NKX
NKX proteins are homeodomain transcription factors related to Drosophila tinman and vnd, and they are the main exception to the rule that homeodomains read a TAAT core. NKX2-1 (TTF-1), NKX2-2 to NKX2-8, NKX3-1 and NKX3-2 prefer TAAGTG (read as CACTTA on the other strand), a specificity recognized as unusual when TTF-1 was first cloned. The family collects 51 motifs for 16 genes, including NKX1-1/1-2, NKX6-1 to NKX6-3 and NANOG. It is the least cohesive homeodomain family in this atlas because the NKX6 proteins in fact read a conventional TAAT core rather than TAAGTG, so the label covers two different sites.
The factors
NKX genes are the vertebrate NK-class homeobox genes, orthologues of Drosophila tinman (heart) and vnd/NK-2 (ventral nervous system). Sixteen symbols and 51 motifs are grouped here: NKX1-1 and NKX1-2; NKX2-1 (also called TTF-1 or TITF1), NKX2-2, NKX2-3, NKX2-4, NKX2-5, NKX2-6 and NKX2-8; NKX3-1 and NKX3-2; NKX6-1, NKX6-2 and NKX6-3; plus NANOG and unassigned NKX archetypes. Each is restricted to a small number of lineages: NKX2-1 to thyroid and lung, NKX2-5 to heart, NKX2-2 and NKX6-1 to pancreatic endocrine progenitors and ventral neural tube, NKX3-1 to prostate epithelium, NKX3-2 (BAPX1) to axial skeleton, NANOG to pluripotent cells. All bind DNA as monomeric homeodomains; none of them needs a dedicated DNA-binding partner, although NKX2-5 and GATA4 co-activate shared targets.
The motif
The consensus is TAAGTG. Randomized-site selection with the NKX2-5 homeodomain gave high-affinity sites of the form 5'-TNNAGTG-3' and only intermediate or weak binding to 5'-C(A/T)TTAATTN-3', the conventional TAAT-type site, which is the cleanest demonstration that NKX proteins prefer a different core 7797561. The TTF-1/NKX2-1 site is often described by a CAAG core (CAAGTG) 8890172. That is not TAAGTG read on the other strand, whose reverse complement is CACTTA, but a one-base variant at the first core position, where NK-2 homeodomains differ among themselves according to N-terminal-arm residues 6-8 8890172,18585360.
The structural reason is not position 50. Every NKX homeodomain has glutamine there, the residue that in other families specifies TAATTA 8692913; the NMR structure of the vnd/NK-2 homeodomain on DNA confirms Gln50 together with the canonical Asn51 hydrogen bond and minor-groove contacts from Lys3 and Arg5 10356327. What differs is elsewhere in the recognition helix, notably position 54, which is tyrosine in NKX2-1, NKX2-5 and NKX3-1 but methionine in Antp-type and alanine in En-type TAATTA readers. In the fly NK-2 specificity group, Tyr54 specifies the G of TAAG through the complementary C 18585360, and mutational comparison of the TTF-1, Antennapedia and Engrailed homeodomains ties residue 54 to the base at the 3' end of the core 8890172. This is why the position-50 rule, which explains TAATTA versus TAATCC, does not explain TAAGTG.
The family is internally split, and this is its most important practical feature. NKX2-x and NKX3-x models are TAAGTG: NKX3-2 (the medoid, HOCOMOCO NKX32.H14CORE.0.PSM.A) TAAGTG, NKX3-1 TTAAGTGG, NKX2-3 TTAAGTG, NKX2-8 TTAAGTG, NKX2-1 and NKX2-2 TTGAGTG. The NKX6 models are not: NKX6-1 and NKX6-2 give TTAATTA and NKX6-3 gives ATTAATGAC, ordinary TAAT-core sites. NANOG models are TAATGG-like 18177668. Family cohesion is accordingly 0.74, the lowest of the homeodomain families here, and NKX6-2's CIS-BP model is flagged as resembling anterior HOX - correctly, since TAATTA is what it is. Within each subgroup the paralogues are indistinguishable: no matrix separates NKX2-1 from NKX2-4, or NKX3-1 from NKX3-2. Several models here are dimeric or extended, such as NKX3-2 TTAAGTGGCACTTAA (two half-sites in inverted orientation) and NKX6-1 ATTAATGGGHWTTTAAT.
