Super-resolution mapping of scaffold nucleoporins in the nuclear pore complex

Nuclear pore complexes (NPCs) play a central role in nucleocytoplasmic transport of macromolecules between the nucleus and the cytoplasm of eukaryotic cells (Köhler and Hurt, 2007; Grünwald et al., 2011; Hinshaw et al., 1992; Mattaj and Englmeier, 1998). Human NPCs are composed of ∼30 different nucleoporins (Nups), together having a molecular mass of over ∼110 MDa (Reichelt et al., 1990; Rabut et al., 2004; Doye and Hurt, 1997). Among these Nups, approximately one third are intrinsically disordered proteins containing multiple phenylalanine-glycine (FG) repeats and the rest are folded proteins forming the architecture of the NPC (Terry and Wente, 2009; Yamada et al., 2010; Milles and Lemke, 2011). By using electron microscopy (EM), the cylinder-shaped structure of NPCs has been shown to have an eightfold scaffold geometry with rotational symmetry (Frenkiel-Krispin et al., 2010; Maimon et al., 2012; Brohawn et al., 2008; Akey, 1989). Along the axial dimension, the NPC is made up of cytoplasmic fibrils, a central scaffold region and the nuclear basket, which all together are approximately 200 nm in length. In the radial dimension, three layers are described: the outermost transmembrane layer, the outer scaffold ring and the inner scaffold spoke ring. These layers have three separate distinct diameters and location sites as determined by EM (Reichelt et al., 1990; Stoffler et al., 1999; Vasu and Forbes, 2001; Suntharalingam and Wente, 2003). However, it remains impossible to exactly determine the specific location and arrangement of individual species of Nups at high resolution using EM data (von Appen and Beck, 2016). Additionally, the copy number of each Nup remains imprecisely determined in native NPCs within cells, having thus far only been estimated by using western blot analysis (Cronshaw et al., 2002), EM (Alber, 2007; von Appen and Beck, 2016), photobleaching steps of fluorescently labeled Nups (Ries et al., 2012; Mi et al., 2015) and proteomic approaches (Ori et al., 2013a,b). Although the abovementioned methods can provide an approximation of spatial location and copy number for Nups, a new method for directly imaging specific Nups and counting the copy number of each Nup within native NPCs is still needed (Hurt and Beck, 2015).

Aside from EM, fluorescence techniques have begun to be employed to probe the structure of the NPC and continue to provide great advances into understanding the structure and distribution of NPC components. The spatial orientation of some Nups has been determined through polarized light microscopy (Kampmann et al., 2011) and the relative position of several NPC subcomplexes has been shown through 3D structured illumination microscopy (Schermelleh et al., 2008; Bilokapic and Schwartz, 2012). More recently, single-molecule localization-based super-resolution imaging techniques have been applied towards the goal of resolving the structure and positions of individual Nups within the NPC. So far, stochastic optical reconstruction microscopy (STORM) and photo-activated localization microscopy (PALM) have been applied in this manner (Löschberger et al., 2012; Rust et al., 2006; Xie et al., 2016). These aforementioned microscopy techniques normally require the application of large antibodies (>100 kDa) and/or photoswitchable probes in chemically fixed cells. These requirements may perturb the native distribution inside of NPCs. As an alternative, nanobodies possessing a molecular weight of ∼13 kDa can be used because they are small enough to enter native NPCs and can specifically bind to Nups. By combining these nanobodies with fluorescent dyes, a Nup-specific probe is created. In this manner, the concentration of nanobody can be controlled to ensure optimal single-molecule detection (Szymborska et al., 2013; Pleiner et al., 2015).

In this paper, we have developed a new approach of combining single-point edge-excitation sub-diffraction (SPEED) microscopy and nanobody techniques to map both the radial and axial spatial distributions and the copy numbers for scaffold Nups in the native NPCs of mammalian cells. Prior to this study, only the radial dimension has been provided for several Nups by using super-resolution techniques (Löschberger et al., 2012; Szymborska et al., 2013). In our approach, nanobodies against GFP and labeled with Alexa-Fluor dyes were used to recognize individual GFP-fused Nups within native human NPCs. SPEED microscopy was used to provide a spatial distribution for each tested Nup in both the radial and axial dimension, which was achieved through illumination of Nups within single NPCs at bottom of nuclear envelope (NE) for the radial view, and at the edge of the equator of the NE for the axial view (Ma and Yang, 2010; Ma et al., 2012, 2013, 2016). Single-point laser illuminations were used to illuminate single NPCs in contrast to wide-field setups, such as STORM and PALM, in order to reduce background noise and increase precision in identifying the location of single molecules. Here, we have revealed spatial data and inferred copy number for various scaffold Nups found within three separate scaffold layers of the NPC, including the transmembrane-layer Nup Pom121; an outer ring Nup, Nup37; and an inner spoke ring Nup, Nup35.

