HSF
Consensus: nTTCnnGAAnnTTCn
Heat shock factors bind the heat shock element (HSE), an array of the 5-base unit nGAAn in alternating orientation: nTTCn-nGAAn-nTTCn and longer runs. The DNA-binding domain is a winged helix, and HSF proteins trimerize through a coiled coil so that three subunits each read one 5-base unit, with further trimers loading cooperatively onto longer arrays. Humans have HSF1, HSF2, HSF4, HSF5 and the Y-linked HSFY1/HSFY2. HSF1 is the stress-inducible member, held inactive as a monomer until proteotoxic stress triggers trimerization. HSEs drive heat shock protein genes, and HSF1 activity also supports malignant transformation; HSF4 mutations cause congenital cataract.
The factors
Human HSF1 is the stress-responsive member, kept as an inactive monomer by intramolecular coiled-coil interactions and by chaperone binding, and converted on proteotoxic stress into a DNA-binding trimer that is further tuned by phosphorylation, acetylation and SUMOylation 20628411. HSF2 is developmentally and differentiation-regulated, binds many of the same elements, and forms heterotrimers with HSF1 17213196. HSF4 is expressed mainly in lens and other epithelia and, unlike HSF1, is constitutively trimeric and can act as a repressor at HSF1 targets 20628411. HSF5 is poorly characterized and largely testis-restricted. HSFY1 and HSFY2 lie in multicopy on the Y chromosome within the AZFb interval and are expressed in testis 22158087. All members share an N-terminal winged-helix DNA-binding domain followed by hydrophobic repeats (HR-A/B) that mediate trimerization; the domain is among the most conserved DNA-binding domains in eukaryotes.
The motif
The recognition unit is 5 bases, nGAAn, and a functional heat shock element is an inverted array of at least three of them 3146692,2590940. The classical rendering is nTTCnnGAAnnTTCn; the two permutations, head-to-head and tail-to-tail repeats of the 5-bp unit, bind with similar affinity and with striking two-fold symmetry, which was the original evidence that the unit and not a longer palindrome is the recognition module 2590940. HSF binds arrays cooperatively, so four, five or six units give progressively higher affinity and are common in strongly inducible promoters 1899357. The crystal structure of an HSF DNA-binding domain showed a three-helix bundle capped by a four-stranded antiparallel beta sheet, a winged-helix variant of the helix-turn-helix motif 8284672, and the human HSF1 structure in complex with DNA showed how the trimer positions three such domains around the helix 26727489. Each domain reads GAA in the major groove, so most of the information sits on the GAA/TTC triplets, but the flanks are not free: the two positions just 5' of each GAA affect whether an element works 3146692, and the in vitro models here fix them as the TA of TTCTAGAA, with a weaker CG preference in GAAcgTTC. That is why database models for HSF1, HSF2 and HSF4 look like TTCTAGAA-cg-TTC or aGAA-cg-TTCTAGAA: they are the same three-unit array written with different phase and strand. The HSFY and HSF5 models are different in a real way, favoring TTCGAA and CGTTCGAA arrangements, so the family separates loosely into an HSF1/2/4 group and an HSFY/HSF5 group; overall cohesion (0.66) is the lowest in this batch. Within the HSF1/2/4 group no member is distinguishable from sequence.
Where it acts
HSEs are promoter-proximal at the classic targets: HSPA1A/HSPA1B, HSP90AA1, DNAJB1, HSPB1 and the other chaperone and co-chaperone genes carry clustered HSEs within a few hundred bases of the start site, usually with a paused RNA polymerase already engaged so that induction is a release step rather than a recruitment step 35687139. HSF1 is also found at thousands of distal enhancers, and the enhancer repertoire differs between stress types: heat shock and oxidative stress drive HSF1 and HSF2 to partly distinct gene and enhancer sets 35687139. GAGA-type factors and, in mammals, SP1 and NF-Y occupy the same promoters; NF-Y binds the CCAAT boxes of several heat shock promoters. In cancer cells, HSF1 occupies a large set of non-chaperone genes that is distinct from the classical heat shock program 22863008, and in tumor stroma it drives a separate program in cancer-associated fibroblasts. HSF1 also localizes to sex chromatin during meiotic repression.
