Structures of mammalian GLD-2 proteins reveal molecular basis of their functional diversity in mRNA and microRNA processing

Abstract The stability and processing of cellular RNA transcripts are efficiently controlled via non-templated addition of single or multiple nucleotides, which is catalyzed by various nucleotidyltransferases including poly(A) polymerases (PAPs). Germline development defective 2 (GLD-2) is among the first reported cytoplasmic non-canonical PAPs that promotes the translation of germline-specific mRNAs by extending their short poly(A) tails in metazoan, such as Caenorhabditis elegans and Xenopus. On the other hand, the function of mammalian GLD-2 seems more diverse, which includes monoadenylation of certain microRNAs. To understand the structural basis that underlies the difference between mammalian and non-mammalian GLD-2 proteins, we determine crystal structures of two rodent GLD-2s. Different from C. elegans GLD-2, mammalian GLD-2 is an intrinsically robust PAP with an extensively positively charged surface. Rodent and C. elegans GLD-2s have a topological difference in the β-sheet region of the central domain. Whereas C. elegans GLD-2 prefers adenosine-rich RNA substrates, mammalian GLD-2 can work on RNA oligos with various sequences. Coincident with its activity on microRNAs, mammalian GLD-2 structurally resembles the mRNA and miRNA processor terminal uridylyltransferase 7 (TUT7). Our study reveals how GLD-2 structurally evolves to a more versatile nucleotidyltransferase, and provides important clues in understanding its biological function in mammals.


Supplementary Figures
Amino acid sequences of human (hs)GLD-2 (UniProt accession Q6PIY7), mouse (mm)GLD-2 (Q91YI6), rat (rn)GLD-2 (Q5U315) and C. elegans (ce)GLD-2 (O17089) are aligned. Residues with a conservation of 100% are in red shades, greater than 70% in green shades and 45% in grey shades, respectively. α-helices are shown as cylinders and β-strands as arrows for rnGLD-2 above the sequences. The secondary structure signs are coloured and labelled as in Fig. 1B. Regions not resolved in the crystal structure are indicated by dashed lines. Important residues that are discussed in the paper are specified by different symbols according to their functional roles.

(B)
The consensus NTase motif in rnGLD-2. The key catalytic residues are indicated as ball-and-stick models, and secondary structural elements surrounding the catalytic center are specified.
(C) Structure comparison of rnGLD-2 with mmGLD-2 and other PAPs (btPAPα: PDB code 1f5a; hsPAPγ: 4lt6; ggMTPAP: 5a30) with native catalytic residues showing that the D279A mutation of rnGLD-2 does not affect the overall and local folding of rnGLD-2. The catalytic domain (upper) and catalytic site (lower) of rnGLD-2 is individually superimposed. Figure S4, Structure of rnGLD-2 represents an active-like state (A) Structural comparisons between the β6β9 region in rnGLD-2 and equivalent regions in Cid1 (complexed with UMPNPP, left) and in ceGLD-2 (complexed with RNP-8, right). In left panel, the UTP-selecting histidine (His336) is shown as ball-and-stick models. In right panel, residues of rnGLD-2 β9 and ceGLD-2 β9 neighbouring the nucleotide binding pocket are shown as ball-and-stick models for comparison. Note that the N-terminal portion of a RNP-8 molecule from the neighbouring asymmetric unit of the crystal (coloured yellow) inserts into the active site of ceGLD-2.
(B) rnGLD-2 showed a more closed conformation as compared to ceGLD-2, as indicated by the width of the cleft between the catalytic and central domains. The width is defined by the distance between the Cα atoms of residues Arg236 and Lys450 in rnGLD-2, and of their corresponding residues Val625 and Val891 in ceGLD-2.
(C) Crystal packing of rnGLD-2 viewed along a, b and c-axes of the unit cell. Two rnGLD-2 molecules (a copy of which is coloured orange) are in the asymmetric unit of the crystal.   Binding affinities (dissociation constant, Kd) between rnGLD-2(D279A) and ATP, ATPγS, or AMP were measured by isothermal titration calorimetry (ITC). rnGLD-2(D279A) lacks binding affinity to all these nucleotides. N/D, not determined.

Figure S9, Results of Dali search for rnGLD-2, mmGLD-2, and ceGLD-2
Structural homologs of rnGLD-2, mmGLD-2, and ceGLD-2, excerpted from the output of search using Dali Server. 10 candidates with highest Z-score for each search is shown. atURT1, Arabidopsis thaliana UTP:RNA uridylyltransferase 1. (C) Structural comparison between rnGLD-2 and TUT7 CM /UTP/dsRNA at the hydrophobic platform of the latter that is crucial in positioning the first base pair of the substrate RNA duplex.

Figure S11, Positively charged surface residues of rnGLD-2 potentially involved in substrate binding
Positively charged surface residues of rnGLD-2 that may be involved in the binding with RNA substrates. rnGLD-2 is superimposed individually with UMPPNP/U 2 -bound TUT7 CM (left, PDB code 5w0n) and UTP/dsRNA-bound TUT7 CM (right, PDB code 5w0o). Residues of rnGLD-2 that may be involved in the interaction with the RNA substrate are shown as ball-and-stick models.

Figure S12, Summary of the cellular function of mammalian GLD-2
Mammalian GLD-2, as an intrinsically potent NTase, is involved in cytoplasmic polyadenylation of mRNAs and adenylation/uridylation of certain mature miRNAs and pre-miRNAs. For mRNA polyadenylation, GLD-2 forms a so-called cytoplasmic polyadenylation complex with other partners including CPEB, CPSF, symplekin, and carry out polyadenylation with the presence of the eukaryotic initiation factor 4A/4E/4G (eIF4A/4E/4G) complex. QKI-7 recruits GLD-2 to miR-122 for mono/oligo-adenylation. It is not clear whether there are other partners involved in the adenylation/uridylation function of mammalian GLD-2 on miRNAs. The deadenylase PARN is known to inhibit the activity of the cytoplasmic polyadenylation complex during mRNA polyadenylation. PARN is also involved in the inhibition of GLD-2-mediated miRNA 3'-end processing together with CUGBP1. NTR denotes the N-terminal region (residues 1146, see Figure 1) of rnGLD-2.