USF
USF1 and USF2 are ubiquitously expressed bHLH-LZ proteins, first purified as the adenovirus major late transcription factor (MLTF), that bind DNA as homodimers and heterodimers. They read a canonical E-box core, but with a characteristic extended palindromic preference, GTCACGTGAC: the T and A flanks are what separate a USF site from a MYC-MAX site, which prefers C and G there. Because both proteins are constitutive, USF motifs turn up in a very large number of accessible promoters and CpG islands in every cell type, making USF one of the most frequently recovered motifs in promoter and footprint analyses. Human genetics links USF1 variants to familial combined hyperlipidemia.
The factors
USF1 and USF2 are ubiquitous basic helix-loop-helix leucine zipper (bHLH-LZ) proteins of about 43 and 44 kDa, originally purified as the factor that stimulates transcription from the adenovirus major late promoter and therefore also called MLTF 4075392,3403559. They bind DNA as homodimers 2249772 and as USF1-USF2 heterodimers; besides the bHLH-LZ they carry a USF-specific region (USR) immediately N-terminal to the basic domain that is required for activation but not for binding. A third gene, USF3 (KIAA2018), encodes a much larger protein with a divergent bHLH and no leucine zipper, and is not a simple USF1/USF2 paralogue. Both USF1 and USF2 are expressed in essentially all tissues, which is one reason USF sites appear in so many promoters. In mice, loss of either gene alone is tolerated and the two cross-regulate asymmetrically, with Usf2 loss reducing total USF DNA-binding activity more than Usf1 loss; combined loss was reported to be embryonic lethal 9520440. The atlas assigns 8 motifs to this family (2 JASPAR, 3 HOCOMOCO, 2 CIS-BP, 1 Vierstra archetype) over the symbols USF1, USF2 and USF3, with high cohesion (0.94) and USF1.H14CORE.0.PSM.A as the representative.
The motif
USF reads the canonical E-box CACGTG, but what defines the family in practice is the extended, strongly palindromic form GTCACGTGAC: a pyrimidine-purine flank on each side that USF prefers and that MYC-MAX does not 4075392,8052536. The bHLH-LZ fold places one basic helix in each half of the major groove, where a conserved histidine, glutamate and arginine read the CAC triplet; the crystallographic analysis of the USF b/HLH/Z domain showed the same recognition geometry as Max, with the leucine zipper fixing dimer choice 8306960. Because core contacts are shared across the bHLH-ZIP class, almost all of the discriminating information in a USF model sits in the two flanking positions, and in vitro base-preference measurements confirmed that USF and MYC family proteins differ mainly there rather than in the core 8052536. Flanking preference of this kind is largely mediated by local DNA shape rather than by new base contacts 23562153. Variants matter: the JASPAR USF1 model MA0093.4 is a CATGTG-type site (GTCATGTGAC), so USF tolerates the class-A-like half-site as well. The central CG of the core is a CpG, so occupancy can be modulated by DNA methylation; the HOCOMOCO USF1 and USF2 models carry both HT-SELEX and methyl-HT-SELEX support 37971293. The HOCOMOCO USF3 model is an outlier: its consensus GgCCACGTGGg has MYC-like C/G flanks and its nearest neighbours across the four databases sit outside this family.
Where it acts
USF1 and USF2 are promoter factors. They occupy tens of thousands of accessible, mostly promoter-proximal and CpG-island regions in essentially every cell type examined, which makes USF one of the most commonly recovered motifs in promoter analyses and in DNase footprint atlases. Because both proteins are constitutive, USF occupancy is not a cell-type marker; the dependence is on accessibility and on methylation state rather than on a tissue-restricted partner. USF sites are frequently found close to TATA-less initiators, in housekeeping and metabolic gene promoters, and in composite arrangements with other promoter-proximal factors. The Vierstra footprint archetype covering these motifs is labelled AC0066:USF/BHLHE:bHLH - that is, USF motifs cluster with BHLHE40/41 motifs and were not separable in the footprint data 32728250.