Our epifluorescence microscopy imaging experiments have shown that GFP-tagged Nups in native NPCs can be recognized well by nanobodies at micro-molar concentrations that target GFP (Fig. S1). Furthermore, to obtain the radial distributions of Nups within a single NPC, an overlapped laser beam comprising 488-nm and 633-nm lasers was directed to the bottom of the cell nucleus and focused as a diffraction-limited illumination point spread function (iPSF) on a single NPC at its cross section (Fig. 1A). For the axial distribution of Nups within a singe NPC, the overlapped laser beam was focused on the edge of the nuclear equator by forming an inclined iPSF at an angle of 45° to the z direction (Fig. 1B). The inclided iPSF possesses projected PSFs with same sizes in the xy and xz planes (Ma and Yang, 2010; Ma et al., 2012; 2013; 2016). Based on the Rayleigh criterion, the vertical and inclined iPSF in the x,y and x,y planes, respectively, is ≈210 nm and ≈320 nm for the 488-nm laser (Fig. 1A), and ≈270 nm and ≈360 nm for the 633-nm laser (Fig. 1B). These PSFs enabled us to illuminate a single NPC at the nuclear bottom or equator because they are smaller than the average nearest neighbor distance for NPCs at 400–600 nm in HeLa cells (Ma and Yang, 2010, 2012; Kubitscheck et al., 2005). In this way, both single GFP-tagged NPCs and individual Alexa-Fluor-647-labeled nanobodies were illuminated. Of note, our localization approach is reliant on GFP tags being present on each Nup of interest, and it cannot be guaranteed that for every NPC all copies of that Nup will be tagged with GFP (particularly in transiently transfected cells). Thus, only individual GFP-tagged NPCs with fluorescence intensity values corresponding to at least eight copies of GFP were selected for subsequent single-molecule localization. This pre-selection is to avoid those NPCs that may be severely under-sampled. To ensure nanobody–Nup binding events were infrequent, well resolvable and well isolated, nanobody concentrations were reduced down to 0.1 nM. We found that approximately 40% of imaged nanobodies against GFP firmly bound within NPCs to GFP-tagged Nups. These binding events lasted many frames with dwell times ranging from 250 ms to 8 s before photobleaching occurred and the signal diminished. The other 60% of imaged nanobodies diffused quickly through NPCs within 2 ms, presumably without binding to a Nup. Given the size of the used nanobodies, this quick transport time agrees with previously recorded data for passively diffusing 10-kDa dextran transport through NPCs (Ma et al., 2012). By selecting binding events lasting at least 250 ms and setting a minimum photon requirement of 2000 (Fig. 2A), transient diffusing or binding events were avoided in the final data.

A typical single-molecule trajectory for a nanobody–Nup binding event is shown in Fig. 2A. By combining all images recorded before photobleaching, the location of this GFP–Nup can be determined with localization precisions of ∼1.6 nm (Fig. 2B; Fig. S1). Also, only GFP-tagged NPCs providing three or more separate location sites of Nups were included in in our analyses (Materials and Methods). Previous cryo-EM studies have been used to suggest that the cross-section of the scaffold of an NPC forms a circle shape with negligible error (Beck et al., 2007). Radial data for each Nup was fitted with a circle function to reveal the central point (Fig. S2; Materials and Methods). Fitting data for each Nup to a circle and rotating the data around the centroid generated symmetry of eight distinct positions at the cross section of the NPC.

For Pom121, the merging of spatial location sites from 16 NPCs provides the final radial distribution as shown in Fig. 2C-E. Eight distinct grouped location sites are shown, reminiscent of the eightfold symmetry detected in EM images of the NPC structure. This ring-shaped spatial distribution of Pom121 has a diameter of 118±4 nm (mean±s.e.m.). For Nup37, merging data from eight NPCs revealed eight clustered location sites with a circular diameter of 90±4 nm. For Nup35, data obtained from 20 NPCs revealed eight grouped location sites forming a ring with a diameter of 64±5 nm (Fig. 3A-C,G-I). Our above results not only reaffirmed that these scaffold Nups are distributed at three distinct and separate layers (Alber et al., 2007), but also provided the accurate spatial location for each of them in the radial dimension of native NPC. A montage of data from individual NPCs can be seen in Fig. S2.

