Molecular insights into mRNA export regulation by the human TREX-2 complex

Abstract

The nuclear export of mRNA represents a critical regulatory node in eukaryotic gene expression. This process is orchestrated by two conserved multi-subunit assemblies: the transcription-and-export complex (TREX) and TREX-2. While TREX facilitates mRNP packaging through multivalent RNA-protein interactions, the precise mechanism by which TREX-2 contributes to mRNA export has remained elusive. Here, we report a functional interaction between UAP56 and TREX-2 and resolve the structures of TREX-2 in both apo and UAP56-bound states. UAP56 engages TREX-2 via its N-terminal region, positioning its RecA domains on the V-shaped surface of the complex. A conserved loop from TREX-2 inserts between the RecA domains of UAP56, stabilizing an open conformation. Biochemical assays demonstrate that TREX-2 significantly stimulates the ATPase activity of UAP56, thereby promoting RNA release. These findings provide structural and mechanistic insights into TREX-2-mediated regulation of mRNA export through UAP56 remodeling.

The nuclear export of mRNA is a crucial regulatory step in eukaryotic gene expression. Here, the authors show that TREX-2 interacts with UAP56, enhances its ATPase activity, and promotes mRNA release from UAP56. These findings provide key insights into the mechanisms underlying mRNA export.

Introduction

In eukaryotes, gene expression involves a series of tightly coordinated nuclear events, such as transcription, 5’-capping, splicing, and 3’-end processing^1–6. During these processes, messenger RNA (mRNA) becomes associated with multiple RNA-binding proteins, forming a messenger ribonucleoprotein (mRNP) complex. The proper assembly of mRNP is essential for the efficient export of mRNA through the nuclear pore complex (NPC)^7–13.

Two evolutionarily conserved transcription-export complexes, TREX and TREX-2, play essential roles in mRNA nuclear export^14–19. TREX is assembled via association of THO complex with ATPase/helicase UAP56 (also known as DDX39B) and the mRNA-binding protein ALYREF^10,12,16,20. It forms multivalent interactions with exon-junction complex (EJC) and mRNA, compacting them into export-competent globules^11,21,22. In contrast, TREX-2 consists of GANP, PCID2, DSS1, ENY2, and CETN2/CETN3, and is stably anchored to the NPC^15,23–28. Both TREX and TREX-2 deliver their mRNA cargo to the principal export receptor NXF1-NXT1 heterodimer, which directly interacts with FG-nucleoporins lining the NPC channel to mediate mRNA translocation^29–33. Beyond their mRNA export functions, TREX and TREX-2 are also involved in transcriptional regulation and maintenance of genome integrity^20,34,35.

The structure and function of TREX in mRNA export have been extensively characterized^{9,21,22,36–42}; however, the mechanism governing TREX-2-mediated export remain insufficiently understood^28,43–53. Two recent studies revealed a crucial functional link between mRNP-engaged UAP56 and TREX-2^54,55. Key questions persist regarding the nature of mRNP-TREX-2 interactions and the mechanistic details of mRNA handover from TREX-2 to NXF1-NXT1 (Fig. 1a). Moreover, it is still unclear whether TREX and TREX-2 mediate mRNA export through independent pathways or function synergistically to enable selective mRNA export^2,56.

Fig. 1. UAP56 interacts with the TREX-2 complex.

a A simplified schematic illustrating two key complexes involved in mRNA export, TREX and TREX-2. While both play critical roles in directing mRNA from the nucleus to the cytoplasm, key questions persist regarding the nature of mRNP-TREX-2 interaction and the mechanistic details of mRNA handover from TREX-2 to NXF1-NXT1. Furthermore, it remains unclear whether TREX and TREX-2 mediate mRNA export through independent pathways or function synergistically to enable selective mRNA export. b Volcano plot illustrating proteomics data of proteins enriched in Strep immunoprecipitations from Strep-PCID2-expressing cells compared to Strep-only controls (n = 3). Log₂ fold Label-Free Quantification (LFQ) changes of interactor signals in the individual IPs over their maternal cell line control were plotted against -log₁₀ of Student’s t-test p-values calculated over triplicate data sets. c Schematic of domain organization of the TREX-2 complex and UAP56. TREX-2^M includes PCID2, DSS1, and the Sac3 domain of GANP. d UAP56 interacts with the TREX-2^M complex. Strep-tagged TREX-2^M were immobilized on Strep-Tactin beads and incubated with purified THO, UAP56, and ALYREF, respectively. Proteins were washed, eluted, and analyzed by SDS-PAGE. Experiments were independently repeated three times with similar results. The negative controls are shown in Supplementary Fig. 1g. Source data are provided as a Source Data file.

In this study, we explored the potential functional crosstalk between TREX and TREX-2, revealing UAP56 as a co-factor of the TREX-2 complex. Through an integrated approach combining biochemical analyses, cryo-EM studies, and functional assays, we provide key structural and functional insights into the mechanisms underlying TREX-2-mediated mRNA export.

