NeuroD / Atonal / ASCL
Class II basic helix-loop-helix (bHLH) proteins of the Atonal, achaete-scute, Olig and NSCL groups — NEUROD1/2, NEUROG1-3, ATOH1/7/8, ASCL1/2, OLIG1-3, PTF1A, BHLHA15, FIGLA, MSGN1, NHLH1/2 and relatives — bind DNA only as heterodimers with the broadly expressed class I E-proteins TCF3, TCF4 and TCF12. The dimer reads a six-base E-box, CANNTG; this group prefers CAGATG (CATCTG on the opposite strand), with CATATG and CAGCTG as the two common variants. These factors are commitment switches for progenitors in the nervous system, inner ear, retina, pancreas, intestine and presomitic mesoderm, and several of them — ASCL1, NEUROD1, NEUROG2, ATOH1 — are the active ingredients of direct reprogramming protocols that convert fibroblasts into neurons.
The factors
This family collects the tissue-restricted ("class II") bHLH proteins whose DNA-binding domains are closest to Drosophila atonal and achaete-scute. In the atlas it spans NEUROD1/2 (and NEUROD4/6), NEUROG1/2/3, ATOH1/7/8, ASCL1/2, OLIG1/2/3, PTF1A, FERD3L, FIGLA, BHLHA9, BHLHA15 (MIST1), BHLHE22/23, MSGN1 and NHLH1/2 — around twenty human genes. The binding unit is a ~60-residue basic helix-loop-helix domain: a basic segment that lies in the major groove, followed by two helices separated by a loop that pack into a parallel four-helix bundle with the partner subunit. Almost none of these proteins binds DNA usefully on its own. The functional species is a heterodimer with a class I E-protein — TCF3 (E12/E47), TCF4 (ITF2) or TCF12 (HEB) — which are expressed in nearly every cell type. The crystal structure of the E47-NEUROD1 bHLH dimer on the insulin-gene E-box shows why heterodimers dominate: the E47-NEUROD1 interface is as extensive as the E47 homodimer interface, while NEUROD1 homodimers are unstable 18069799. Expression of the class II partners is narrow and largely non-overlapping: NEUROG1/2 in neural progenitors, NEUROD1 in newly born neurons and pancreatic islets, ATOH1 in cerebellar granule, inner-ear hair-cell and intestinal secretory progenitors, ASCL1 in ventral telencephalon and autonomic and neuroendocrine lineages, OLIG2 in motor-neuron and oligodendrocyte precursors, PTF1A in pancreas and cerebellum, BHLHA15 in exocrine secretory cells, FIGLA in oocytes, MSGN1 in presomitic mesoderm. "Neural" describes most of the family, not all of it.
The motif
The site is a six-base E-box, CANNTG, an imperfect palindrome in which each monomer reads one CAN half-site. A glutamate in the basic region hydrogen-bonds the cytosine and adenine of its half-site and an adjacent arginine contacts the guanine on the complementary strand; these contacts are what make CA...TG invariant. The central two positions are the variable ones, and in the E47-NEUROD1 structure no side chain contacts them directly — specificity there comes from backbone contacts, DNA deformability and flanking sequence 18069799. That is the single most important fact for interpreting hits: CAGATG, CAGCTG and CATCTG (the reverse complement of CAGATG) differ at positions the protein barely touches.
The data reflect this. The representative model is gaCAGATGG, and the 82 motifs split into three overlapping groups: CAGATG (mostly ChIP-derived, for NEUROD1/2, NEUROG2, ATOH1, PTF1A), CATATG (almost all HT-SELEX or methyl-SELEX models, covering NEUROG1/2/3, OLIG1/2/3, ATOH1/7, BHLHE22/23, NEUROD1/2, BHLHA15) and CAGCTG (ASCL1/2, NHLH1/2, PTF1A, FERD3L, ATOH8, BHLHA9), with a CACCTG minority for FIGLA and some ASCL1 and PTF1A models. Genome-wide comparison of NEUROD2 and MYOD binding found CAGCTG shared between them and CAGATG as the NEUROD2-preferred site, so CAGCTG is the least diagnostic of the three 22445365. The flanks add a little: FERD3L models read GTAACAGCTGTTGC, NHLH models CGCAGCTGCG, and PTF1A ChIP gives an E-box followed after about five bases by a TTC, the TC-box of the trimeric PTF1 complex with RBPJ or RBPJL. Overall cohesion is 0.81 and no member is flagged as belonging elsewhere, but the CATATG and CAGCTG subgroups are separated by more than the usual paralogue-level noise.
