Stoichiometry and compositional plasticity of the yeast nuclear pore complex revealed by quantitative fluorescence microscopy

NuRIM: A Quantitative Live-Cell Imaging Assay to Determine the Relative Stoichiometry of GFP-Tagged Nups Within Intact NPCs.

Relative abundances (i.e., stoichiometries) of Nups were comprehensively evaluated both in vertebrate cells and in budding yeast using bulk biochemical methods such as mass spectrometry (18) or quantitative SDS/PAGE (4). These methods rely on biochemical purifications where it is difficult to ensure and assess the integrity of the NPC preparations. By contrast, fluorescence-imaging approaches have the potential to determine Nup abundances within intact NPCs in living cells. If the tags do not interfere with NPC incorporation, comparisons of average intensities over many NPCs within many NEs expressing different GFP-tagged Nups should provide very accurate relative abundance information. For example, a Nup present at twice the copy number of another should yield an average intensity value at the NE differing by a factor of 2 (SI Material and Methods).

We, therefore, set out to develop a live-cell fluorescence imaging workflow to measure fluorescence signal intensities of endogenously labeled Nups specifically at the NE. To implement it, we created a library of strains endogenously expressing C-terminally tagged Nups with yeast-optimized enhanced green fluorescent protein (yEGFP) preceded by a short flexible linker. To ensure that labeling does not interfere with Nup function and NPC incorporation, the logarithmic growth rates of all labeled stains were compared with wild-type cells (Fig. S1A) and the Nup-yEGFP localization together with the cellular morphology of the tagged strains were carefully examined. For the Nup Nic96, we also included an additional N-terminal GFP-tagged variant since the C-terminally tagged strain displayed a fitness defect manifesting in a delayed recovery from stationary phase (Fig. S3A). In total, we tagged and analyzed 26 different Nups.

To accurately quantify Nup-yEGFP intensities at the NE in an unbiased manner, we developed an automated image-processing pipeline (NuRIM). The NE was marked using the endoplasmic reticulum marker dsRed-HDEL. Yeast cells coexpressing various yEGFP-tagged Nups together with dsRed-HDEL were grown in parallel under the same conditions, and imaged side-by-side in an automated manner using identical microscope settings, both in bright field and for the dsRed and GFP channels (Fig. 1A). Bright-field images were used to avoid highly crowded areas (Fig. 1A, i and ii), while the dsRed channel was used to produce binary masks outlining the NE contours by an algorithm based on Laplacian edge detection (29) (Fig. 1A, iii and iv, and SI Material and Methods). These masks closely overlapped spatially with the Nup-yEGFP distribution, and their medial line (morphological skeleton) was used to compute corresponding background-corrected pixel intensities from the yEGFP channel (30) (Fig. 1A, v and vi). These values were averaged for each cell, image and yeast strain over thousands of nuclei to produce a final population average intensity value. SI Material and Methods provide a more systematic description of the rationale and the assumptions that underlie NuRIM.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Analysis of the relative NPC stoichiometry. (A) Outline of the nuclear rim intensity measurement (NuRIM) technique. (A, i–vi) Cells coexpressing dsRed-HDEL and various GFP-tagged Nups were imaged sequentially in bright-field, dsRed, and GFP channels. (A, i) Bright-field channel shows localization of immobilized yeast cells. (A, ii) Exclusion of the overcrowded regions (light areas) based on the bright-field images. (A, iii) dsRed-HDEL channel shows localization of dsRed-HDEL marking the NEs. (A, iv) Production of binary NE contour masks based on the dsRed images. (A, v) GFP channel showing localization of endogenously GFP-tagged Nups. (A, vi) Background-subtracted Nup-GFP intensities are measured and averaged across thousands of binary NE contours. (Scale bars: 2 μm.) (B) Linearity of GFP-based intensity measurements. Cells expressing indicated yEGFP-tagged Nups alone or in pairwise combinations were imaged in parallel. Corresponding NE signal intensities were quantified using NuRIM. Gray bars represent corresponding arithmetic sums of the NE signal intensities of singly labeled strains. Mean ± SD. n, number of image frames analyzed. (C) Robustness of GFP-based intensity measurements against microenvironmental variability. Intensity values obtained using NuRIM for the indicated Nups with C-terminal yEGFP tag separated by 6- or 102-aa linker sequence. Intensity values were normalized to Nup84-yEGFP containing a short linker. Mean ± SD. n, number of image frames analyzed. (D) Schematic representation of the NPC. Nup subgroups are depicted as colored cartoons. (E) Relative abundance of yeast Nups. NE intensities for the indicated yEGFP-tagged Nups were quantified using NuRIM and normalized for the average value obtained for the six components of the Y complex. *Relative abundance for GFP-Nic96 was obtained from independent measurements normalized to Nup120-GFP. Mean ± SD. n, number of independent experiments performed on different days.