Although radial imaging of the NPC allowed us to obtain distributions representing the radial organization of scaffold Nups, this view is limited and cannot alone reveal the complete distribution and copy number for each Nup. There may be more than one ‘layer’ of Nups along the axial direction of the NPC (Fig. S3). The predicted copy number for a single Nup species within the NPC is assumed to be 8, 16 or 32 due to eightfold symmetry. To account for these numbers, there may be 1, 2 or even 4 layers of the Nup present along the NPC axis. Data along the axial dimension would predictably distribute in 1, 2 or 4 isolated lines (axial view in Fig. S3). To obtain a view along the axial dimension of the NPC, we illuminated a single NPC located on the equator of the nucleus on either the left or right side (Fig. 1B). Similar to radial view imaging, only GFP-tagged NPCs providing three or more separate locations of Nups were included in in our analyses (Materials and Methods). As shown in Fig. 2F-H and Table 1, the spatial organization of Pom121 revealed a single line along the axial dimension of the NPC. However, two isolated line distributions were found for both Nup37 and Nup35 per NPC, separated by greatly varying distances of 36±3 nm and 7±3 nm, respectively (Fig. 3E-F,J-L). Thus, Pom121 was determined to have a single layer while Nup37 and Nup35 were determined to have two layers along the axial dimension of the NPC.

Owing to Pom121 having displayed an eightfold grouping in the radial distribution and a single layer in the axial dimension, a total copy number of eight is suggested corresponding to the model in Fig. S3i. Furthermore, the GFP intensity distribution of NPCs that expressed GFP–Pom121 confirmed this number, as shown in Fig. S4. In detail, the GFP intensity values from 24 NPCs of 24 measured cells revealed 8±1 (mean±s.e.m.) copies of GFP–Pom121 per NPC. Data for both Nup37 and Nup35 revealed eight groupings in the radial dimension and two layers in the axial dimension. From the data alone, we cannot fully distinguish between models iii, iv and v in Fig. S3 as it is possible that there could be 8 or 16 copies in each Nup layer in the axial dimension. Similar to the intensity-based determination of copy number for Pom121, GFP intensities were recorded for randomly selected NPCs containing either GFP-tagged Nup37 or Nup35 to estimate the copy number (Fig. S4B,C). For Nup37, ∼49% of tested NPCs had GFP intensity values corresponding to more than 16 copies of GFP. This provides evidence favoring 32 copies of Nup37 structurally incorporated into native NPCs and suggests a final spatial distribution corresponding to model v in Fig. S3. For Nup35, the GFP intensity distribution reveals that ∼50% of measured NPCs contained intensities corresponding to more than eight copies of GFP, while 100% of measured NPCs had fewer than 32 copies. This information is consistent with 16 copies of Nup35 being present in native NPCs and with models iii and iv in Fig. S3. Instead of a single-peak position indicating eight copies of GFP in NPCs that comprise GFP-POM121, NPCs comprising GFP-Nup37 or GFP-Nup35 do not have single-peak positions that indicate 8, 16 or 32 copies of GFPs. In contrast to the stable cell line used for analysis of GFP–Pom121, transiently transfected cells were used to analyze Nup35 and Nup37, in which it is likely that there could be some unlabeled endogenous Nups in the NPCs. Meanwhile, since no distribution pattern was found to have four lines in the axial dimension, it can be concluded that none of these tested scaffold Nups has a final distribution matching those suggested in model vi of Fig. S3.

An accurate copy number and spatial distribution for Nups within native NPCs are urgently needed to produce a better three-dimensional architectural maps of NPCs. Moreover, precise information on subunit stoichiometry can further advance our understanding of the function, biogenesis and compositional dynamics of NPCs. Here, we have demonstrated that the combination of SPEED microscopy and nanobody-specific labeling can provide both accurate spatial location and presumed copy number for individual scaffold Nups in native NPCs. By analyzing scaffold Nups Pom121, Nup37 and Nup35 in native NPCs, we recorded their separate spatial locations with distinct diameters of ∼118 nm, ∼90 nm and ∼64 nm at the scaffold rings of NPC, corresponding to the transmembrane ring, outer ring and inner spoke ring, respectively. These distributions in the radial dimension of NPCs reveal similar distributions to those obtained in recent studies (Löschberger et al., 2012; Bui et al., 2013; von Appen and Beck, 2016). For instance, the transmembrane ring diameter of 118 nm for Pom121 agree well with a diameter of 114 nm determined by cryo-EM (Bui et al., 2013; von Appen and Beck, 2016). An earlier super-resolution paper by Löschberger et al. describes a diameter of 161±17 nm (mean±s.d.) obtained for the radial dimension of Gp210 (Löschberger et al., 2012). Given the facts that Pom121, Gp210 and Ndc1 together form the transmembrane ring Nup complex, and that different diameters have been determined for Pom121 and Gp210, these two Nups may be located at different sites within the transmembrane ring. Our suggestion is that Pom121 may locate at the inner edge of the ring facing towards the NPC center and Gp210 could stay around the outer edge of the ring. Some previous work on the location of Pom121 within the NPC supports this suggestion (Hallberg et al., 1993; von Appen and Beck, 2016).