Results

UAP56 association with TREX-2

To investigate the potential interactions between TREX-2 and mRNP, we established two Flp-In T-REx 293 cell lines that stably express either Strep-tagged GANP or Strep-tagged PCID2 – the scaffold components of human TREX-2, whose depletion causes growth impairment and mRNA export defects⁵⁶. Immunopurification followed by mass spectrometry revealed enrichment of TREX-2 components and the export receptor NXF1 in Strep-GANP pull-downs (Supplementary Fig. 1a, b). Notably, UAP56 peptides were abundantly detected, whereas other components of TREX complex (e.g., THO subunits) showed minimal association. This selectivity was corroborated in Strep-PCID2 immunoprecipitations, where UAP56 – but not other TREX complex factors – exhibited significant enrichment (Fig. 1b). These mass spectrometry analyses suggest a direct physical interaction between UAP56 and TREX-2.

To validate this interaction, we initially purified the recombinant five-subunits TREX-2 complex (containing GANP^580-1264, PCID2, DSS1, ENY2, and CETN2/3). However, its poor biochemical behavior impeded further structural and functional analysis (Supplementary Fig. 1c). Consequently, we focused on the core module of TREX-2 (hereafter referred to as TREX-2^M), comprising the Sac3 domain of GANP (residues 580-1000), PCID2, and DSS1⁴⁸ (Fig. 1c; Supplementary Fig. 1d). THO complex (THOC1, -2, -3, -5, -6, -7), UAP56 and ALYREF were also expressed and purified to homogeneity (Supplementary Fig. 1e, f). Pull-down assays demonstrated weak but specific binding between TREX-2^M and UAP56, with no detectable specific interaction with THO or ALYREF (Fig. 1d; Supplementary Fig. 1g). Strikingly, stoichiometric UAP56-TREX-2^M complexes formed exclusively in the presence of ADP, ATP or its non-hydrolysable analog AMPPNP, suggesting that the nucleotide-bound state of UAP56 governs its association with TREX-2 (Supplementary Fig. 1h). Notably, DDX39A (also known as URH49⁵⁷), a paralog of UAP56, was also enriched in PCID2 immunoprecipitates (Fig. 1b). However, pulldown assays revealed substantially weaker binding of DDX39A to TREX-2 compared to UAP56 (Supplementary Fig. 1i). These findings suggest a model wherein UAP56 acts as a functional adaptor bridging TREX-2 and export-competent mRNP.

Structure of the human UAP56-TREX-2^M complex

To gain structural insights, we first reconstituted UAP56-RNA complex by incubating UAP56 with poly-U RNA and AMPPNP. Subsequent addition of TREX-2^M enabled assembly of UAP56-TREX-2^M complex. Single particle cryo-EM analysis yielded structures of TREX-2^M in the apo form and UAP56-bound state at resolutions of 3.1 and 3.3 Å, respectively (Supplementary Figs. 2, 3). Both reconstructions exhibited well-defined density for key structural motifs, enabling unambiguous modeling of side chains (Fig. 2; Supplementary Figs. 4–6; Supplementary Table 1).

Fig. 2. Cryo-EM structures of the human TREX-2^M and UAP56-TREX-2^M complexes.

a Structure of the human TREX-2^M. Shown here is the cryo-EM map (upper panel) and model (bottom panel) of TREX-2^M. The map and model are colored by subunit. b Cryo-EM reconstitution of the UAP56-TREX-2^M complex. The cryo-EM density map is colored by subunit. Two views are shown. c Cartoon representation of the structure of UAP56-TREX-2^M complex. UAP56 (colored red) binds TREX-2^M via two conserved loops (Loop-1 and Loop-2) from its N-terminal region. The UAP56 RecA1 domain is anchored onto the V-shaped surface of TREX-2^M. d Superposition of apo TREX-2^M and UAP56-TREX-2^M structures. Conformations of PCID2 and DSS1 remain unchanged, but UAP56 binding induced a rotation of the GANP winged-helix (WH) domain toward the RecA1 of UAP56.

The human TREX-2^M complex adopts a V-shaped architecture formed by GANP and PCID2, mirroring the structural organization of its yeast homolog^47–53 (Fig. 2a; Supplementary Fig. 5a). DSS1, while predominantly unstructured, stably associates with the PCID2 (Fig. 2a). In the UAP56-TREX-2^M structure, conformations of PCID2-DSS1 remain largely unchanged, but the N-terminal half of GANP Sac3 domain rotates slightly towards PCID2 due to the binding of UAP56 (Fig. 2b–d; Supplementary Fig. 5b). Intriguingly, despite reconstituting the complex with RNA-bound UAP56, the cryo-EM map revealed UAP56 in an open conformation with no observable RNA density. The RecA1 of UAP56, positioned within the cleft of the V-shaped GANP-PCID2 scaffold, was resolved with high confidence (Fig. 2b), whereas the RecA2 domain exhibited lower resolution, permitting only partial docking into the cryo-EM map (Supplementary Fig. 4f). Crucially, two N-terminal UAP56 loops anchor at the GANP-PCID2 interface, forming key interaction nodes (Fig. 2c; Supplementary Fig. 4a–c). These structural observations suggest that TREX-2 not only interacts with UAP56 but also remodels its RNA-binding conformation, potentially priming the helicase for ATP hydrolysis and subsequent mRNA release.

