RFX
RFX proteins recognize the X-box, a pseudo-palindrome built from two GTTRCC-like half-sites separated by zero to three bases. The DNA-binding domain is a 76-residue winged helix that reads bases with its beta-hairpin wing rather than with the usual recognition helix, and two domains bind one X-box cooperatively. Humans have eight members, RFX1-RFX8. RFX1-3 form homodimers and heterodimers and drive ciliary gene programs; RFX5 works in a complex with RFXAP and RFXANK to switch on MHC class II genes; RFX6 is restricted to pancreatic islet and enteroendocrine cells. The full 13-14 bp palindrome is one of the most information-rich motifs in vertebrate collections, which makes RFX hits unusually trustworthy.
The factors
The human RFX family has eight genes. All share a highly conserved 76-residue DNA-binding domain (DBD) 10706293; most also carry dimerization domains (B, C and DIM regions) that RFX1, RFX2 and RFX3 use to form homodimers and heterodimers with each other. RFX1-3 are the ciliogenic branch and are broadly expressed, with RFX2 and RFX3 concentrated in tissues that build motile cilia. RFX4 is largely brain- and testis-restricted and produces isoforms lacking parts of the dimerization region. RFX5 does not dimerize with other RFX proteins; it functions in the RFX (also called RFXANK-RFXAP-RFX5) complex that, together with the coactivator CIITA, drives MHC class I and class II transcription 10779326. RFX6 is restricted to pancreatic islet and gut enteroendocrine cells and works downstream of NEUROG3 20148032. RFX7 is broadly expressed and acts downstream of p53 34197623. RFX8 is poorly characterized in humans, and the CIS-BP human RFX8 entry is not a human measurement at all: it is transferred from a louse (Pediculus humanus) protein at 55% DBD identity 29510665,25215497.
The motif
The literature consensus, from the RFX1 crystal structure, is 5'-GTNRCC(N0-3)RGYAAC-3' 10706293. In practice vertebrate RFX models converge on the tighter form GTTGCCATGGCAAC, which is what the motif data in this family reports (GTTGCCATGG). The two forms are not in conflict: the databases are dominated by high-affinity selection experiments that pick GTTGCC as the half-site and a 2-3 bp spacer, while the literature consensus is written loosely enough to cover the weaker, variably spaced sites found in real promoters. The structural explanation is unusual. RFX1 is a winged-helix (helix-turn-helix) protein, but instead of inserting a recognition helix into the major groove it uses a beta-hairpin wing to make the base-specific contacts; two DBDs bind cooperatively on opposite half-sites of a symmetrical X-box 10706293. This is why the informative positions are the GTT/AAC ends and the CC/GG pair inside them, and why the central spacer bases are almost free. Half-sites alone are real sites: the RFX6 and RFX7 models in these databases are essentially single half-sites (GTTGCTA, GTTGCCAGG, CCTAGCAAC) with 7-12 bits, whereas the JASPAR models for RFX1-5 are full palindromes of 20-23 bits. RFX5 is in the palindrome group: its HT-SELEX model (JASPAR MA0510.3, 20.2 bits) and its grade-A HOCOMOCO model (18.6 bits) both read GTTGCCATGGCAAC, so RFX1-5 are not separable by these models. Only RFX6 and RFX7 stand apart, by the absence of a strong second half-site, and only the RFX7 half-site rests on in vitro data.
Where it acts
X-boxes are promoter-proximal more often than not, typically within a few hundred bases of the transcription start site of ciliary genes (IFT and BBS genes, dynein subunits, ciliary membrane proteins) and of MHC class II genes, where the X box sits next to the X2 box and about 15-20 bp from the Y box. The Y box is a CCAAT element bound by NF-Y, and RFX5 and NF-Y bind the class II promoter cooperatively 10779326; this is the best-characterized composite element containing an X-box. In motile ciliated epithelium RFX2 and RFX3 co-occupy promoters with FOXJ1, the forkhead factor that specifies motile cilia, and in Drosophila the equivalent partnership is Rfx with Fd3F 22698283. RFX6 in islets co-occupies enhancers with NEUROD1, PDX1 and NKX2-2. X-box counts are modest for a vertebrate motif precisely because the full palindrome is long: genome-wide scans typically recover a few thousand strong matches, most of them at ciliary and immune loci.