Biology and disease
Hsf1-null mice are viable but lack thermotolerance, show tissue-specific failure of stress-inducible chaperone expression 12112007, and develop demyelination and protein aggregate accumulation in the CNS; Hsf1 is also required for oocyte meiosis. HSF1 activity extends lifespan in C. elegans together with chaperones, linking the pathway to proteostasis in aging 20628411. In cancer, HSF1 supports malignant transformation through a transcriptional program distinct from the heat shock response 22863008, high nuclear HSF1 predicts poor outcome in breast cancer, and loss of Hsf1 suppresses lymphoma in p53-deficient mice. HSF4 mutations in the DNA-binding domain cause autosomal dominant lamellar and Marner cataract 12089525, and a homozygous splice mutation causes recessive congenital cataract; Hsf4-null mice develop cataract with reduced gamma-S-crystallin. HSF2 contributes to inducible hsp expression through heterotrimerization with HSF1 rather than acting alone 17213196. HSFY copy loss accompanies AZFb deletions in azoospermia, though targeted studies argue HSFY genes are not themselves required for spermatocyte maturation 22158087, so the association may be positional.
Reading a hit
A three-unit HSE is reasonably informative, and models here reach 17-21 bits, so a good match is worth following up. Two specific traps make this family riskier than that number suggests. First, nGAAn arrays overlap other purine-rich motifs. A GGAA core is the ETS recognition sequence and GGAAA is the NFAT core, so loose HSF models score on ETS and NFAT sites and vice versa; one Vierstra archetype in this family is explicitly labelled HSF/NFATC for that reason. A single nGAAn unit is worthless as evidence. Second, and more concretely, several ChIP-derived HSF1 models in these collections are not HSEs at all but resemble the tRNA internal B-box promoter element (GTTCGAANNC) 2109308: consensus strings such as GGGTTCGAATCCC appear in CIS-BP, HOCOMOCO and Vierstra entries labelled HSF1 or HSF, and these are the entries whose label agreement with the rest of the family is lowest. Treat a GGGTTCGAATCC-like match as a possible tRNA gene rather than an HSE. The evidence codes help: HSF1, HSF2 and HSF4 grade-A or grade-B models built on SELEX plus methyl-SELEX (PSM, SM) give the canonical three-unit HSE, while the B-box-like models are ChIP-only (P). Finally, a genuine HSE hit does not mean the gene is stress-inducible, because HSF2 and HSF4 occupy the same elements constitutively and HSF4 can repress there.
How this was worked out
- 1982mediumPromoter deletion mutants, synthetic element transfer, transfection, S1 mapping6814763 6327271
Deletion analysis of the Drosophila hsp70 promoter in COS cells shows that residues -47 to -66, upstream of the TATA box, are needed for heat-inducible transcription; the region carries a sequence shared by heat shock genes, and a synthetic copy of the inverted-repeat consensus CT-GAA-TTC-AG makes a heterologous promoter heat-inducible.
Still open then: What bound the element, and whether the palindrome or a smaller unit was the recognition module.
Heat shock elements are modular, built from GAA blocks in alternating orientation at 2-nucleotide intervals, giving a 5-nucleotide periodicity; functional elements need three or more blocks, and the two nucleotides upstream of each GAA affect competence.
The minimal site for stable binding by Drosophila HSF is an inverted repeat of the 5-bp unit nGAAn; head-to-head and tail-to-tail arrangements bind equally, footprints grow with each added unit, and cross-linking indicates HSF binds as a trimer.
HSF binding is cooperative both within a trimer and between trimers: occupying one of two adjacent trimer sites raises binding to the second by more than 2000-fold, explaining why strong heat shock promoters carry multiple elements.
The 1.8 A structure of the HSF DNA-binding domain is a three-helix bundle capped by a four-stranded antiparallel beta sheet, a variant of the helix-turn-helix fold.
Still open then: The domain was solved without DNA, so base contacts were inferred.