The outer ring Nup, Nup37, has previously been studied using super-resolution microscopy in the radial dimension revealing a diameter of ∼91 nm (Szymborska et al., 2013), which is nearly identical to the 90-nm diameter obtained from our study. Our radial determination also agrees with the radial diameter of the Nup107 complex (containing Nup37) at ∼84 nm in the NPC previously obtained from EM measurements (Bui et al., 2013). Moreover, our axial dimension data reveals two layers separated by 36±3 nm, which is consistent with recent EM data locating the Nup107 complex (containing Nup37) at both the nuclear and cytoplasmic outer ring structures (Bui et al., 2013; von Appen and Beck, 2016). Finally, our determination of 32 copies – two layers each containing 16 copies – for Nup37 also agrees well with the results obtained from previous EM measurements and proteomic approaches (von Appen and Beck, 2016; Ori, 2013a,b; Cronshaw et al., 2002; Alber et al., 2007). Additionally, previous EM data has shown that the inner spoke ring Nup complex, comprising Nup205, Nup188, Nup155 and Nup35, has a radial diameter of ∼41 nm (Bui et al., 2013; von Appen and Beck, 2016). This is significantly smaller than our determined diameter of ∼64 nm for Nup35. Given the facts that the EM measurement did not specifically look at individual Nup35 molecules and that Nup35 is capable of binding to nuclear membranes, perhaps Nup35 could be present predominantly on the outer edge of the inner spoke ring Nup complex. As a result, Nup35 could possess a larger diameter than the centroid of the entire inner spoke ring Nup complex. Prior to our study here, the copy number of Nup35 has been only estimated by using bulk protein and cell lysate approaches (Cronshaw et al., 2002; Alber et al., 2007). Here, our high-resolution data provide a more complete distribution for 16 copies of Nup35, distributed as two layers that are ∼7 nm apart.

The relative spatial locations of these scaffold Nups in both the radial and axial dimensions of NPCs mapped here can provide experimental evidence and have implications regarding NPC assembly and nucleocytoplasmic transport. First, the determined single layer middle position of Pom121 with a radial diameter bigger than that for Nup35 and Nup37 supports the proposal that the transmembrane Nup complex (formed by Pom121, Ndc1 and Gp210) could anchor the NPC into the NE during the early stages of NPC assembly, and then subsequently recruit the inner spoke ring Nup155 complex (containing Nup35) and the outer ring Nup107-160 complex (containing Nup37) to the NPC (Talamas and Hetzer, 2011). Second, the double-layer arrangements of Nup35 and Nup37 agree with the geometry that almost all structural Nups are symmetrically distributed on the cytoplasmic and nuclear sides of NPCs. Meanwhile, the double-layer distributions for the inner ring and outer spoke ring Nups, together with linker Nups (Nup88 and Nup99), could also be capable of providing attachment sites that are specifically adaptable for either symmetrically or asymmetrically located FG-containing Nups along the NPC transport pathway (Alber et al., 2007).

Conclusively, our high-resolution data firstly reaffirm the previously estimated copy numbers for Pom121 at eight copies and of 32 copies for Nup37, and also clarify that the copy number for Nup35 is 16 copies (Bui et al., 2013; Cronshaw et al., 2002; Alber et al., 2007). Also, accurate spatial locations of these scaffold Nups in both the axial and radial dimensions have been provided in native NPCs. We believe that this multi-dimensional approach to determine subunit stoichiometry will enable us to provide a complete map for the entirety of Nups in the native NPCs, and eventually help to refine understanding of the 3D architecture and function of NPCs.