Biology and disease
Loss of the RFX complex causes MHC class II deficiency (bare lymphocyte syndrome, complementation groups C, D and E): RFX5, RFXAP or RFXANK mutations remove X-box binding, class II expression fails, and patients present with combined immunodeficiency 10779326. The original RF-X activity was defined by its absence in such patients' cells 3133120. RFX1-3 are the transcriptional core of ciliogenesis across animals 10882127,32725242; when X-box control of ciliary genes arose is disputed, with one survey placing it before the origin of metazoans 20441589 and another early in the animal lineage 20615967. Rfx3 mutant mice have laterality defects, hydrocephalus and impaired nodal cilia 15121860; combined Rfx1/2/3 loss ablates motile ciliogenesis 32725242. RFX6 loss of function causes Mitchell-Riley syndrome (neonatal diabetes with intestinal atresia) 20148032; heterozygous truncating RFX6 variants cause maturity-onset diabetes of the young with reduced penetrance 29026101, and Rfx6 is also required to maintain adult beta-cell identity. RFX7 is recurrently mutated in lymphoid cancers and mediates part of the p53 response, controlling genes including DDIT4 and PDCD4 34197623. An estrogen-receptor-RFX4 fusion has been reported in breast tumors 10706293.
Reading a hit
A full 13-14 bp X-box match is strong evidence, and among vertebrate motifs it is one of the few where a single high-scoring hit is worth acting on: the palindrome carries about 20 bits, so chance occurrences in the human genome are rare. What a hit does not tell you is which RFX protein. RFX1, RFX2, RFX3 and RFX4 are indistinguishable from sequence, and the practical question is which of them is expressed in your cell type. Half-site-only matches (GTTGC plus a purine) are a different matter: at 7-9 bits they occur by chance constantly, and they are the form in which the RFX6, RFX7 and RFX8 models are represented here, so hits on those models deserve much more skepticism than hits on the full-palindrome RFX1-5 models. Two provenance problems affect this family in CIS-BP: the RFX8 motif comes from a louse protein at 55% DBD identity and the RFX6 motif from guinea pig at 71%, so neither is a human measurement. One HOCOMOCO RFX3 secondary model (evidence code S, grade B) does not resemble an X-box at all and should not be used.
How this was worked out
An X-box-binding activity on the HLA-DRA promoter, named RF-X, is missing from B-cell lines of patients with congenital MHC class II deficiency, the first link between the X box and disease.
Still open then: Which protein or proteins made up RF-X, and whether the patients' defect lay in DNA binding itself.
The single C. elegans RFX factor, DAF-19, is expressed in all ciliated sensory neurons and is required for their cilia; genes shared by all ciliated neurons carry RFX sites in their promoters and depend on DAF-19.
Still open then: Whether the vertebrate paralogues had kept the ciliary role.
The 1.5 A structure of two RFX1 DNA-binding domains on a symmetrical X-box shows cooperative binding and base recognition by the beta-hairpin wing instead of the recognition helix, and frames the consensus GTNRCC(N0-3)RGYAAC.
- 2000mediumDomain mutagenesis, in vitro and in vivo complex assembly, EMSA, complementation of patient cells10779326
RFX5 is the subunit that assembles the RFX5-RFXAP-RFXANK complex on the MHC class II X box, and a separate C-terminal domain makes the complex bind cooperatively with NF-Y on the neighboring Y box.
Rfx3-deficient mice have stunted nodal cilia and frequent left-right asymmetry defects, extending the RFX-cilia link from nematodes to mammals.
Rfx6 is required for islet cell differentiation downstream of Neurog3 in mice, and recessive human RFX6 mutations cause a neonatal diabetes syndrome, giving the family its first endocrine role.
Two comparative-genomics surveys agree that RFX genes arose independently of cilia, but disagree on when X-box regulation of ciliary genes began: one places the convergence before the first metazoans, the other early in the animal lineage, with X-box footprints found only in animal ciliary genes.
Still open then: Whether choanoflagellate RFX proteins regulate ciliary genes through X-boxes.