In highly malignant human cancer cells HSF1 drives a transcriptional program distinct from the heat shock response, including many non-chaperone genes, and the program is active in primary breast, colon and lung tumors and associated with metastasis and death.
Cocrystal structures of the human HSF1 DNA-binding domain with heat shock element DNA, compared with the paralogue Skn7, show that single amino acid changes in the domain can switch DNA-binding specificity, and support a model of the trimer embracing the element.
Still open then: Whether natural paralogues such as HSFY use such substitutions to read different sites.
HSF1 and HSF2 bind distinct genes and enhancers in oxidative stress and heat shock, release paused RNA polymerase II at promoters, and HSF1 occupies many chaperone genes under oxidative stress without activating them.
Still unresolved
- The GGGTTCGAATCC-like models labelled HSF1 or HSF are ChIP-only and match the tRNA B-box, and whether they reflect genuine HSF1 occupancy at tRNA genes or contamination of HSF1 peak sets by heavily occupied tRNA loci has not been established.
- Whether HSFY and HSF5 have their own specificity is not resolved: the in vitro HSFY1 models favor TTCGAA-type arrays unlike HSF1/2/4, and the two low-grade HSF5 models disagree with each other (ACGTTCYA versus AACATTCCA, the latter resembling a TEAD site); single-residue changes can switch HSF specificity, but in vivo targets of HSFY and HSF5 are unknown.
The 5-base unit nGAAn is useless alone; require at least three alternating units. nGAAn arrays cross-score with ETS (GGAA) and NFAT (GGAAA) sites, and one archetype here is labelled HSF/NFATC for that reason. Several ChIP-only models labelled HSF1 or HSF match the tRNA B-box (GGGTTCGAATCC) rather than an HSE and have the lowest label agreement in the family; a hit of that shape may be a tRNA gene. The family also splits into an HSF1/2/4 group and a distinct HSFY/HSF5 group, and cohesion is the lowest in this batch. A hit does not imply stress-inducibility, since HSF2 and HSF4 bind the same elements constitutively.
References
- Perisic O, Xiao H, Lis JT. Stable binding of Drosophila heat shock factor to head-to-head and tail-to-tail repeats of a conserved 5 bp recognition unit. Cell 1989;59:797-806. PMID 2590940
- Xiao H, Perisic O, Lis JT. Cooperative binding of Drosophila heat shock factor to arrays of a conserved 5 bp unit. Cell 1991;64:585-93. PMID 1899357
- Harrison CJ, Bohm AA, Nelson HC. Crystal structure of the DNA binding domain of the heat shock transcription factor. Science 1994;263:224-7. PMID 8284672
- Neudegger T, Verghese J, Hayer-Hartl M, et al. Structure of human heat-shock transcription factor 1 in complex with DNA. Nat Struct Mol Biol 2016;23:140-6. PMID 26727489
- Akerfelt M, Morimoto RI, Sistonen L. Heat shock factors: integrators of cell stress, development and lifespan. Nat Rev Mol Cell Biol 2010;11:545-55. PMID 20628411
- Himanen SV, Puustinen MC, Da Silva AJ, et al. HSFs drive transcription of distinct genes and enhancers during oxidative stress and heat shock. Nucleic Acids Res 2022;50:6102-6115. PMID 35687139