Where it acts
Sites sit in lineage-specific enhancers and in promoters of tissue-defining genes: TTF-1 was purified on thyroid-specific gene promoters 1976511, NKX2-5 acts on cardiac structural gene enhancers, and NKX6-1 acts downstream of NKX2-2 in beta-cell formation 11076772. TAAGTG is a common hexamer, expected on the order of two million times in 3.1 Gb at human base composition, so genome-wide match counts are large and uninformative; accessibility and expression do the work. NKX2-5 sites frequently co-occur with GATA and TBX motifs in cardiac enhancers, NKX6-1 with PDX1 and other pancreatic homeodomain sites.
Biology and disease
Heterozygous NKX2-5 mutations cause atrial septal defect with progressive atrioventricular conduction block 9651244, and Nkx2-5 null mouse embryos arrest at the heart-tube stage 7628699. NKX2-1 haploinsufficiency causes brain-lung-thyroid syndrome (choreoathetosis, congenital hypothyroidism, pulmonary disease) 11854319, while the NKX2-1 locus at 14q13 is focally amplified in lung adenocarcinoma 17982442. NKX3-1 is a prostate tumor suppressor whose expression is lost early in prostate carcinogenesis 14648854. NKX6-1 is required for pancreatic beta-cell formation 11076772, NKX3-2 for axial skeletal development, and NANOG for maintenance of pluripotency in the epiblast and in embryonic stem cells.
Reading a hit
A TAAGTG match indicates an NKX2/NKX3-type site and no particular gene. A TAAT-type match labelled NKX6 is a generic homeodomain site and should be read as such. This family also carries the clearest indirect-binding problems in the homeodomain set: 6 of its 17 HOCOMOCO models are ChIP-only, and two of them are flagged as matching other families entirely. The NANOG ChIP-derived model reads CCTTTGTNATGCAAA - the octamer ATGCAAAT, which is the POU/OCT4 motif, exactly what indirect capture of a co-occupied partner looks like for a factor that shares enhancers with OCT4 and SOX2 25070602. An NKX2-5 ChIP model reads GTGHTGACAG and matches the TBX family. Six CIS-BP entries in this family are transferred from mouse orthologues, one of them (NKX2-6) inferred at 93% DNA-binding-domain identity rather than measured. Prefer the HT-SELEX-derived models when testing whether a sequence is an NKX site.
How this was worked out
TTF-1 (NKX2-1), a thyroid nuclear factor that binds thyroid-specific promoters, is cloned. Its DNA-binding domain is a homeodomain related to Drosophila NK-2, with a DNA-binding specificity unlike that of other known homeodomains.
Still open then: The site was known from a few thyroid promoters; whether it was general to NK proteins was not.
Random-oligonucleotide selection with the NKX2-5 homeodomain gives high-affinity TNNAGTG sites and only intermediate or weak binding to TAAT-type sites, so the NK core differs from the TAATTA consensus.
Heterozygous NKX2-5 mutations cause congenital heart disease with atrioventricular conduction abnormalities; two of the three mutations are predicted to impair DNA binding.
The NMR structure of the vnd/NK-2 homeodomain on its 16-bp consensus site confirms the canonical contacts, Asn51 to the second adenine and Lys3 and Arg5 in the minor groove, with Gln50 at the interface. The homeodomain sits on DNA with small but significant differences in orientation from other homeodomains.
Two genome-scale surveys, protein-binding microarrays for 168 mouse homeodomains (at least 65 distinct activities) and bacterial one-hybrid selection for all 84 Drosophila homeodomains (11 specificity groups), resolve NK-type TAAGTG binding as a distinct activity; the fly survey ties it to recognition-helix positions beyond 50, including Tyr54.
The crystal structure of the NANOG homeodomain, with binding assays, shows a preference for TAAT(G/T)(G/T) and places NANOG as a distant Q50-type homeodomain, not a TAAGTG reader.
Across hundreds of ChIP-seq datasets, peaks carrying motifs of factors other than the one immunoprecipitated make up as much as 45% of a dataset. This is the background for the NANOG and NKX2-5 ChIP models here that read POU and TBX-type sequences.
Still unresolved
- NANOG does not read the NK site: its homeodomain behaves as a distant Q50 type and its in vitro models are TAATGG-like. Whether it belongs with the NK family, as its gene classification suggests, or with TAAT-core homeodomains, as its specificity suggests, is a real split in this family's data (cohesion 0.74).
- How much NKX2-5 occupancy in the heart depends on its own TAAGTG site, and how much on co-bound partners, is unsettled: the HT-SELEX-supported model is clean TAAGTG, while one ChIP-only NKX2-5 model reads a TBX-type sequence (GTGHTGACAG).
- Whether the inverted-repeat models (NKX3-2 TTAAGTGGCACTTAA, NKX6-1 ATTAATGGGHWTTTAAT) reflect cooperative dimer binding used in vivo, or two independent monomer sites recovered together in selection, has not been established.
