Atypical / composite homeodomain
This family collects homeodomains that do not behave like homeodomains. Instead of one small domain in a compact protein, these factors bury one to five divergent homeodomains inside large multidomain proteins, usually alongside many C2H2 zinc fingers: ZHX1-3, ZFHX2 and ZFHX3, HOMEZ, TSHZ1-3, HMBOX1, HDX, HHEX, and the SATB-type CUT-plus-homeodomain proteins. Several were discovered as something else entirely, an AT-rich enhancer-binding protein, a telomere repeat-binding protein, a chromatin scaffold protein. Their motifs are correspondingly heterogeneous: there is no single family consensus, several database models carry almost no information, and where a good motif exists it may be read by the zinc fingers rather than by the homeodomain. Useful as a bucket, not as an identifier.
The factors
What unites these proteins is architecture, not sequence preference. Domain annotation gives the flavour. ZHX1 and ZHX3 are 873 and 956 residues with five annotated homeobox regions each plus two C2H2 zinc fingers; ZHX2 has four homeoboxes. ZFHX3, originally named ATBF1, is 3,703 residues carrying four homeodomains and 23 zinc fingers; ZFHX2 is 2,572 residues with three homeodomains and 12 zinc fingers. HOMEZ is a 550-residue protein with three homeodomains and a leucine-zipper-like region and no zinc fingers at all 12925734. TSHZ2 and TSHZ3, vertebrate teashirt proteins, have five zinc fingers and a single homeodomain annotated as atypical. HDX has two homeodomains; HMBOX1 has one, embedded in a POU-like arrangement. HHEX is the outlier in the other direction, a 270-residue protein with a single homeodomain, but it homo-oligomerises through a proline-rich region rather than working as a monomer 18498250. SATB1 and SATB2 pair a divergent homeodomain with two CUT domains 26464018. Several of these proteins were characterised before anyone called them homeodomain factors: ZFHX3/ATBF1 as the AT-motif binding factor of the human alpha-fetoprotein enhancer 1719379, HMBOX1 as HOT1, a direct double-stranded telomere repeat-binding protein 23685356, SATB1 as a matrix-attachment-region binding protein with "unusual binding site recognition" 1505028, and ZHX1 as an NF-Y-interacting repressor 10441475.
The motif
There is no single consensus, and the entry deliberately does not offer one. The three best-supported models are as follows. HOMEZ, from HT-SELEX, reads TAATCGTTT, about 13 bits: a recognisable TAAT core with an unusual CGTTT extension, and it is reproduced independently in CIS-BP as TATAATCGTTTT and as a footprint archetype, so it is probably real 23332764,25215497. HMBOX1 gives CTAGTTA, equivalently TAACTAG, supported by both in vitro selection and ChIP 23332764; this is a TAAT-like core with an inverted 3' half. Notably it does not resemble the TTAGGG telomeric repeat that HOT1/HMBOX1 was reported to bind directly 23685356, and that discrepancy is not resolved in the literature. TSHZ2 gives a 15-base CATCAAATGGAATC with high information content and grade A evidence, and it contains no TAAT core whatever; the most economical reading is that the zinc fingers do the binding, as they do for Drosophila teashirt, and the atypical homeodomain contributes little. The remaining models are close to empty: ZFHX2 is a bare TAATTA at 5.5 bits, ZFHX3 reduces to AAT at 2.5 bits, HHEX to AT at 1.9, ZHX1 to AC at 1.3, HDX to a single T. Family cohesion is 0.55, the lowest of the homeodomain groups, which is the expected signature of a bucket.
Where it acts
Genuinely varied, and for several members not well mapped. HMBOX1/HOT1 localises to telomeres and associates with actively processed telomere ends. ZHX proteins act as promoter-proximal repressors in complexes with NF-Y 10441475. ZFHX3/ATBF1 was defined at an AT-rich enhancer element and binds AT-rich sequence generally 1719379, which is one reason its motif carries so little information. SATB1 occupies base-unpairing regions and matrix attachment regions and organises chromatin loops in T cells rather than acting at a short consensus 1505028,10716941. TSHZ3 works at neuronal target genes in cortical projection neurons 27668656. The general pattern is that occupancy for these proteins is determined by protein-protein interaction and by regional DNA character, AT-richness or repeat content, more than by a short high-information consensus, which is exactly why motif-based prediction performs poorly for them.
Biology and disease
Several members have solid human genetics. Somatic mutation and deletion of ZFHX3/ATBF1 occur recurrently in prostate cancer 15750593. Deletion of TSHZ3 causes an autism syndrome with defects in cortical projection neurons 27668656. De novo mutations in ADNP, another large protein with a homeodomain in unusual company, cause a SWI/SNF-related autism syndrome 24531329. HHEX, as PRH/Hex, is required for formation of the vertebrate body axis, the haematopoietic and vascular systems, liver, thyroid and forebrain, and it can both repress and activate targets and additionally regulate nuclear export of specific mRNAs 8096636,18498250. HMBOX1/HOT1 is a positive regulator of telomere length: depletion shortens and overexpression lengthens telomeres, and it promotes chromatin association of telomerase 23685356. SATB1 controls thymocyte development 10716941 through higher-order chromatin organisation.
