Motif Tools

SOX (HMG box)

SoxAACAAT87 motifsaka HMG-box site, SRY box, Sox element
TCTATTGTGACT
representative: SOX10.H14CORE.1.PSM.A · data consensus CATTGTT
members tight · 0.78stated AACAAT · data CATTGTT27 in vitro9 ChIP-only2 cross-species12 references, all PMIDs verified

SOX proteins are the 20 human transcription factors built around a single SRY-type HMG box, a 78-residue L-shaped domain that binds DNA entirely in the minor groove, wedges a hydrophobic residue between base pairs and bends the double helix by 70-80 degrees. They read a six- to seven-base core usually written AACAAT, or CATTGTT/ATTGTT on the opposite strand. A SOX protein bound alone does little; transcriptional output comes from a partner factor on an adjacent site, most famously OCT4 in the SOX-OCT composite element of pluripotency enhancers, and from dimeric palindromic sites used by SOX8, SOX9 and SOX10. The family covers sex determination (SRY, SOX9), cartilage (SOX9, SOX5, SOX6), neural stem cells (SOX2), neural crest and melanocytes (SOX10) and endothelium (SOX17, SOX18).

The factors

Humans have 20 SOX genes: SRY (group A), SOX1/2/3 (B1), SOX14/21 (B2), SOX4/11/12 (C), SOX5/6/13 (D), SOX8/9/10 (E), SOX7/17/18 (F), SOX15 (G) and SOX30 (H). Each carries one HMG box of about 78 residues folded into a twisted L: three alpha-helices whose concave face lies against DNA. Unlike almost all other transcription factors, recognition happens entirely in the minor groove. Side chains intercalate between base pairs - isoleucine 13 in SRY - widening the groove, unwinding the helix and bending it by 70-80 degrees 7774012. That distortion is the point: SOX proteins are as much architectural as regulatory, and one bound alone generally neither activates nor represses. Output depends on a partner factor at an adjacent site, which is why the same SOX protein behaves differently in different tissues 24086078. Two subgroups add dimerization hardware. SOXD proteins (SOX5, SOX6, SOX13) carry a coiled-coil that makes them obligate dimers or heterodimers; SOXE proteins (SOX8, SOX9, SOX10) carry a short DIM domain immediately N-terminal to the HMG box that supports cooperative homo- and heterodimers 26013289. SOX2 and SOX11 can also bind DNA on the nucleosome surface and are bona fide pioneer factors 32350470.

The motif

The core is six to seven bases, conventionally written AACAAT and equivalently ATTGTT or CATTGTT on the other strand; the medoid of the 87 models here is CATTGTT, the reverse complement of AACAATG. The SRY consensus determined by site selection is (A/T)(A/T)CAA(A/T)G 8190643. Information is concentrated on the CAA/TTG centre, which is where the intercalating residue inserts and where groove width is most constrained 7774012; the outer positions are AT-preferring largely because AT-rich DNA is easier to bend. Site selection on Sox4, Tcf1 and yeast Ste11 showed the functional element is really 10-12 base pairs, with residues C-terminal to the HMG box reading the 3' flank 10867006. Many models here are dimeric: CIS-BP SOX8 AACAATGTGCAGTGTT (28.8 bits), CIS-BP SRY, SOX15, SOX7 and SOX2 AACAAT(A/G)(n)CATTGTT, HOCOMOCO SOX9 ATSAATTGCAGTSAT. All are head-to-head inverted repeats of the core with a two- to four-base spacer, consistent with the finding that SOXE dimers tolerate a range of spacings rather than a fixed one 26013289. Within the family sequence does not separate members: SOX1, SOX2 and SOX3 are interchangeable on DNA, as are SOX8, SOX9 and SOX10 apart from their preference for dimeric sites, and the Vierstra archetypes are labelled "SOX" or "SOX/SRY", not a protein. One overlap reaches outside the family: HOCOMOCO SOX4, SOX11, SOX13 and one SOX2 model read CCTTTGTT, whose reverse complement AACAAAGG also satisfies the TCF/LEF Wnt response element CTTTGWW. SOX and TCF/LEF share the HMG fold and their consensus sequences differ at essentially one position, so CTTTG-type hits are genuinely ambiguous between the two families.

