An ancient protein-DNA interaction underlying metazoan sex determination

Abstract

DMRT transcription factors are deeply conserved regulators of metazoan sexual development. They share the DM DNA-binding domain, a unique intertwined double zinc-binding module followed by a C-terminal recognition helix, which binds a pseudopalindromic target DNA. Here we show that DMRT proteins use a unique binding interaction, inserting two adjacent antiparallel recognition helices into a widened DNA major groove to make base-specific contacts. Versatility in how specific base contacts are made allows human DMRT1 to use multiple DNA binding modes (tetramer, trimer and dimer). Chromatin immunoprecipitation with exonuclease treatment (ChIP-exo) indicates that multiple DNA binding modes also are used in vivo. We show that mutations affecting residues crucial for DNA recognition are associated with an intersex phenotype in flies and with male-to-female sex reversal in humans. Our results illuminate an ancient molecular interaction underlying much of metazoan sexual development.

Figure 1. DMRT1 binds similar sites in vivo and in vitro

(a) Examples of ChIP-Seq data showing binding of DMRT1 to the Lrh1 (Nr5a2) gene in mouse and human testes. (b) Consensus DMRT1 DNA binding motif derived in vitro, compared to motifs associated with in vivo binding in mouse and human testes. (c) DNase I footprinting showing protection by DMRT1^67–136 of the in vitro binding consensus top strand (Site 1, upper) and a modified DNA (Site 2, lower). Diagram at bottom summarizes protection by DMRT1^67–136. Solid bars indicate strong and dashed bars indicate weaker protection. (d) Top: predicted and observed minor groove width for DMRT1 binding sites. Horizontal line indicates width of canonical B form DNA minor groove, black trace is minor groove width of Site 1 observed in the structure of DMRT1^67–136 bound to Site 1 shown in Fig. 2a; red and blue lines are minor groove width of unbound Site 1 and Site 2 DNAs predicted using DNAshape. Bottom: major groove width observed in structure of DMRT1^67–136 bound to Site 1. (e) EMSA analysis showing slower migrating complex (upper arrowhead) formed between full length DMRT1 and Site 2. Uncropped gels for all figures are shown in Supplementary Data Set 1.

Figure 2. Major groove interactions and use of multiple DNA binding modes by DMRT proteins

(a) Overview of structure, showing two DMRT1^67–136 protomers (A and B) inserted in DNA major groove on one side of Site 1 and one (C) in the major groove on the other side. Binding site symmetry is indicated (−6 and +6). Red oval: central basepair. Grey spheres: zinc ions. (b) Interaction of protomers A, B and C from back side, highlighting insertion of R72 sidechains of protomer B and C (dashed ovals) into minor groove. Amino acids labeled with solid ovals make major groove DNA contacts. (c) Overlay showing different orientation of R72 and different angle between zinc binding module and recognition helix of protomer A relative to those of B and C. (d) DNA base contacts. Middle diagram summarizes contacts made by each protomer. (e) Overlaid views of DMRT1 recognition helices bound to DNA, aligned at R111 and R123 and viewed from front and back. Right and left side DNAs are color-coded: pink is bound by protomers A and B; green by protomer C. (f) Major groove and minor groove interactions on left side of Site 1. (g) Major groove and minor groove hydrogen bond interactions on right side of Site 1. (h) Protomer B R72 interactions with minor groove. (i) Major groove interactions by protomer C. In h and I, blue mesh shows the sigma-A weighted 2Fo-Fc electron density map contoured at 1.5 σ Black dashed lines in f-i indicate hydrogen bonds. Red dashed line: arginine-thymine stacking interaction. Stippled spheres: van der Waals radii.

Figure 3. DNA backbone contacts, protein-protein interactions and binding summary

(a) Molecular surface of DMRT1^67–136 bound to DNA with charged groups contacting DNA phosphate backbone indicated in yellow. (b) Amino acids mediating protomer-protomer contacts. Interdigitating hydrophobic zipper and a Q to K hydrogen bond link protomers A and B. Two Q to R hydrogen bonds link B and C. (c) Close-up view of interaction between protomers A and B. Leucines and valines of interdigitating hydrophobic zipper are shown. In addition, R113 of protomer A and E110 of protomer B appear to form a salt-bridge (dashed line). Blue mesh shows the sigma-A weighted 2Fo-Fc electron density map contoured at 1.0 σ (d) Overlay of DMRT1 protomer B structure with Dsx NMR structure showing similar fold of zinc-binding domain. (e) Summary of DMRT1^67–136-DNA interactions. Colors indicate which protomer makes each contact. Thin-lined ovals with arrowheads identify amino acids that make DNA backbone contacts and thick-lined ovals identify amino acids that contact DNA bases. (f) Conservation of metazoan DM domains. Structural motifs and functional amino acids revealed by DMRT1 structural analysis are indicated for the region resolved by crystallography. Additional interactions could exist, particularly those bridged by water molecules. Amino acids are colored according to their chemical properties: polar amino acids (G,S,T,Y,C) are green, basic (K,R,H) blue, acidic (D,E) red, hydrophobic (A,V,L,I,P,W,F,M) black and neutral amino acids (Q,N) are purple.

