Mutations in nucleoporin NUP88 cause lethal neuromuscular disorder

Nucleoporins build the nuclear pore complex (NPC), which, as sole gate for nuclear-cytoplasmic exchange, are of outmost importance for normal cell function. Defects in the process of nucleocytoplasmic transport or in its machinery have been frequently described in human diseases, such as cancer and neurodegenerative disorders, but only in a few cases of developmental disorders. Here we report biallelic mutations in the nucleoporin NUP88 as a novel cause of lethal fetal akinesia deformation sequence (FADS) in two families. FADS comprises a spectrum of clinically and genetically heterogeneous disorders with congenital malformations related to impaired fetal movement. We show that genetic disruption of nup88 in zebrafish results in pleiotropic developmental defects reminiscent of those seen in affected human fetuses, including locomotor defects as well as defects at neuromuscular junctions. Phenotypic alterations become visible at distinct developmental stages, both in affected human fetuses and in zebrafish, whereas early stages of development are apparently normal. The zebrafish phenotypes caused by nup88 deficiency are only rescued by expressing wild-type nup88 and not the disease-linked mutant forms of nup88. Furthermore, using human and mouse cell lines as well as immunohistochemistry on fetal muscle tissue, we demonstrate that NUP88 depletion affects rapsyn, a key regulator of the muscle nicotinic acetylcholine receptor at the neuromuscular junction. Together, our studies provide the first characterization of NUP88 in vertebrate development, expand our understanding of the molecular events causing FADS, and suggest that variants in NUP88 should be investigated in cases of FADS.


Introduction
The nucleoporin NUP88 [MIM 602552] is a constituent of the nuclear pore complex (NPC), the gate for all trafficking between the nucleus and the cytoplasm (1).
Fetal movement is a prerequisite for normal fetal development and growth.
Intrauterine movement restrictions cause a broad spectrum of disorders characterized by one or more of the following features: contractures of the major joints (arthrogryposis), pulmonary hypoplasia, facial abnormalities, hydrops fetalis, pterygia, polyhydramnios and in utero growth restriction (15). The unifying feature is a reduction or lack of fetal movement, giving rise to the term fetal akinesia deformations sequence (FADS [OMIM 208150]) (16). FADS is a clinically and genetically heterogeneous condition of which the traditionally named Pena-Shokeir subtype is characterized by multiple joint contractures, facial abnormalities, and lung hypoplasia resulting from the decreased in utero movement of the fetuses (15).
Affected fetuses are often lost as spontaneous abortions (in utero fetal demise) or stillborn. Many of those born alive are premature and die shortly after birth. In the past, the genetic basis for these disorders was frequently unknown, but due to the recent availability of next generation sequencing, the molecular etiology is becoming increasingly understood. Many cases of FADS result from impairment along the neuromuscular axis and from mutations in genes encoding components of the motor neurons, peripheral nervous system, neuromuscular junction and the skeletal muscle.
Here, we report a Mendelian, lethal developmental human disorder caused by mutations in NUP88. We demonstrate that biallelic mutations in NUP88 are associated with fetal akinesia of the Pena-Shokeir-like subtype. We confirm in zebrafish that loss of NUP88 impairs locomotion behavior and that the human mutant alleles are functionally null. We show that loss of NUP88 affects protein levels and localization of rapsyn in cell lines and subject samples. Consistent with altered rapsyn, AChR clustering in zebrafish is abnormal. We propose that defective NUP88 function in FADS impairs neuromuscular junction formation.

Identification of NUP88 mutations in individuals affected by fetal akinesia
We performed exome sequencing and Sanger sequencing on genomic DNA from individuals affected with FADS from two families ( Figure 1A). Clinical and genetic findings are summarized in Table1 (Table 1). In Family B, one affected son was born to healthy unrelated parents of European descent. Exome sequencing in the affected individual, his parents and his two unaffected sibs ( Figure S1B) revealed that the individual is compound heterozygous for two NUP88 mutations, i.e. a nonsense c.1525C>T (p.R509*) and a frameshift c.1899-1901del (p.E634del; Figure 1A; B.II.2), originally absent in relevant databases. Variants were entered in dbSNP and gnomAD. Parents and healthy siblings were heterozygous carriers of the one or the other of the mutations, thus confirming correct segregation consistent with recessive inheritance ( Figure 1A).