Where it acts

SOX sites are enhancer-biased and usually appear as one half of a composite element. The best characterized is the SOX-OCT element, where a SOX core abuts the octamer ATGCAAAT: SOX2 and OCT4 occupy it at the FGF4, UTF1, NANOG and POU5F1 enhancers 7590241, and the crystal structure of the POU/HMG/DNA complex showed that the two proteins assemble in different conformational arrangements depending on which enhancer's spacing they are given 12923055. Other named partnerships include SOX2 with PAX6 at the lens DC5 element, SOX9 with SOX5/SOX6 at the COL2A1, COL9A2 and COL11A2 enhancers, SOX10 with MITF at melanocyte enhancers, and SOXF factors with ERG and FLI1 in endothelium. Because AACAAT is only six bases it is expected roughly 1.5 million times in a 3.1 Gb double-stranded genome, far more often than it is occupied; nucleosome position matters, since SOX2 and SOX11 bind at superhelical location 2, distort the DNA there and help detach the terminal nucleosomal DNA 32350470.

Biology and disease

SRY is the testis-determining gene 1695712, and HMG-box point mutations that abolish DNA binding or bending cause 46,XY sex reversal 7774012. SOX9 mutations cause campomelic dysplasia with XY sex reversal 7990924; Sox9 is required for cartilage formation, and a mutation in the dimerization domain alone gives skeletal disease without sex reversal, showing that the dimeric and monomeric activities are separable 12837698. SOX10 mutations cause Waardenburg syndrome type 4 (Waardenburg- Hirschsprung disease) and peripheral demyelinating neuropathy 9462749, and SOX10 is the lineage- dependency factor of melanoma. SOX2 mutations cause bilateral anophthalmia, SOX2 is one of the four Yamanaka reprogramming factors, and its locus is amplified in squamous carcinomas of lung and esophagus. SOX17 and SOX18 are required for endothelial and lymphatic development, and SOX11 is a diagnostic marker in mantle cell lymphoma.

Reading a hit

A CATTGTT/AACAAT match identifies a SOX-family site, not a SOX protein: all 20 members read it with the same fold and the same contacts, and the archetype labels ("SOX", "SOX/SRY") say so. Expression data is the practical discriminator. Even the family assignment is soft at the boundary with TCF/LEF, because CTTTG-type variants fit both. The core is short enough to occur by chance more than a million times, so an isolated match means little; look for a partner site

  • an octamer, a second SOX core in inverted orientation, a MITF E-box - before treating it as

functional. Of the 36 HOCOMOCO models here, 27 carry HT-SELEX or methyl-HT-SELEX evidence and nine are ChIP-only; the ChIP-only ones (SOX5, SOX6, SOX11, SOX13, SOX30, some SOX2 and SOX17 models) are the likeliest to have absorbed a partner's preference. Two CIS-BP entries (SOX1, SOX11) are mouse-derived at 100% DNA-binding-domain identity, which does not affect specificity; no member of the family has out-of-family nearest neighbors.

How this was worked out

  1. 1990mediumPositional cloning, sequence comparison1695712

    SRY is identified in the sex-determining region of the human Y chromosome and found to carry a DNA-binding motif shared with the nuclear HMG proteins HMG1 and HMG2, defining the family the other SOX genes were later named for.

    Still open then: Whether SRY bound DNA sequence-specifically, and what sequence, was not known.

  2. 1994mediumBinding-site selection8190643

    Binding-site selection defines the SRY consensus as (A/T)(A/T)CAA(A/T)G, a six- to seven-base AT-rich core.

    Still open then: Left open whether the other SOX proteins read the same core, and how such a short site could be selective.

  3. 1994highTranslocation breakpoint cloning, mutation analysis7990924

    SOX9 mutations are shown to cause campomelic dysplasia with autosomal sex reversal, establishing that SOX proteins outside SRY control both skeletal development and testis determination.

  4. 1995highMultidimensional NMR7774012

    The NMR structure of the SRY HMG box on its site shows an L-shaped domain binding entirely in the minor groove, with a side chain intercalated between base pairs, the helix unwound and bent by 70-80 degrees, and explains sex-reversal point mutations at the atomic level.

  5. 1995mediumcDNA cloning, EMSA, reporter assays7590241

    Sox2 is identified as the factor that acts with Oct-3 on the FGF-4 enhancer, and only the Sox2-Oct-3 pair activates transcription: the first demonstration that a SOX protein works through an adjacent partner site.

  6. 2000mediumBinding-site selection, footprinting with deletion mutants10867006

    Site selection on Sox4, Tcf1, MatMc and Ste11 shows that sequence-specific HMG-box factors read 10-12 base pairs, not the 6-8 assumed, with residues C-terminal to the box contacting the 3' bases.