Treat a TAAGTG hit as "an NKX2/NKX3-type site", not as a named factor; paralogues within each subgroup are indistinguishable. The family label spans two different motifs - NKX2 and NKX3 proteins read TAAGTG, but NKX6-1/6-2/6-3 read an ordinary TAAT core - which is why cohesion is only 0.74 and why one NKX6-2 model is flagged as HOX-like. ChIP-only evidence is a real risk here: 6 of 17 HOCOMOCO models are ChIP-derived, and two are flagged as belonging elsewhere, including a NANOG model that is actually the OCT4 octamer. Six CIS-BP motifs are inherited from mouse proteins. Expect roughly two million genome-wide matches, so require accessibility and expression.
References
- Guazzi S, Price M, De Felice M, et al. Thyroid nuclear factor 1 (TTF-1) contains a homeodomain and displays a novel DNA binding specificity. EMBO J 1990;9:3631-9. PMID 1976511
- Chen CY, Schwartz RJ. Identification of novel DNA binding targets and regulatory domains of a murine tinman homeodomain factor, nkx-2.5. J Biol Chem 1995;270:15628-33. PMID 7797561
- Gruschus JM, Tsao DH, Wang LH, et al. The three-dimensional structure of the vnd/NK-2 homeodomain-DNA complex by NMR spectroscopy. J Mol Biol 1999;289:529-45. PMID 10356327
- Wilson DS, Sheng G, Jun S, et al. Conservation and diversification in homeodomain-DNA interactions: a comparative genetic analysis. Proc Natl Acad Sci U S A 1996;93:6886-91. PMID 8692913
- Berger MF, Badis G, Gehrke AR, et al. Variation in homeodomain DNA binding revealed by high-resolution analysis of sequence preferences. Cell 2008;133:1266-76. PMID 18585359
- Schott JJ, Benson DW, Basson CT, et al. Congenital heart disease caused by mutations in the transcription factor NKX2-5. Science 1998;281:108-11. PMID 9651244
- Lyons I, Parsons LM, Hartley L, et al. Myogenic and morphogenetic defects in the heart tubes of murine embryos lacking the homeo box gene Nkx2-5. Genes Dev 1995;9:1654-66. PMID 7628699
- Krude H, Schütz B, Biebermann H, et al. Choreoathetosis, hypothyroidism, and pulmonary alterations due to human NKX2-1 haploinsufficiency. J Clin Invest 2002;109:475-80. PMID 11854319
- Shen MM, Abate-Shen C. Roles of the Nkx3.1 homeobox gene in prostate organogenesis and carcinogenesis. Dev Dyn 2003;228:767-78. PMID 14648854
- Sander M, Sussel L, Conners J, et al. Homeobox gene Nkx6.1 lies downstream of Nkx2.2 in the major pathway of beta-cell formation in the pancreas. Development 2000;127:5533-40. PMID 11076772
- Jauch R, Ng CK, Saikatendu KS, et al. Crystal structure and DNA binding of the homeodomain of the stem cell transcription factor Nanog. J Mol Biol 2008;376:758-70. PMID 18177668
- Worsley Hunt R, Wasserman WW. Non-targeted transcription factors motifs are a systemic component of ChIP-seq datasets. Genome Biol 2014;15:412. PMID 25070602
- Noyes MB, Christensen RG, Wakabayashi A, et al. Analysis of homeodomain specificities allows the family-wide prediction of preferred recognition sites. Cell 2008;133:1277-89. PMID 18585360
- Weir BA, Woo MS, Getz G, et al. Characterizing the cancer genome in lung adenocarcinoma. Nature 2007;450:893-8. PMID 17982442
