ZEB
ZEB1 and ZEB2 are large C2H2 zinc-finger repressors that bind the E-box-like sequence CACCTG, written CAGGTG when read on the other strand. Each protein carries two widely separated zinc-finger clusters, either of which can engage a site, so one molecule can hold two boxes at once. ZEB proteins contain no basic helix-loop-helix domain, yet they compete for the same CAGGTG boxes that bHLH activators use, so an E-box hit does not imply a bHLH factor. Both paralogues repress epithelial genes, E-cadherin (CDH1) above all, and are central to epithelial-mesenchymal transition. Loss-of-function mutations cause posterior polymorphous corneal dystrophy (ZEB1) and Mowat-Wilson syndrome (ZEB2), and high ZEB1 marks poorly differentiated carcinomas.
The factors
Humans have two paralogues: ZEB1 (also TCF8, deltaEF1, AREB6, ZFHX1A) and ZEB2 (SIP1, ZFHX1B, SMADIP1), each roughly 1,100-1,200 residues. Their architecture is atypical. Instead of one contiguous finger array, each protein carries two widely separated C2H2 clusters - three to four fingers near the N terminus and three near the C terminus - flanking a central region that includes a degenerate homeodomain which does not itself bind DNA. Remacle and colleagues showed that each cluster binds a target site on its own, so a single molecule works with two hands and can occupy two sites simultaneously 10487759. ZEB proteins do not dimerize through a leucine zipper or a helix-loop-helix; repression is delivered by recruited partners, chiefly CtBP through a central interaction domain, plus NuRD and HDAC complexes. ZEB2 was independently isolated as a SMAD-interacting protein in a yeast two-hybrid screen 10400677, and both paralogues couple TGF-beta signaling to transcription. ZEB1 is expressed in mesenchymal tissue, T cells and corneal endothelium; ZEB2 in neural crest derivatives, the central nervous system and blood. Both are induced during epithelial-mesenchymal transition (EMT) and are restrained post-transcriptionally by the miR-200 microRNA family 18483486.
The motif
The literature writes the site CACCTG, or equivalently CAGGTG on the opposite strand: a 6 bp member of the CANNTG E-box class, sometimes called the E2-box or Z-box 8065305,9544704. Essentially all the information sits in those six positions. Each three-finger cluster reads about one base triplet per finger from the major groove, which accounts for a footprint of roughly 6 bp and for the absence of strong flanking preference 9544704. The family data matches this: the CIS-BP and JASPAR models are rendered CACCTG and CTCACCTG while the HOCOMOCO models read CAGGTG and CAGGTGAG, which is the same motif on the other strand (the reverse complement of CAGGTGAG is CTCACCTG). This is a strand convention, not a disagreement. Information content is low, 10-12 bits, and the ZEB1 and ZEB2 models are effectively interchangeable, with cross-database similarities of 0.94-0.99, so sequence alone cannot tell the two paralogues apart. Because each protein has two independent hands, tandem CACCT(G) boxes bind more strongly than a single box 10487759, and validated targets such as CDH1 carry two or more 11430829; no fixed spacing is required.
Where it acts
ZEB sites occur in proximal promoters of epithelial structural genes - the E-pal element of CDH1 is the textbook case 11430829 - and in large numbers of distal enhancers in mesenchymal and carcinoma cells. CACCTG contains no CpG, so it is not depleted by genome-wide CpG loss: a strict 6 bp match occurs roughly once every 2 kb counting both strands, on the order of a million times in the human genome. Occupancy is therefore set by chromatin state and competition, not by site availability. ZEB1 ChIP-seq peaks overlap extensively with SNAI1/SNAI2 peaks and with bHLH factor peaks, because all of them use the same box. Functional partners include SMAD2/SMAD3 (especially for ZEB2), CtBP, NuRD and, in some settings, YAP; the best-characterized regulatory circuit is the mutual repression between ZEB1 and the miR-200 family 18483486.
Biology and disease
ZEB1 and ZEB2 repress epithelial genes including CDH1, EPCAM, CRB3 and LGL2 11430829,17486063, and are required for EMT in gastrulation, neural crest migration and organ fibrosis. Mouse deltaEF1 (Zeb1) mutants show skeletal defects and impaired thymocyte development 9126927. Zeb1 also couples EMT to cellular senescence through the CDKN2A locus, so loss of Zeb1 in mouse embryonic fibroblasts triggers premature senescence 18192284. In humans, heterozygous ZEB1 mutations cause posterior polymorphous corneal dystrophy type 3, with ectopic COL4A3 expression in corneal endothelium 16252232, while truncating mutations and deletions of ZEB2 cause Mowat-Wilson syndrome (intellectual disability, microcephaly, Hirschsprung disease and a distinctive face) 11279515. In carcinomas, ZEB1 expression tracks with loss of apicobasal polarity and dedifferentiation 17486063. One honest qualification: ZEB-driven EMT is firmly established in culture and in development, but in human tumors the transition is usually partial, and how much full EMT contributes to metastasis is still argued; current consensus statements describe a spectrum of hybrid epithelial-mesenchymal states rather than a binary switch 27368099.