Reading a hit
This is the family where a motif hit licenses the least. Of the 13 models, five carry under 6 bits and three carry under 3 bits, which means they cannot be scanned meaningfully at all; a "hit" for the CIS-BP ZHX1 or HDX model is not information. The CIS-BP HHEX model is inherited across species: it was measured on a Caenorhabditis elegans protein at 77% DNA-binding-domain identity and transferred to human HHEX, not measured on the human protein 25215497. The CIS-BP ZFHX2 model is a bare TAATTA whose nearest neighbours across all four databases are Q50 homeodomain motifs, and the ZHX1 model's nearest neighbours are CpG-binding MBD/CXXC motifs; both are flagged in the underlying data as probably belonging elsewhere. Only HOMEZ, HMBOX1 and TSHZ2 rest on direct in vitro selection for the human protein, and for TSHZ2 the motif is probably the zinc fingers' rather than the homeodomain's. Read a hit here as a statement about sequence composition, and go to expression and ChIP data before naming a protein.
How this was worked out
ATBF1, later renamed ZFHX3, is cloned as the protein that binds an AT-rich motif in the human alpha-fetoprotein enhancer and turns out to be a 306-kDa protein with four homeodomains and 17 annotated zinc fingers, the first example of homeodomains embedded in a very large multidomain factor.
Still open then: The protein was defined by binding an AT-rich region rather than a defined consensus, and which of its domains contacts DNA was not determined.
SATB1 is cloned as a thymus-enriched matrix-attachment-region binding protein that recognizes an AT-rich context indirectly, contacting the minor groove and the sugar-phosphate backbone rather than specific bases, and losing binding when the unwinding potential of the site is removed.
Still open then: Established that a protein in this group can select sites with no conventional consensus, which no position weight matrix can represent.
ZHX1 is isolated in a two-hybrid screen for NF-YA partners and found to be an 873-residue protein with five homeodomain motifs and two zinc fingers that is expressed ubiquitously, placing it at promoters through a partner rather than through its own site.
HOMEZ is described as a vertebrate-specific protein combining three atypical homeodomains with leucine-zipper-like motifs and an acidic domain, and is grouped with the ZHX proteins as a distinct architectural subset of the homeobox superfamily.
Still open then: No DNA-binding data; the homeodomains were called atypical on sequence grounds alone.
HT-SELEX on recombinant human proteins gives the first directly measured specificities for members of this group, including the HOMEZ site TAATCGTTT and the HMBOX1 site TAACTAG, both of which deviate from the canonical homeodomain TAAT consensus.
Still open then: Most members of the group yielded no usable model, and the HMBOX1 site does not resemble a telomeric repeat.
- 2013mediumSILAC-based DNA-protein interaction screen with ChIP, depletion and overexpression23685356
A quantitative DNA-protein interaction screen identifies HMBOX1, renamed HOT1, as a direct binder of double-stranded TTAGGG telomere repeats that associates with the telomerase complex and acts as a positive regulator of telomere length, a sequence preference different from the one in vitro selection reports for the same protein.
Still open then: The discrepancy between the telomeric repeat and the selected TAACTAG site has not been reconciled.
- 2014highprotein-binding microarrays across 131 eukaryotes with inference by DNA-binding-domain identity25215497
A survey of more than 1,000 transcription factors establishes that closely related DNA-binding domains usually share preferences and uses that rule to infer motifs for about a third of known eukaryotic factors; several models carried in this family, including the human HHEX motif, are inferred from a non-human protein rather than measured.
Still open then: Whether the identity-based inference rule holds for divergent and atypical homeodomains was not tested separately.
- 2016mediumhuman deletion mapping with mouse knockout transcriptomics, electrophysiology and behavior27668656
Heterozygous deletions at 19q12-q13.11 are narrowed to TSHZ3 as the critical gene for an autism syndrome, and Tshz3 mutant mice show altered expression of cortical projection neuron markers, synaptic changes and autism-like behavior, giving one member of this architecturally defined group a defined developmental function.
Still open then: Does not say whether TSHZ3 acts through its zinc fingers, its atypical homeodomain or through partners.
Still unresolved
- The HMBOX1 specificity is contradictory: an interaction screen reports direct binding to double-stranded TTAGGG telomere repeats, while in vitro selection and the database models give TAACTAG, and no study has established whether the telomeric association is direct binding to a second site class or recruitment by a telomere-bound partner.
- It is unresolved whether the high-information TSHZ2 site is read by the zinc fingers or by the atypical homeodomain: the 15-base selected sequence contains no TAAT core, which points to the fingers as in Drosophila teashirt, but there is no structure or domain-deletion binding experiment for the human protein.
- For the multi-homeodomain proteins it is unknown whether any of the homeodomains binds DNA in cells at all: ATBF1 was defined by AT-rich binding and ZHX1 was found as an NF-Y partner, both consistent with the homeodomains working as protein-interaction modules, and no domain-by-domain binding data exist.
