Structure of an active human histone pre-mRNA 3′-end processing machinery

Abstract

The 3′-end processing machinery for metazoan replication-dependent histone pre-mRNAs contains the U7 snRNP and shares the key cleavage module with the canonical cleavage/polyadenylation machinery. We reconstituted an active human histone pre-mRNA processing machinery using 13 recombinant proteins and 2 RNAs, and determined its structure by cryo-electron microscopy. The overall structure is highly asymmetrical and resembles an amphora with one long handle. We captured the pre-mRNA in the CPSF73 endonuclease active site, poised for cleavage. The endonuclease and the entire cleavage module undergo extensive rearrangements for activation, triggered through the recognition of the duplex between the authentic pre-mRNA and U7 snRNA. Our studies also have significant implications for understanding canonical and snRNA 3′-end processing.

The 3′-end processing machineries for polyadenylated (1, 2) and histone pre-mRNAs (3, 4) both utilize CPSF73 to cleave pre-mRNA (5, 6), but the molecular mechanism for their functions is still poorly understood. CPSF73, CPSF100, symplekin and CstF64 comprise the histone pre-mRNA cleavage complex (HCC) (Figs. 1A,B, table S1), which is equivalent to the mammalian cleavage factor (mCF) for polyadenylated pre-mRNAs (7, 8). The cleavage site in histone pre-mRNAs is located between a conserved stem-loop (SL) that is recognized by SL binding protein (SLBP) and a histone downstream element (HDE) that base pairs with the 5′ end of U7 snRNA, forming an HDE-U7 duplex (Fig. 1A). The U7 snRNP is critical for this processing, and the Lsm11-FLASH complex recruits the HCC to the machinery (9–12) (Supplementary text).

Figure 1. Overall structure of the human histone pre-mRNA 3′-end processing machinery.

(A) Schematic drawing of the histone pre-mRNA 3′-end processing machinery. (B) Domain organizations of the subunits of HCC, and Lsm10 and Lsm11. The domains in CPSF100 are given slightly darker colors compared to their homologs in CPSF73. The vertical bar in the symplekin CTD marks its N-terminal half that interacts with CPSF73. Abbreviations are defined in table S1. (C) Cryo-EM density at 3.2 Å resolution for the core of the machinery. (D) Schematic drawing of the structure of the core of the machinery, viewed after a 150° rotation around the vertical axis from panel C. The proteins are colored as in Figs. 1A,1B. The U7 snRNA is in dark green, and H2a* in orange. (E) Cryo-EM density for the entire machinery (gray), low-pass filtered to 8 Å resolution to show the density of FLASH and SLBP. The possible density for CTD3 of CPSF73 is indicated with the asterisk. Structure figures are produced with PyMOL (www.pymol.org) unless noted otherwise. Panels C and E produced with Chimera (30).

To prepare a fully recombinant machinery, we reconstituted human U7 snRNP (13) and mixed it with purified human HCC, FLASH (14), and SLBP (15). Using a modified mouse histone H2a pre-mRNA (H2a*, fig. S1) as substrate, we observed robust cleavage activity generating the authentic product (Supplementary text, figs. S2,3). Strikingly, the N-terminal domain (NTD) of symplekin was essential for processing, and its binding partner Ssu72 (16) inhibited the cleavage reaction. A mutation in the active site of CPSF73 abolished the cleavage.

We purified the active machinery (fig. S3F) and obtained a cryo-EM reconstruction at 3.2 Å resolution for its core (Figs. 1C,D), and a reconstruction at 4.1 Å resolution for the entire machinery (tables S2,3, figs. S4–6). The overall structure of the machinery resembles an amphora with one long handle (Fig. 1E, fig. S6B, movie S1). The machinery core constitutes the body of the amphora, with the U7 snRNA 3′-end SL and the Sm ring at the base and the CTDs of CPSF73 and CPSF100 and the first few helical repeats of the symplekin CTD forming the mouth. CPSF73 and symplekin NTD are positioned opposite each other on the Sm ring (Fig. 1D, fig. S6A). CPSF100 interacts with both CPSF73 and symplekin, but does not directly contact the Sm ring (Fig. 1C). The symplekin CTD, FLASH dimer (14), SLBP, pre-mRNA SL, and residues 20–65 of Lsm11 form the handle of the amphora (Fig. 1E). The FLASH dimer makes an 80 Å long connection from the symplekin CTD to the SLBP-SL complex. CstF64 was not observed in the EM density, and is not required for cleavage in vitro (Supplementary text, fig. S2F).

Twelve consecutive Watson-Crick base pairs in the HDE-U7 duplex were observed in the center of the amphora (Fig. 1D, fig. S1). The metallo-β-lactamase domain of CPSF73, the β-CASP domain of CPSF100, and the concave face of the symplekin NTD (fig. S6C) surround the duplex on three sides (Figs. 1D,1E,2A). The interactions are ionic and hydrophilic in nature, but involve none of the bases in the duplex (Fig. 2B), explaining that base pairing rather than sequence is important for processing (3, 4, 13). The structure revealed an extra, U-U base pair at the bottom of the duplex (Fig. 2C, fig. S1), and analysis of histone pre-mRNA sequences suggested that U-U base pairs are common in HDE-U7 duplexes (fig. S7).