The N-terminal region of UAP56 is crucial for TREX-2 binding

The N-terminal region (NTR) of UAP56 was previously shown to be essential for mRNA export⁵⁸, and a recent study reveals structural differences at the NTR between UAP56 and DDX39A shape selective mRNA export pathways⁵⁷. Despite these tantalizing clues, the exact role of the NTR remains enigmatic. Our structure suggests the NTR of UAP56 may be responsible for TREX-2 binding (Fig. 2c). Two fragments of the NTR were structurally resolved in our UAP56-TREX-2^M reconstitution: Loop-1 (residues 8–16) and Loop-2 (residues 36–45) (Fig. 3a, b; Supplementary Fig. 4a, b).

Fig. 3. The N-terminal region of UAP56 is crucial for TREX-2 binding.

a UAP56 engages TREX-2^M primarily through two conserved N-terminal loops (Loop-1 and Loop-2). Shown here is a close-up view on the interface between Loop-1 and TREX-2^M. b A close-up view on the interface between Loop-2 and TREX-2^M. c Two auxiliary interaction sites between UAP56 and TREX-2^M. d Validation of UAP56-TREX-2 binding by pull-down assays. Strep-TREX-2^M was incubated with wild-type (WT) or mutant UAP56 (AAA: triple Ala substitution of Glu14-Asp15-Asp16; ∆16: deletion of residues 1-16). Complexes were affinity-purified using Strep-Tactin beads and analyzed by SDS-PAGE. Experiments were independently repeated three times with similar results. e The pulldown efficiency, expressed as a ratio of pulldown over input, was averaged from three independent repeats (Supplementary Fig. 7) and normalized to the corresponding WT protein. Data are shown as the means of three measurements ± SD. Source data are provided as a Source Data file.

Loop-1 (residues 8–16) is enriched with negatively-charged residues and engages a positively-charged cleft at the PCID2-GANP interface, a region previously assigned as a putative mRNA binding site of TREX-2⁴⁸ (Fig. 3a; Supplementary Figs. 5c, d & 6a). This interface features several hydrogen bonds: side chains of Glu9, Asp12, and Asp15 from UAP56 form H-bonds with Gln337, Lys388, and Gln389 of PCID2, respectively. The conserved Tyr13 from UAP56 makes Van der Waals contacts with Tyr376 and Val385 from PCID2 and V891 from GANP (Fig. 3a; Supplementary Fig. 4a). In addition, Loop-2 (residues 36-45) further stabilizes the interaction via H-bonds and a cation-π interaction between Lys817 of GANP and His43 of UAP56 (Fig. 3b; Supplementary Fig. 4b). Two auxiliary interfaces likely reinforce UAP56 anchoring on TREX-2^M (Fig. 3c; Supplementary Fig. 4c).

To validate our structural findings, we conducted in vitro pull-down assays using MBP-tagged TREX-2^M and UAP56 variants. Wild-type (WT) UAP56 robustly bound TREX-2^M, with input-normalized complex formation set as 100% (Fig. 3d, lane 1; Supplementary Fig. 7). Mutations in Loop-1 significantly impaired binding: L10A and Y13A reduced relative binding to nearly 20% (Fig. 3d, lanes 3 & 4; Fig. 3e); triple alanine substitution E14A/D15A/D16A (AAA) and entire Loop-1 deletion (∆16) also abolished TREX-2^M interaction (Fig. 3d, lanes 5 & 6; Fig. 3e). Similarly, the H43A/H73A double mutation, which disrupts the interface between Loop-2 and TREX-2 (Fig. 3b), significantly reduced UAP56-TREX-2 binding (Fig. 3d, lane 7 & 8; Fig. 3e). These results, combined with our structural data, establish the NTR of UAP56 as essential for TREX-2 recognition, with both Loop-1 and Loop-2 acting as critical interaction motifs.

TREX-2 stimulates the ATPase activity of UAP56

We next asked what is the biological relevance of the UAP56-TREX-2 interaction. Previous studies in yeast suggested that TREX-2 facilitates the exchange of Sub2 (UAP56 homolog) for Mex67 (NXF1 homolog) on mRNP, yet the molecular basis remained elusive⁵⁹. Our structure of UAP56-TREX-2^M complex reveals a potential regulatory mechanism: despite being reconstituted with RNA-bound UAP56 and AMPPNP, the structure exhibits UAP56 in an open conformation unclamped from RNA (Fig. 2c; Supplementary Fig. 4f). Structural overlay with RNA-bound UAP56⁶⁰ reveals that a conserved loop from GANP (residues 672-688, hereafter referred to as UAP56-activating loop (UAL)) inserts between the RecA1 and RecA2 domains of UAP56, displacing the RecA2 lobe from its RNA-bound position (Fig. 4a, b). Thus, UAL likely plays a role during the close-to-open conformational change of UAP56. Importantly, AMPPNP binds at the UAP56 ATP-binding pocket and is jointly coordinated by residues from UAP56 and UAL of GANP (Fig. 4c; Supplementary Fig. 4d, e). Specifically, the adenine ring of AMPPNP is recognized by Phe65 from UAP56 and Arg678 from UAL, with Arg678 forming a potential H-bond with the adenine moiety (Fig. 4c). Arg678 shares overlapping position with Phe381 of UAP56 – a residue critical for recognizing ATP in RNA-clamped UAP56⁶⁰. Notably, sequence conservation analysis identifies Arg678 as highly conserved from yeast to human (Supplementary Fig. 6b). This repurposing of nucleotide-binding interactions suggests that GANP UAL acts as a structural cofactor, reconfiguring the conformation of UAP56 to prime RNA release.