Figure 4. Confirmation of critical protein-DNA contacts

(a-c) Confirmation of major groove DNA contacts by chemical substitution in vitro. (a) Chemical structures of base pair analogs, with shaded circles indicating atoms altered in modified bases. (b) EMSA assay of DNAs modified at −6 and +6 positions, showing that major groove but not minor groove changes reduce DMRT1 binding. c, EMSA assay of DNAs modified at −2 and +2 positions, showing that major groove but not minor groove modifications reduce binding. (d,e) Confirmation of critical protein contacts by amino acid substitution. (d) EMSA assays showing effects of alanine substitution of DMRT1 amino acids making base contacts. (e) EMSA assay showing that substituting K92, which interacts with the DNA phosphate backbone, reduces DMRT1 DNA binding.

Figure 5. DMRT1 binds DNA with multiple stoichiometries in vitro and in vivo

(a) EMSA showing binding of DMRT1 tetramer, trimer and dimer to Sites 2, 1, and 3, respectively. (b) EMSA of in vitro translated SUMO-DMRT1^67–136 binding to Sites 1 and 4, showing monomer through tetramer binding. (c) Protein crosslinking showing interaction of DMRT1 to form higher-order complexes. (d) Workflow testing DMRT1 binding stoichiometry by ChIP-Exo. Sites under ChIP peaks were grouped based on bilateral or unilateral minor groove narrowing and sequence at positions −6, +5, and +6 (see Methods). (e) Left, diagrams comparing ChIP-Seq and ChIP-Exo. Crosslinked protein blocks exonuclease digestion in ChiP-Exo within several bases of the crosslink. Colors indicate binding by different protomers; “X” illustrates potential crosslinks. Right, compilation of 5′ ends of ChIP-Seq (top) and ChIP-Exo (bottom) reads aligned to DMRT1 tetramer-binding consensus, showing higher resolution of ChIP-Exo. (f) ChIP-Exo analysis of DMRT1 binding in the mouse testis at sites sorted as indicated in panel d, showing three distinct patterns. Structural diagrams interpret DMRT1 binding modes based on ChIP-exo patterns and indicate potential crosslinks (red balls: crosslinkable DNA residues; yellow balls, crosslinkable protein residues). Note shared pattern in left side but differences on right side. Stars highlight prominent differences in ChIP-Exo pattern in putative trimer binding sites relative to tetramer sites. Based on the structure, trimers have more potential crosslinks with protomer C near center and right side of binding site due to its different position in the major groove.

Figure 6. Modeling DNA interaction by Dsx and MAB-3 suggests different binding modes

(a) In vitro DNA binding motifs for DMRT1, Dsx and MAB-3 showing that the Dsx site is symmetrical but lacks selection at −6 and +6 positions while the MAB-3 motif resembles the left side of DMRT1 motif. (b) EMSA assay showing that binding by the female Dsx isoform Dsx(f) requires specific DNA basepairs at the −2 and +2 but not the −6 and +6 DNA positions, consistent with the in vitro consensus. (c) Docking model of DMRT1 binding as a dimer to a previously determined Dsx binding site DNA structure, illustrating likely Dsx binding mode. (d) A model of proposed interaction of MAB-3 DM domains with DNA illustrating binding of MAB-3 as a covalently-joined “internal dimer”. MAB-3 (center and right) is proposed to form a structure on its consensus element similar to DMRT1 protomers A and B bound to the left side of the DMRT1 consensus element (left). The first DM domain of MAB-3 (DMa) is predicted to have a truncated recognition helix, with the remainder forming a linker joining DMa to DMb (Supplementary Fig. 5).

Figure 7. Disruption of crucial DMRT1 DNA contacts by a sex-reversing human mutation

(a) EMSA assay showing importance of R123, R111 and three amino acids that position R111 of protomer A for sequence-specific DNA interaction. (b) Sanger sequencing chromatograms showing de novo A to G heterozygous sequence change causing R111G coding change in 46,XY female. (c) EMSA showing reduced binding affinity and altered binding specificity of R111G mutation. Binding to Site 1 is reduced, but DMRT1^R111G binds Site 1 substituted at −6 and +6 better than wild-type DMRT1. (d) EMSA comparing DMRT1 and DMRT1^R111G. Left lanes, wild type DMRT1 alone; middle, DMRT1 plus DMRT1^R111G; right, DMRT1^R111G alone. EMSAs contained 2 ul in vitro translated of protein; wedges indicate added increments up to a total of 6 ul. DMRT1^R111G can convert wild type trimers on Site 1 into slower-migrating tetramers (arrowheads), likely by occupying the right side of the binding site. (e) Same experiment as in panel d, except DMRT1 binding site is from the Sox9 gene. (f) Protomer A R111 (pink) is positioned for hydrogen bonding with −6 guanine by M115 and Q118 of protomer B (blue) and M115 of protomer A (pink). Methionine methyl groups make van der Waals contacts with each other and R111. (g) In protomer C R111 can recognize +5 or +6. (h) Walleye stereo view of protomer A R111 interacting with −6G and protomer B R111 interacting with DNA backbone. Blue mesh: sigma-A weighted 2Fo-Fc electron density map contoured at 1.5 σ.