- Zhang Y, Huang L, Zhang J, et al. Targeted disruption of hsf1 leads to lack of thermotolerance and defines tissue-specific regulation for stress-inducible Hsp molecular chaperones. J Cell Biochem 2002;86:376-93. PMID 12112007
- Mendillo ML, Santagata S, Koeva M, et al. HSF1 drives a transcriptional program distinct from heat shock to support highly malignant human cancers. Cell 2012;150:549-62. PMID 22863008
- Bu L, Jin Y, Shi Y, et al. Mutant DNA-binding domain of HSF4 is associated with autosomal dominant lamellar and Marner cataract. Nat Genet 2002;31:276-8. PMID 12089525
- Ostling P, Björk JK, Roos-Mattjus P, et al. Heat shock factor 2 (HSF2) contributes to inducible expression of hsp genes through interplay with HSF1. J Biol Chem 2007;282:7077-86. PMID 17213196
- Kichine E, Rozé V, Di Cristofaro J, et al. HSFY genes and the P4 palindrome in the AZFb interval of the human Y chromosome are not required for spermatocyte maturation. Hum Reprod 2012;27:615-24. PMID 22158087
- Gaëta BA, Sharp SJ, Stewart TS. Saturation mutagenesis of the Drosophila tRNA(Arg) gene B-Box intragenic promoter element: requirements for transcription activation and stable complex formation. Nucleic Acids Res 1990;18:1541-8. PMID 2109308
- Pelham HR. A regulatory upstream promoter element in the Drosophila hsp 70 heat-shock gene. Cell 1982;30:517-28. PMID 6814763
- Amin J, Ananthan J, Voellmy R. Key features of heat shock regulatory elements. Mol Cell Biol 1988;8:3761-9. PMID 3146692
- Pelham HR, Bienz M. A synthetic heat-shock promoter element confers heat-inducibility on the herpes simplex virus thymidine kinase gene. EMBO J 1982;1:1473-7. PMID 6327271
24 motifs in this family — JASPAR 3 · HOCOMOCO 10 · CIS-BP 6 · Vierstra 5
| motif | database | consensus | bits |
|---|---|---|---|
| M05515_2.00 | CIS-BP | tTCGAACGTTTCGAACG | 22.7 |
| M07977_2.00 | CIS-BP | GTCccGGGTTCGAaTCCC | 22.1 |
| HSF1.H14CORE.0.PSM.APSMA | HOCOMOCO | aGAAcgTTCTAGAA | 20.9 |
| AC0664:HSF:HSF | Vierstra | TCccGGGTTCGAaTCC | 20.1 |
| HSF1 (MA0486.2) | JASPAR | TTCTAGAAcgTTC | 18.7 |
| AC0688:HSF:HSF | Vierstra | ACaTTCTAGAACGT | 18.3 |
| HSFY1.H14CORE.1.S.BSB | HOCOMOCO | CgTTCGAAaCGTTCGA | 18.2 |
| M02744_2.00 | CIS-BP | GAAtGTTCTAGA | 18.2 |
| HSF4.H14CORE.0.SM.BSMB | HOCOMOCO | taGAAcgTTCTAGAA | 17.9 |
| AC0685:HSF:HSF | Vierstra | gGAAcgttvnnnnnvncAACgTTCc | 17.9 |
| HSF2.H14CORE.0.PSM.APSMA | HOCOMOCO | GAAcgTTCTAGAA | 17.6 |
| M05519_2.00 | CIS-BP | ACgTTCTAGAACGT | 17.4 |
| HSF2 (MA0770.1) | JASPAR | TTCTAGAAcdTTC | 17.2 |
| HSF1.H14CORE.2.P.BPB | HOCOMOCO | CgcGGGTTCGAdtCCc | 14.9 |
| HSF4 (MA0771.1) | JASPAR | TTCtaGAAbnTTC | 13.6 |
| AC0660:HSFY:HSF | Vierstra | TTCGAAacgTTCGA | 13.6 |
| HSFY1.H14CORE.0.S.BSB | HOCOMOCO | CGTTCGAA | 13.5 |
| HSF5.H14CORE.0.M.CMC | HOCOMOCO | AACATTCcA | 13.4 |
| M05509_2.00 | CIS-BP | GAAcgTTCtAGAA | 13.2 |
| HSFY1.H14CORE.2.M.BMB | HOCOMOCO | CaTTCbAanCgTTCbA | 12.6 |
| HSF5.H14CORE.1.S.CSC | HOCOMOCO | ACGTTCyA | 11.8 |
| M03326_2.00 | CIS-BP | TTCGAAhcgtTCGAA | 11.7 |
| HSF1.H14CORE.1.P.BPB | HOCOMOCO | TTCTGGAA | 11.6 |
| AC0686:HSF/NFATC:HSF | Vierstra | GAAngTTCtaGA | 11.5 |
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.