Biology and disease
USF1/USF2 targets are enriched for lipid and glucose metabolism, stress response and cell-cycle genes, and the pair have been characterized as stress-responsive general regulators rather than as a developmental switch 16162174. The clearest human genetic link is metabolic: variants in USF1 are associated with familial combined hyperlipidemia, one of the more common inherited dyslipidemias, in Finnish and subsequently other cohorts 14991056,15976322. USF3 is a separate story; a germline compound heterozygous poly-glutamine deletion in USF3 has been reported in predisposition to heritable and sporadic epithelial thyroid carcinoma 28011713. Because USF binding is methylation-sensitive, USF sites are also discussed in the context of promoter CpG island hypermethylation.
Reading a hit
A GTCACGTGAC match is better evidence for USF than a bare CACGTG, but it is not proof. MITF, TFE3, TFEB and TFEC prefer exactly the same extended palindrome, and BHLHE41 and CLOCK-BMAL1 models look very similar; a strong T/A-flanked E-box narrows the candidate set to USF plus MITF/TFE plus BHLHE41 rather than to one factor. USF1 and USF2 are not distinguishable from each other by sequence at all, so a hit identifies USF, not a paralogue. Do not read the USF3 model as a USF site: it is MYC-flanked and clusters away from USF1/USF2. Given the number of genomic E-boxes, treat a single hit as a hypothesis.
How this was worked out
USF is partially purified from HeLa nuclei as a factor that stimulates the adenovirus major late promoter and footprints the palindrome GGCCACGTGACC upstream of the TATA box.
Still open then: The protein's structure and its relationship to other factors were unknown.
Purified 43- and 44-kDa USF forms bind the major late site with identical affinity, and binding is strongly influenced by neighbouring sequence.
cDNA cloning places USF in the MYC-related bHLH-leucine-zipper group and shows that it binds DNA as a dimer.
The USF bHLH co-crystal on CACGTG reproduces the MAX recognition geometry, and the zipper-containing domain forms a bivalent tetramer able to bind two sites.
Binding-site selection defines the optimum USF site as RYCACGTGRY and shows that the base at ±4 discriminates among E-boxes: a T at -4 inhibits MYC binding but not USF or MAX.
Still open then: Why a base outside the contacted core changes affinity.
Usf1 and Usf2 knockout mice are viable with asymmetric cross-regulation of the remaining paralogue, while combined loss is embryonic lethal.
Familial combined hyperlipidemia is linked and associated with USF1 in Finnish families, and the association is replicated in Mexican families in 2005.
Still open then: No causal variant or mechanism was identified.
- 2013mediumGenomic-context protein-binding microarrays, DNA shape analysis (yeast bHLH factors)23562153
Flanking sequence beyond the E-box changes bHLH binding through DNA shape, offering a mechanism for flank preferences such as USF's.
Still open then: Shown for yeast Cbf1 and Tye7, not directly for USF.
Still unresolved
- Whether USF1 variants cause familial combined hyperlipidemia is not settled: linkage and association were found in Finnish and Mexican families, but no causal variant or mechanism has been established, and one of the associated Mexican SNPs lies in the neighbouring F11R gene, which could not be excluded.
- Whether anything in the sequence separates USF sites from MiT/TFE and BHLHE41 sites is unresolved: all prefer GTCACGTGAC, and the Vierstra footprint archetype AC0066 does not split USF from BHLHE40/41.
- The databases disagree about the USF1 half-site: the JASPAR USF1 model MA0093.4 reads GTCATGTGAC while the HOCOMOCO and CIS-BP USF1 models read GTCACGTG, and the models do not settle whether USF1 genuinely prefers the CATGTG form at some sites.
- Whether USF3 belongs with USF1/USF2 as a DNA-binding factor is unclear: it lacks the leucine zipper, and its well-supported HOCOMOCO model reads a MYC-like GgCCACGTGGg whose nearest neighbours lie outside this family.