ChIP-seq and RNA-seq in mouse ependymal cells show that RFX1, RFX2 and RFX3 occupy many of the same loci, while only RFX2 and RFX3 act, redundantly, in motile ciliogenesis.
Still open then: What makes shared occupancy produce paralogue-specific function.
Still unresolved
- When X-box control of ciliary genes arose is disputed: Chu and colleagues argue the RFX-ciliary gene convergence predates metazoans because both co-occur in choanoflagellates, while Piasecki and colleagues find X-box footprints only in animal ciliary genes and place the rewiring early in the animal lineage.
- How RFX5 reads DNA in its native complex is not settled by the motif data: the in vitro-supported RFX5 models here (JASPAR MA0510.3, HOCOMOCO grade A) are full GTTGCCATGGCAAC palindromes like those of RFX1-3, yet RFX5 does not dimerize with other RFX proteins and binds the MHC class II X box as part of the RFX5-RFXAP-RFXANK complex.
- Whether RFX6 and RFX7 genuinely prefer single half-sites is unclear: the RFX7 in vitro model is a half-site (GTTGCYANG), but the RFX6 models are either ChIP-only or inferred from guinea pig, so a weak second half-site could be a property of the protein or of the data.
Trust full-palindrome X-box hits; treat half-site hits (GTTGC alone) as near-worthless on their own, because they carry under 9 bits. Nothing in the sequence separates RFX1, RFX2, RFX3 and RFX4, so assign a member from expression, not from the match. The RFX5 models are full palindromes with in vitro support, although in cells RFX5 binds MHC class II X boxes as part of a three-protein complex. In CIS-BP the human RFX8 motif is transferred from a louse protein (55% DBD identity) and the RFX6 motif from guinea pig (71%); neither was measured on the human protein. One HOCOMOCO RFX3 secondary model is not X-box-like and looks unreliable.
References
- Gajiwala KS, Chen H, Cornille F, et al. Structure of the winged-helix protein hRFX1 reveals a new mode of DNA binding. Nature 2000;403:916-21. PMID 10706293
- Reith W, Satola S, Sanchez CH, et al. Congenital immunodeficiency with a regulatory defect in MHC class II gene expression lacks a specific HLA-DR promoter binding protein, RF-X. Cell 1988;53:897-906. PMID 3133120
- Villard J, Peretti M, Masternak K, et al. A functionally essential domain of RFX5 mediates activation of major histocompatibility complex class II promoters by promoting cooperative binding between RFX and NF-Y. Mol Cell Biol 2000;20:3364-76. PMID 10779326
- Lemeille S, Paschaki M, Baas D, et al. Interplay of RFX transcription factors 1, 2 and 3 in motile ciliogenesis. Nucleic Acids Res 2020;48:9019-9036. PMID 32725242
- Piasecki BP, Burghoorn J, Swoboda P. Regulatory Factor X (RFX)-mediated transcriptional rewiring of ciliary genes in animals. Proc Natl Acad Sci U S A 2010;107:12969-74. PMID 20615967
- Newton FG, zur Lage PI, Karak S, et al. Forkhead transcription factor Fd3F cooperates with Rfx to regulate a gene expression program for mechanosensory cilia specialization. Dev Cell 2012;22:1221-33. PMID 22698283
- Smith SB, Qu HQ, Taleb N, et al. Rfx6 directs islet formation and insulin production in mice and humans. Nature 2010;463:775-80. PMID 20148032