Where it acts
These are enhancer factors. Sites are distal, AT-rich rather than CpG-island-associated, and binding is largely confined to E-boxes that are already in accessible chromatin, so the same motif is used by different family members in different cell types with little overlap. Typical genome-wide occupancy is thousands to tens of thousands of sites per factor, a small fraction of the available E-boxes. Frequent neighbours are SOX (SOX2, SOX10), POU (POU3F2/BRN2), LHX and RFX motifs in neural enhancers, and in the pancreas FOXA, PDX1 and RBPJ/RBPJL. The named composite element in this family is the PTF1 site: an E-box and a TC-box about one helical turn apart, bound by PTF1A together with RBPJ (PTF1-J) or RBPJL (PTF1-L).
Biology and disease
These factors are commitment switches. Atoh1 loss removes cerebellar granule neurons, inner-ear hair cells and the intestinal secretory lineage 10364557; Neurog3 loss removes all four pancreatic endocrine cell types 10677506; Olig1 and Olig2 couple motor-neuron and oligodendrocyte specification, and Olig2 loss removes both 11955447; Neurog1 is required for the proximal cranial sensory ganglia, whose precursors are never determined in its absence 9539122, and Neurog1 and Neurog2 are required for the two successive waves of dorsal root ganglion neurogenesis, Neurog2 mainly for the early trkB- and trkC-positive neurons and Neurog1 for most trkA-positive ones 10398684; the proneural framework these knockouts established is set out in a review 12094208. ASCL1 also drives progenitor proliferation, not only differentiation, which is why proneural activity and cell-cycle exit are not synonymous 21536733. In humans, biallelic PTF1A mutations cause pancreatic and cerebellar agenesis 15543146. ASCL1 and NEUROD1 define transcriptional subtypes of small-cell lung cancer and OLIG2 marks oligodendroglial tumours. The same commitment function underlies direct reprogramming: ASCL1 with POU3F2 and MYT1L converts fibroblasts to functional neurons 20107439, and NEUROD1, NEUROG2 and ATOH1 are used in comparable protocols.
Reading a hit
A CAGATG, CATATG or CAGCTG match tells you a class II bHLH-E-protein heterodimer could bind, and nothing about which one. All ~20 members are indistinguishable from sequence; the Vierstra archetype AC0084:OLIG/NEUROG:bHLH says so in its name. Chance matters: a fixed six-mer is expected roughly 750,000 times per strand of a 3.1-Gb genome, so a bare E-box hit is close to uninformative without conservation, accessibility or expression evidence. Which motif version you scan with changes the answer — the in vitro (HT-SELEX) models say CATATG while the ChIP models say CAGATG or CAGCTG 23332764. Members with only ChIP evidence (HOCOMOCO fourth-field P, e.g. NDF1.H14CORE.0.P.B) support "this factor occupies these regions", not "this factor binds this sequence"; the S and M models do the latter. One outlier deserves naming: the Vierstra archetype AC0461:ASCL:bHLH has the consensus ACACGACGC, which is not an E-box and matches nothing else in the family.
How this was worked out
NeuroD is identified as a neuronal differentiation factor: ectopic expression converts Xenopus ectoderm into neurons, showing that one tissue-restricted bHLH protein can impose neuronal identity.
Still open then: Which E-box variant NeuroD reads, and with which partner, was not addressed.
Neurogenin is placed upstream of NeuroD as a determination gene, giving the vertebrate family the same two-tier organization as the Drosophila proneural genes.
Still open then: Whether family members differ in target sequence or only in timing and expression domain.
Knockouts show single members are required for whole lineages: Atoh1 for inner-ear hair cells (1999), Neurog3 for all four pancreatic endocrine cell types (2000), Olig1/Olig2 together for motor neurons and oligodendrocytes (2002).
Still open then: The non-redundant phenotypes could not be explained by DNA-binding differences, since the proteins read the same E-box.
Biallelic truncating PTF1A mutations are shown to cause pancreatic and cerebellar agenesis in humans, carrying the family's lineage-commitment role into monogenic disease.
The crystal structure of the E47-NEUROD1 bHLH heterodimer on the insulin-gene E-box CATCTG shows a heterodimer interface as extensive as the E47 homodimer interface and no side-chain contact with the two central base pairs.
Still open then: With nothing contacting the central bases, what sets the CAGATG/CATATG/CAGCTG preference measured in solution was left unexplained.
ASCL1 with POU3F2 and MYT1L converts fibroblasts directly into functional neurons, showing these factors act as commitment switches outside their normal lineage.