Figure 2. Recognition of the HDE-U7 duplex and the U7 Sm site.

(A) The HDE-U7 duplex is surrounded by CPSF73, CPSF100 and symplekin NTD, shown as a transparent surface. Lsm11 has interactions with the bottom of the duplex. (B) Electrostatic surface of the proteins in the duplex binding site, showing charged interactions with the backbone of the duplex. (C) A U-U base pair at the bottom of the duplex, flanked on the other face by A19 of U7 snRNA. (D) A C-G base pair in the 3′ CUAG sequence of the U7 Sm site. The base pair is flanked on one side by Arg34 of Lsm10, and on the other by Arg174 of Lsm11.

The structure also revealed a Watson-Crick base pair between C28 and G31 of the CUAG sequence at the 3′ end of the U7 Sm site (Fig. 2D, fig. S1). It is flanked by residues from Lsm10 and Lsm11 and assumes a different backbone conformation compared to other Sm sites (Fig. 2D, fig. S8A). In addition, G26 is hydrogen-bonded with C33 of H2a*, providing a direct connection between the Sm site and the pre-mRNA (figs. S1,S8B). The recognition of the first five Sm site nucleotides (21-AAUUU-25) and U27 is similar to that in spliceosomal Sm rings (figs. S1,S8B) (17, 18), although there are substantial differences in the extensions of the Sm proteins and the positions of the RNA outside the Sm ring (figs. S8C–E).

The pre-mRNA substrate (Fig. 3A) is bound in the active site of CPSF73. The correct scissile phosphate, after A26 (fig. S1), is coordinated to the two zinc ions in the active site (Fig. 3B). The A26 base has hydrogen-bonding interactions to its N1 and N6 atoms, consistent with the preference for an adenine at the cleavage site (3, 4) (figs. S9A–C). C25 has weak density (Fig. 3A) and is not recognized by CPSF73. This binding mode of the pre-mRNA clearly illuminated the molecular mechanism for the cleavage reaction. The hydroxide ion that is a bridging ligand between the two zinc ions (6) is the nucleophile that initiates the cleavage reaction (Fig. 3B), and the 3′ oxyanion of A26, the leaving group, is protonated by His396, which is activated by Glu204. Glu204, His396 and the ligands to the zinc ions are conserved among CPSF73 homologs (19, 20) including IntS11, the endonuclease for snRNA cleavage (21). Therefore, the conformation of the machinery observed here is likely poised for the cleavage reaction. Except for the brief moment during EM grid preparation, the sample was kept at 4 °C or on ice, which slowed the reaction (12) and allowed us to observe the pre-mRNA in the CPSF73 active site. There are substantial differences in the orientation of the β-CASP domain compared to that in RNase J (22, 23) (fig. S9D), and especially in the binding modes of the RNA substrate (fig. S9E).

Figure 3. CPSF73 is in an active state, poised for the cleavage reaction.

(A) Cryo-EM density for H2a* nucleotides bound in the CPSF73 active site. The scissile phosphate is indicated with the black arrow. (B) The endonuclease mechanism of CPSF73. The positions of the zinc ions (gray spheres) and the bridging hydroxide (red sphere) are based on the crystal structure of CPSF73 alone (6) (PDB entry 2I7V). The position of the sulfate ion observed in the earlier structure is shown in thin sticks. (C) Overlay of the structure of CPSF73 in the active state observed here (in color) with the inactive, closed state reported earlier (gray) (6). The metallo-β-lactamase domain was used for the overlay. The rearrangement of the β-CASP domain is indicated with the red arrow, corresponding to a rotation of 17°. (D) Molecular surface of the active site region of CPSF73, colored by the domains. Lsm10 is located at the rim of the canyon, contacting nucleotides downstream of the cleavage site.

The reported structures of CPSF73 (6) and its yeast homolog Ysh1 (24) are in a closed, inactive conformation. We observed here an open, active conformation of CPSF73. A large rearrangement of its β-CASP domain relative to the metallo-β-lactamase domain, corresponding to a rotation of ~17° (Fig. 3C, fig. S9F), is necessary to create a narrow, deep canyon that is only large enough to accommodate single-stranded RNA (Fig. 3D, figs. S9A,F,G).

The N- and C-terminal extensions of Lsm10, highly conserved among vertebrate homologs (fig. S10A), have a crucial role in this conformational change for CPSF73. These extensions are placed directly against the β-CASP domain (fig. S11A) and have extensive steric clashes with its closed conformation (Fig. 3C, fig. S9F), likely helping to trigger the activation of CPSF73. In addition, a segment in the C-terminal extension of Lsm10 (residues 107–110) is positioned at the rim of the canyon (fig. S9A) and forms a part of the binding site for the 3′ portion of the substrate (Fig. 3D).