Structural modelling of NUP88
The missense substitution p.D434Y and deletion p.E634del affect evolutionary highly conserved NUP88 residues ( Figure 1D) indicating functional relevance.
Accordingly, SIFT/Provean, Polyphen-2, and MutationTaster predicted both mutations to be disease causing or potentially pathogenic (Table S1). The crystal structure of NUP88 is not known, therefore we performed structural modelling.
Models obtained (see Methods) predicted the N-terminal domain (NTD) to form a 7bladed ß-propeller, set up in a (4,4,4,4,4,4,3) arrangement of ß-strands and no Velcro lock as typical for classical ß-propellers ( Figure 2A). Around 60 residues precede the ß-propeller and are located at the bottom or side of the propeller thereby shielding 2-4 blades in their vicinity ( Figure 2A). The model reveals high similarity to the PDB deposited structures of Nup82 from Baker's yeast and Nup57 from Chaetomium thermophilum ( Figure 2B). The most prominent differences are a loop region and a helix-turn-helix (HTH) motif emanating from blades 4 and 5, respectively ( Figure 2B). Models obtained for NUP88's C-terminal domain (CTD) exhibited low reliability, but the CTD, in analogy to its yeast homolog, is likely composed of extended α -helices ( Figure 2C) that form trimeric coiled-coils, either in cis or in trans. In this context, an arrangement with its complex partners NUP214 and NUP62 in trans is most likely, as described for the yeast counterpart of the complex (23,24).
According to the model structure, the p.D434Y mutation is located in the loop of a HTH motif between the two outermost ß-strands of blade 6 ( Figure 2B, overall view; Figure 2D, magnification). The mutation likely leads to a decrease in the interaction with one of the neighboring proteins, thereby leading to a destabilization of the complex. The non-sense mutation c.1525C>T resulting in p.R509* is located just after the ß-propeller in the linker region to the CTD resulting in a complete loss of all α helices. Thus, the interaction of NUP88 with its complex partners is likely reduced to only propeller interactions, if the protein is not completely lost due to non-sense mediated decay of the mRNA. The p.E634del mutation is located in the middle of the CTD sequence and predicted to lie in the last fifth of an extended helix. The deletion results in a frame-shift of the remainder of the α -helix, which shifts the following residues by about a third of a helical turn and thus disrupts the interaction pattern of all following residues, which, as a consequence, decreases the overall stability of the interactions within this helix bundle.  (25,26).

Genetic disruption of nup88 affects zebrafish development
To study the impact of nup88 deletion on zebrafish development, we used the nup88 sa2206 allele generated by the Zebrafish Mutation Project (27,28). Heterozygous nup88 sa2206 carriers were outcrossed for four generations with wild-type AB zebrafish prior to phenotypic analysis. The nup88 sa2206 allele is characterized by a nonsense mutation, c.732T>A ( Figure 3A), resulting in a premature stop codon at amino acid 244. nup88 mRNA levels are reduced by about 90% in nup88 mutants (see Figure   7E), suggesting that the mRNA is subjected to nonsense-mediated decay. For the purpose of this study, nup88 sa2206/sa2206 is therefore referred to as nup88 -/-. During early stages of development and up to 3 days post fertilization (dpf), no marked differences in morphological features of nup88 -/compared to nup88 +/+ and nup88 +/siblings were observed. Starting at 4 dpf, phenotypic alterations became visible: smaller head and eyes, lack of a protruding mouth, downwards curvature of the anterior-posterior axis, abnormal gut and aplastic swim-bladder ( Figure 3B). Further analyses of the cranial abnormalities revealed that nup88 -/larvae exhibit severe defects in the ventral viscerocranium formed by seven cartilaginous pharyngeal arches (29,30). In nup88 -/larvae, the posterior pharyngeal arches 3-7 were dramatically reduced, distorted or even absent ( Figure 3C). The reduced size of head and eyes likely correlated with an increase in apoptosis in the head of nup88 -/embryos ( Figure   3D). Apoptotic cells, as assessed by acridine orange staining, were readily detected in the eyes, the brain and the anterior trunk of 35 hpf mutant embryos, but not in other parts of the body ( Figure S2C). Together, these data indicate that nup88 mutants are phenotypically similar to the large class of jaw and branchial arch zebrafish mutants, designated the flathead group (31,32). Disruption of nup88 furthermore led to impaired survival with most death of nup88 -/larvae occurring at or after 9 dpf ( Figure   3E).

