The Renpenning syndrome–associated protein PQBP1 facilitates the nuclear import of splicing factor TXNL4A through the karyopherin β2 receptor

Renpenning syndrome belongs to a group of X-linked intellectual disability disorders. The Renpenning syndrome–associated protein PQBP1 (polyglutamine-binding protein 1) is intrinsically disordered, associates with several splicing factors, and is involved in pre-mRNA splicing. PQBP1 uses its C-terminal YxxPxxVL motif for binding to the splicing factor TXNL4A (thioredoxin like 4A), but the biological function of this interaction has yet to be elucidated. In this study, using recombinant protein expression, in vitro binding assays, and immunofluorescence microscopy in HeLa cells, we found that a recently reported X-linked intellectual disability–associated missense mutation, resulting in the PQBP1-P244L variant, disrupts the interaction with TXNL4A. We further show that this interaction is critical for the subcellular location of TXNL4A. In combination with other PQBP1 variants lacking a functional nuclear localization signal required for recognition by the nuclear import receptor karyopherin β2, we demonstrate that PQBP1 facilitates the nuclear import of TXNL4A via a piggyback mechanism. These findings expand our understanding of the molecular basis of the PQBP1–TXNL4A interaction and of the etiology and pathogenesis of Renpenning syndrome and related disorders.

Renpenning syndrome, a group of X-linked intellectual disability, is caused by mutations in human PQBP1 (polyglutamine-binding protein 1, also known as Npw38) gene on Xp11.2 (1)(2)(3). PQBP1 gene encodes an intrinsically disordered protein predominantly expressed in the central nervous system during development and playing important roles in neurodevelopment and neuronal functions (4 -7). PQBP1 is composed of three interacting domains: a WW domain interacting with the nucleocytoplasmic shuttling splicing factor WBP11 (also known as SIPP1/NPWBP/SNP70) (8 -10); a polar amino acid-rich domain (PRD) 2 interacting with polyglutamine tracts (11,12); a proline-tyrosine nuclear localization signal (PY-NLS) recognized by nuclear import receptor karyopherin ␤2 (Kap␤2) (13); and an unstructured C-terminal domain (CTD) interacting with splicing factor TXNL4A (14 -18). Despite its interaction with splicing factors, PQBP1 is also found present in the precatalytic B complex (19,20), suggesting the involvement of PQBP1 in alternative pre-mRNA splicing. Wang et al. (21) reported that depletion of Pqbp1 in mouse primary cortical neurons changed alternative splicing of mRNAs enriched for neuron projection functions and resulted in reduced outgrowth and branching of dendrites. However, most mutations of PQBP1 found in Renpenning syndrome patients, except for the missense mutation Y65C in the WW domain (22), are frameshift mutations in the PRD or CTD region, resulting in premature termination (1)(2)(3). How different mutations affect PQBP1 function remains unclear. We previously discovered that PQBP1-⌬459 -462 and ⌬463-464 mutants encode a new epitope that specifically associates with FMRP (fragile X mental retardation protein) and promotes its degradation (23). It points out the fact that frameshift mutations may gain unexpected functions with the new sequences, instead of just losing the original ones. It makes the mechanistic study more complicated.
Human TXNL4A (also known as U5-15kD/snRNP15/ hDim1) is an evolutionarily highly conserved component of the U5 and U4/U6.U5 small nuclear ribonucleoprotein particles (24). Its yeast ortholog Dim1/Dib1/Snu16 is essential for pre-mRNA splicing in vivo (25,26). TXNL4A has a thioredoxin-like fold formed by four-stranded ␤-sheets flanked by three ␣-helices (24) and binds a continuous 23-residue segment of the C-terminal domain of PQBP1 (14,16). The structure of the binary complex of TXNL4A and PQBP1 shows that the hydrophobic groove of TXNL4A formed by Val 14 , Ile 18 , Phe 69 , Met 72 , Tyr 73 , Met 82 , and Phe 84 recognizes the 245 YxxPxxVL motif of PQBP1 (17). This complex is mainly stabilized by these hydrophobic interactions as well as two hydrogen bonds (PQBP1-Y245: TXNL4A-E74; PQBP1-N255:TXNL4A-D68) (17). PQBP1 frameshift mutants that lack the C-terminal domain cannot interact with TXNL4A, which is believed to affect the function of PQBP1 in splicing (17). The interaction of PQBP1 and TXNL4A seems important for the localization of TXNL4A. Exogenously expressed TXNL4A is predominately diffused in the cells but will form large punctate nuclear structure if co-expressed with PQBP1 (15). Whether PQBP1 promotes the nuclear import of TXNL4A has yet to be elucidated.
Recently, Redin et al. (27) identified a new missense mutation of PQBP1 (c.731CϾT, P244L) in a targeted high-throughput sequencing test in intellectual disability or autism patients. This mutation is right before the 245 YxxPxxVL 252 motif of PQBP1 for TXNL4A binding. In this study, we found that the PQBP1-P244L mutant disrupted the interaction with TXNL4A. In combination with other PQBP1 NLS mutants, we clearly showed that PQBP1 facilitates the nuclear import of TXNL4A in a piggyback mechanism.

