CDC5L

The factors

CDC5L (also CDC5, CEF1) was named for Schizosaccharomyces pombe cdc5+, cloned in 1994 as an essential gene whose product has homology to c-MYB ^8313892. The protein carries two Myb/SANT-like repeats near the N terminus - fewer than the three repeats of c-MYB itself - followed by a long low-complexity C-terminal region with a latent transcriptional activation domain ^11082045. A mammalian homologue was shown to regulate G2 progression and mitotic entry ^9468527. There is a single human gene and no close paralogue, so within this family there is no member ambiguity; the ambiguity is about whether the protein functions as a DNA-binding transcription factor at all.

The best-established function is in splicing. CDC5L is the scaffold of a multiprotein complex in nuclear extract - the CDC5L/Prp19 complex, containing PRPF19, PLRG1, BCAS2, CTNNBL1, HSPA8 and others - that joins the spliceosome in an ATP-dependent step and is required for the second catalytic step of pre-mRNA splicing ^11101529. Immunodepleting it blocks splicing product formation without preventing spliceosome assembly ^11101529. The complex is resolved in cryo-EM structures of the human catalytic spliceosome ^29301961. CDC5L also interacts with ATR and is required for the S-phase cell-cycle checkpoint ^19633697. Orthologues behave similarly: Arabidopsis CDC5 (AtCDC5) is part of the MOS4-associated complex (MAC) with MOS4 and PRL1, which is required for plant innate immunity and for splicing and microRNA biogenesis ^{17575050,28947490}; AtCDC5 was shown to have DNA-binding activity ^{8917598,17575050}, which is where the plant-derived motif in these databases comes from ^25215497. CDC5L is broadly and constitutively expressed, as befits a core splicing factor.

The motif

There is no reliable consensus for human CDC5L, and this entry does not assert one. A 12-bp sequence bound specifically and with high affinity by the N-terminal region of human Cdc5 was identified in 2000, together with human genomic fragments containing similar sequences that bound with comparable specificity, and the interaction could activate transcription in a reporter ^11082045. That paper explicitly noted that previous attempts by others had failed to find a preferential binding site for Cdc5-family proteins ^11082045, which sets the expectation for how well-determined this specificity is.

The two models in this atlas disagree with one another, and the disagreement is the single most important fact about this family. The CIS-BP model, M02698, has consensus GCGCTGAGC - and it is not measured on the human protein: it is transferred from Arabidopsis thaliana CDC5 at 87% DNA-binding-domain identity ^25215497. The Vierstra archetype AC0311:CDC5L:Myb/SANT has consensus GATTTAACATA ^32728250, an AT-rich sequence with no resemblance to the CIS-BP model. Family cohesion is 0.21, essentially no agreement, and neither model's best match in any other family exceeds about 0.72. There are no JASPAR or HOCOMOCO models. So the honest description is that this family contains two mutually inconsistent models, one of which is a cross-species inference and the other of which is a footprint-derived archetype named for a factor whose footprints may reflect something else entirely.

Where it acts

There is no well-mapped genomic distribution for human CDC5L binding sites. The 2000 study recovered human genomic fragments containing sequences related to its 12-bp site and suggested these are specific, low-abundance sequences rather than a common promoter element ^11082045, but genome-wide occupancy data for CDC5L as a DNA-binding factor are essentially absent. Because CDC5L is abundant, nuclear and associated with co-transcriptional splicing machinery, it is highly likely to crosslink to actively transcribed gene bodies and promoters in ChIP experiments without binding DNA sequence-specifically. That is precisely the situation in which a footprint- or ChIP-derived "CDC5L motif" is likely to be the motif of whatever else occupies those regions. In Arabidopsis, AtCDC5 acts through the MAC complex on defence gene expression and on splicing ^{17575050,28947490}, again a mixture of direct and indirect routes.

Biology and disease

Loss of Cdc5 function is lethal in fission yeast ^8313892. In mammalian cells, overexpressing CDC5L shortened G2, and a dominant-negative mutant lacking the C-terminal activation domain slowed G2 progression and delayed entry into mitosis ^9468527. The S-phase checkpoint function requires the CDC5L-ATR interaction ^19633697, linking it to the DNA damage response. In the Arabidopsis system, MAC components including AtCDC5 are required for SNC1-mediated innate immunity ^17575050 and for correct splicing under stress ^28947490. In human disease, CDC5L is reported as overexpressed in several cancers, but the mechanistic accounts are inconsistent and mostly correlative, and it is not established whether any such effect runs through DNA binding rather than through splicing. There is no monogenic human disease attributed to CDC5L. The literature is thin on the transcription-factor side and rich on the splicing side, and motif databases reflect the former more confidently than the evidence supports.

Reading a hit

Do not read a CDC5L hit as evidence that CDC5L binds your sequence. The two available models are mutually contradictory (cohesion 0.21), which alone should prevent a specific call. The CIS-BP model is inherited from Arabidopsis CDC5 at 87% DNA-binding-domain identity, so it reports a plant protein's preference, and a GCGCTGAGC match in human sequence is at least as likely to be a coincidence or another Myb/SANT-type or CRE-adjacent element. The Vierstra archetype is a footprint cluster labelled CDC5L; footprint archetypes are named after the factor group that dominated a cluster, not after a demonstrated binder, and CDC5L's abundance at transcribed regions makes an indirect explanation plausible. There is no paralogue problem, but there is a class problem: CDC5L's demonstrated essential activity is in splicing, and a DNA match gives you no purchase on that. If a CDC5L hit matters to your analysis, verify it directly rather than relying on the database label.