nup88 mutants show strongly impaired locomotor behavior
To further investigate the implication of NUP88 in the etiology of FADS, we next determined whether locomotor function was impaired in nup88 -/zebrafish using locomotion and touch-evoked escape assays. Zebrafish embryos develop spontaneous muscle contractions at 18 hpf (33), therefore we first analyzed the coiling behavior of nup88 -/embryos as compared to nup88 +/+ and nup88 +/embryos at 22-24 hpf. We did not detect problems in coiling behavior in nup88 -/embryos at this developmental stage ( Figure S3A, Supplementary Movie 1). Next, we analyzed spontaneous swimming activity at 4 dpf ( Figure 4A) and found that only about 35% of the nup88 -/larvae showed spontaneous movement as compared to ~83% of nup88 +/+ and about 73% of nup88 +/larvae ( Figure 4B). Moreover, those nup88 -/larvae moving displayed drastically reduced motor activity, traveled shorter distance ( Figure 4C) and initiated less often swim bouts ( Figure 4D). In contrast, statistically significant differences in the mean velocity were not observed in nup88 mutant larvae ( Figure 4E).
We next performed touch-evoked escape response assays at 3 dpf and 4 dpf. At 3 dpf, percentages of responsive animals and response duration were not significant different between wild-type and nup88 mutant larvae ( Figure S3B). At 4 dpf percentages of responsive animals were also not significant different between wildtype and nup88 mutant larvae ( Figure 4G), however, the response duration among the larvae that moved was significantly reduced in homozygous nup88 mutants in comparison to wild-type zebrafish. Interestingly also heterozygous nup88 mutants showed a shortened response duration, although less significant (Figures 4F, 4H; Supplementary Movies 2-4).

FADS-related mutations in nup88 lead to a loss-of-function phenotype
To address the question whether NUP88 mutations identified in the familial cases of FADS affect NUP88 function, we performed phenotypic rescue experiments in zebrafish. Two of the three mutated residues in the uncovered FADS cases are conserved between human and zebrafish ( Figure 1D 5C). In addition, the arches largely resembled the morphologically wild-type structures ( Figures 5A and 5D). In contrast to WT nup88 mRNA, injection of nup88 D414Y mRNA, nup88 H490* mRNA as well as nup88 E613del mRNA failed to suppress the nup88 -/phenotype ( Figures 5A-5D), indicating that the resulting nup88 mutant proteins are functionally null.

FADS-related mutations in NUP88 alter NUP88 interactions
Consistent with data from higher vertebrates, NUP88 is not essential for NPC integrity: NPCs remain unaffected by the loss of nup88 in D. rerio brain sections ( Figure S4A) and in muscle histology sections of individual B.II.2 ( Figure S4B) as revealed by immunolabelling with a monoclonal antibody recognizing NPCs (mAb414). To assess whether the NUP88 mutations identified in the individuals with FADS affect the recruitment of NUP88 to NPCs, we performed immunofluorescence microscopy of GFP-tagged NUP88 protein. Upon expression in HeLa cells, wild-type NUP88 and all mutants were co-localizing with the NPC marker mAb414, although recruitment of the NUP88 p.R509* and p.E634del mutants to NPCs appeared reduced compared to wild-type NUP88 and the p.D434Y mutant ( Figure S4C). Moreover, all forms of NUP88 also localized partially to the cytoplasm, as previously seen for NUP88 overexpression (2,12), and NUP88 p.R509* to the nucleoplasm ( Figure   S4C). To define the effect of the mutations in NUP88 on the interface with its binding partners NUP214, NUP98 and NUP62, we employed GFP trap affinity purification in combination with Western blot analysis of lysates from HeLa cells expressing the GFP-NUP88 mutants. We found that NUP88 and the p.D434Y mutant co-purified NUP214 and NUP62, while the p.R509* and the p.E634del mutant did not so ( Figure   6A). Binding of NUP214 to GFP alone was similar as compared to the p.R509* and the p.E634del mutant, indicating some unspecific binding of the NUP214 and/or the antibodies to GFP. The disrupted interaction between NUP88 and NUP214, however, did not impair NUP214 localization at NPCs ( Figure 6B), whereas NUP62 association with NPCs was reduced in cells expressing NUP88 E634del ( Figure 6C). Our GFP trap assays further showed that NUP98 associated with NUP88 and all mutant forms ( Figure 6A). Consequently, NUP98 association with NPCs appeared unaffected in cells expressing NUP88 mutants ( Figure 6D). Furthermore, the disease-related mutations in NUP88 did not affect the organization of the nuclear lamina as revealed by immunofluorescence analysis of lamin A/C (LA/C; Figure S4D), Western blots analysis of protein levels of lamin A/C and GST-pull-down down assays with NUP88 and the p.D434Y mutant ( Figures S4E and S4F). The integrity of the nuclear envelope (NE) was further evidenced by immunofluorescence analyses of the inner nuclear membrane marker Sun1 and Sun2 as well as the outer nuclear membrane marker proteins Nesprin 1 and Nesprin 2 (not shown): the organization of these NE proteins is indistinguishable between HeLa cells overexpressing wild-type NUP88 or any of the disease-related mutants. Depletion of NUP88 from HeLa cells using siRNAs had no visible effect on the distribution of lamin A/C (LA/C) and the NE marker proteins emerin, Nesprin-1, Nesprin-2, Sun1, and Sun2 ( Figure S5).
As NUP88 is critically involved in CRM1-dependent nuclear export of proteins, we asked next whether the mutations in NUP88 affect nuclear import and/or export, but we observed no defects in general nuclear protein import or export ( Figure S6A) or the three CRM1 targets mTOR, p62/SQSTM and TFEB ( Figure S6B).

