The NHL domain of BRAT is an RNA-binding domain that directly contacts the hunchback mRNA for regulation

Abstract

The Drosophila protein brain tumor (Brat) forms a complex with Pumilio (Pum) and Nanos (Nos) to repress hunchback (hb) mRNA translation at the posterior pole during early embryonic development. It is currently thought that complex formation is initiated by Pum, which directly binds the hb mRNA and subsequently recruits Nos and Brat. Here we report that, in addition to Pum, Brat also directly interacts with the hb mRNA. We identify Brat-binding sites distinct from the Pum consensus motif and show that RNA binding and translational repression by Brat do not require Pum, suggesting so far unrecognized Pum-independent Brat functions. Using various biochemical and biophysical methods, we also demonstrate that the NHL (NCL-1, HT2A, and LIN-41) domain of Brat, a domain previously believed to mediate protein-protein interactions, is a novel, sequence-specific ssRNA-binding domain. The Brat-NHL domain folds into a six-bladed β propeller, and we identify its positively charged top surface as the RNA-binding site. Brat belongs to the functional diverse TRIM (tripartite motif)-NHL protein family. Using structural homology modeling, we predict that the NHL domains of all TRIM-NHL proteins have the potential to bind RNA, indicating that Brat is part of a conserved family of RNA-binding proteins.

Keywords: BRAT; RNA binding; TRIM-NHL; gene regulation; hunchback; translational repression.

Figure 1.

Direct RNA binding of the Brat-NHL domain. (A) Schematic representation of Drosophila melanogaster Brat domain organization. The two B-boxes (B1 and B2) are shown in orange, the coiled coil (CC) domain is in yellow, and the NHL domain, composed of six NHL repeats, is in green. Also indicated and colored in gray are serine-rich (S), glutamine-rich (Q), and histidine-rich (H) stretches. The numbers below indicate domain boundaries. (B) Schematic presentation of the maternally derived hb mRNA (NM_169234) and hb RNA fragments used in this study. The ∼100-nt-long hb RNA fragment (for simplicity referred to here as hb RNA) contains two NREs, each composed of one BoxA and one BoxB motif. Nucleotides mutated in hb 55–81 are indicated by red letters. (C) Native gel analysis to probe for RNA binding of the Brat-NHL domain. Increasing amounts of recombinant Brat-NHL (amino acids 756–1037) were incubated with 500 pM ³²P-labeled hb RNA, and complexes were analyzed by native gel electrophoresis. (D) Brat-NHL binds sequences that contain the NRE. Short, 27-nt, ³²P-labeled RNA probes that span the hb RNA (as depicted in A) were incubated with the indicated amounts of recombinant Brat-NHL, and complexes were analyzed by native gel electrophoresis. (E) MST measurements. (Top) Representative binding curves for the interaction of Brat-NHL with hb RNA (green triangles) or hb fragment 55–81 (red triangles). (Bottom) Summary of independent MST measurements. Brat-NHL binds the hb RNA with a K_d of 137.4 nM ± 36.8 nM (six independent repeats); the short hb (fragment 55–81) is bound with lower affinity and a K_d of 2.0 µM ± 0.8 µM (five independent repeats). No binding was detected for hb mutants that lack the Brat-binding motif (NRE2_BoxA or NRE1+2_BoxA) (see Fig. 2). (F) The Brat-NHL domain is a sequence-specific, ssRNA-binding domain. Complex formation of recombinant Brat-NHL with ssRNA (lanes 1–3), ssRNA of antisense sequence (lanes 4–6), dsRNA (lanes 7–9), or ssDNA (lanes 10–12) was analyzed by native gel electrophoresis. In all cases, nucleotide sequences corresponding to hb fragment 55–81 were used. DNA oligonucleotides contained dT instead of dU. (G) Mutations in BoxA but not in BoxB abrogate Brat-NHL binding to the NRE2. Native gel analysis to test Brat-NHL binding to hb 55–81 or the indicated mutants. Free RNA or RNA–protein complexes are indicated by black arrows. Asterisk (*) denotes a less well-defined Brat-NHL–hb RNA complex appearing at high protein concentrations.

Figure 2.