Fig. 4. A conserved loop from TREX-2 recognizes the ATP-binding pocket of UAP56.

a Structural alignment of UAP56-TREX-2^M (this study) and RNA-bound UAP56 (PDB: 8ENK⁶⁰). UAP56 is in an open conformation in the structure of UAP56-TREX-2^M complex. In contrast, the two RecA domains of UAP56 clamp the RNA, forming a closed conformation in the RNA-bound structure. b Close-up view highlights steric clash between the UAP56-activating loop (UAL) of GANP and RecA2 of RNA-bound UAP56. Only key structural elements are shown. c Close-up view of the ATP-binding site in the UAP56-TREX-2^M complex. The non-hydrolysable ATP analog AMPPNP is coordinated by conserved residues from UAP56. Arg678 of “UAL” recognizes the adenine moiety of AMPPNP (left), occupying the position of Phe381 of RecA2 in the structure of RNA-bound UAP56 (right). This structural feature implicates TREX-2 in regulating the ATPase activity of UAP56.

To test this hypothesis, we first investigated whether TREX-2 regulates the ATPase activity of UAP56. In vitro ATPase assays revealed that WT TREX-2^M robustly stimulates the ATPase activity of UAP56 by ~10-fold in the presence of RNA (Fig. 5a, lanes 1, 3, 4). Michaelis–Menten equation fitting showed TREX-2^M binding increased the maximum ATPase/Pi release velocity (V_max) of UAP56 from 1.43 ± 0.48 nmol^-1min^-1mg^-1 to 12.36 ± 0.46 nmol^-1min^-1mg^-1 protein. (Fig. 5b; Supplementary Table. 2). This activation depends critically on the UAP56-TREX-2 interaction: the GANP R678A mutant – which disrupts AMPPNP coordination at the UAL interface (Fig. 4c) – reduced the stimulatory effect (Fig. 5a, lanes 2, 3, 5). Similarly, the UAP56 Δ16 truncation, which impairs TREX-2 binding (Fig. 3d, e), exhibited diminished ATPase activity when combined with WT or R678A mutant of TREX-2^M (Fig. 5a, lanes 6-8). Together, these results demonstrate that TREX-2 directly activates the ATPase activity of UAP56 through a physical interaction mediated by UAL.

Fig. 5. TREX-2 stimulates the ATPase activity of UAP56, promoting RNA release.

a TREX-2^M stimulates the ATPase activity of UAP56. The ATPase/Pi release rate of WT and ∆16 mutant of UAP56 were measured in the presence of WT or R678A mutant TREX-2^M using malachite green assay. Data are shown as the means of three measurements ± SD. Significant differences were determined with a paired two-tailed Student’s t-test. **P = 0.0089, *P = 0.0273. b Measurement of ATPase activities of UAP56 both alone and in complex with TREX-2^M. TREX-2^M binding induces a ~ 10-fold increase in V_max. Data points were nonlinear fitted using the Michaelis–Menten equation. Data are means of three measurements ± SD. c TREX-2 promotes RNA release from UAP56. First, poly-U RNA was incubated with UAP56 in the presence of ATPγS, forming UAP56/RNA/ATPγS complex. Then WT or R678A mutant of TREX-2^M were added to the mixture. RNA release was analyzed by native-PAGE. WT TREX-2^M robustly displaces RNA (lanes 1–5), while the R678A mutant (lanes 6–8) shows compromised activity. The ∆16 mutants of UAP56 retain RNA-binding capacity (lanes 9–10) but fail to respond to TREX-2M (lanes 11–16). RNA shift efficiency (bound/free RNA ratio) was quantified from three replicates (Supplementary Fig. 8) and normalized to no-TREX-2 controls. Data are shown as the means of three measurements ± SD. Source data are provided as a Source Data file.

TREX-2 drives ATPase-dependent RNA release from UAP56

Integrating structural and biochemical evidence, we propose that TREX-2 facilitates mRNP handover from UAP56 to NXF1/NXT1 by activating the ATPase activity of UAP56. To test this model, we developed an in vitro RNA loading-release assay: poly-U RNA was first loaded onto UAP56 in the presence of ATPγS (an ATP analog with slow hydrolysis rate), forming stable UAP56-RNA complexes (Fig. 5c, lane 1–2). Subsequent addition of TREX-2^M triggered progressive RNA release, quantified by native gel electrophoresis (Fig. 5c; Supplementary Fig. 8). WT TREX-2^M dose-dependently displaced RNA from UAP56 (Fig. 5c, lanes 3–5). In contrast, the R678A mutant of UAL, which is defective in nucleotide coordination and ATPase activation, shows compromised activity (Fig. 5c, lanes 6-8). The ∆16 mutant of UAP56 retains RNA-binding capacity but fails to respond to TREX-2^M (Fig. 5c, lanes 9-16), aligning with its impaired binding to TREX-2 (Fig. 3d, e).