Still open then: Why only a small subset of genomic E-boxes is engaged during conversion.
Side-by-side genome-wide mapping of NEUROD2 and MYOD finds a shared CAGCTG site plus factor-specific sites (CAGATG for NEUROD2, CAGGTG for MYOD); transcription tracks only the factor-specific sites, and occupancy is confined to E-boxes already in accessible chromatin.
Still open then: The relative weight of sequence and chromatin in choosing targets remained contested.
HT-SELEX models for NEUROG1-3, OLIG1-3, ATOH1/7 and NEUROD1/2 read CATATG rather than the ChIP-derived CAGATG, so the in vitro preference of these proteins differs from the site recovered from chromatin.
Still open then: Whether the CATATG/CAGATG difference reflects partner proteins, chromatin or selection conditions was not resolved.
Still unresolved
- In vitro and ChIP-derived models of the same proteins disagree about the two central bases: HT-SELEX and methyl-SELEX models for NEUROG1-3, OLIG1-3 and ATOH1/7 read CATATG, while ChIP models for NEUROD1/2, NEUROG2, ATOH1 and PTF1A read CAGATG or CAGCTG, and it is unsettled whether partner proteins, chromatin context or the selection protocol accounts for the shift.
- What selects member-specific targets in vivo is unresolved, because the one solved class I/class II structure shows no side-chain contact with the variable central bases; flanking sequence, DNA shape, cooperating factors and pre-existing accessibility have all been proposed, and the NEUROD2/MYOD comparison argues that sequence chooses the program while chromatin decides which sites are available.
- Whether proneural activity is separable from proliferative activity is open: ASCL1 drives both cell-cycle exit and progenitor expansion, and which targets or conditions tip that balance has not been settled.
- The Vierstra archetype AC0461:ASCL:bHLH has the consensus ACACGACGC, which is not an E-box and matches nothing else in the family; whether it captures a real secondary binding mode of ASCL factors or is a footprint-clustering artifact is unknown.
A hit identifies the family, not the factor: all twenty-odd members read the same six-base E-box and cannot be told apart from sequence, which is why one Vierstra archetype here is named OLIG/NEUROG rather than after a protein. The three variants CAGATG, CATATG and CAGCTG differ at positions the protein does not directly contact, so a scan tuned to one will miss or misrank the others, and CAGCTG in particular is shared with E-protein homodimers, MYOD, TWIST and TCF21. Six-mers are far too common to be evidence on their own. ChIP-only models (HOCOMOCO evidence letter P) show occupancy, not binding specificity.
References
- Lee JE, Hollenberg SM, Snider L, et al. Conversion of Xenopus ectoderm into neurons by NeuroD, a basic helix-loop-helix protein. Science 1995;268:836-44. PMID 7754368
- Ma Q, Kintner C, Anderson DJ. Identification of neurogenin, a vertebrate neuronal determination gene. Cell 1996;87:43-52. PMID 8858147
- Bertrand N, Castro DS, Guillemot F. Proneural genes and the specification of neural cell types. Nat Rev Neurosci 2002;3:517-30. PMID 12094208