Loss of NUP88 affects rapsyn levels and localization
Impeded formation of AChR clusters at the neuromuscular junction (NMJ) is considered a key defect in FADS. Given the central role of rapsyn in AChR clustering and FADS (17,18,21), we therefore asked whether a loss of NUP88 function would negatively affect rapsyn and depleted NUP88 by siRNAs from HeLa and C2C12 cells and monitored protein levels of rapsyn by Western blotting. As shown in Figure 7A, depletion of NUP88 from HeLa cells in fact coincided with a decrease in rapsyn levels. Quantification revealed that siRNA treatment led to reduction of NUP88 by 80-90% and at the same time a reduction of rapsyn by 60% ( Figure 7C). In contrast to that, NUP88 downregulation had no effect on MuSK levels, another key player in FADS (20), and no effect on known NUP88 targets, such as CRM1 and NF-κB levels.
Similarly, reduced rapsyn levels were observed in C2C12 cells depleted for NUP88 using siRNAs (Figures 7B and C) and shRNAs ( Figure 7C). Muscle biopsy of affected individual B.II.2 and immunohistochemistry on paraffin sections furthermore showed weaker staining and irregular distribution of rapsyn in the cytoplasm in comparison to biopsy samples from a control fetus ( Figure 7D). Rapsyn is known to not only localize to the plasma membrane, but also to the cytoplasm (34)(35)(36). Rapsyn protein levels could not be determined in zebrafish due to a lack of antibodies, but qRT-PCR analyses revealed a reduction of RAPSN mRNA by about 20% ( Figure 7E).
Consistent with reduced rapsyn levels, we observed impaired AChR clustering in fast-twitch muscle fiber synapses, but not in myoseptal synapses of the 5 dpf zebrafish trunk ( Figure 7G). Quantification of the size of individual AChR cluster in WT and mutant zebrafish revealed that the diameter of the AChR clusters was significantly reduced in nup88 -/larvae as compared to nup88 +/+ larvae ( Figure 7H). Interestingly, AChR cluster size was also reduced in nup88 +/larvae, both in comparison to WT and nup88 -/larvae. This reduced size of AChR clusters in the heterozygotes may account for the observed defects in touch-evoke response ( Figure 4H). In accordance with impaired neuromuscular junction formation as a consequence of loss of nup88, muscle organization in zebrafish appeared indistinguishable in electron micrographs from nup88 +/+ and nup88 -/larvae ( Figure 7F). Similarly, in affected fetus B.II.2 skeletal muscle structure was, based on the autopsy report, intact. Thus, both in vitro and in vivo evidence support the notion that loss-of-function of NUP88 has a negative effect of on rapsyn, which likely affects AChR clustering and proper formation of neuromuscular junctions.