Together, these results establish a direct causal link between UAP56-TREX-2 interaction, ATPase activation, and RNA release. By coupling structural remodeling of UAP56 with catalytic activation, TREX-2 liberates mRNP from UAP56 to allow NXF1-NXT1 recruitment. Our study shows the mechanistic basis of export factor exchange conserved from yeast to humans³³.

Discussion

A unified model for UAP56-driven mRNA export

The NPC-anchored TREX-2 serves as a critical gatekeeper for mRNA export^51,61, yet the molecular basis for mRNP-TREX-2 interplay has been a long-standing enigma. Our study unveils UAP56 as the central orchestrator linking nuclear mRNP maturation to TREX-2-mediated export. We demonstrate that TREX-2 engages UAP56-bound mRNP through direct interactions with its N-terminal region (NTR) (Figs. 2 & 3), activating the ATPase activity via a conserved loop (UAL) to drive mRNP dissociation (Figs. 4 & 5). This ATP-dependent release enables transfer of the liberated mRNP to the NXF1-NXT1 export receptor, which docks at the N-terminus of GANP. Direct interactions with both UAP56 and NXF1-NXT1 position TREX-2 as a central platform for mRNP remodeling and NXF1-NXT1 transfer.

Combining with previous data, we proposed a unified model for UAP56-driven mRNA export^41,42 (Fig. 6). First, during mRNA maturation, the THO complex deposits UAP56 onto nascent mRNAs, where it cooperates with CBC, ALYREF⁷, SARNP⁶², and EJCs to compact mRNAs into export-competent globules via multivalent protein-RNA interactions^21,22. THO subsequently dissociates from the packaged mRNP via undefined mechanism, leaving UAP56 stably clamped on mRNA^33,60. At the NPC, TREX-2 recruits UAP56-mRNP by binding to the NTR of UAP56 (Fig. 3), then activates its ATPase activity to trigger mRNP dissociation. The released mRNP is spontaneously captured by NXF1-NXT1 – a decisive checkpoint for nuclear export competence²⁶ (Fig. 6). While the mechanism of THO release from UAP56-mRNP remains unclear, structural alignment demonstrates that UAP56 utilizes overlapping interfaces for interaction with both THO and TREX-2 (Supplementary Fig. 9), suggesting sequential and competitive binding during mRNP export. Additionally, SARNP, which shares a UAP56-binding site with THO, may also play a role in THO release and mRNP maturation^60,62.

Fig. 6. A working model for UAP56 in mRNA nuclear export.

First, during their passage through the nucleus, mRNAs are bound by a spectrum of RNA-binding proteins, including the CBC (colored grey) and ALYREF (orange), to form a protein-decorated mRNP. Then, the THO complex (blue), in association with UAP56 (colored salmon), is recruited to remodel and package the mRNP. THO subsequently dissociates from the packaged mRNP via an undefined mechanism, leaving UAP56 stably clamped on mRNA. Third, the UAP56-bound mRNP diffuses through the nucleoplasm until captured by the NPC-anchored TREX-2 (represented by a pair of green ellipses). TREX-2 engages the N-terminal region (NTR) of UAP56, stimulating its ATPase activity and promoting mRNP release. Finally, the liberated mRNP is transferred to the principal export factor NXF1-NXT1 (colored black), which mediates directional transport through the NPC via interactions with FG-repeat nucleoporins.

Key questions persist regarding temporal control and fidelity assurance. While structural data suggest that overlapping UAP56-THO/TREX-2 interfaces mediate sequential binding, the signals that trigger THO release remain incompletely understood. Furthermore, the molecular basis for mRNP capture by NXF1-NXT1 and quality control mechanisms ensuring error-free adapter exchange demand exploration (Fig. 6). Resolving these unknowns will illuminate how dynamic protein-RNA interactions balance export efficiency with precision.

UAP56, a conserved molecular switch

During the preparation of this manuscript, structures of the yeast Sub2-TREX-2 complex⁵⁰ and human UAP56-TREX-2 complex⁵⁴ were reported by two parallel studies. Structural comparisons revealed a highly conserved mechanism for UAP56/Sub2 regulation by TREX-2 complex (Supplementary Fig. 10a–c). In yeast, Sub2 is tethered to TREX-2 through its N-terminal motif (NTM), which mirrors the Loop-1 of UAP56. Notably, no structural element corresponding to Loop-2 of UAP56 was observed in the yeast structure. Both systems feature a conserved activating loop (also referred to as trigger loop or wedge loop) that inserts between the RecA domains of UAP56/Sub2, positioning it adjacent to the ATP catalytic pocket (Supplementary Fig. 10a–c). Importantly, TREX-2 stimulates UAP56 not by promoting ATP hydrolysis, but by facilitating RNA, ADP, and Pi release, as recently demonstrated⁵⁴.