An important assumption of our approach is that each individual Nup-yEGFP contributes linearly to the total intensity emitted from the NPC. To test this, we quantified the NPC intensities in a set of strains expressing individually labeled Nups, including Gle1, Nup42, Nup120, and Nup133, as well as the NPC-associated nuclear transport factor Mex67, and compared their intensities to the values obtained in double-labeled stains that coexpressed various pairs of these proteins. Within less than 5%, the fluorescence intensities of all tested double-labeled strains matched the values calculated by summing up the intensities of the two corresponding strains expressing a single Nup-yEGFP. This shows that our intensity measurements increase linearly in relation to copy number and can be used to measure relative abundance (Fig. 1B).

Another assumption is that neither crowding nor a different “microenvironment” in the NPC significantly influences fluorescence intensities and affects the accuracy of our measurements. Extreme proximity of EGFPs in oligomers can lead to quenching of fluorescence emission by homo-FRET (31, 32), and tight binding of anti-GFP antibodies to GFP was shown to modify fluorescence intensities (33). While the structural symmetries present in the NPC are expected to prevent such immediate contacts of individual GFP probes, we nevertheless set out to obtain evidence that the microenvironment does not significantly affect our intensity measurements. To this end, we introduced a 102-aa-long, flexible linker upstream of yEGFP in some of the strains. Such a long linker should release the yEGFP probes from their microenvironment—thus leading to a potential alteration of the fluorescence intensities. However, when we analyzed the corresponding signal intensities, no significant changes were observed compared with the results obtained with regular linkers (Fig. 1C).

Furthermore, for the correct interpretation of the NuRIM results, it is important that the spatial probability distribution (denoted p in SI Material and Methods) of the various analyzed Nups is identical. However, some Nups might be present in only a subset of NPCs or in additional, non-NPC pools at the NE. To recognize such unusual behavior, we compared the distribution patterns of all Nups in our library using normalized cumulative intensity distributions and the autocorrelation of intensities along the NE. For the majority of Nups, we observed highly similar signatures (Fig. S1 B and C). The three outliers, Ndc1, Mlp1, and Mlp2, are consistent with the additional presence of Ndc1 at the spindle pole body (34), and with the previously described absence of both Mlp1 and Mlp2 from NPCs in the nucleolar region (35) (Fig. S1F).

Although we were careful to perform all imaging experiments under very similar conditions, we also tested the robustness of NuRIM against inherent variability in culture density and the density of cells plated on the microscope slide (Fig. S1 D and E). To this end, we varied either of these parameters for one of the strains (Nup84-yEGFP) in the range ∼0.5× to 2× around the standard values used in the experiments and analyzed the impact on the NuRIM output. Within this range, we did not observe any significant influence on the readout (Fig. S1 D and E). Therefore, our imaging and quantification strategy is resilient against small unavoidable variations in cell culture and imaging conditions.

In summary, we conclude that NuRIM allows for an accurate and unbiased determination of the relative stoichiometry of endogenously labeled Nups within intact NPCs of living cells.

Determination of the Relative Nup Abundance.

Next, we applied NuRIM to systematically quantify abundances for all Nups in our library (Fig. 1 D and E). To robustly compare sets of measurements performed on separate days, the average yEGFP intensity of six Y-complex components was used for normalization. The intensity readouts were highly reproducible and revealed a striking pattern of abundances (Fig. 1E and Table 1). The vast majority of Nups were present in very similar amounts, displaying values within 15% of each other (Fig. 1E and Table 1). This includes all members of the Y complex and the major components of the NPC scaffold. One exception was Nic96, which displayed an abundance of ∼1.5 for the C-terminally tagged version and an abundance of ∼2 for the N-terminal GFP variant (Fig. 1E, Table 1, and Fig. S3B). Since C-terminal tagging of Nic96 led to a fitness defect, it is likely that Nic96 is twice as abundant as other scaffold Nups (Fig. 1E, Table 1, and Fig. S3B). By contrast, Pom152, Nup42, and Gle1 displayed distinctly lower values clustering around 0.5. The three outliers that are not evenly distributed along the NE, Ndc1, Mpl1, and Mlp2 are included for completeness (Table 1) but need to be viewed with prudence. Accordingly, the majority of Nups are present in the same copy number, and in general, Nups can be found in well-defined quanta, that is, ∼0.5, 1, or 2. Such distinct abundance classes are consistent with the eightfold symmetrical architecture of the NPC in which Nups are expected to be present in multiple copies of eight per structure.

Determination of Absolute Copy Numbers Reveals That Most Yeast Nups Are Present in 16 Copies per NPC.

