Structure and regulation of the nuclear exosome targeting complex guides RNA substrates to the exosome

Summary In mammalian cells, spurious transcription results in a vast repertoire of unproductive non-coding RNAs, whose deleterious accumulation is prevented by rapid decay. The nuclear exosome targeting (NEXT) complex plays a central role in directing non-functional transcripts to exosome-mediated degradation, but the structural and molecular mechanisms remain enigmatic. Here, we elucidated the architecture of the human NEXT complex, showing that it exists as a dimer of MTR4-ZCCHC8-RBM7 heterotrimers. Dimerization preconfigures the major MTR4-binding region of ZCCHC8 and arranges the two MTR4 helicases opposite to each other, with each protomer able to function on many types of RNAs. In the inactive state of the complex, the 3′ end of an RNA substrate is enclosed in the MTR4 helicase channel by a ZCCHC8 C-terminal gatekeeping domain. The architecture of a NEXT-exosome assembly points to the molecular and regulatory mechanisms with which the NEXT complex guides RNA substrates to the exosome.

. NEXT S and NEXT L cryo-EM sample preparation and data processing (related to Figure 1 The MTR4 N-terminus (residues 75-97), the entire coiled-coil ⍺1 helices (residues 45-77), the ZCCHC8 zinc-finger followed by ⍺3 helix (residues 220-260), and two short ZCCHC8 fragments tethering the RBM7-binding module to the MTR4 RecA2 domain, were modelled with the AlphaFold and rigid body fitted in either the overall 4.5 Å NEXT S density, or the 6.8 Å resolution density of the single NEXT L protomer.
The RBM7-binding module (Falk et al., 2016) was rigid-body fitted in the 6.8 Å resolution density of the single NEXT L protomer. The gray, graduated triangle below depicts model confidence in area of the composite model -higher for the rigid-body fitted MTR4 fragments and de novo built ZCCHC8 dimerization module, and lower for the AlphaFold predictions and RBM7-binding module rigid-body fitted in low resolution map. B) and C) Local resolution analysis of the overall (B) and focused (C) NEXT S complex reconstruction. Distribution of local resolution was estimated with RELION and colored accordingly. D) and E) Spherical angular distribution of the NEXT S complex particles used in the final 3D auto-refinement yielding the overall (D) and focused (E) reconstruction. F) and G) The 3D Fourier Shell Correlation plots generated with the Salk Institute software (Tan et al., 2017). The red line represents the estimated global masked half-map FSC curve indicating an overall resolution for the overall NEXT S reconstruction (F) and for the focused NEXT S reconstruction (G), according to the gold standard FSC cut off of 0.143 (Rosenthal and Henderson, 2003).
H) and I) Model vs. map FSC plots for the real space refined model of the overall NEXT S reconstruction (H) and for the focused NEXT S reconstruction (I).
J) AlphaFold prediction of the ZCCHC8 N-terminal homodimerization region 1-260 shows overall structural similarity to the cryo-EM map interpretation performed in this study. The first ~40 residues in each protomer remain unstructured, as predicted by Falk et al., 2016, and are followed by the coiled-coil helices ⍺1. In line with our findings, AlphaFold predicted a ZCCHC8 dimerization module composed of the central β-sheet with swapped β1 strands. Regions colored in red and orange correspond to the ZCCHC8 dimerization module built de novo in the 4.0 Å focused cryo-EM reconstruction of the NEXT S . In the absence of the MTR4 KOW, ZCCHC8 residues 185-220 immediately following the ZCCHC8 AIM domain (highlighted with dotted circles), are predicted to extend the dimerization module through addition of two short anti-parallel β-strands. This is the major difference between the Alpha cerevisiae Nop53 (Falk et al., 2017b) and Air2 (Falk et al., 2014). Sequence alignment reveals the consensus sequence of the AIM to be X-F/W-X-L/I/V/T-D-X-X-G/P. The C-terminal glycine or proline residue allows the ZCCHC8 chain to bend away from the AIM-binding site.
ZCCHC8 and NRDE2 sequence alignment, at the bottom of the panel, shows conserved residues within their KOW-interacting regions.    Particles from the latter class were re-extracted with a larger box size and recentered on the NEXT S ZCCHC8 dimerization module. Upon further rounds of 2D classification, we obtained a small class of particles corresponding to a NEXT S homodimer bound to two exosomes (5,633 particles) and a larger class corresponding to a NEXT S homodimer bound to a single EXO13 (11,273 particles).
3D classification of all particles used in the last 2D classification step led to a 9.5 Å 3D reconstruction that could be interpreted by fitting the known atomic models.