PQBP1-P244L mutant disrupts its interaction with TXNL4A
Previous studies show that TXNL4A recognizes a 245 Yxx-PxxVL 252 motif in the C-terminal domain of PQBP1, which is missing in PQBP1 frameshift mutants found in Renpenning syndrome patients. Our in vitro binding assay also demonstrated that GST-PQBP1-⌬461-462 and ⌬575-576 lost their binding capacity to TXNL4A, whereas the missense mutation Y65C in the WW domain did not affect TXNL4A binding (Fig.  1, A-C), and the RPY mutant in the NLS region only slightly affected TXNL4A binding (Fig. 1, A-C). Redin et al. (27) reported a new missense mutation in the CTD of PQBP1 (c.731CϾT) using targeted high-throughput sequencing, which changes residue 244 from proline to leucine (P244L). Pro 244 is right before the 245 YxxPxxVL 252 motif (Fig. 1A) and was not identified in crystal structures as a "hot spot" for TXNL4A binding (17). We were curious about whether P244L would affect the binding to TXNL4A, so we generated the recombinant mutant protein GST-PQBP1-P244L for in vitro binding assay. The result showed a significant decrease in TXNL4A binding of GST-PQBP1-P244L (ϳ17.6% of GST-PQBP1-WT), which is comparable with that of GST-PQBP1-⌬575-576 (ϳ13.2% of GST-PQBP1-WT) (Fig. 1, B and C).
In the structure of PQBP1-CT43 (238 -260 amino acids) in complex with TXNL4A (Protein Data Bank code 4BWQ), PQBP1 takes an L-shaped structure, and residues 238 -247 are Immobilized GST-PQBP1-WT and mutants were incubated with purified recombinant TXNL4A. Representative results from three independent experiments. C, densitometric analysis of B. The relative density of TXNL4A band against GST-PQBP1 band in each reaction was normalized to that in the reaction with GST-PQBP1-WT (100%). The data show as means Ϯ S.D. from three independent experiments. ****, p Ͻ 0.0001. MW, molecular mass.

PQBP1 facilitates TXNL4A nuclear import
in an extended conformation followed by an ␣-helix formed by residues 248 -259 (17). Pro 244 sits in a small cavity formed by Lys 88 , His 89 , and Met 91 of TXNL4A, and it is not involved in hydrophobic interactions or hydrogen bonds with TXNL4A ( Fig. 2A). Indeed, P244A mutant only showed a fairly small change in TXNL4A binding, indicating that Pro 244 is not a hot spot (Fig. 2, B and C). Mutagenesis analysis using PyMOL (28) showed P244L mutation introduced a long side chain in the small space, and the collision might disrupt the interface between PQBP1 and TXNL4A ( Fig. 2A). To test this, we made an additional mutant P244V that also introduces a long side chain at that site and performed the binding assay. As we expected, both PQBP1-P244V and P244L mutant proteins showed remarkable reduction in TXNL4A binding (Fig. 2, B and C). We demonstrated that a single residue mutation P244L in the CTD of PQBP1 abolished its interaction with TXNL4A.