- Damante G, Pellizzari L, Esposito G, et al. A molecular code dictates sequence-specific DNA recognition by homeodomains. EMBO J 1996;15:4992-5000. PMID 8890172
51 motifs in this family — JASPAR 12 · HOCOMOCO 17 · CIS-BP 14 · Vierstra 8
| motif | database | consensus | bits |
|---|---|---|---|
| AC0064:NKX:Homeodomain | Vierstra | CAaTAACGACAT | 18.4 |
| NKX32.H14CORE.1.S.BSB | HOCOMOCO | TTAAGTGGcactTaA | 16.8 |
| NKX63.H14CORE.0.SM.BSMB | HOCOMOCO | ATTAATgAC | 14.5 |
| NKX31.H14CORE.0.PSM.APSMA | HOCOMOCO | TTAAGTGG | 13.2 |
| NKX61.H14CORE.0.PS.APSA | HOCOMOCO | AttAATggghwtTTAAT | 12.4 |
| Nkx3-2 (MA0122.4) | JASPAR | CACTTA | 11.9 |
| NANOG.H14CORE.0.P.BPBlooks like HD_POU | HOCOMOCO | cctTTgtnATGCAAA | 11.9 |
| NKX61.H14CORE.2.S.BSB | HOCOMOCO | TTAATtA | 11.8 |
| M05012_2.00 | CIS-BP | TAAGTGG | 11.1 |
| NKX25.H14CORE.0.PSM.APSMA | HOCOMOCO | TtAAGTGg | 11 |
| AC0134:NKX:Homeodomain | Vierstra | TCAAGga | 11 |
| Nkx3-1 (MA0124.3) | JASPAR | CCACTTA | 10.8 |
| Nkx2-1 (MA1994.2) | JASPAR | CACTTGA | 10.8 |
| NANOG.H14CORE.2.M.CMC | HOCOMOCO | GTTAATGat | 10.8 |
| NKX25.H14CORE.1.P.BPBlooks like TBX | HOCOMOCO | gTGhTGAcAG | 10.8 |
| AC0106:NKX:Homeodomain | Vierstra | TAAGTAT | 10.8 |
| NKX2-3 (MA0672.2) | JASPAR | cCACTTgA | 10.6 |
| NKX2-2 (MA1645.2) | JASPAR | CCACTCAA | 10.6 |
| NKX23.H14CORE.0.SM.BSMB | HOCOMOCO | tTAAGTG | 10.6 |
| NKX62.H14CORE.0.SM.BSMB | HOCOMOCO | TTAATtA | 10.5 |
| M05242_2.00 | CIS-BP | tTAATGAs | 10.5 |
| NKX6-3 (MA1530.2) | JASPAR | sTCATTAa | 10.4 |
| M05255_2.00 | CIS-BP | TAAGTGG | 10.4 |
| NKX2-5 (MA0063.3) | JASPAR | CACTCAA | 10.2 |
| NKX32.H14CORE.0.PSM.APSMA | HOCOMOCO | TAAGTG | 10.2 |
| NKX28.H14CORE.0.SM.BSMB | HOCOMOCO | TtAAGTG | 9.8 |
| M05184_2.00looks like HD_HOX_ANT | CIS-BP | TAATTa | 9.6 |
| Nanog (MA2339.1) | JASPAR | GCAATCA | 9.5 |
| NANOG.H14CORE.1.P.BPB | HOCOMOCO | TGATGGcT | 9.3 |
| NKX61.H14CORE.1.P.BPB | HOCOMOCO | tTAATGA | 9.2 |
| NKX6-1 (MA0674.2) | JASPAR | TaATTAA | 9.1 |
| NKX2-4 (MA2003.2) | JASPAR | cCACTTgA | 9 |
| NKX2-8 (MA0673.2) | JASPAR | CACTTgA | 8.9 |
| M05042_2.00 | CIS-BP | AAGTG | 8.9 |
| AC0027:NKX:Homeodomain | Vierstra | AAGTG | 8.6 |
| AC0029:NKX/MSANTD:['Homeo_domain_factors'] | Vierstra | CACTCAa | 8.5 |
| NKX21.H14CORE.0.P.BPB | HOCOMOCO | TtGAGtG | 7.7 |
| NKX22.H14CORE.0.P.BPB | HOCOMOCO | TtgAGtG | 7.1 |
| AC0062:NKX:Homeodomain | Vierstra | TcAACGa | 6.9 |
| NKX6-2 (MA0675.2) | JASPAR | taATTA | 6.8 |
| AC0031:NKX:Homeodomain | Vierstra | ggAGaG | 6.4 |
| M05015_2.00looks like HD_Q50 | CIS-BP | TaAtbG | 5.1 |
| M00466_2.00Mus musculus 1 | CIS-BP | TAATda | 4.9 |
| M00320_2.00 | CIS-BP | AAGTG | 4.7 |
| M00442_2.00Mus musculus 1 | CIS-BP | TAAT | 4.5 |
| M00497_2.00Mus musculus 0.93 | CIS-BP | aAGtg | 4 |
| M00503_2.00Mus musculus 1 | CIS-BP | TAAT | 3.7 |
| M00385_2.00Mus musculus 1 | CIS-BP | aAGtg | 3.5 |
| AC0030:NKX:Homeodomain | Vierstra | aAGtg | 3.5 |
| M00438_2.00Mus musculus 1 | CIS-BP | aAGtg | 3.4 |
| M00301_2.00 | CIS-BP | aAGtg | 3.3 |
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.