Mutations in BoxA, but not BoxB, abrogate binding of Brat-NHL to hb RNA. (A) Sequence of the hb RNA and its mutants used in binding assays. Mutated nucleotides are indicated by red letters. (B–E) Recombinant Brat-NHL was incubated with ³²P-labeled hb RNA or mutant RNAs as indicated and analyzed by native gel electrophoresis. Mutation of either BoxA site is sufficient to greatly impair RNA binding of Brat-NHL. Free RNA or RNA–protein complexes are indicated by black arrows. Asterisk (*) denotes a less well-defined Brat-NHL–hb RNA complex appearing at high protein concentrations.

Figure 3.

Binding of the Pum-HD to hb RNA facilitates Brat-NHL binding but does not overcome the requirement for two Brat-binding sites. (A) Binding of the Pum-HD and Brat-NHL to the hb RNA is not mutually exclusive, and preincubation of hb RNA with the Pum-HD facilitates Brat-NHL binding. Indicated amounts of Brat-NHL were mixed with ³²P labeled hb RNA alone (lanes 1–7) or hb RNA preincubated with 10 nM Pum-HD (lanes 8–14). Complexes were separated by native gel electrophoresis. (B) Binding of the Pum-HD to hb RNA does not overcome the requirement for two Brat-binding sites. Experiment was done as described in A except that the hb mutant NRE2_BoxA was used as a substrate for complex formation.

Figure 4.

Repression by Brat is independent of Pum. (A) Schematic representation of the FL reporter constructs used in this study. The reporter constructs contain the FL coding sequence fused to a 3′ UTR of interest. In B and D–G, the 100-nt hb 3′ UTR fragment and its various mutants (as depicted in Fig. 2A) were analyzed. In F and G, the 3′ UTRs of myc and mad were studied. (B) Dmel2 cells were cotransfected with plasmids expressing the indicated FL 3′ UTR reporter constructs, a RL control, and the indicated HA fusion proteins. HA-GW, not known to regulate hb translation, served as an additional control. Values represent means of three independent experiments, each performed in triplicate, and error bars show standard error of mean. A representative experiment, including raw FL and RL values and normalization steps, is shown in Supplemental Figure 5. Expression of HA-Brat or HA-Pum, but not of HA-GW, led to an ∼1.5-fold repression of the hb 3′ UTR reporter. Mutations in BoxA abrogated repression by HA-Brat but not by HA-Pum, while mutations in BoxB impaired repression by HA-Pum but not by HA-Brat. (C) Efficacy of dsRNA treatment was assayed by Western blotting. Due to the lack of an antibody against endogenous Pum, Pum depletion was assayed by cotransfected HA-Pum. (D) Knockdown of endogenous Pum but not Brat relieves hb repression. Dmel2 cells were treated with dsRNA to Pum and Brat as detailed in the Supplemental Material. The indicated FL reporter constructs were cotransfected with a RL control vector, and reporter gene expression was analyzed as described in B. (E) Knockdown of endogenous Pum does not affect repression by Brat. Reporter gene assay was performed as described in D except that plasmids expressing the indicated HA fusion proteins were cotransfected along with the luciferase constructs. (F) Repression of myc and mad 3′ UTRs by HA-Pum and HA-Brat. Experiment was performed as described in B. Expression of Brat leads to repression of both the myc and the mad 3′ UTR reporter, while expression of Pum results in repression of myc but not of mad. (G) Knockdown of endogenous Pum relieves repression of hb and myc but not of mad. Experiment was performed as described in D.

Figure 5.