A wild-type HeLa cell (American Type Culture Collection) line was used. Freshly split cells were grown overnight on coverslips in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS. N-terminal GFP-fused Pom121, Nup35 or Nup37 was transfected into the HeLa cell line separately. In our experiments, a stable GFP–Pom121 HeLa cell line and transiently transfected cells expressing GFP–Nup35 or GFP–Nup37 were utilized. For microscopy imaging, flow chambers were constructed with a top coverslip and two lines of silicone grease as spacers. Cells were washed with transport buffer (20 mM HEPES, 110 mM KOAc, 5 mM NaOAc, 2 mM MgOAc, 1 mM EGTA, pH 7.3), permeabilized for 2 min with 40 μg/ml digitonin in transport buffer, and washed again with transport buffer supplemented with 1.5% polyvinylpyrrolidone (PVP; 360 kDa). PVP was included in all transport buffer solutions after digitonin treatment to prevent osmotic swelling of nuclei.

The Nup plasmids used in this experiment contain N-terminal GFP fusions. GFP-targeting nanobodies were purchased from Chromotek. Succinimidyl-esters of Alexa-Fluor-647 (Invitrogen) were dissolved at 10 mg/ml in DMSO, stored at −80°C and, for labeling, added in twenty-fold molar excess to the nanobodies. The mixture was incubated for 2 h at room temperature and the products were dialyzed into 0.2 M NaHCO₃, pH 8.2, in a mini-dialysis unit to remove free unbound dyes. The final labeling ratio is approximately two dyes per nanobody. For single-molecule measurements, 0.1 nM of labeled nanobody was added with an oxygen scavenging system comprising 5 mM protocatechuic acid (PCA) and 100 µM protocatechuate dioxygenase (PCD) in transport buffer with 1.5% PVP.

The SPEED microscope includes an Olympus IX81 equipped with a 1.4 NA 100× oil-immersion apochromatic objective (UPLSAPO 100×, Olympus), a 35-mW 633-nm He–Ne laser (Melles Griot), a 50-mW solid-state 488-nm laser (Obis), an on-chip multiplication gain charge-coupled device (CCD) camera (Cascade 128+, Roper Scientific) and the Slidebook software package (Intelligent Imaging Innovations) for data acquisition and processing. An optical chopper (Newport) was used to generate an on–off mode of laser excitation. The two lasers were combined with an optical filter (FFF555/646 Di01, Semrock), collimated and focused into an overlapped illumination volume in the focal plane. The green and red fluorescence emissions were collected by the same objective, filtered by a dichroic filter (Di01- R405/488/561/635-25×36, Semrock) and an emission filter (NF01- 405/488/561/635-25×5.0, Semrock) and imaged by an identical CCD camera.

The position of the NE was determined at super accuracy by fitting the fluorescence of GFP–Nup as follows. The pixel intensities within a row or a column approximately perpendicular to the NE were fitted with a Gaussian function. The peak position of the Gaussian function for a particular set of pixel intensities was considered the NE position for that row and column. The peak positions of a series of such Gaussian functions were then fit with a second-degree polynomial, yielding the orientation of the NE within the entire image. The following rules were then used to select a single NPC and determine its orientation, which needs to be perpendicular to the NE at the bottom or on the equator of the nucleus and to the x or y direction of the Cartesian coordinates (x,y) in the CCD camera: first, to find a GFP-tagged NPC with 8, 16 or 32 copies of a GFP-tagged Nup. In detail, individual GFP-tagged NPCs on the NE were selected when their fluorescence intensity corresponded with at least 8 copies of GFP. Second, we chose a fluorescent NPC at the bottom or on the equator of the nucleus such that the tangent of the NE at the location of this NPC was parallel to the x or y direction of the Cartesian coordinates (x,y) in the CCD camera. Third, we examined the ratio of Gaussian widths in the long (L) and short (S) axes of the chosen GFP-tagged NPC fluorescence spot, which needed to be ∼L:S=1 for an NPC that was located at the bottom or fall between L:S=1.74 and L:S=1.82 for a NPC at the equator (within this range, an illuminated NPC only has a free angle of 1.4^o to the perpendicular direction to the NE) (Ma et al., 2012; Alber et al., 2007).

To justify the precision obtained by the standard deviation of multiple measurements and the algorithm, both methods were used to determine the location of immobile fluorescent molecules (Alexa-Fluor-647-labeled GFP absorbed on the surface of a cover-slip). A total of 220 immobile labeled GFP molecules was measured. The two methods yielded a difference of 0.5±0.1 nm (mean±s.e.m.).

In detail, a program code in Wolfram Mathmatica 10.0. was utilized to calculate the above parameters, which will be available from the corresponding author upon request.

Experimental measurements are reported as s.e.m. unless otherwise noted.

We thank the support from multiple grants as shown under Funding.