- Longo A, Guanga GP, Rose RB. Crystal structure of E47-NeuroD1/beta2 bHLH domain-DNA complex: heterodimer selectivity and DNA recognition. Biochemistry 2008;47:218-29. PMID 18069799
- Fong AP, Yao Z, Zhong JW, et al. Genetic and epigenetic determinants of neurogenesis and myogenesis. Dev Cell 2012;22:721-35. PMID 22445365
- Jolma A, Yan J, Whitington T, et al. DNA-binding specificities of human transcription factors. Cell 2013;152:327-39. PMID 23332764
- Castro DS, Martynoga B, Parras C, et al. A novel function of the proneural factor Ascl1 in progenitor proliferation identified by genome-wide characterization of its targets. Genes Dev 2011;25:930-45. PMID 21536733
- Bermingham NA, Hassan BA, Price SD, et al. Math1: an essential gene for the generation of inner ear hair cells. Science 1999;284:1837-41. PMID 10364557
- Gradwohl G, Dierich A, LeMeur M, et al. neurogenin3 is required for the development of the four endocrine cell lineages of the pancreas. Proc Natl Acad Sci U S A 2000;97:1607-11. PMID 10677506
- Sellick GS, Barker KT, Stolte-Dijkstra I, et al. Mutations in PTF1A cause pancreatic and cerebellar agenesis. Nat Genet 2004;36:1301-5. PMID 15543146
- Zhou Q, Anderson DJ. The bHLH transcription factors OLIG2 and OLIG1 couple neuronal and glial subtype specification. Cell 2002;109:61-73. PMID 11955447
- Vierbuchen T, Ostermeier A, Pang ZP, et al. Direct conversion of fibroblasts to functional neurons by defined factors. Nature 2010;463:1035-41. PMID 20107439
- Ma Q, Chen Z, del Barco Barrantes I, et al. neurogenin1 is essential for the determination of neuronal precursors for proximal cranial sensory ganglia. Neuron 1998;20:469-82. PMID 9539122
- Ma Q, Fode C, Guillemot F, et al. Neurogenin1 and neurogenin2 control two distinct waves of neurogenesis in developing dorsal root ganglia. Genes Dev 1999;13:1717-28. PMID 10398684
82 motifs in this family — JASPAR 29 · HOCOMOCO 29 · CIS-BP 21 · Vierstra 3
| motif | database | consensus | bits |
|---|---|---|---|
| FER3L.H14CORE.0.SM.BSMB | HOCOMOCO | GTAACAGCTGTtgC | 20.5 |
| HEN2.H14CORE.0.SM.BSMB | HOCOMOCO | gGGaCGCAGCTGCG | 18.5 |
| M04177_2.00 | CIS-BP | GnCGCAGCTGCGtCc | 18.4 |
| FERD3L (MA1485.1) | JASPAR | GCaaCAGCTGTtAC | 18.1 |
| M04140_2.00 | CIS-BP | GTaACAGCTGttGC | 18.1 |
| NHLH2 (MA1529.2) | JASPAR | GnCGCAGCTGCGtCc | 18 |
| OLIG3.H14CORE.0.SM.BSMB | HOCOMOCO | ACCATATGTT | 17.4 |
| M04156_2.00 | CIS-BP | AACAGCTGACg | 16.3 |
| OLIG1.H14CORE.0.S.BSB | HOCOMOCO | AaCATATGgT | 16.2 |
| NGN1.H14CORE.0.SM.BSMB | HOCOMOCO | gtGACATATGtt | 16.1 |
| NGN2.H14CORE.1.SM.BSMB | HOCOMOCO | AACATATGTC | 15.7 |
| ATOH1.H14CORE.1.SM.BSMB | HOCOMOCO | GaCATATGtT | 15.4 |
| BHA15.H14CORE.1.SM.BSMB | HOCOMOCO | AcCATATGG | 15.3 |
| M02804_2.00 | CIS-BP | AaCATATGtT | 15 |
| BHLHE23 (MA0817.2) | JASPAR | AaCATATGtT | 14.7 |
| OLIG1 (MA0826.1) | JASPAR | AaCATATGtT | 14.7 |
| M02806_2.00 | CIS-BP | AaCATATGtT | 14.7 |
| NEUROG2 (MA0669.1) | JASPAR | aACATATGtC | 14.5 |
| M04166_2.00 | CIS-BP | AACAGCTGTC | 14.2 |
| Msgn1 (MA1524.3) | JASPAR | gACAAATGGt | 14.1 |
| M02796_2.00 | CIS-BP | CGCAGCTGC | 14 |
