Fig 2: TREX–mRNP model and protein crosslinking.a. The TREX–mRNP complex model in a left side view shows how ALYREF could simultaneously (i) recognize the mature mRNP through its WxHD and RRM domains, (ii) bind adjacent UAP56 helicases through its N- and C-UBMs, and (iii) guide mRNA to UAP56 through its RBD1 and -2 domains. Note that N- and C-UBM assignments show one of many possible arrangements and in native complexes not all ALYREF UBMs would need to be engaged with THO–UAP56. The shown model illustrates the concept of multivalent ALYREF UBM to UAP56 interactions (see Supplementary Video 5). To obtain this model, we superimposed the UAP56–ALYREF-mRNA model from TREX monomer B on monomer A. The THO complex is shown as a transparent cartoon. An ALYREF–EJC–RNA dimer was placed into the mRNP density.b. TREX–mRNP protein-protein crosslinks agree with the model. 3,125 crosslinked residues were detected using the UV-activatable crosslinker sulfo-SDA (2% protein-protein interaction-level false discovery rate). Inter- and intra-protein crosslinks are shown for selected TREX–mRNP proteins (see Extended Data Fig. 7, Supplementary Table 3). Crosslinks within TREX are shown in dark green lines, crosslinks from ALYREF to mRNP proteins in light green, within the EJC in orange, and within mRNP proteins in grey. Protein colors as in panel a.

Fig 3: Endogenous TREX–mRNP complex structure.a. Purification scheme (left) and SDS-PAGE analysis (right) of endogenous TREX-mRNP complexes from human K562 cells (see Methods, Extended Data Fig. 4). For gel source data, see Supplementary Figure 2.b. Denoised cryo-EM micrograph of TREX–mRNPs (see Methods). TREX complexes are indicated with arrows heads. Scale bar, 500 Å.c. Single cryo-EM particles of TREX complexes on TREX–mRNPs are shown (left) next to corresponding 2D averages (right). A curved dashed line (white) separates TREX from mRNP densities in the 2D averages. A schematic of the 2D averages is shown underneath with protein colors as in panel a.d. TREX–mRNA structure shown from front and left side views. The UAP56 RecA1 (light pink) and RecA2 (pink) domains are observed in TREX monomer B, while monomer A is more mobile, precluding structural modelling of the UAP56 RecA1 lobe (see Extended Data Fig. 5d, g, h). Since ALYREF N- or C-terminal UAP56-binding motifs (UBMs) cannot be distinguished at low resolution, we label these peptides as ‘UBM’, and modelled them based on the C-UBM in the homologous yeast Sub2–Yra1–RNA crystal structure41 and AlphaFold2 multimer. In the inset (top center) TREX–mRNA complex dimers are labelled 1 and 2, and the constituent monomers A and B. THOC1 (green), -2 (light green), -3 (dark green), -5 (light blue), -6 (blue), -7 (light blue), UAP56 RecA1, -2 (light pink, pink), ALYREF (purple), putative mRNA (black).e. Details of the TREX–mRNA structure. THOC2 is colored in shades of green, other subunits and mRNA colored as in d. THOC6, and THOC5 and -7 C-terminal regions were omitted for clarity (see also Extended Data Fig. 5d, g, h). Domain organization (bottom) of THOC1, THOC2, and UAP56. Solid and dashed black lines indicate atomic and backbone regions, respectively. CD, charged domain; Death, death domain.

Fig 4: Structure of an ALYREF-exon-junction complex oligomer.a. Assembly scheme (left) and domain organization (right) of ALYREF–EJC–RNA complex components (see Extended Data Fig. 1 and Methods for details). Solid lines indicate regions included in the atomic model, dotted lines indicate protein construct boundaries. N- and C-terminal UAP56-binding motifs (N- and C-UBM); RNA-binding domains 1 and 2 (RBD1 and RBD2); RNA-recognition motif (RRM). The protein color code is used throughout.b. ALYREF55-182–EJC–RNA complex structure shown from top and side views. The structure shows an ALYREF55-182–EJC–RNA hexamer, one of the various oligomers observed in vitro (Extended Data Fig. 1c, f). In the top view (left), every second protomer is rendered transparent for clarity. In the side view (right) protomers 3 to 6 are transparent for clarity. In the cartoon inset ALYREF55-182–EJC–RNA protomers are labelled, 1 to 6, and the dimer is outlined in black.c. One ALYREF molecule bridges three EJCs, labelled 1, 2, and 3, through its WxHD (interface b) and RRM domains (interfaces c and d), suggesting mechanisms for mRNP recognition and packaging. The conserved ALYREF R144 arginine-finger (R-finger) of interface d wedges in between EIF4A3 and MAGOH of protomer 3. See Extended Data Fig. 2 for details.d. Assembly scheme (left), SDS-PAGE analysis (center), and cryo-EM (right) of the in vitro reconstituted TREX–EJC–RNA complex. Addition of recombinant THO–UAP56 to ALYREFN–EJC–RNA yields the TREX–EJC–RNA complex (center) (see also Extended Data Fig. 1l). Representative cryo-EM particles and their cartoon interpretations are shown on the right. Below, 2D averages of the ALYREF–EJC–RNA complex bound to THO–UAP56 or in isolation show an indistinguishable complex organization (see Extended Data Fig. 2d). THO complex (green), UAP56 (pink), all other protein colors as in panel a. For gel source data, see Supplementary Figure 1.