PQBP1 contains a PY-NLS recognized by Kap␤2
PQBP1 is dominantly localized in nuclei (6,8,29,30). Previous reports show that PQBP1 contains a PY-NLS and is a substrate of nuclear import receptor Kap␤2 (13). The PY-NLS of PQBP1 is within the residues 170 -187 region right before the C-terminal domain (Fig. 3A). It is composed of a N-terminal basic residue-enriched epitope, a conserved Arg residue in the middle, and PY residues at the C terminus. To disrupt the interaction between PQBP1 and Kap␤2, we mutated 180 RR 181 and 186 PY 187 to alanines by site-directed mutagenesis (Fig. 3A) and tested its binding ability to Kap␤2. An in vitro binding assay showed that Kap␤2 bound to the GST-PQBP1-RPY mutant reduced to less than 10% of that bound to GST-PQBP1-WT (Fig. 3, B and C). Meanwhile, we compared the interactions of other PQBP1 pathogenic mutants with Kap␤2. The missense mutation in the WW domain Y65C did not affect the interaction between PQBP1 and Kap␤2, whereas the ⌬461-462 frameshift mutant lacking the partial PRD, NLS, and the entire CTD lost its binding to Kap␤2 (Fig. 3, B and C). The frameshift mutant ⌬575-576 that changes the sequence right after the NLS caused ϳ50% loss of Kap␤2 binding, and the missense mutant in CTD P244L slightly affected Kap␤2 binding (Fig. 3, B and C). These data verify that the interaction between PQBP1 and nuclear import receptor Kap␤2 is mainly mediated by the PY-NLS within the region of residues 170 -187.

PQBP1 can simultaneously bind Kap␤2 and TXNL4A
PQBP1 interacts with Kap␤2 and TXNL4A through different regions, which made us suspect that PQBP1 might bind Kap␤2 and TXNL4A simultaneously. We immobilized GST-PQBP1-WT proteins on GSH resins and incubated with a mixture of TXNL4A and Kap␤2. The results clearly showed that PQBP1 bound Kap␤2 and TXNL4A together (Fig. 4A), indicating that PQBP1 might function as a connector between Kap␤2 and TXNL4A. Furthermore, when we immobilized MBP-TXNL4A to test its interaction with soluble Kap␤2 in the presence or absence of different PQBP1 proteins, we found that MBP-TXNL4A bound Kap␤2 only in the presence of PQBP1-WT (Fig. 4B). In the presence of PQBP1-P244L, MBP-TXNL4A could not bind Kap␤2 because its interaction with PQBP1 was disrupted; whereas in the presence of PQBP1-RPY mutant, MBP-TXNL4A associated with PQBP1-RPY, but PQBP1-RPY could not bind Kap␤2 because of the loss of PY-NLS (Fig. 4B). Our data confirm that PQBP1 mediates the interaction between TXNL4A and nuclear import receptor Kap␤2.

Discussion
Both PQBP1 and TXNL4A are evolutionarily conserved molecules and components of spliceosome (15,19,20,24,(31)(32)(33). The molecular interaction between them is well-established (14 -18), but what the biological function of this interaction is remains unclear. TXNL4A is a member of the U5 spliceosomal complex and mainly functions in the nucleus. There is no obvious NLS in its sequence, and its molecular mass is only 15 kDa. So, it is easy to believe that it gets into the nucleus through passive diffusion by default. However, our study using exogenously expressed proteins demonstrates that receptor-mediated active transport is also involved in TXNL4A nuclear import. It is transported with PQBP1 through Kap␤2-mediated pathway in a piggyback mechanism. Interestingly, we noticed that TXNL4A-GFP was more accumulated in the nucleus (32) than EGFP-TXNL4A (15), indicating the EGFP tag position might affect the subcellular distribution of TXNL4A. Given that the N terminus of TXNL4A is responsible for PQBP1 binding (17), N-terminal tagging could affect its interaction with PQBP1 and reduce its import efficiency. Because of the very low expression level and inaccessibility to antibodies of endogenous TXNL4A, we have not been able to test the import of endogenous TXNL4A yet. Our data show that PQBP1 and TXNL4A are not just associated in the spliceosome, but they may be transported together. Llorian et al. (34) showed that splicing