Although the pulldown assay revealed a substantially weaker binding between DDX39A and TREX-2 (Supplementary Fig. 1i), our data show that TREX-2 is able to stimulate the ATPase activity of URH49 to a similar extent as observed with UAP56 (Supplementary Fig. 10d; Supplementary Table 2). Moreover, the ATPase stimulation is linked to RNA release: although URH49 binds poly-U RNA less efficiently, TREX-2 can nevertheless promote RNA release from URH49-clamped complexes (Supplementary Fig. 10e). These findings suggest that URH49, like UAP56, is regulated by TREX-2 through weak but specific interactions. Given previous reports that URH49 is responsible for exporting a subset of mRNA cargo, our data suggest that both URH49 and UAP56 share a common mechanism for mRNP release upon docking to the NPC-anchored TREX-2 complex, and the cargo specificity is likely determined at earlier stages of mRNP assembly.

Emerging GANP-like factors

Intriguingly, two GANP-like factors—LENG8 and SAC3D1—were highly enriched in our PCID2 immunoprecipitates (Fig. 1b). LENG8 was recently reported to form a complex with PCID2 and DSS1, mirroring the TREX-2 complex architecture, and to functions in mRNA export⁵⁵ (Supplementary Fig. 10d). Additionally, LENG8 has been reported to mediate RNA nuclear retention and degradation by acting as a dominant negative factor for TREX-2; depletion of LENG8 leads to the leakage of misprocessed mRNAs^63,64. We further demonstrate that SAC3D1 similarly forms complexes with PCID2-DSS1 and interacts with both UAP56 and DDX39A (Supplementary Fig. 10g, h). The physiological significance of these interactions remains to be characterized.

The binding mode of UAP56 with TREX-2 is structurally recapitulated in spliceosomal RNA helicases like DDX46 and DDX42^65,66, where acidic N-terminal motifs anchor to basic surfaces while flexible helicase domains enable RNA adaptability. This evolutionary parallel underscores a conserved paradigm for helicase regulation during RNA processing.

Methods

GANP and PCID2 immunoprecipitation

The open reading frames of full-length GANP (Uniport ID: O60318-1) and PCID2 (Uniprot ID: Q5JVF3-1) were individually cloned into the pcDNA5-FRT-TO vector, generating N-terminal GFP and C-terminal Strep-tag fusions. The resulting plasmids were transfected into HEK293 Flp-In T-REx cells (Thermo Fisher Scientific) to establish stable cell lines. Protein expression was induced with 0.5 μg/mL tetracycline. Cells were lysed in lysis buffer (25 mM HEPES-KOH pH 7.9, 150 mM NaCl, 3 mM MgCl₂, 0.05% NP40, 1 mM DTT) supplemented with 2 U/mL Benzonase (Beyotime) and 1× protease inhibitor cocktail (Yeasen). After centrifugation at 11,000× g for 30 min at 4 °C, supernatants were incubated with Strep-Tactin beads (IBA) for 2 h at 4 °C on a head-over-tail rotor. After six washes with lysis buffer, proteins were eluted with lysis buffer containing 50 mM biotin (Yeasen) and analyzed on SDS-PAGE gels.

Mass spectrometry analysis

Immunoprecipitates of Strep-GANP, Strep-PCID2 and Strep controls were performed label-free and in triplicates. Proteins were precipitated with 15% trichloroacetic acid (TCA) and vacuum-dried. The resulting pellet was dissolved in 8 M urea, 100 mM Tris 8.5, followed by TCEP reduction, IAA alkylation, and trypsin digestion. Tryptic peptides were resuspended in 0.1% formic acid (w/w), desalted using Pierce C18 spin column (Thermo Fisher), and separated on an EASY-nLC1200 liquid chromatography (LC) system by applying a gradient of 0-80% acetonitrile (ACN) in 0.1% formic acid over 120 minutes. Eluates were analyzed by LC-MS using an Easy-nLC system directly coupled to a Thermo Scientific Orbitrap Exploris 480 mass spectrometer. Mass spectrometry data were acquired in Data-Dependent Acquisition (DDA) mode and searched against the UniProt protein sequence database using MaxQuant, with “match between runs” and “Label-Free Quantification (LFQ)” enabled. The resulting MaxQuant protein group output was further analyzed using the DEP package for downstream processing, including normalization, imputation, and differential analysis. Log₂ fold LFQ changes of interactor signals in the individual immunoprecipitates over cell line controls were plotted against -log₁₀ of Student’s t-test p-values calculated over triplicate data sets.

Expression and purification of TREX-2^M

The cDNA constructs of GANP^580-1000, PCID2, DSS1 were individually cloned into a modified pCAG mammalian expression vector⁶⁷. To facilitate protein purification, GANP^580-1000 and DSS1 were each fused to an N-terminal Flag tag; PCID2 was fused to an N-terminal Twin-Strep tag. HEK293F cells (Invitrogen) were cultured in SMM 293T-II medium (Sino Biological Inc.) under 6% CO₂ at 37 °C with constant shaking. Equal amounts of the plasmids for the five subunits of TREX-2^M were co-transfected into cells using 40-kD linear polyethylenimines (Yeasen). Transfected cells were incubated for 72 h and harvested by centrifugation at 4000 × g for 10 minutes. Cell pellets were resuspended in lysis buffer (50 mM HEPES-KOH pH 7.9, 150 mM NaCl, 3 mM MgCl₂, 0.05% NP40, 1 mM DTT) and disrupted by sonication. Clarified cell lysates were loaded into columns containing anti-Flag M2 affinity gel (Sigma) with gravity flow. After extensive washing using lysis buffer, proteins were eluted with lysis buffer containing 0.3 µg/mL Flag peptide. The eluent was loaded to a column containing 2 mL of Strep-Tactin beads (IBA). After five washes with lysis buffer, target proteins were eluted with 50 mM biotin and applied to size exclusion chromatography (SEC, Superdex−200 Increase, Cytiva) with GF75 running buffer (25 mM HEPES-KOH, pH 7.9, 75 mM NaCl, 1.5 mM MgCl₂). Peak fractions were analyzed on SDS-PAGE gels followed by Coomassie staining.