Reading a hit
A CACCTG/CAGGTG match is among the least specific hits a scan can return. Plausible binders include the bHLH E-proteins TCF3, TCF4 and TCF12, MYOD and MYF, TWIST and HAND, NEUROD and ATOH, MAX and USF, and the C2H2 repressors SNAI1-3, SCRT1/2 and ZBTB18. An E-box hit therefore does not imply a bHLH factor. Nothing in the 6 bp core distinguishes ZEB from these alternatives; only context does. Two boxes within a short window, in a mesenchymal or TGF-beta-stimulated cell type, and near an epithelial gene, is a reasonable circumstantial case for ZEB. A single isolated box is not. Note also the evidence types: the HOCOMOCO ZEB1 and ZEB2 models are ChIP-seq-derived (evidence code P, grade B) 37971293 and the JASPAR ZEB1 matrix MA0103.4 is ChIP-seq-derived 37962376, so they report where ZEB was found, not necessarily what ZEB binds directly; the CIS-BP entries for both genes are direct rather than inferred. Low label agreement in the family data (0.2-0.4) means the nearest neighbours of these matrices across databases are often non-ZEB E-box models, which is expected rather than a sign of error.
How this was worked out
Random-oligonucleotide selection with deltaEF1 (ZEB1) returned CACCT, and deltaEF1 was shown to compete with the bHLH activator E47 at the E2 box CACCTG and to block MyoD-driven myogenesis - the first evidence that a zinc-finger protein reads an E-box.
Still open then: Whether repression worked only by occluding the box or also through an intrinsic repression domain.
Deletion analysis separated two repression mechanisms - binding-site competition, which needs both finger clusters, and active repression through the N-terminal NR domain - and deltaEF1 mutant mice showed that the factor is required in vivo, with a 100-fold reduction in thymocyte number.
Still open then: How a protein with two separated finger clusters engages DNA.
Each zinc-finger cluster was shown to bind a CACCT site independently, and the highest-affinity natural sites are bipartite, one CACCT plus one CACCTG with variable spacing and orientation; a single molecule binds with two hands.
Still open then: Whether bipartite sites are the rule at endogenous targets genome-wide.
- 2001mediumpromoter mutagenesis, conditional expression in MDCK cells, patient sequencing11430829 11279515
ZEB2/SIP1 was shown to repress E-cadherin through both E2 boxes of the CDH1 promoter and to induce invasion, and SIP1 mutations were found to cause Hirschsprung disease with intellectual disability, linking the motif to an epithelial program and to human disease.
Heterozygous truncating mutations in TCF8/ZEB1 were shown to cause posterior polymorphous corneal dystrophy, with ectopic COL4A3 expression in corneal endothelium, providing a human loss-of-function phenotype for ZEB1 itself.
ZEB1 was shown to repress the miR-200 family, which in turn targets ZEB1, defining a double-negative feedback loop; in parallel, Zeb1 mutant mouse fibroblasts were found to senesce prematurely through p15INK4b and p21 rather than through the Ink4a pathway.
Reviews and subsequent consensus guidelines reframed EMT as a set of partial, reversible states rather than a binary switch, which revised how a ZEB-driven program in a tumor should be described and cautioned against inferring full EMT from marker or motif evidence.
Still open then: How much, if any, complete EMT occurs in human carcinomas and whether it is needed for metastasis.
Still unresolved
- It is unresolved how often ZEB proteins actually use both finger clusters in the genome: the highest-affinity in vitro sites are bipartite CACCT plus CACCTG elements with variable spacing, but every database matrix for ZEB1 and ZEB2 models a single 6 bp box, and no survey has reported what fraction of occupied sites carries two boxes.
- Whether ZEB1 and ZEB2 differ at all in sequence preference is unknown rather than settled: their models are 0.94-0.99 similar, so paralogue-specific target sets reported in different tissues may reflect expression and cofactors rather than DNA recognition, and no assay has yet separated the two possibilities.
- The contribution of ZEB-driven EMT to human metastasis remains disputed, with cell-culture and developmental work supporting a full transition and consensus statements on human tumors favoring partial, hybrid states that are hard to distinguish from other sources of mesenchymal gene expression.
