Telobox (TERF1/TERF2/ZBTB48)

The factors

Human telomeres end in several kilobases of tandem TTAGGG repeats, bound by the six-protein shelterin complex. Two shelterin subunits supply the double-stranded DNA contact: TERF1 (TRF1) and TERF2 (TRF2) ^{7502076,9326950}. Each has an N-terminal domain (acidic in TRF1, basic in TRF2), a central TRFH homodimerization domain, and a C-terminal Myb/SANT-type helix-turn-helix DNA-binding domain ^{7502076,9326950}. The Myb domain here is a single repeat, not the three-repeat array of c-MYB, and crystal and NMR structures show that it adopts a homeodomain-like fold, the same solution the budding-yeast telomere protein Rap1 uses ^{9739097,15608617}. Both proteins bind DNA as preformed homodimers in which the two DNA-binding domains are independent structural units on a flexible tether ^{15608617,10523316}. The family name comes from the telobox, a Myb-related telomeric DNA-binding motif recognized as a conserved module across yeast, plants and humans ^8614633.

ZBTB48, also called TZAP or HKR3, is grouped here because it binds the same sequence, but it is a different protein class: a BTB/POZ zinc-finger factor, not a telobox protein. It binds telomeric and subtelomeric variant repeats directly ^28500257, competes with TRF1 and TRF2 for long telomeres ^28082411, and also acts as a conventional transcriptional activator at a small set of promoters ^28500257. TRF1 and TRF2 are broadly expressed ^9326950 and essential; ZBTB48 is a more transient telomere visitor.

The motif

The literature consensus is the telomeric repeat itself, TTAGGG, and in practice the unit the protein reads is the 7-9 bp half-site YTAGGGTTR ^10523316. High-resolution crystal structures of the TRF1 and TRF2 DNA-binding domains on telomeric DNA show that both recognize the same TAGGGTT site, that two of the three G-C base pairs characteristic of the repeat are read specifically, and that an unusually large number of ordered water molecules mediate the protein-DNA interface ^15608617. SELEX with dimeric TRF1 defined the site as bipartite: two YTAGGGTTR half-sites with essentially no constraint on distance or relative orientation, simultaneously contacted by one dimer, with the intervening DNA looped out ^10523316. A tandem repeat array therefore does not have one canonical binding register.

The matrices in this atlas all describe fragments of this repeat in one orientation or the other: AGGGTT (HOCOMOCO TERF1), AACCCTA (CIS-BP TERF2), TCCCTAA and GGTCCCTDA (ZBTB48/TZAP), GGTCCCT (a Vierstra ZBTB archetype). These are the same repeat unit read on opposite strands and phased differently, which is why the atlas medoid, TCCCTAA, looks unlike TTAGGG at first glance - it is CCCTAA, the reverse complement, shifted by one. Family cohesion is about 0.81, and the residual spread is phase and strand, not genuine specificity differences. None of these short matrices captures the dimeric bipartite logic, so a scan reports individual half-sites rather than functional dimer footprints. The members are not distinguishable from sequence: TRF1, TRF2 and ZBTB48 all read the same repeat.

Where it acts

Almost everywhere this motif occurs it is a telomeric repeat. ChIP-seq for TRF1 and TRF2 recovers telomeres, subtelomeric regions, and a defined subset of interstitial telomeric sequences (ITSs) - internal blocks of degenerate TTAGGG repeats scattered through the genome, many of them near or within gene introns - plus some centromeric and pericentromeric satellite DNA ^21423270. ITSs are themselves unstable and are recombination-prone. ZBTB48 additionally binds subtelomeric variant repeats ^28500257. So the honest description of the genomic distribution is: chromosome ends, subtelomeres, ITSs and satellite arrays, with only a small number of genuine promoter sites, such as the ZBTB48 target MTFP1 ^28500257. This motif is not a promoter element in the ordinary sense, and CpG-island promoter enrichment is not expected.

Biology and disease

TRF1 is a negative regulator of telomere length: overexpression shortens telomeres progressively and a dominant-negative form lengthens them, in telomerase-positive cells ^9034193. TRF2 protects chromosome ends from being processed as double-strand breaks; its inhibition causes end-to-end chromosome fusions ^9476899 and triggers an ATM-dependent damage response, and TRF2 is implicated in the t-loop structure at telomeres, although the crystal structures show no DNA distortion that would explain that activity ^15608617. ZBTB48/TZAP binds preferentially to long telomeres, where shelterin density is low, and initiates telomere trimming - rapid deletion of telomeric repeats - which sets an upper limit on telomere length ^28082411; it is also a negative regulator of telomere length in cancer cells regardless of their telomere-maintenance mechanism ^28500257. Telomere dysfunction through these routes connects to senescence, genome instability in cancer, and the telomere-biology disorders, and TRF1 and TRF2 are reviewed as shelterin components in that context.

Reading a hit

The practical rule for this family is different from every other family in the atlas: a match is a telomeric repeat, so ask where it is before asking who bound it. If the hit is at a chromosome end or in a subtelomere, it is telomeric. If it is internal, the most likely explanation is an interstitial telomeric sequence or a satellite array, and a TTAGGG-rich internal block will score highly whether or not any protein binds it. Only after excluding those should a hit be read as a regulatory element. The members cannot be separated by sequence - TRF1, TRF2 and ZBTB48 all read TTAGGG - and the short matrices available represent single half-sites rather than the flexible bipartite site a TRF dimer actually occupies, so scores will not track dimer affinity. The CIS-BP TERF2 model is inferred from mouse Terf2 at 92% DNA-binding-domain identity rather than measured on the human protein, the HOCOMOCO TZAP model is ChIP-only (grade B), and the Vierstra archetype is labelled for the broad ZBTB group rather than for ZBTB48 specifically.