Expression and purification of THO

cDNA constructs of THOC1, THOC2^1-1195, THOC3, THOC5, THOC6 and THOC7 were individually cloned into a modified pCAG expression vector. THOC2^1-1195, THOC5 and THOC6 were each fused to an N-terminal Flag tag; THOC1 was fused to an N-terminal MBP tag. Plasmids for the six subunits of THO were co-transfected into HEK293F cells and cultured for 72 hours for protein production. Cells were harvested by centrifugation, resuspended in lysis buffer, and lysed by sonication. Protein complexes were sequentially purified through affinity selection using anti-Flag M2 affinity gel (Sigma) and Dextrin beads (Smart-LifeSciences). The eluent was further purified using SEC (Superose-6 Increase, Cytiva) with GF75 running buffer (25 mM HEPES-KOH, pH 7.9, 75 mM NaCl, 1.5 mM MgCl₂). Peak fractions were analyzed on SDS-PAGE gels followed by Coomassie staining.

Preparation of UAP56 and ALEREF

UAP56 was fused with an N-terminal 8×His tag and sub-cloned into pET-21b expression vector. Protein was expressed in E. coli Rosetta (DE3) cells grown in LB medium. Protein expression was induced at OD600 of 0.8 by adding 0.25 mM IPTG and incubated at 16°C for 16 hours. Cells were collected in the H300 buffer (50 mM Tris-HCl, pH 8.0, 300 mM NaCl, 1.5 mM MgCl₂, 25 mM imidazole, 0.5 mM PMSF) and lysed by sonication. After centrifugation, the supernatant was loaded to columns containing Ni-NTA agarose gel (Smart-Lifesciences) and allowed to flow by gravity. After six washes with H300 buffer, protein was eluted with GF75 buffer (25 mM HEPES-KOH, pH 7.9, 75 mM NaCl, 1.5 mM MgCl₂) containing 300 mM imidazole. The eluent was further purified through ion-exchange chromatography using a heparin column (HiTrap Heparin 5 mL, Cytiva). Peak fractions of target protein were pooled, concentrated with 30 kDa Amicon Filter Unit (Millipore) and applied to SEC (Superdex-200 Increase, Cytiva) with GF75 running buffer. Purified protein was concentrated, flash frozen and stored at -80°C. Mutants and truncations of UAP56 were generated by Quick-change mutagenesis (Vazyme) and purified identically.

8×His-MBP tagged ALYREF was expressed and purified similarly with following exceptions: first, the NaCl concentration of the running buffer for SEC was adjusted to 150 mM; second, proteins were used for pull-down or cryo-EM sample preparation immediately after purification, without frozen and thaw.

TREX-2^M pull-down assay

To assess the potential interactions between TREX-2 and TREX components, 1 µM purified TREX-2^M were individually mixed with 1.5 µM of THO, 2 µM of UAP56 and 2 µM of ALYREF in GF75 buffer (25 mM HEPES-KOH, pH 7.9, 75 mM NaCl, 1.5 mM MgCl₂). The solutions were incubated at 4 °C for 2 h to allow complex formation. Proteins were captured relying on the Strep-tag at the N-termini of PCID2. After extensive washes with GF75 buffer, proteins were eluted with 50 mM biotin and analyzed on SDS-PAGE gels.

Cryo-EM sample preparation and data collection

Purified TREX-2^M were concentrated to ~1.2 mg/mL and directly used for cryo-EM specimen preparation. The UAP56-TREX-2^M complex were reconstituted in the presence 2 mM AMP-PNP and 10 µM of 20 nt poly-U RNA. Samples were applied to glow-discharged amorphous NiTi foil grids (Au300-1.2/1.3, Nanodim Tech.), blotted for 3 seconds and rapidly plunged into liquid ethane using Vitrobot Mark IV operating at 8 °C and 100% humidity. Cryo-EM data was collected with EPU on a 300-kV Titan Krios electron microscope (Thermo Fisher Scientific) using a normal magnification of 105,000×. Movies were recorded using a Gatan K3 detector equipped with a GIF Quantum energy filter (slit width 20 eV) at pixel size of 0.82 Å. The defocus value varied from -1.2 to -1.8 μm. 3,651 and 9848 micrographs were collected for TREX-2^M and UAP56-TREX-2^M, respectively.