- The inferred human HHEX motif rests on an untested assumption: it was measured on a Caenorhabditis elegans protein sharing 77% DNA-binding-domain identity, and whether identity-based transfer is reliable for divergent homeodomains of this kind has not been evaluated for this group.
Treat this family as a bucket label, not an identification. Five of the 13 models carry under 6 bits and three under 3 bits, so they are not scannable; matches to them are noise. The CIS-BP HHEX model is transferred from a C. elegans protein at 77% DNA-binding-domain identity rather than measured on human HHEX. The CIS-BP ZFHX2 and ZHX1 models sit closer to Q50 homeodomain and MBD/CXXC motifs respectively than to anything in this family. Only HOMEZ, HMBOX1 and TSHZ2 have credible human in vitro motifs, and the TSHZ2 site contains no TAAT core and is likely read by its zinc fingers. For most members, occupancy is driven by partners and by AT-rich or repeat context, not by the consensus.
References
- Bürglin TR, Affolter M. Homeodomain proteins: an update. Chromosoma 2016;125:497-521. PMID 26464018
- Bayarsaihan D, Enkhmandakh B, Makeyev A, et al. Homez, a homeobox leucine zipper gene specific to the vertebrate lineage. Proc Natl Acad Sci U S A 2003;100:10358-63. PMID 12925734
- Kappei D, Butter F, Benda C, et al. HOT1 is a mammalian direct telomere repeat-binding protein contributing to telomerase recruitment. EMBO J 2013;32:1681-701. PMID 23685356
- Morinaga T, Yasuda H, Hashimoto T, et al. A human alpha-fetoprotein enhancer-binding protein, ATBF1, contains four homeodomains and seventeen zinc fingers. Mol Cell Biol 1991;11:6041-9. PMID 1719379
- Sun X, Frierson HF, Chen C, et al. Frequent somatic mutations of the transcription factor ATBF1 in human prostate cancer. Nat Genet 2005;37:407-12. PMID 15750593
- Caubit X, Gubellini P, Andrieux J, et al. TSHZ3 deletion causes an autism syndrome and defects in cortical projection neurons. Nat Genet 2016;48:1359-1369. PMID 27668656
- Helsmoortel C, Vulto-van Silfhout AT, Coe BP, et al. A SWI/SNF-related autism syndrome caused by de novo mutations in ADNP. Nat Genet 2014;46:380-4. PMID 24531329
- Yamada K, Printz RL, Osawa H, et al. Human ZHX1: cloning, chromosomal location, and interaction with transcription factor NF-Y. Biochem Biophys Res Commun 1999;261:614-21. PMID 10441475
- Bedford FK, Ashworth A, Enver T, et al. HEX: a novel homeobox gene expressed during haematopoiesis and conserved between mouse and human. Nucleic Acids Res 1993;21:1245-9. PMID 8096636
- Soufi A, Jayaraman PS. PRH/Hex: an oligomeric transcription factor and multifunctional regulator of cell fate. Biochem J 2008;412:399-413. PMID 18498250
- Dickinson LA, Joh T, Kohwi Y, et al. A tissue-specific MAR/SAR DNA-binding protein with unusual binding site recognition. Cell 1992;70:631-45. PMID 1505028
- Weirauch MT, Yang A, Albu M, et al. Determination and inference of eukaryotic transcription factor sequence specificity. Cell 2014;158:1431-1443. PMID 25215497
- Jolma A, Yan J, Whitington T, et al. DNA-binding specificities of human transcription factors. Cell 2013;152:327-39. PMID 23332764
- Alvarez JD, Yasui DH, Niida H, et al. The MAR-binding protein SATB1 orchestrates temporal and spatial expression of multiple genes during T-cell development. Genes Dev 2000;14:521-35. PMID 10716941
13 motifs in this family — JASPAR 1 · HOCOMOCO 3 · CIS-BP 7 · Vierstra 2
| motif | database | consensus | bits |
|---|---|---|---|
| TSH2.H14CORE.0.SG.ASGA | HOCOMOCO | CATCaAATGGAATCg | 25.8 |
| HOMEZ.H14CORE.0.S.BSB | HOCOMOCO | aaATCGTTT | 13.3 |
| M03242_2.00 | CIS-BP | taTAATCGTTTT | 12.6 |
| HMBX1.H14CORE.0.PSM.APSMA | HOCOMOCO | CTAGTTAvc | 11.8 |
| AC0627:HOMEZ:Homeodomain | Vierstra | aTAATCGTTT | 11.7 |
| M05180_2.00 | CIS-BP | TAACTAG | 10 |
| AC0472:HMBOX:Homeodomain | Vierstra | TAACTAG | 9.4 |
| HMBOX1 (MA0895.2) | JASPAR | TAGTTAa | 8.9 |
| M02089_2.00looks like HD_Q50 | CIS-BP | TAATTA | 5.5 |
| M01849_2.00 | CIS-BP | aAT | 2.5 |
| M02184_2.00Caenorhabditis elegans 0.77 | CIS-BP | AT | 1.9 |
| M02093_2.00looks like MBD_CXXC | CIS-BP | ac | 1.3 |
| M01246_2.00 | CIS-BP | t | 1.2 |
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.