGT elements at all: the ChIP-derived GMEB models are GCGTCAGCGCG-like, the HOCOMOCO SP100 model is AACGGTTA and the SP140 model is GTCGTCACTA, each with zero label agreement, and the Vierstra AIRE archetype (AC0160) is a CCAAT-containing element rather than anything resembling a SAND site.
Where it acts
GMEB sites are promoter-proximal and, by construction, CpG-containing, so they concentrate in CpG-island promoters where their methylation sensitivity could make them responsive to DNA methylation state. The glucocorticoid modulatory element itself modulates the sensitivity of a glucocorticoid-responsive promoter rather than driving it 7665613, so GMEB acts as a modifier of another factor's output. AIRE acts in a completely different regime: in medullary thymic epithelial cells it associates with super-enhancers 28135252 and with thousands of promoters of tissue-restricted antigen genes, and its principal mechanism is releasing stalled RNA polymerase II rather than recruiting it to a specific sequence 22203960. SP140 occupies and maintains repressed loci in macrophages, including lineage-inappropriate developmental genes 32386862.
Biology and disease
Biallelic loss-of-function mutations in AIRE cause autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy (APECED, also APS-1), a monogenic autoimmune disease 9398839; Aire-deficient mice lose promiscuous expression of thousands of tissue-restricted antigens in the thymus and develop multi-organ autoimmunity, which established the central-tolerance model of negative selection 12376594. Missense mutations in the SAND domain of DEAF1 cause intellectual disability with severe speech impairment and behavioral problems 24726472. Variants at SP140 and SP110 are associated with Crohn disease, multiple sclerosis and chronic lymphocytic leukemia, and SP110 mutations cause hepatic venoocclusive disease with immunodeficiency 32386862. GMEB proteins have no established monogenic disease link, but they are required for parvovirus DNA replication 10523663.
Reading a hit
The central caution is that this key mixes proteins that read sequence with proteins that do not. A match to a GMEB model is a real, if weak, claim: look for the second CpG half-site within nine bases and check the methylation state, because an unmethylated CpG is part of the site rather than incidental. A match to an AIRE, SP100 or SP140 model is much harder to interpret. The three are not alike. AIRE's SAND domain lacks the KDWK residues, and its only motif here is a Vierstra archetype that reads as a CCAAT-type element. SP100's SAND domain does carry KDWK, since the SAND structure itself was solved from Sp100b 11427895, and the SP100 and SP140 models are grade A in HOCOMOCO with in vitro selection support (HT-SELEX, GHT-SELEX and SMiLE-seq, plus PBM for SP140) as well as ChIP. Those two are measurements of direct binding, even though their nearest neighbours lie in other families. The general risk for the ACGT-centered GMEB models is confusion with bZIP sites, which share the same four-base core and are far more abundant 10523663,7665613; a TACGTA match in a promoter is more likely to be a CRE-like or MAF-like element than a GMEB site.
How this was worked out
The factor binding the glucocorticoid modulatory element (GME) far upstream of the tyrosine aminotransferase gene is a novel, ubiquitous complex of two proteins, GMEB1 and GMEB2, distinct from the CREB/ATF proteins that also bind the element.
Still open then: The proteins' sequences, family and recognition rules were unknown.
Parvovirus initiation factor is cloned as the p96/p79 heterodimer (GMEB1/GMEB2), placed in the KDWK family, and shown to bind two PuCGPy half-sites that can be spaced 1 to 15 nucleotides apart.
Still open then: The exact consensus and spacing optimum.
NMR structure of the Sp100b SAND domain defines a new alpha/beta DNA-binding fold, with the conserved KDWK motif on a positively charged helical surface that mutagenesis and chemical-shift mapping identify as the DNA-binding surface; the NUDR (DEAF1) SAND domain alone suffices for DNA binding.
Selection and dissociation-rate analysis define the GMEB consensus as ACGY-N(1-9)-RCGY with an optimal spacer of six; CpG methylation of the half-sites abolishes binding, yet over half of 100 human promoters carry a site.
Still open then: Whether methylation regulates GMEB occupancy in cells.
AIRE is reported to be a sequence-specific DNA-binding protein whose dimers and tetramers bind ATTGGTTA and TTATTA motifs.
Still open then: The motifs were not tied to AIRE target genes in thymic epithelium.
Aire-deficient mice lose ectopic expression of peripheral tissue-restricted antigens in medullary thymic epithelial cells and develop autoimmunity that depends on Aire in thymic stroma, establishing AIRE's role in central tolerance.
In medullary thymic epithelial cells AIRE binds essentially all Pol II-rich start sites and activates its targets by releasing stalled RNA polymerase II, not by recognizing specific promoter sequences. This is at odds with the earlier report of sequence-specific AIRE binding.
Still unresolved
- Whether CpG methylation regulates GMEB binding in cells is open: methylating the half-site CpGs abolishes binding in vitro, but the proposal that this controls occupancy in vivo rests on that biochemistry.
- Whether AIRE has any intrinsic sequence preference is disputed: an early in vitro study reported AIRE oligomers binding ATTGGTTA and TTATTA motifs, and the Vierstra AIRE archetype AC0160 carries ATTGGTTA on its reverse strand, whereas genome-wide work finds AIRE at essentially all Pol II-rich start sites, acting by releasing stalled polymerase rather than by sequence recognition.
- Whether SP100 and SP140 bind DNA directly through their SAND domains is unsettled: their HOCOMOCO models are grade A and supported by in vitro assays (HT-SELEX, GHT-SELEX and SMiLE-seq, plus PBM for SP140) as well as ChIP, yet the motifs (AACGGTTA, GTCGTCACTA) share little with the GMEB half-site beyond a CpG, and their nearest neighbours lie in other families.