Discussion
Here, we have identified biallelic homozygous and compound heterozygous mutations in NUP88 as cause of fetal akinesia. We demonstrate that the mutations in NUP88 lead to a loss-of-function phenotype, which coincides with reduced spontaneous motor activity and touch-evoked escape response in zebrafish. Consistent with the fact that the mutations in NUP88 affect different regions of the protein, we observed distinct effects of the mutants on binding to NUP214 and NUP62 in GFPtrap assays, whereas binding to NUP98 appears indistinguishable between wild-type and mutant forms of NUP88 ( Figure 6A). This suggests that impaired interaction with partner nucleoporins may contribute, but are unlikely to be causative for NUP88 malfunction in FADS. Our data further suggest that NUP88 malfunction in FADS is at least in part due to dysfunctional rapsyn, a known key player in FADS, and consequently impaired AChR clustering and neuromuscular junction formation.
Muscle integrity, in contrast, appears unaffected by a loss of NUP88.
Genetic disruption of nup88 in zebrafish led to pleiotropic morphological defects,  Table S3). The reduced head and eye size likely originates from increased apoptosis of neuronal cells ( Figure 3C) and it will be interesting to identify the affected subset of neurons. Reduced rapsyn levels might partly cause fetal akinesia upon loss of NUP88.
Rapsyn is one of the many contributing proteins required for the correct assembly of the AChR and is particularly involved in AChR assembly and localization to the cell membrane (37,38). We observed reduced rapsyn protein levels in the absence of functional NUP88 in cellular assays using human and mouse cell lines in combination with siRNA and shRNA-mediated depletion of NUP88 (Figures 7). Due to a lack of cell lines derived from affected individuals, we could analyze rapsyn only in histological sections from muscle of one affected fetus, which revealed a weaker staining for and a perturbed intracellular localization of rapsyn ( Figure 7D).
Consistent with aberrant rapsyn expression, AChR clustering in trunk regions of nup88 +/and nup88 -/zebrafish was impaired ( Figures 7G and 7H). Rapsyn protein levels could not be determined in zebrafish due to a lack of antibodies, but qRT-PCR analyses revealed a reduction of its mRNA by about 20% ( Figure 7E). Our data therefore suggest that absence of functional NUP88 causes fetal akinesia at least in part through misregulation of rapsyn expression. How loss of NUP88 results in reduced rapsyn levels on a mechanistic level remains to be seen. Moreover, this likely is not the only pathway by which NUP88 acts in NMJs, as effects on only one cellular pathway would be indeed (i) very surprising for a nucleoporin per se, (ii) irreconcilable with the cranial defects observed in human and zebrafish, and (iii) not in line with the broad central nervous system expression pattern of nup88 in zebrafish.
Our data, however, demonstrate (i) that the nup88 spectrum of phenotypes indeed include locomotor defects and that therefore NUP88 deficiencies might result in FADS in humans, and (ii) that human alleles are dysfunctional. We further observed that (iii) NMJ defects correlate to this phenotype. Whether this is the primary cause of akinesia is impossible to determine given the pleiotropic effects of nup88 deficiency, but at least they are sufficient to explain the phenomenon. Further mechanistic details will be subject for future studies.

RNA expression constructs
Capped messenger RNA was synthesized using the mMESSAGE mMACHINE kit (Ambion). The following expression plasmids were generated and used in this study: the full-length zebrafish nup88 ORF was cloned from 48 hpf cDNA and recombined into BamHI-XhoI digested pCS2 using In-fusion cloning (Takara). nup88 mutants corresponding to the sequences identified in human fetal akinesia cases were generated by site-directed mutagenesis using QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies). All primer sequences are listed in Table S2.
mRNAs (300 pg) were injected at the one-cell stage.

Genotyping and RT-PCR
nup88 sa2206 genotyping was performed by RFLP assay using MseI restriction of a 150 bp PCR product. Real-time PCR was done at various stages of embryonic and larval development (4-cell to 5 dpf). Primer sequences are listed in Table S2. Staging of embryos was performed according to (58).

Alcian blue staining
Alcian Blue staining and histology were performed as described elsewhere (59).
All images were acquired using an Olympus SZX16 stereomicroscope and an Olympus XC50 camera using the imaging software Cell* after embryo anesthesia with a low dose of tricaine.

Apoptosis assay using acridine orange in live embryos
Live embryos at stages 24 hpf, 36 hpf, 48 hpf were dechorionated and immersed in egg water containing 5 μ g/ml acridine orange. They were incubated at 28.5°C for 15 minutes in the dark and then thoroughly washed with egg water. Embryos were mounted in low-melting agarose for positioning and immediately imaged using an Axio Observer Z1-1 microscope. Images were processed using Zeiss Zen TM software.