Fig 5: Biochemical characterization of TREX–EJC–RNA and ALYREF–EJC–RNA complexes.a. Domain architecture of ALYREF constructs and their nomenclature used throughout. N- and C-UBM, N- and C-terminal UAP56-binding motif; RBD1 and RBD2, RNA-binding domain 1 and 2; RRM, RNA-recognition motif; MBP, Maltose Binding Protein; 3C, PreScission protease cleavage site; His, Histidine-tag.b. ALYREFN reconstitutes the EJC in vitro. Pulldown assay with MBP-ALYREFN (bait) incubated with EIF4A3, MAGOH–Y14 (residues 66-154), or both, with or without a 15 nucleotide long single stranded (ss) RNAs and/or AMP-PNP. Complex formation was determined by SDS-PAGE analysis with Coomassie blue staining. This exact experiment was done once, but similar results were obtained in two additional experiments either without AMP-PNP or without RNA.c. ALYREFN–EJC–RNA complexes form multimers. ALYREFN–EJC–RNA was assembled on 50 (top) or 15 nucleotides (nt) long single stranded RNAs (ssRNAs) (bottom) and analyzed in sucrose density gradients. SDS-PAGE analysis with Coomassie blue staining of gradient fractions indicates multiple oligomeric states. The sucrose gradient sedimentation profile (bottom) is based on quantification of MAGOH band intensities. The sedimentation coefficients were estimated in CowSuite based on the predicted molecular weights of the different oligomeric states (an ALYREFN–EJC–RNA monomer is ~150 kDa). The sedimentation range of one to six ALYREFN–EJC–RNA complexes is indicated. We analyzed even fraction numbers and included fractions 7 and 15 to better resolve monomer and hexamer peaks using SDS-PAGE. Gradient conditions are specified on top. This exact experiment was done once, but ALYREFN–EJC–RNA multimerization was similarly observed in an experiment with different gradient ultracentrifugation parameters.d. The ALYREF WxHD domain is sufficient for EJC reconstitution. Pulldown assay with different MBP-ALYREF truncation constructs (see panel a) or MBP-CASC3SELOR as a bait and EIF4A3 and MAGOH–Y14 (residues 66-154) to probe EJC-reconstitution efficiency. Complex formation was determined by SDS-PAGE analysis with Coomassie blue staining. This experiment was done twice. For gel source data, see Supplementary Figure 3.e. ALYREF55-182–EJC–RNA oligomers form in vitro, are resistant to RNase treatment, and do not require the ALYREF RED1 and UBM domains. The ALYREF55-182–EJC– RNA complex was assembled on 15 nt ssRNA and treated (bottom) or not treated (top) with 20 µg benzonase mL-1 to digest protein-unbound RNA. The complexes were then analyzed in sucrose density gradients. SDS-PAGE analysis with Coomassie blue staining of gradient fractions indicates indistinguishable oligomeric sedimentation profiles of the ALYREF55-182–EJC–RNA complex, with or without benzonase digestion. The sucrose gradient sedimentation profile (bottom) is based on quantification of MAGOH band intensities. The hexamer peak (confirmed by negative staining, see panel f) is indicated with a grey box. Gradient conditions are specified. This experiment was done twice, the second time with 2 µg benzonase mL-1.f. Negative stain 2D class averages show that MBP-ALYREF55-182–EJC–RNA (15nt) complexes form trimeric (top) and hexameric (bottom) complexes. Cartoon interpretations are shown on the right. Scale bar, 250 Å.g. Ribbon model showing the location of mutated residues in ALYREF in the ALYREF-EJC interface. Mutated residues are shown as sticks and Ca-spheres colored by the ALYREF–EJC interface.h. ALYREF–EJC interface mutations in the ALYREF55-182 constructs reduce the efficiency of EJC reconstitution. The pulldown assay was carried out as in panel d. Mutated residues are indicated in panel g. Complex formation was determined by SDS-PAGE analysis with Coomassie blue staining. This experiment was done twice.i. Mutation of ALYREF in the ALYREF–EJC interfaces