- The permitted GMEB half-site spacing differs between studies: the cloning work reported spacings of 1 to 15 nucleotides, whereas selection combined with dissociation-rate analysis gave 1 to 9 with an optimum of six.
Split this family in two before trusting anything. GMEB1 and GMEB2 have the KDWK motif and a real, if unusual, site: two CpG-containing half-sites with a 1-9 base spacer, abolished by CpG methylation. But the matrices compress this to a six- or seven-base ACGT core that overlaps bZIP space, and two GMEB1 models here match an out-of-family motif better than any relative. AIRE is the real exception: its SAND domain lacks the KDWK residues and its only model is a Vierstra archetype that reads as a CCAAT-type element. The SP100 and SP140 models are grade A with in vitro selection support, so they do record sequences those proteins bind, but their nearest neighbours lie in other families. Treat the AIRE model as a record of where AIRE was found rather than of a sequence it reads. DEAF1 has no motif here at all.
References
- Bottomley MJ, Collard MW, Huggenvik JI, et al. The SAND domain structure defines a novel DNA-binding fold in transcriptional regulation. Nat Struct Biol 2001;8:626-33. PMID 11427895
- Christensen J, Cotmore SF, Tattersall P. Two new members of the emerging KDWK family of combinatorial transcription modulators bind as a heterodimer to flexibly spaced PuCGPy half-sites. Mol Cell Biol 1999;19:7741-50. PMID 10523663
- Burnett E, Christensen J, Tattersall P. A consensus DNA recognition motif for two KDWK transcription factors identifies flexible-length, CpG-methylation sensitive cognate binding sites in the majority of human promoters. J Mol Biol 2001;314:1029-39. PMID 11743720
- Oshima H, Szapary D, Simons SS Jr. The factor binding to the glucocorticoid modulatory element of the tyrosine aminotransferase gene is a novel and ubiquitous heteromeric complex. J Biol Chem 1995;270:21893-901. PMID 7665613
- Huggenvik JI, Michelson RJ, Collard MW, et al. Characterization of a nuclear deformed epidermal autoregulatory factor-1 (DEAF-1)-related (NUDR) transcriptional regulator protein. Mol Endocrinol 1998;12:1619-39. PMID 9773984
- Vulto-van Silfhout AT, Rajamanickam S, Jensik PJ, et al. Mutations affecting the SAND domain of DEAF1 cause intellectual disability with severe speech impairment and behavioral problems. Am J Hum Genet 2014;94:649-61. PMID 24726472
- Nagamine K, Peterson P, Scott HS, et al. Positional cloning of the APECED gene. Nat Genet 1997;17:393-8. PMID 9398839
- Anderson MS, Venanzi ES, Klein L, et al. Projection of an immunological self shadow within the thymus by the aire protein. Science 2002;298:1395-401. PMID 12376594
- Org T, Chignola F, Hetényi C, et al. The autoimmune regulator PHD finger binds to non-methylated histone H3K4 to activate gene expression. EMBO Rep 2008;9:370-6. PMID 18292755
- Giraud M, Yoshida H, Abramson J, et al. Aire unleashes stalled RNA polymerase to induce ectopic gene expression in thymic epithelial cells. Proc Natl Acad Sci U S A 2012;109:535-40. PMID 22203960
- Bansal K, Yoshida H, Benoist C, et al. The transcriptional regulator Aire binds to and activates super-enhancers. Nat Immunol 2017;18:263-273. PMID 28135252
- Fraschilla I, Jeffrey KL. The Speckled Protein (SP) Family: Immunity's Chromatin Readers. Trends Immunol 2020;41:572-585. PMID 32386862
- Kumar PG, Laloraya M, Wang CY, et al. The autoimmune regulator (AIRE) is a DNA-binding protein. J Biol Chem 2001;276:41357-64. PMID 11533054
15 motifs in this family — JASPAR 2 · HOCOMOCO 8 · CIS-BP 2 · Vierstra 3
| motif | database | consensus | bits |
|---|---|---|---|
| M05735_2.00 | CIS-BP | cTACGTAcggTACGTA | 21.2 |
| AC0268:GMEB:SAND | Vierstra | ACGTGgTtACGT | 12.4 |
| GMEB2.H14CORE.2.SM.BSMB | HOCOMOCO | gTACGTaAcG | 12 |
| GMEB1.H14CORE.2.SM.BSMB | HOCOMOCO | TACGTAA | 11.7 |
| GMEB2.H14CORE.0.P.CPC | HOCOMOCO | GCGTcagcGCG | 10.3 |
| AC0160:AIRE:SAND | Vierstra | taaCCaatawaACC | 10.1 |
| GMEB1.H14CORE.1.P.CPC | HOCOMOCO | gCGTcagcGCGc | 10 |
| GMEB2 (MA0862.1) | JASPAR | TACGTAA | 9.9 |
| SP140.H14CORE.0.PSGIB.APSGIBA | HOCOMOCO | GtCGtCAcTa | 9.5 |
| GMEB2.H14CORE.1.P.CPC | HOCOMOCO | tGCGTgag | 9.2 |
| GMEB1.H14CORE.0.P.CPC | HOCOMOCO | TGCGTg | 9 |
| SP100.H14CORE.0.PSGI.APSGIA | HOCOMOCO | aaCGGTtA | 9 |
| AC0272:GMEB:SAND | Vierstra | tACGTA | 8.9 |
| Gmeb1 (MA0615.2) | JASPAR | tACGTa | 7.2 |
| M02460_2.00 | CIS-BP | aCGt | 5 |
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.