- Patel KA, Kettunen J, Laakso M, et al. Heterozygous RFX6 protein truncating variants are associated with MODY with reduced penetrance. Nat Commun 2017;8:888. PMID 29026101
- Coronel L, Riege K, Schwab K, et al. Transcription factor RFX7 governs a tumor suppressor network in response to p53 and stress. Nucleic Acids Res 2021;49:7437-7456. PMID 34197623
- Sugiaman-Trapman D, Vitezic M, Jouhilahti EM, et al. Characterization of the human RFX transcription factor family by regulatory and target gene analysis. BMC Genomics 2018;19:181. PMID 29510665
- Weirauch MT, Yang A, Albu M, et al. Determination and inference of eukaryotic transcription factor sequence specificity. Cell 2014;158:1431-1443. PMID 25215497
- Swoboda P, Adler HT, Thomas JH. The RFX-type transcription factor DAF-19 regulates sensory neuron cilium formation in C. elegans. Mol Cell 2000;5:411-21. PMID 10882127
- Bonnafe E, Touka M, AitLounis A, et al. The transcription factor RFX3 directs nodal cilium development and left-right asymmetry specification. Mol Cell Biol 2004;24:4417-27. PMID 15121860
- Chu JS, Baillie DL, Chen N. Convergent evolution of RFX transcription factors and ciliary genes predated the origin of metazoans. BMC Evol Biol 2010;10:130. PMID 20441589
31 motifs in this family — JASPAR 7 · HOCOMOCO 12 · CIS-BP 8 · Vierstra 4
| motif | database | consensus | bits |
|---|---|---|---|
| RFX3 (MA0798.3) | JASPAR | CGTTGCCATGGCAACg | 23.2 |
| RFX1 (MA0509.3) | JASPAR | GTTGCCATGGCAAC | 23.1 |
| RFX5 (MA0510.3) | JASPAR | GTTGCCATGGcAAC | 20.2 |
| RFX4 (MA0799.3) | JASPAR | GTTGCttaGCAAC | 20 |
| RFX2 (MA0600.3) | JASPAR | GTTgCCATGGcAAC | 19.9 |
| M05723_2.00 | CIS-BP | GTTgCCATGGcAAC | 19.8 |
| RFX5.H14CORE.0.PSG.APSGA | HOCOMOCO | GTTGCCATGGcaACgg | 18.6 |
| AC0292:RFX:RFX | Vierstra | GTTGCTAgGCAAC | 18.6 |
| RFX1.H14CORE.1.PSM.APSMA | HOCOMOCO | GTTGCCATGGhaAC | 17.7 |
| RFX2.H14CORE.1.PSM.APSMA | HOCOMOCO | GTTGCCATGGtaAC | 17.7 |
| RFX3.H14CORE.2.S.BSB | HOCOMOCO | GTTGTAGGttAGGGagg | 17.1 |
| AC0289:RFX/MXI:RFX | Vierstra | GTTGCCATGGcAAC | 15.9 |
| M03456_2.00 | CIS-BP | GTtgCCATGGcaAC | 15.6 |
| M02750_2.00 | CIS-BP | GTTgCCATGG | 14.7 |
| RFX3.H14CORE.1.PS.APSA | HOCOMOCO | GTTGCCATGG | 14.6 |
| RFX3.H14CORE.0.PSM.APSMA | HOCOMOCO | GTTGCCAtGGnaAC | 14.4 |
| RFX4.H14CORE.0.SM.BSMB | HOCOMOCO | GTTGCCATGGtaAC | 14.4 |
| RFX2.H14CORE.0.PS.APSA | HOCOMOCO | GTTGCCATGGC | 13.9 |
| RFX1.H14CORE.0.PM.APMA | HOCOMOCO | GTTGCCATGG | 13.6 |
| M08009_2.00 | CIS-BP | TCAbctGTTGCtAGG | 12.6 |
| RFX6.H14CORE.0.P.CPC | HOCOMOCO | GTTGCCaGG | 11.7 |
| Rfx6 (MA1724.2) | JASPAR | CCTAGCAAC | 11.6 |
| AC0291:RFX:RFX | Vierstra | ctGTTGCtAGG | 11.1 |
| RFX1.H14CORE.2.S.CSC | HOCOMOCO | cGTTGCT | 11 |
| M05729_2.00 | CIS-BP | cGTTGCtA | 9.3 |
| RFX7 (MA1554.2) | JASPAR | cGTTGCtA | 9.1 |
| RFX7.H14CORE.0.SM.BSMB | HOCOMOCO | GTTGCyang | 7.3 |
| M02449_2.00 | CIS-BP | GTtgCtAdG | 7.3 |
| AC0290:RFX:RFX | Vierstra | GTTGCtA | 7 |
| M02454_2.00Pediculus humanus 0.55 | CIS-BP | GTtgCtAnG | 6.4 |
| M02448_2.00Cavia porcellus 0.71 | CIS-BP | GTtgCcanG | 4 |
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.