Residues on the top, electropositive surface of the NHL domain contact RNA. (A) Structure-based sequence alignment of the six NHL repeats that form the Brat-NHL domain (based on the crystal structure of Brat [PDB ID 1Q7F]) (Edwards et al. 2003). Secondary structure elements are depicted above the alignment. Residues that make up the β strands are shown in bold and accentuated in yellow, pink, orange, and purple for β strands βa, βb, βc, and βd, respectively. (B) Structure of the Brat-NHL domain looking from the top (left) or the side (right). Each blade of the six-bladed β propeller is composed of four anti-parallel β strands (termed βa to βd) that are connected by flexible loop regions. By definition, the loops that connect βb with βc and βd with βa form the top surface of the molecule, while loops connecting βa with βb and βc with βd make up the bottom surface. β Strands of blade V are colored according to the sequence alignment shown in A. (C) Electrostatic calculations reveal an electropositive top surface and an electronegative bottom surface. Negative surface potential is shown in red, and positive surface potential is shown in blue. (D) Mutation of the top surface residues greatly impairs BRAT-NHL RNA binding. Recombinant Brat-NHL or the indicated point mutants were incubated with ³²P-labeled hb RNA and analyzed by native gel electrophoresis. (E) Electrostatic surface potential, with the residues tested in D indicated in yellow. (F–H) Summary of in vitro cross-linking experiments. Brat-NHL–hb RNA complexes were UV-cross-linked, and, following isolation, peptide–oligonucleotide cross-links were analyzed by liquid chromatography (LC)/MS. Identified peptides are shown in purple, while residues sequenced as RNA adducts are highlighted in yellow. Five out of six peptides span the top surface.

Figure 6.

Mutation of the top surface residues abrogates Brat-mediated repression. (A) Repression of the hb 3′ UTR by HA-Brat or the indicated HA-Brat point mutants in Dmel2 cells. (Top) Dmel2 cells were cotransfected with plasmids expressing the FL hb 3′ UTR reporter, the indicated HA fusion proteins, and a RL control plasmid. FL was normalized to RL, and values of normalized FL produced in the presence of an empty control vector were set to 1. (Bottom) Protein expression was analyzed by Western blotting. (B) Tethering experiment. Mutations that impair Brat-mediated hb repression have no effect when Brat is artificially tethered to the RNA via fusion to the λ phage N-peptide (N), targeting the fusion protein to hairpin structures in the 3′ UTR of the reporter (see the inset). Dmel2 cells were cotransfected with plasmids expressing FL-5boxB, the indicated NHA or HA fusion proteins, and a RL control plasmid. FL was normalized to RL, and values of normalized FL produced in the presence of an empty control vector were set to 1. Tethering of GW served as a positive control. Values represent means of three independent experiments, each performed in triplicate, and error bars show standard error of the mean. (C,D) Summary of mutagenesis studies. Mutations that affect Brat RNA binding are shown in yellow, while mutations that have no effect on Brat RNA binding are depicted in green. All mutations that lie on the top surface of the molecule, except C820A, either impair Brat-mediated hb repression (A) or in vitro binding of the Brat-NHL domain to hb RNA (Fig. 5D). (F) Model of the Brat:Pum:RNA repressor complex. The Brat-NHL and the Pum-HD contact the RNA directly. A ribbon representation of the Brat-NHL domain (Protein Data Bank [PDB] ID: 1Q7F, chain A) is shown colored in blue, and a ribbon representation of the Pum-HD (PDB ID: 3H3D, chain X) domain is shown colored in green. A ribbon representation of the NRE1 RNA is shown in orange and purple. Binding of the D. melanogaster Pum-HD to the NRE1 sequence was modeled by superposition of the domain on the structure of the Homo sapiens Pum-HD of Pumilio1 bound to the NRE1 sequence (PDB ID: 1M8X, chain A) using the align algorithm implemented in Pymol. Protein domains and ribbon representations of the RNA are drawn to scale. The dotted orange line indicates the 5′ region of the NRE1 RNA containing BoxA and, in the absence of structural information, is not drawn to scale.

Figure 7.

Phylogeny of six-bladed NHL domains. The neighbor-joining tree was derived from a structure-based multiple sequence alignment as described in the Materials and Methods. Characteristic proteins are represented by their known three-dimensional structure (name in red) or by homology models (name in bold and black). The subfamilies of putative RNA binders and of enzymes are separated from the rest by a bootstrap value of 85% or 100%, respectively. For each structure, Particle Mesh Ewald long-range electrostatic calculations were performed in YASARA and used to color-code the solvent-accessible surface: A negative charge is indicated by a red surface, and a positive charge is indicated by a blue surface. Abbreviations for protein names and species are given next to the Uniprot ID. (Bb) Borrelia burgdorferi; (Ce) Caenorhabditis elegans; (Dm) Drosophila melanogaster; (Hs) Homo sapiens; (Mt) Mycobacterium tuberculosis; (Rn) Rattus norvegicus; (Tp) Treponema pallidum. The length of the horizontal bar corresponds to 0.1 substitutions per site.