PQBP1 facilitates TXNL4A nuclear import
factor WBP11, another PQBP1 interactor, could be piggybacktransported into the nucleus through the interaction with the PQBP1 WW domain. It seems that the full function of PQBP1 in splicing factor transport requires both the WW domain (35) and the C-terminal domain to be intact, suggesting that PQBP1 is a "two-pronged" adaptor for the transport of splicing factors.
PQBP1 is an intrinsically disordered protein with a small folded WW domain and long disordered regions encompassing the PRD and CTD. The unstructured C-terminal region of PQBP1 undergoes disorder-to-order transition upon complex formation with TXNL4A, and the evolutionarily conserved 245 YxxPxxVL 252 motif in this region is critical for PQBP1-TXNL4A binding (17,36). Residues Tyr 245 , Pro 248 , Val 251 , and Leu 252 are involved in the hydrophobic interactions with TXNL4A to stabilize the complex. Residue Pro 244 not involved in any hydrophobic interactions or hydrogen bonds with TXNL4A sits in a small cavity on the surface of TXNL4A that can only accommodate short side chains. This accommodation is critical for the proper position of Tyr 245 in the hydrophobic groove of TXNL4A because P244V and P244L mutants do not bind TXNL4A, but P244A does. In the structure of a fusion protein of TXNL4A residues 4 -137 and PQBP1-CT43, residues 223-243 and 260 -265 of PQBP1 are missing because of conformational flexibility, whereas residue Pro 244 is still present in the structure (17), which provides another line of evidence to support the importance of proper accommodation of residue 244.
Most mutations of PQBP1 found in Renpenning syndrome patients are frameshift mutations in the PRD and CTD regions, resulting in premature terminations (1)(2)(3). They share the common phenotypes like intellectual deficiency, microcephaly, short stature, and microrchidia (2,37). PQBP1-P244L missense mutation was recently identified in intelligent disability patients using targeted high-throughput sequencing (27). These patients show moderate intellectual disability, poor autonomy, communication and social interaction disorders, learning difficulties, and obvious autistic behaviors (27). The quite different phenotypes in P244L patients indicate a distinct molecular pathogenesis. Our study shows that P244L mutation has a slight effect on its own subcellular localization and structure but significantly affects the subcellular distribution and potential function of its partner TXNL4A. It suggests a new pathogenic mechanism of Renpenning syndrome.

Recombinant protein expression
All GST or MBP fusion proteins were expressed in Escherichia coli BL21 cells. Bacteria were cultured in LB broth (Generay Biotech, Shanghai, China) and induced by 1 mM of isopropyl ␤-D-thiogalactopyranoside for 4 h at 25°C or 12-16 h at 4°C. The proteins were purified with GSH-Sepharose (G-Biosciences) and amylose resins (New England Biolabs, Ipswich, MA) as previous described (23).

In vitro binding assays
GST-PQBP1-WT and mutant fusion proteins immobilized to GSH resins were incubated with purified TXNL4A or Kap␤2 proteins at 4°C for 30 min and extensively washed with TB buffer (20 mM HEPES, pH 7.3, 110 mM KOAc, 2 mM MgOAc, 2 mM DTT, 1 mM EGTA, 10% glycerol). Bound proteins were visualized using SDS-PAGE and Coomassie Blue staining. Approximately 5 g of immobilized GST-PQBP1 proteins were incubated with ϳ20 g of purified TXNL4A or ϳ20 g of purified Kap␤2. Approximately 25% of the bound proteins was loaded for gel analysis.

Cell culture, transfection, and immunofluorescence
HeLa cells were transfected with plasmids using X-treme GENE HP DNA transfection reagent (Roche, Basel, Switzer-

PQBP1 facilitates TXNL4A nuclear import
land) according to the manufacturer's protocol. After 16 h, the transfected cells were fixed with 4% paraformaldehyde in PBS for 20 min followed by permeabilization in 0.5% PBST for 10 min. After three washes, the cells were briefly blocked with 3% BSA in PBS, followed by incubation with anti-FLAG antibody (1:300, mouse, Abmart, Shanghai, China) at room temperature for 2 h. Then the samples were washed in 0.1% PBST and incubated with Alexa 555-labeled anti-mouse antibody (1:400, Abcam, Cambridge, UK) for 2 h at room temperature before mounting using the mounting medium with 4Ј,6Ј-diamino-2phenylindole (Vector Laboratories Inc., Burlingame, CA). The cells were examined under a Leica DM5000 B (Leica, Germany). The fluorescence images were analyzed with ImageJ software (National Institutes of Health), and the data from three independent experiments were averaged. The quantitative data are means Ϯ S.D.