Motif Tools
Family atlas / C2H2 zinc finger

ZNF384 / ZNF362

C2H2 zinc fingerAAAAAAAA5 motifsaka CIZ, Nmp4, NMP4
GCATATGTTACTATTCGAT
representative: AC0215:ZNF:C2H2_ZF · data consensus TTTTTTTT
members tight · 0.9stated AAAAAAAA · data TTTTTTTT2 in vitro11 references, all PMIDs verified

ZNF384 and its near-identical paralogue ZNF362 are C2H2 zinc-finger proteins that bind homopolymeric runs of A on one strand and T on the other. They are unusual among zinc fingers in reading the minor groove of an A-tract rather than the major groove of a mixed sequence, and only three of their fingers contact DNA. ZNF384 was cloned twice, as CIZ, a partner of the adaptor p130Cas that activates matrix metalloproteinase genes, and as Nmp4, a nuclear matrix factor at the type I collagen alpha-1 promoter. In human leukaemia ZNF384 is a recurrent fusion partner: rearrangements joining its C-terminal region to EP300, TCF3, TAF15 and others define a subtype of B-cell precursor acute lymphoblastic leukaemia.

The factors

ZNF384 and ZNF362 are paralogues whose zinc-finger arrays are close enough that no assay distinguishes their sequence preferences; the family therefore has two members and one motif. Splice variants of ZNF384 contain between five and eight C2H2 fingers, together with an SH3-binding region overlapping a putative AT-hook and a polyglutamine-alanine repeat 11867614. Mapping the DNA-binding function showed that only fingers 2, 3 and 6 are needed for association with the A-tract element 11867614, so most of the array does something other than read bases. ZNF384 was described independently as CIZ, a protein that binds the focal-adhesion adaptor p130Cas and activates matrix metalloproteinase promoters 10669742, and as Nmp4, one of a set of nuclear matrix proteins occupying the rat COL1A1 promoter 10972987. The N-terminus carries a strong activation domain that is masked in the full-length protein and becomes productive once the fingers engage DNA, and the polyglutamine-alanine repeat then contributes to activation, suggesting an allosteric switch triggered by binding 11867614. Both proteins are broadly expressed.

The motif

This is the atlas entry whose consensus is least word-like. The element is a homopolymeric (dA:dT) tract: eight or more consecutive adenines on one strand, equivalently eight or more thymines on the other. The literature usually writes it as poly(dA:dT) rather than as an IUPAC string, and the AT-rich promoter sites in COL1A1 and the MMP genes are described as TAAAAA-type runs 10669742,11112339. ZNF384 and ZNF362 are, as far as is known, the only C2H2 zinc-finger proteins that bind within the minor groove of such a tract 11867614, which is why only three fingers are involved: there are no distinguishing major-groove bases to read. All five matrices in this family are A- or T-runs of seven to twelve positions, differing mainly in run length and in whether they specify a flanking base; the HOCOMOCO ZNF362 matrix adds a 3' C and the ZNF384 matrix a 3' GT. Information content is therefore low in the usual sense while being high in a positional sense, because the requirement is a structural one: an uninterrupted, narrow-minor-groove A-tract. ZNF384 and ZNF362 cannot be distinguished from sequence at all.

Where it acts

Occurrences are promoter-proximal in the characterized cases, notably the COL1A1 and MMP promoters in osteoblasts, and genome-wide mapping of Nmp4 in bone cells recovered a large target set biased toward extracellular-matrix and secretory-pathway genes 26244796,37147466. A-tracts are, however, everywhere: they are among the most abundant simple sequence features in mammalian genomes, occur heavily in intergenic and intronic DNA, and are also intrinsically nucleosome-disfavouring, which means they influence chromatin accessibility independently of any protein binding them. That double life is the central interpretive problem for this family. There is no named composite element with a partner factor, and the reported cofactor interactions are protein-protein rather than DNA-adjacent: p130Cas for CIZ, and the nuclear matrix for Nmp4.

Biology and disease

Nmp4/ZNF384 is best understood in bone as a brake. Nmp4-null mice respond more strongly than wild type to anabolic parathyroid hormone therapy, gaining more trabecular bone 19189321, and their osteoblasts show enhanced secretory capacity 37147466; the factor has accordingly been described as closing the PTH anabolic window 19766748. Nmp4 loss also alters innate immune responses to influenza A 32152414 and modifies serum-induced arthritis 37147466, so the effects are not skeletal alone. The prominent human disease link is oncogenic and structural rather than transcriptional-regulatory in the ordinary sense: ZNF384 is a recurrent fusion partner in B-cell precursor acute lymphoblastic leukaemia, with EP300, TCF3, TAF15, CREBBP and other 5' partners fused to a nearly complete ZNF384 coding region 27634205,34529760. These ZNF384-rearranged leukaemias form a distinct transcriptional subtype with a characteristic immunophenotype 27634205, they include mixed-phenotype acute leukaemia cases, and additional classes of ZNF384 aberration have since been described 34529760. Because the fusion retains the zinc-finger array, the chimeric protein is thought to redirect a strong heterologous activation domain to ZNF384's A-tract sites.