Behavioral assays in zebrafish embryos and larvae
Spontaneous tail coiling of 22-24 hpf embryos, placed in the grooves of an agarose chamber, was recorded for 2 minutes at 30 frames per second. Coiling events were scored manually. To analyze spontaneous locomotion, 4 dpf larvae were placed into 96-well plates (one larva per well). Their behavior was recorded at 30 frames per second for 5 minutes and quantified using EthoVision XT 8.5 software (Noldus).
Touch-induced escape responses were analyzed in 4 dpf head-restrained larvae.
Larvea were first embedded in 2% low melting point agarose, and then the agarose surrounding their tail was removed with a blade. Escape behavior was induced by touching the tail of larvae with a plastic pipette tip and recorded at 300 frames per second. Duration of behavioral responses was quantified with ImageJ.

qRT-PCR on 5 dpf zebrafish larvae
Total RNA of 5 dpf nup88 mutant and WT zebrafish larvae was extracted using

Transmission electron microscopy of zebrafish skeletal muscle
Skeletal muscle of 5 dpf nup88 +/+ and nup88 -/zebrafish embryos were analysed by transmission electron microscopy on longitudinal and transversal ultrathin sections.
Briefly, embryos were fixed in 2.

Immunohistochemistry-paraffin (IHC-P) labelling of fetal muscle sections
Brain and skeletal muscle paraffin-embedded blocks were obtained from autopsy of individual B.II.2. Written consent form for use of paraffin samples for functional analysis was signed by the parents. Control tissues of a fetus of the same age without neuromuscular disorder were used for IHC-P. Blocks were cut with a microtome and 5 µm sections were mounted. Sections were deparaffinized and rehydrated. Heatinduced epitope retrieval was performed using Tris/EDTA, pH 9.0 for all tissues when using antibodies against rapsyn and with sodium citrate, pH 6.0 for anti-nucleoporin antibody mAb414. 0.3% H 2 O 2 for 20 minutes was used to block samples endogenous peroxidase followed by washes with PBS. Tissues were permeabilized using PBS containing 0.5% Triton X-100 for 5 minutes at RT and subsequently blocked with After several PBS washes, the samples were counterstained with hematoxylin for 15 seconds, washed thoroughly under running tap water to remove excess staining agent and mounted for subsequent observations with Aquatex. Images were acquired with an Olympus BX41 microscope and were processed using Olympus cellSense TM software.

Plasmids for studies in human cell lines
For all constructs, human NUP88 was amplified by PCR. All constructs were verified by DNA sequencing. GFP-NUP88 was produced as described previously (2).
FLAG-NUP88 was cloned into KpnI/XbaI cut pFLAG-CMV2 (Sigma-Aldrich). GFP-Nup88 and FLAG-NUP88 mutants were generated by site-directed mutagenesis using the QuikChange Lightning site-directed mutagenesis kit (Agilent Technologies) following the manufacturer's instructions. Primers are listed in Table S2.

Antibodies
The following polyclonal antibodies were used in this study for Western blotting

Immunofluorescence microscopy of HeLa cells
HeLa cells were grown on glass coverslips, transfected, fixed in 4% PFA in PBS for 5 min, permeabilized with 0.5% Triton-X-100 in PBS for 5 min and then fixed again. Blocking was performed with 2% BSA/0.1% Triton-X-100 in PBS for 30 min at RT. Primary antibodies were incubated at 4°C over-night in a humidified chamber.
Secondary antibodies were incubated 1 hour at RT in the dark. Excess antibodies after primary and secondary antibody staining were removed by three washing steps using 0.1% Triton-X-100 in PBS for 5 min. Cells were imaged using a Zeiss LSM 710 (Zeiss, Oberkochen, Germany) confocal laser scanning microscope with Zeiss Plan-Apochromat 63x/1.4 oil objective. Images were acquired using the microscope system software and processed using Image J and Adobe Photoshop (Adobe Systems, Mountain View, CA).

GFP-trap assay
HeLa cells, grown in a 10 cm dish, were transfected with GFP and GFP-NUP88 wild-type and mutant variants, respectively. Cells were grown for 48 hours at 37°C in a humidified atmosphere with 5% CO 2 . To harvest cells, growth medium was aspirated off, 1 ml of ice-cold PBS was added to cells and cells were scraped from dish. The cells were transferred to a pre-cooled tube, spinned at 500 g for 3 min at Tween 20 and 5% non-fat dry. The membranes were next incubated with secondary antibodies for 1 hour, washed 3x in TBS and developed. X-ray films were scanned and processed using ImageJ. BF, BV, EB, AF, GR, KH, RJ, MH, and LP wrote the manuscript. BF, BV, and KH  designed and supervised the study. EB, PC, AF, RJ, SK, MH, AD, VM, MV, LS, NM,  LP, GR, ML, SP, CS, SJ, ER, JVD, ML, RHK, RF, NL, KH, BV, and BF acquired, analyzed and interpreted the data.