  7. 2003highX-ray crystallography; human genetics with EMSA and reporter assays12923055 12837698

    Two results separate SOX functions that one motif cannot: the crystal structure of a POU/HMG/DNA complex shows Oct4 and Sox2 adopting different arrangements on different enhancers, and a SOX9 mutation outside the HMG box abolishes cooperative dimer binding at chondrocyte enhancers while leaving monomeric binding and sex determination intact.

  8. 2020highCryo-electron microscopy32350470

    Cryo-EM structures of SOX2 and SOX11 DNA-binding domains on nucleosomes show them binding and distorting DNA at superhelical location 2 and detaching terminal nucleosomal DNA, giving a structural basis for SOX pioneer activity.

    Still open then: Did not establish which genomic SOX sites are engaged in this nucleosomal mode in cells.

Still unresolved

How much to trust a hit

A hit tells you a SOX-family HMG box could bind; it does not tell you which of the 20 proteins. SOX1/2/3, SOX8/9/10 and SOX4/11/12 are mutually indistinguishable from sequence, and the Vierstra archetypes are named "SOX" or "SOX/SRY" for that reason. The core is six bases and occurs over a million times by chance, so isolated matches are weak evidence; SOX proteins also need a partner factor to do anything, so a lone SOX site rarely predicts regulation. CTTTG-type variants overlap the TCF/LEF Wnt response element and cannot be confidently assigned to either family. Nine of the 36 HOCOMOCO models are ChIP-only and may reflect a partner's specificity.