The recognition of the HDE-U7 duplex may be the critical event to initiate the conformational rearrangement in CPSF73, consistent with the requirement of symplekin NTD for cleavage. On the other hand, the NTD-Ssu72 complex is incompatible with the structure observed here, as Ssu72 would clash with the duplex as well as CPSF73 (fig. S6D), explaining the inhibitory effect of Ssu72 (Supplementary text).

Besides the rearrangement in CPSF73, an extensive change in the architecture of HCC is required for activation. We recently showed that mCF/HCC in an inactive state has a trilobal structure and is highly dynamic (25). In contrast, the HCC structure observed here in the active state shows dramatic differences compared to the inactive state (Fig. 4A). There are intimate contacts between CPSF73 and CPSF100 in the current structure; in fact they form a pseudo dimer (figs. S11B,C). These may be hallmarks of the active state for HCC/mCF.

Figure 4. Schematic of histone pre-mRNA 3′-end processing cycle.

(A) Significant structural differences of HCC in an active state compared to an inactive state. The structure of HCC observed here is docked into the EM density for mCF (gray surface) (25), using the symplekin CTD as the reference. (B) Schematic drawing of the CTD2 domain complex of CPSF73 (light green) and CPSF100 (darker green) and the N-terminal segment of the symplekin CTD (magenta). The CTD complex of IntS9 and IntS11 (28) was docked into the EM density at 4.1 Å resolution (transparent surface) using Chimera. Panels A and B produced with Chimera. (C) A putative model for histone pre-mRNA 3′-end processing cycle. The machinery is assembled from the U7 snRNP (state I) with the recruitment of FLASH (II) and HCC (III), followed by the recognition of the pre-mRNA for CPSF73 and HCC activation and pre-mRNA cleavage (IV). The machinery is likely highly dynamic before the binding of the authentic pre-mRNA, and the possible flexible regions are indicated with the curved arrows and dashed lines. After the cleavage (V), the downstream product is degraded by an exonuclease activity and the machinery can be recycled directly (solid arrow), or possibly disassembled followed by re-assembly. State IV corresponds to the structure reported here, with the scissors indicating cleavage by CPSF73, and the other states are models.

The conformational dynamics of HCC/mCF is due to flexibility in its core, formed by the CTDs of CPSF73, CPSF100 and symplekin (10–12, 26, 27) (Figs. 1E,4B). The CTD of CPSF73 likely has three sub-domains (CTD1–3), and that of CPSF100 has two sub-domains (Figs. 1B,E, fig. S12A). The CTD1 sub-domains of CPSF73 and CPSF100 form a six-stranded β barrel-like structure (fig. S12B). The CTD2 sub-domains form a separate complex, which makes only a small contact with the CTD1 complex, contributing to the flexibility in HCC/mCF. The overall structure of the CTD2 complex is similar to that of IntS11 and IntS9 (28). The first two helices of the symplekin CTD pack against the helices in the CTD2 complex (Fig. 4B) in the core of HCC/mCF.

The structure showed that HCC is recruited to the machinery directly by both FLASH and Lsm11 through two tethering contacts. Residues in FLASH prior to the coiled-coil domain, including the LDLY motif (10), interact with the symplekin CTD (fig. S10C), and residues 107–118 of Lsm11 interact with CPSF73 (figs. S10B,S11D) (Supplementary text). These observations explain earlier data showing the importance of the LDLY motif in FLASH and residues 65–130 in Lsm11 for HCC recruitment (10, 11, 29).

Mutagenesis and biochemical experiments supported the structural observations (Supplementary text). HCC recruitment was abolished by mutating the FLASH LDLY motif or symplekin CTD (fig. S13A). Removing the N- and C-terminal extensions of Lsm10 greatly reduced the cleavage activity without affecting U7 snRNP or machinery assembly (figs. S13B–D). Moreover, the Lsm10 mutants showed mis-processing of the pre-mRNA. Therefore, these extensions may also have a crucial role in correctly positioning CPSF73 for the cleavage reaction. Mutating as few as two symplekin NTD residues that interact with the HDE-U7 duplex (fig. S6C) greatly reduced the cleavage activity (fig. S13E). Finally, the experiments also provided evidence for an Lsm11-FLASH-SLBP-SL quaternary complex (Fig. 1E, fig. S13F).

The structure of the machinery suggests how it may be assembled for processing (Fig. 4C, movie S2, Supplementary text) and provides a molecular foundation to understand and explain the large body of biochemical and functional data on histone pre-mRNA 3′-end processing (3, 4). The structure also has significant implications for understanding canonical pre-mRNA and snRNA 3′-end processing. The binding mode of the histone pre-mRNA in CPSF73 is likely similar for canonical pre-mRNAs and snRNAs, and the active conformation of mCF for canonical pre-mRNAs is likely to be the same as that of HCC observed here. The comparison to the structure of mCF in an inactive state suggested that its correct architecture is another critical requirement for the activation of the processing machineries.

Supplementary Material

movie S1

A movie showing the composite cryo-EM map and atomic model of the overall histone pre-mRNA 3′-end processing machinery.

movie S2

A movie showing the putative histone pre-mRNA 3′-end processing cycle. The subunits in the machinery are represented by their structures. Conformational changes in the HCC and CPSF73 for their activation are also indicated.