- What decides whether a given CAGGTG box is occupied by a ZEB protein or by a bHLH factor is not established; competition for the same core was demonstrated biochemically in the 1990s, but whether in vivo selectivity comes from flanking sequence, chromatin state or relative protein abundance has not been resolved.
Trust the family, not the member, and not even the family without context. The 6 bp CACCTG core is shared with bHLH activators, SNAIL/SCRATCH, INSM and ZBTB18, so a hit licenses "E-box" and nothing more; ZEB1 and ZEB2 motifs are indistinguishable from each other at 0.94-0.99 similarity. The site is common enough to occur roughly a million times by chance in the genome, so single matches carry almost no evidence. HOCOMOCO ZEB1/ZEB2 and JASPAR MA0103.4 are ChIP-derived and can reflect indirect or co-binding. Paired boxes, an epithelial target gene and a mesenchymal or TGF-beta context are what make a ZEB call defensible.
References
- Remacle JE, Kraft H, Lerchner W, et al. New mode of DNA binding of multi-zinc finger transcription factors: deltaEF1 family members bind with two hands to two target sites. EMBO J 1999;18:5073-84. PMID 10487759
- Sekido R, Murai K, Kamachi Y, et al. Two mechanisms in the action of repressor deltaEF1: binding site competition with an activator and active repression. Genes Cells 1997;2:771-83. PMID 9544704
- Comijn J, Berx G, Vermassen P, et al. The two-handed E box binding zinc finger protein SIP1 downregulates E-cadherin and induces invasion. Mol Cell 2001;7:1267-78. PMID 11430829
- Higashi Y, Moribe H, Takagi T, et al. Impairment of T cell development in deltaEF1 mutant mice. J Exp Med 1997;185:1467-79. PMID 9126927
- Krafchak CM, Pawar H, Moroi SE, et al. Mutations in TCF8 cause posterior polymorphous corneal dystrophy and ectopic expression of COL4A3 by corneal endothelial cells. Am J Hum Genet 2005;77:694-708. PMID 16252232
- Wakamatsu N, Yamada Y, Yamada K, et al. Mutations in SIP1, encoding Smad interacting protein-1, cause a form of Hirschsprung disease. Nat Genet 2001;27:369-70. PMID 11279515
- Aigner K, Dampier B, Descovich L, et al. The transcription factor ZEB1 (deltaEF1) promotes tumour cell dedifferentiation by repressing master regulators of epithelial polarity. Oncogene 2007;26:6979-88. PMID 17486063
- Burk U, Schubert J, Wellner U, et al. A reciprocal repression between ZEB1 and members of the miR-200 family promotes EMT and invasion in cancer cells. EMBO Rep 2008;9:582-9. PMID 18483486
- Liu Y, El-Naggar S, Darling DS, et al. Zeb1 links epithelial-mesenchymal transition and cellular senescence. Development 2008;135:579-88. PMID 18192284
- Nieto MA, Huang RY, Jackson RA, et al. EMT: 2016. Cell 2016;166:21-45. PMID 27368099
- Vorontsov IE, Eliseeva IA, Zinkevich A, et al. HOCOMOCO in 2024: a rebuild of the curated collection of binding models for human and mouse transcription factors. Nucleic Acids Res 2024;52:D154-D163. PMID 37971293
- Rauluseviciute I, Riudavets-Puig R, Blanc-Mathieu R, et al. JASPAR 2024: 20th anniversary of the open-access database of transcription factor binding profiles. Nucleic Acids Res 2024;52:D174-D182. PMID 37962376
- Sekido R, Murai K, Funahashi J, et al. The delta-crystallin enhancer-binding protein delta EF1 is a repressor of E2-box-mediated gene activation. Mol Cell Biol 1994;14:5692-700. PMID 8065305
- Verschueren K, Remacle JE, Collart C, et al. SIP1, a novel zinc finger/homeodomain repressor, interacts with Smad proteins and binds to 5'-CACCT sequences in candidate target genes. J Biol Chem 1999;274:20489-98. PMID 10400677
5 motifs in this family — JASPAR 1 · HOCOMOCO 2 · CIS-BP 2
| motif | database | consensus | bits |
|---|---|---|---|
| ZEB1.H14CORE.0.P.BPB | HOCOMOCO | CAGGTGaG | 12.5 |
| M07842_2.00 | CIS-BP | CtCACCTG | 11.9 |
| M08080_2.00 | CIS-BP | CACCTG | 11 |
| ZEB1 (MA0103.4) | JASPAR | CACCTG | 10.6 |
| ZEB2.H14CORE.0.P.BPB | HOCOMOCO | CAGGTG | 9.7 |
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
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.