References

87 motifs in this family — JASPAR 19 · HOCOMOCO 36 · CIS-BP 20 · Vierstra 12
motifdatabaseconsensusbits
M03483_2.00CIS-BPAACAATgTGCAGTGTT28.8
AC0249:SOX:SoxVierstraAACAATGTGCAGTGTT27.5
M03508_2.00CIS-BPAACAATAaCATTGTT26.7
M03525_2.00CIS-BPAACAATAnCATTGTT25.5
AC0099:SOX:SoxVierstraTGAATGTgCAGTCA25.2
M03519_2.00CIS-BPAACAATGgTATTGTT24.2
M03516_2.00CIS-BPAACAATgtncATTGTT23.6
SOX21 (MA0866.1)JASPARAACAATGgtAGTGTT23.4
M03497_2.00CIS-BPAACAATGgtAGTGTT23.4
AC0248:SOX/SRY:SoxVierstraAACAATgddyATTGTT22.3
SOX17.H14CORE.2.S.CSCHOCOMOCOAACAATtnCATTGTT22.1
AC0294:SOX/SRY:SoxVierstraAACAATggcAgTGTT21.5
Sox1 (MA0870.1)JASPARaACAATAnCATTGTT20.9
AC0293:SOX/SRY:SoxVierstraTCAATggcAtTGA19.4
AC0659:SOX/SRY:SoxVierstraTGAATtnCAtTCA19.3
SOX14.H14CORE.0.SM.BSMBHOCOMOCOACAATgnCATTGTTc18.9
SOX1.H14CORE.0.S.BSBHOCOMOCOTGAATAnCATTCAt17.6
SOX10.H14CORE.0.PSM.APSMAHOCOMOCOAaaAatdrvCCATTGTTC17.1
SRY.H14CORE.1.S.BSBHOCOMOCOaACAATAhCATTGTT17
SOX9.H14CORE.2.S.BSBHOCOMOCOATsAATTgCAGTsAT16.8
SOX7.H14CORE.1.SM.BSMBHOCOMOCOACAATtnCATTGT15.5
SOX9.H14CORE.0.PS.APSAHOCOMOCOAatdgnctATTGTTC15.1
AC0252:SOX:SoxVierstraACAATGgnncCATTGT13.9
SOX4.H14CORE.0.PSM.APSMAHOCOMOCOCCTTTGTTC13.8
Sox7 (MA2095.1)JASPARAGAACAATGG13.5
SOX11.H14CORE.0.P.BPBHOCOMOCOcCTTTGTtC13.4
SOX12 (MA1561.2)JASPARACCGAACAAT13.3
M05789_2.00CIS-BPATTGTTCGGT13.3
Sox6 (MA0515.1)JASPARcCATTGTttt13.2
SOX2.H14CORE.3.S.BSBHOCOMOCOAATahCATTCA12.8
SOX10.H14CORE.2.S.CSCHOCOMOCOaTTGTTCGGtg12.7
SOX2.H14CORE.1.P.BPBHOCOMOCOCCaTTGTtaTGCaaA12.7
SOX14 (MA1562.2)JASPARCGAACAATG12.5
M01026_2.00CIS-BPAACAAAGg12.2
SOX10.H14CORE.1.PSM.APSMAHOCOMOCOCATTGTT12.1
SOX8.H14CORE.1.PS.APSAHOCOMOCOcCATTGTTc12
SOX3.H14CORE.0.PM.APMAHOCOMOCOccATTGTT11.9
SOX12.H14CORE.0.SM.BSMBHOCOMOCOcATTGTTC11.8
SOX7.H14CORE.0.PS.APSAHOCOMOCOCCATTGTTC11.8
SOX2.H14CORE.0.P.BPBHOCOMOCOCCtTTGTtmT11.6
Sox11 (MA0869.3)JASPARGAACAAAG11.5
SOX17.H14CORE.1.P.BPBHOCOMOCOcCATTGTaTg11.5
Sox17 (MA0078.3)JASPARAgAACAATGG11.4
Sox5 (MA0087.3)JASPARAACAATGG11.4
SOX5.H14CORE.0.P.BPBHOCOMOCOCCATTGTTc11.4
SOX8.H14CORE.0.PSM.APSMAHOCOMOCOAAaggnncCaTTGTtt11.4
SOX11.H14CORE.1.S.BSBHOCOMOCOTGTTCGG11.3
SOX3.H14CORE.1.S.CSCHOCOMOCOaatgcCATTGT11.3
M08165_2.00CIS-BPACAAAG11.2
AC0254:SOX:SoxVierstraCGAACAAT11.1
M05892_2.00CIS-BPAACAATag11
SOX4 (MA0867.3)JASPARAACAAAGg10.9
M05896_2.00CIS-BPaACAAT10.9
SOX18.H14CORE.0.SM.BSMBHOCOMOCOCATTGT10.8
SOX30.H14CORE.1.SM.BSMBHOCOMOCOcATTGTt10.8
SOX10 (MA0442.3)JASPARACAAAG10.7
SOX13.H14CORE.0.P.BPBHOCOMOCOcCtTTGTT10.7
SOX17.H14CORE.0.PS.APSAHOCOMOCOCCATTGTtT10.5
SOX6.H14CORE.0.P.BPBHOCOMOCOcCATTGTt10.1
M08166_2.00CIS-BPACAATGg10.1
Sox3 (MA0514.3)JASPARACAATGG10
SOX13 (MA1120.2)JASPARACAATGg10
SOX18 (MA1563.2)JASPARaACAATdc10
SOX21.H14CORE.0.S.BSBHOCOMOCOAaTahCATTgtt10
SOX9.H14CORE.1.PS.APSAHOCOMOCOATTGT10
SOX2 (MA0143.5)JASPARACAATGG9.8
SOX30.H14CORE.0.P.CPCHOCOMOCOCCatTGTGAC9.8
SOX8 (MA0868.3)JASPARAACAATg9.6
M03501_2.00CIS-BPAACAATg9.6
SOX2.H14CORE.2.S.BSBHOCOMOCOTanCATT9.5
SRY.H14CORE.0.S.BSBHOCOMOCOCATTct9.5
M05894_2.00CIS-BPAACAATa9.5
SOX15 (MA1152.2)JASPARctATTGT9
SOX9 (MA0077.2)JASPARCcATTGTT8.9
SRY (MA0084.2)JASPARaAACAAT8.9
M05753_2.00CIS-BPAACAAT8.9
M05769_2.00CIS-BPcCATTGTnvg8.7
SOX1.H14CORE.1.S.BSBHOCOMOCOCATTC8.2
AC0255:SOX/SRY:SoxVierstraACAATg8.2
AC0253:SOX/SRY:SoxVierstraACAAaG8
SOX15.H14CORE.0.P.BPBHOCOMOCOCTTTGT7.5
M05783_2.00CIS-BPCATTGT7.3
M05791_2.00CIS-BPACAATGc7
AC0041:SRY:SoxVierstraAAACAaa6.9
AC0251:SOX:SoxVierstraACAATGc6.9
M00210_2.00Mus musculus 1CIS-BPaTTGT6.5
M00214_2.00Mus musculus 1CIS-BPAAT3.7

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

how this page was made Drafted by a language model reading the primary literature, then checked: every cited PMID is resolved against PubMed and its citation text regenerated from PubMed's own record, and every claim about the motifs themselves is checked against the matrices in the four source databases. Methods and limits →

Other Sox families