The extended GTCACGTGAC form is shared with MITF, TFE3, TFEB, TFEC and BHLHE41, so it narrows a hit to a T/A-flanked E-box group, not to USF. USF1 and USF2 are indistinguishable by sequence. The Vierstra footprint archetype that carries these motifs is labelled USF/BHLHE, which is an explicit statement that USF and BHLHE40/41 motifs did not separate. The HOCOMOCO USF3 model has MYC-like CCACGTGG flanks and its nearest neighbours lie outside this family, so it should not be used as a USF site definition. USF occupancy is also methylation-sensitive because the core contains a CpG, so a sequence match says nothing about binding in a methylated promoter.
References
- Sawadogo M, Roeder RG. Interaction of a gene-specific transcription factor with the adenovirus major late promoter upstream of the TATA box region. Cell 1985;43:165-75. PMID 4075392
- Sawadogo M. Multiple forms of the human gene-specific transcription factor USF. II. DNA binding properties and transcriptional activity of the purified HeLa USF. J Biol Chem 1988;263:11994-2001. PMID 3403559
- Gregor PD, Sawadogo M, Roeder RG. The adenovirus major late transcription factor USF is a member of the helix-loop-helix group of regulatory proteins and binds to DNA as a dimer. Genes Dev 1990;4:1730-40. PMID 2249772
- Ferré-D'Amaré AR, Pognonec P, Roeder RG, et al. Structure and function of the b/HLH/Z domain of USF. EMBO J 1994;13:180-9. PMID 8306960
- Bendall AJ, Molloy PL. Base preferences for DNA binding by the bHLH-Zip protein USF: effects of MgCl2 on specificity and comparison with binding of Myc family members. Nucleic Acids Res 1994;22:2801-10. PMID 8052536
- Sirito M, Lin Q, Deng JM, et al. Overlapping roles and asymmetrical cross-regulation of the USF proteins in mice. Proc Natl Acad Sci U S A 1998;95:3758-63. PMID 9520440
- Pajukanta P, Lilja HE, Sinsheimer JS, et al. Familial combined hyperlipidemia is associated with upstream transcription factor 1 (USF1). Nat Genet 2004;36:371-6. PMID 14991056
- Corre S, Galibert MD. Upstream stimulating factors: highly versatile stress-responsive transcription factors. Pigment Cell Res 2005;18:337-48. PMID 16162174
- Gordân R, Shen N, Dror I, et al. Genomic regions flanking E-box binding sites influence DNA binding specificity of bHLH transcription factors through DNA shape. Cell Rep 2013;3:1093-104. PMID 23562153
- Ni Y, Seballos S, Fletcher B, et al. Germline compound heterozygous poly-glutamine deletion in USF3 may be involved in predisposition to heritable and sporadic epithelial thyroid carcinoma. Hum Mol Genet 2017;26:243-257. PMID 28011713
- Vierstra J, Lazar J, Sandstrom R, et al. Global reference mapping of human transcription factor footprints. Nature 2020;583:729-736. PMID 32728250
- 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
- Huertas-Vazquez A, Aguilar-Salinas C, Lusis AJ, et al. Familial combined hyperlipidemia in Mexicans: association with upstream transcription factor 1 and linkage on chromosome 16q24.1. Arterioscler Thromb Vasc Biol 2005;25:1985-91. PMID 15976322
8 motifs in this family — JASPAR 2 · HOCOMOCO 3 · CIS-BP 2 · Vierstra 1
| motif | database | consensus | bits |
|---|---|---|---|
| USF3.H14CORE.0.PSGIB.APSGIBA | HOCOMOCO | GgCCACGTGGg | 16.3 |
| USF2 (MA0526.5) | JASPAR | GTCACGTGAC | 16.1 |
| M07807_2.00 | CIS-BP | GTCACGTGacc | 15.9 |
| M07785_2.00 | CIS-BP | GTCACGTGgcc | 15.6 |
| USF2.H14CORE.0.PSM.APSMA | HOCOMOCO | GTCACGTG | 15.5 |
| USF1.H14CORE.0.PSM.APSMA | HOCOMOCO | GTCACGTG | 14.8 |
| USF1 (MA0093.4) | JASPAR | GTCATGTGAC | 12.8 |
| AC0066:USF/BHLHE:bHLH | Vierstra | TCACGTGA | 10.7 |
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.