Cryo-EM data processing

For TREX-2^M, 3651 micrographs were collected and imported into Cryo-SPARC v4.2 for further processing. After template picking, a total of 4,635,704 particles were picked, extracted and split into 10 subsets. Two rounds of two-dimensional (2D) classification were performed for each subset of the particles, resulting a total of 2,442,528 good particles. This data set of particles was then used for generation of initial models using Ab-initio reconstitution. Through several rounds of hetero-refinement, non-uniform refinement and 3D classification, a final reconstruction at 3.1 Å for TREX-2^M was obtained from 108,904 particles.

For UAP56-TREX-2^M complex, 9848 micrographs were collected, from which 7,994,980 particles were picked using Topaz picking⁶⁸. After particle extraction, two rounds of 2D classification were performed, generating 3,073,993 good particles. Six classes of initial model were generated using Ab-initio reconstitution, of which class 1 and class 4 were selected for further processing. After three rounds of hetero-refinement, NU-refinement and 3D classification, a final reconstruction at 3.3 Å for UAP56-TREX-2^M was obtained from 147,130 particles. Reported resolution limits were calculated on the basis of the Fourier shell correlation (FSC) 0.143 criterion⁶⁹. Local resolution variations were estimated using CryoSPARC⁷⁰.

Model building

The AlphaFold2 predicted structures for GANP^580-1000, PCID2, and DSS1 were fitted into the cryo-EM map of TREX-2^M in ChimeraX to generate an initial model. The model was then manually adjusted in Coot⁷¹ based on the fine features of EM map. The final models of TREX-2^M were refined against the EM maps using phenix.real_space_refine⁷², with secondary structure restraints. For model building of UAP56-TREX-2^M, the model of TREX-2^M (this study) and UAP56 (PDB: 8ENK⁶⁰) were fitted to the 3.3 Å EM map for the UAP56-TREX-2^M complex. The atomic models were validated through examination of the Molprobity scores and statistics of the Ramachandran plots⁷³.

Binding assay for TREX-2^M and UAP56

A total of 1 µM of TREX-2^M were mixed with 2 µM of wildtype or mutant of UAP56 in GF75 buffer (25 mM HEPES-KOH, pH 7.9, 75 mM NaCl, 1.5 mM MgCl₂). The solutions were incubated at 4 °C for 2 h to allow complex formation. Proteins were captured relying on the MBP-tag at the N-termini of GANP^580-1000. After six washes with GF75 buffer, proteins were eluted with 20 mM maltose and analyzed on SDS-PAGE gels. Each group were repeated for three times.

ATPase activity assay

ATPase activities of WT and truncated (∆16) UAP56 in the presence of RNA were measured by the ATPase Colorimetric Assay Kit (Innova Biosciences). 5 μM of purified protein was added to the reaction buffer (25 mM HEPES-KOH, pH 7.9, 150 mM NaCl, 5 mM MgCl₂ and 10 μM 20 nt poly-U RNA) and pre-incubated on ice for 20 min. Hydrolysis was initiated by adding 3 mM ATP and incubating at 37 °C for 10 min. Released inorganic phosphate was measured quantitatively with the ATPase Colorimetric Assay Kit by determining the light absorbance at 650 nm with the BioTek Synergy H1 Multimode Reader. To assess the effect of TREX-2M on the ATPase activity of UAP56, the assay was performed in parallel in the presence or absence of 5μM WT or mutant (R678A) TREX-2M. Each of the reaction group were repeated three times. Statistical analysis was done using GraphPad Prism 8.

Electrophoretic mobility shift assay

First, 40nM of 5’-FAM labeled poly-U RNA were incubated with 4μM of WT or ∆16 mutants of UAP56 in buffer containing 25mM HEPES-KOH (pH 7.9), 75mM NaCl, 1.5mM MgCl2, 0.5 U/μL RNase inhibitor (Yeasen), and 2mM ATPγS (Sigma). The mixtures were incubated at room temperature for 30 min to allow RNA binding. Then, increasing concentrations of WT or R678A mutant TREX-2^M were added to the pre-formed UAP56/RNA complexes and incubated for another 30 min at room temperature. Samples were resolved on 6% native polyacrylamide gels by electrophoresis in 0.5× Tri Borate EDTA (TBE) running buffer at room temperature. Gels were imaged using Amersham Imager 680 (GE Healthcare).

Statistical analysis

Data were analyzed with GraphPad Prism 8. Data were considered significant at P  <  0.05 by paired two-tailed Student’s t test. The sample sizes were chosen to ensure experiment reproducibility and to obtain meaningful results.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

X.G., R.T., H.Z., M.L., and Y.C. designed and performed all the experiments. X.G. and Y.G. prepared cryo-EM samples and collected the cryo-EM data. Xiaofei G., X.Z., and J.H. processed the cryo-EM data, calculated the cryo-EM map, and built the atomic model. All authors contributed to data analysis. X.Z. supervised the project and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Eugene Valkov and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The atomic coordinates for human TREX-2^M and UAP56-TREX-2^M have been deposited in the Protein Data Bank (https://www.rcsb.org/search) under the accession codes 9UPC and 9UPB, respectively. The cryo-EM maps for TREX-2^M and UAP56-TREX-2^M have been deposited in the EMDB with accession codes EMD-64391 and EMD-64390, respectively. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (https://proteomecentral.proteomexchange.org) via the iProX partner repository⁷⁴ with the dataset identifier PXD068408. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70088-w.