Reading a hit

Treat a hit here as a statement about DNA structure first and about protein occupancy second. Eight-base A-tracts occur many hundreds of thousands of times in the human genome, so a match on its own is close to uninformative; conversely, a long A-tract found in an accessibility or ChIP dataset may reflect nucleosome depletion or an alignment artefact rather than ZNF384 binding. The family also overlaps other AT-rich motif families, and any tool that scans for homopolymer runs will find these sites regardless of which factor is labelled. ZNF384 and ZNF362 are indistinguishable, so a hit identifies the pair. HOCOMOCO's evidence codes for this family include in vitro data (S and M), so direct binding to A-tracts is real; what is not licensed by a hit is the conclusion that a particular tract is a regulated element.

How this was worked out

  1. 2000mediumFar-Western screening, cyclic amplification and selection of targets (CASTing), reporter assays10669742

    CIZ is cloned as a ligand of the p130Cas SH3 domain; target-site selection gives the DNA-binding consensus (G/C)AAAAA(A), which occurs in matrix metalloproteinase promoters, and CIZ binds the MMP-1 promoter and raises MMP-1 and MMP-3 transcription.

    Still open then: Whether the protein bound the short consensus or a longer A-tract, and how, was not addressed.

  2. 2000mediumNuclear matrix fractionation, imaging, overexpression and reporter assays10972987 11112339

    The nuclear matrix proteins NP/Nmp4, which bind the rat COL1A1 promoter in osteoblast-like cells, are reported to include isoforms identical to CIZ; CIZ overexpression raises type I collagen mRNA through a consensus site in the collagen promoter.

  3. 2002mediumDNA-binding assays with deletion mutants, GAL4 fusion transactivation assays11867614

    Nmp4 binds within the minor groove of homopolymeric (dA:dT) DNA, the first C2H2 proteins known to do so; fingers 2, 3 and 6 mediate binding, and a strong N-terminal activation domain that is masked in the full-length protein becomes active once the protein is bound to the A-tract.

    Still open then: No structure was determined, so how C2H2 fingers engage a minor groove remained unknown.

  4. 2009mediumKnockout mice, densitometry and micro-CT19189321 19766748

    Nmp4-null mice gain more femoral trabecular bone than wild type under anabolic PTH treatment, establishing Nmp4 as an attenuator of the PTH bone-building response.

  5. 2015mediumChIP-seq, gene ontology analysis, ovariectomy model, micro-CT26244796

    Nmp4 ChIP-seq in preosteoblasts, embryonic stem cells and blood cell lines gives a genome-wide target set; ovariectomized Nmp4-null mice keep the improved PTH response, with more osteoprogenitors.

    Still open then: ChIP alone could not separate direct A-tract binding from occupancy at accessible, nucleosome-depleted tracts.

  6. 2017mediumWhole-transcriptome sequencing, RT-PCR, immunophenotyping27634205

    ZNF384 fusions, with TCF3, EP300 and other 5' partners, are found in 4.1% of 291 childhood B-cell precursor ALL cases; the fusion-positive leukemias share weak CD10 and aberrant CD13/CD33 expression and a distinct expression signature.

  7. 2021lowTargeted reanalysis of sequencing data in three patients34529760

    Leukemias with the ZNF384-fusion expression signature but no canonical fusion are found to carry a new class of aberration in the last exon of ZNF384 that disrupts the C-terminal portion of the protein, showing that these 'ZNF384r-like' cases are the same genetic subtype.

    Still open then: How a C-terminal disruption phenocopies fusions that keep the entire ZNF384 coding sequence was not explained.

Still unresolved

How much to trust a hit

This is a structural motif, not a word, and its abundance makes single hits nearly uninformative: eight-base A-tracts appear hundreds of thousands of times genome-wide. A-tracts are also intrinsically nucleosome-disfavouring, so their enrichment in open chromatin and in ChIP peak sets can arise without any ZNF384 binding, and an A-tract hit in an accessibility dataset should not be read as occupancy. ZNF384 and ZNF362 are indistinguishable from sequence, and the motif overlaps other AT-rich families. In vitro evidence supports direct minor-groove binding, so the protein-motif link itself is sound; only the interpretation of individual genomic matches is weak. Context, notably promoter position and extracellular-matrix or secretory target genes, does most of the work.

References

5 motifs in this family — JASPAR 1 · HOCOMOCO 2 · CIS-BP 1 · Vierstra 1
motifdatabaseconsensusbits
ZN384.H14CORE.0.PSM.APSMAHOCOMOCOaAAAAAAAAaGt18
ZN362.H14CORE.0.PSGI.APSGIAHOCOMOCOAAAAAAAAC14.9
AC0215:ZNF:C2H2_ZFVierstraTTTTTTTT13.3
M08090_2.00CIS-BPTTTTTTTT13.1
ZNF384 (MA1125.2)JASPARAAAAAAAA12.7

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

how this page was made Drafted by a language model reading the primary literature, then checked: every cited PMID is resolved against PubMed and its citation text regenerated from PubMed's own record, and every claim about the motifs themselves is checked against the matrices in the four source databases. Methods and limits →

Other C2H2 zinc finger families