impairs ALYREF–EJC–RNA complex oligomerization in vitro. ALYREFM-b and ALYREFM-c+?d mutants were made in the ALYREF55-182 construct (see panel g for mutant details). Wild-type or mutant ALYREF55-182 or the isolated CASC3SELOR were used to assemble EJC–RNA complexes on a 15 nt long RNA and analyzed in sucrose density gradients for their multimerization. SDS-PAGE analysis with Coomassie blue staining of gradient fractions indicates loss of high-order oligomers in the sedimentation profiles of ALYREF mutants, which resemble the pattern of the monomeric CASC3SELOR. The sucrose gradient sedimentation profile (bottom) is based on quantification of MAGOH band intensities. The hexamer peak is indicated with a grey rectangle. Gradient conditions are specified on top. This exact experiment was done once.j. The in vivo mutation of ALYREF in the ALYREF–EJC interface (mutant ALYREFM-c+? d; see panel g for details) impairs its interaction with mRNP components. Wild-type FLAG-tagged ALYREFWT or the FLAG-ALYREFM-c+? d mutant were ectopically overexpressed in K562 cells, which also ectopically overexpressed THOC1-GFP. The two cell lines were used to prepare nuclear extract (NE), which were then treated with benzonase for 16 h at 4°C, including a final concentration of 5 mM MgCl2. The benzonase-treated extracts were then applied to anti-FLAG M2 resin for purification. Western blot analysis shows wild-type ALYREF or the mutant ALYREFM-c+? d (via their FLAG-tag), NCBP1, and EIF4A3. This experiment was done twice.k. The THO–UAP56 complex does not form a complex with ALYREFN in sucrose density gradients, suggesting that UAP56 binds the ALYREF UBM with low affinity as observed in yeast31. SDS-PAGE stained with Coomassie blue. The sucrose gradient sedimentation profile (bottom) is based on quantification of THOC2 and ALYREF band intensities. Gradient conditions are specified on top. This experiment was done twice. For gel source data, see Supplementary Figure 4.l. In vitro reconstitution of TREX–EJC–RNA. The recombinant proteins or subcomplexes were mixed as shown in Fig. 1d and applied to sucrose density gradient ultracentrifugation. SDS-PAGE analysis with Coomassie blue staining confirms the formation of a complex containing all eleven proteins subunits and a sedimentation coefficient of ~75 S. Gradient conditions are specified on top. This experiment was done four times.m. The ALYREF-UBM–UAP56 interaction is required to form the TREX–EJC–RNA complex in vitro. Sucrose gradient sedimentation profiles of (from top to bottom): ALYREFN–EJC–RNA, ALYREF55-182–EJC–RNA, THO–UAP56, THO–UAP56 with ALYREFN–EJC–RNA, and THO–UAP56 with ALYREF55-182–EJC–RNA. Gradient fractions were analyzed by SDS-PAGE and Coomassie staining. Bellow, sucrose gradient sedimentation profiles are based on quantifications of the EJC subunit MAGOH and THO complex subunit THOC2 band intensities. MAGOH intensities were multiplied by a factor of 3 for better visualization. Gradient fractions containing ALYREF–EJC–RNA (light grey), THO–UAP56 (light grey), or TREX–EJC–RNA (grey) are shown with rectangles. Gradient conditions are specified on top. This experiment was done five times.n. A monomeric THO complex (THOMonomer) does not form TREX–EJC–RNA complexes in vitro. Sucrose gradient sedimentation profiles of THOMonomer–UAP56 (see Methods for details) alone or in presence of ALYREFN–EJC–RNA, assembled on a 15nt ssRNA. Gradient fractions were analyzed by SDS-PAGE and Coomassie staining. THOMonomer–UAP56 did not form TREX–EJC–RNA complexes (compare to panel m). Below, sucrose gradient sedimentation profiles are based on quantifications of the EJC subunit MAGOH and THO complex subunit THOC2 band intensities. MAGOH intensities were multiplied by a factor of 3 for better visualization. Gradient conditions are specified on top. This experiment was done twice.