| NEUROG1 (MA0623.2) | JASPAR | gaCATATGtc | 13.8 |
| BHLHA15 (MA0607.2) | JASPAR | aCCATATGGt | 13.7 |
| BHLHE22 (MA0818.2) | JASPAR | AcCATATGgt | 13.4 |
| M02799_2.00 | CIS-BP | aACATATGtc | 13.4 |
| NDF1.H14CORE.0.P.BPB | HOCOMOCO | gaCAGATGG | 13.2 |
| MSGN1.H14CORE.0.PSM.APSMA | HOCOMOCO | CAaATGGt | 13.1 |
| M04179_2.00 | CIS-BP | aaCATATGtt | 13.1 |
| OLIG2 (MA0678.1) | JASPAR | AcCATATGgT | 13 |
| M02784_2.00 | CIS-BP | AcCATATGyT | 13 |
| M08055_2.00 | CIS-BP | AGCAGCTGg | 13 |
| NHLH1 (MA0048.3) | JASPAR | CGCAGCTGC | 12.9 |
| Ascl2 (MA0816.1) | JASPAR | AgCAGCTGCt | 12.8 |
| BHA15.H14CORE.0.P.BPB | HOCOMOCO | aaCAGCTGG | 12.8 |
| M04201_2.00 | CIS-BP | gaCATATGtc | 12.8 |
| M04205_2.00 | CIS-BP | cCACCTG | 12.7 |
| NDF2.H14CORE.1.S.BSB | HOCOMOCO | CATATGGT | 12.6 |
| HEN1.H14CORE.0.S.BSB | HOCOMOCO | CAGCTGCghC | 12.5 |
| M04171_2.00 | CIS-BP | AaCATATGGt | 12.5 |
| ATOH1.H14CORE.0.P.BPB | HOCOMOCO | AgCAGcTGG | 12.4 |
| FIGLA.H14CORE.0.SM.BSMB | HOCOMOCO | aCACCTG | 12.4 |
| NGN2.H14CORE.0.P.BPB | HOCOMOCO | agCAGATGG | 12.4 |
| M02812_2.00 | CIS-BP | AcCATATGtT | 12.4 |
| ASCL2.H14CORE.0.PSM.APSMA | HOCOMOCO | CAGCTGCt | 12.3 |
| BHE22.H14CORE.0.SM.BSMB | HOCOMOCO | CATATGgT | 12.3 |
| NDF2.H14CORE.0.P.BPB | HOCOMOCO | agCAGATGG | 12.3 |
| BHE23.H14CORE.0.SM.BSMB | HOCOMOCO | CATATGgT | 12.1 |
| PTF1A.H14CORE.0.P.BPB | HOCOMOCO | CAGCTGtbnnnwttC | 12.1 |
| M08057_2.00 | CIS-BP | aCAGATGG | 12.1 |
| OLIG2.H14CORE.0.P.BPB | HOCOMOCO | cCAGCTGyTbnC | 12 |
| Neurod2 (MA0668.3) | JASPAR | ACAGATGG | 11.8 |
| NEUROD1 (MA1109.2) | JASPAR | aCAGATGG | 11.8 |
| BHLHE22 (MA1635.2) | JASPAR | CAGCTG | 11.8 |
| Bhlha15 (MA1472.3) | JASPAR | ACAGCTGT | 11.7 |
| ASCL1 (MA1631.2) | JASPAR | CACCTGC | 11.7 |
| Neurod2 (MA1993.2) | JASPAR | CAGCTG | 11.7 |
| BHA09.H14CORE.0.P.BPB | HOCOMOCO | cCAGcTGT | 11.7 |
| M04160_2.00 | CIS-BP | aaCATATGgt | 11.7 |
| Olig2 (MA1997.2) | JASPAR | CAGCTG | 11.5 |
| OLIG3 (MA0827.1) | JASPAR | AcCATATGkt | 11.4 |
| Ptf1A (MA1619.2) | JASPAR | aCAGCTGt | 11.3 |
| NDF1.H14CORE.1.S.CSC | HOCOMOCO | CATATGG | 11.3 |
| OLIG2.H14CORE.2.SM.BSMB | HOCOMOCO | CATATGtT | 11.3 |
| OLIG2.H14CORE.1.P.BPB | HOCOMOCO | CCAGCTGcT | 11.2 |
| PTF1A.H14CORE.1.PS.APSA | HOCOMOCO | gCAGCTGt | 11.2 |
| ASCL1.H14CORE.0.PSM.APSMA | HOCOMOCO | CAGCTGC | 11.1 |
| Atoh1 (MA1467.3) | JASPAR | CAGATGG | 11 |
| Atoh1 (MA0461.3) | JASPAR | aaCATATG | 10.8 |
| ATOH7 (MA1468.1) | JASPAR | AaCATATGtc | 10.8 |
| Ptf1A (MA1618.2) | JASPAR | aCAGATGtt | 10.8 |
| Ptf1A (MA1620.2) | JASPAR | ACACCTGT | 10.8 |
| NEUROG2 (MA1642.2) | JASPAR | CAGATGG | 10.8 |
| ATOH8.H14CORE.0.P.CPC | HOCOMOCO | aaaCAGCTG | 10.8 |
| M04193_2.00 | CIS-BP | AaCATATGtc | 10.8 |
| AC0084:OLIG/NEUROG:bHLH | Vierstra | acCATATG | 10.6 |
| M04207_2.00 | CIS-BP | gCAGCTGc | 10.2 |
| M04144_2.00 | CIS-BP | CAtaTGGc | 10.1 |
| AC0078:NHLH/TFAP:bHLH | Vierstra | CAGCTG | 9.8 |
| AC0461:ASCL:bHLH | Vierstra | ACACGACGC | 9.8 |
| M04195_2.00 | CIS-BP | CAGCTGt | 9.7 |
| FIGLA (MA0820.2) | JASPAR | CACCTG | 9.5 |
| ASCL1 (MA1100.3) | JASPAR | gCAGCTGc | 9.3 |
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.