If the absolute NPC copy number was available for even a single Nup, our systematic ratiometric analysis would allow the determination of absolute copy numbers for all other Nups as well. We thus set out to quantify the total fluorescence intensity of a single NPC in cells expressing one of the fluorescently tagged Nups. Of note, endogenous tagging of Nups principally results in 100% labeling efficiency since all copies of the protein carry the fusion tag (but see also Discussion). We took advantage of the observation that single NPCs can occasionally be detected in dividing yeast cells where the part of the NE bridging the mother and daughter nuclei elongates and narrows (Movie S1) (36). These NPCs are more separated from their neighbors, allowing us to image and quantify the intensity of individual NPCs in several thousand cells using the Diatrack particle tracking software (37). As an absolute intensity reference, we also imaged in parallel, yeast-expressed rotavirus-like VP2 particles that contain exactly 120 copies of yEGFP (38). Furthermore, we recorded the intensity of single molecules of yEGFP adsorbed on glass coverslips (Movie S2). The intensity distributions of single yEGFP, Nup84-yEGFP NPCs, and VP2- yEGFP particles were fitted with a normal distribution. Comparison of their means showed that VP2 particles were 116 ± 7 times brighter than single yEGFP molecules, yielding an estimated copy number for Nup84-yEGFP of 13.9 ± 2.8 per NPC (Fig. 2A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Absolute stoichiometry of yeast Nups. (A) Representative intensity histograms and Gaussian fits (smooth lines) obtained for single yEGFP molecules, single NPCs in regions between separating nuclei in Nup84-yEGFP–expressing yeast cells (arrowhead), and for the yeast-expressed 120-mer VP2-yEGFP particles (arrowhead). (Scale bars: 2 μm.) (B) Intensity histograms and fits are displayed similar to A for single NPCs in Nup84-sfGFP–expressing yeast cells, and for purified SF12 and SF24 particles containing 12 and 24 sfGFP molecules, respectively. (C, i) Representative fluorescence microscopy image showing localization of Nup84-yEGFP–labeled NPCs. (C, ii) Computational model of the NE. NPCs are represented as green dots, dsRed-HDEL as red dots, and dim sources simulating background fluorescence as blue dots. (C, iii) Simulated image of the NE produced computationally from the model ii. Note good resemblance to genuine Nup84-yEGFP localization (compare with i). Inset compares computed PSF with experimentally acquired PSF. (C, iv) Representative 3D rendering of image stack shows heat map of intensities in Nup84-yEGFP cells illustrating high-intensity spots or “speckles” (red) produced due to stochastic overlap of the ∼120 NPCs. (Scale bar: 200 nm.)

To validate this result, we performed an independent analysis using a different set of reference intensity markers: tetrahedral particles that contain precisely 12 or 24 GFP moieties (39). These particles contain “superfolder” GFP (sfGFP), a GFP variant with fast folding kinetics (40), and we therefore endogenously tagged several Nups including Gle1, Nup84, Nup120, and Nic96 also with sfGFP. As expected, tagging with sfGFP did not lead to any significant changes in the relative abundance readouts compared with yEGFP (Fig. S2A). Using the same methodology as before, we then recorded in parallel the absolute intensity distributions for 12× sfGFP and 24× sfGFP tetrahedral particles purified from Escherichia coli, and for the Nup84-sfGFP and Nup120-sfGFP NPCs in yeast. Based on the comparison of the means of the distributions, we determined absolute NPC copy numbers of 14.7 ± 2.5 for Nup84 and 15.8 ± 2.3 for Nup120 (Fig. 2B and Fig. S2B).

In accordance with these results, we assigned absolute stoichiometries for all Nups measured in Table 1. We conclude that the yeast NPC contains very close to 16 copies for most Nups (Table 1). These include major scaffold Nups constituting the framework of the NPC: the Y-complex members (Nup133, Nup120, Nup85, Nup84, Nup145C, Seh1) and the inner ring Nups (Nup188, Nup192, Nup157, Nup170). For Nic96, our relative abundance measurements led to an assignment of either 24 or 32 copies per NPC, depending on the position of the GFP tag (Fig. 1E and Table 1). This result was also confirmed by single NPC measurements where Nic96-yEGFP was 1.37 ± 0.1 times brighter than Nup120-yEGFP, and the N-terminal GFP-Nic96 variant displayed a 1.93 ± 0.14 higher brightness compared with Nup120-GFP (Fig. S3 C and D). Overall, the copy numbers obtained here for budding yeast are in stark contrast to the human NPC where corresponding proteins were reported to be generally twice as abundant, including 32 copies of the Y complex (16, 21⇓–23).

The astonishing differences in NPC composition between yeast and human prompted us to develop another independent method to examine NPC stoichiometry. The characteristic intensity spots (i.e., speckles) that are observed by fluorescence microscopy in the yeast interphase NE are produced by randomly overlapping, diffraction-limited images of several closely positioned NPCs (Fig. 2C, iv). Precise determination of a characteristic “NPC clustering” factor within such spots would allow us to estimate an absolute stoichiometry by comparing their intensities with established external fluorescence standards. We thus developed a 3D mathematical model of the NE based on NPC density data derived from 3D EM reconstructions (36) and on classical image formation theory (SI Material and Methods). This model allowed us to estimate the characteristic NPC clustering factor predicting that on average 3.7 NPCs are present in such NPC intensity spots (SI Material and Methods). Three-dimensional stacks acquired for Nup84-yEGFP– and VP2-yEGFP–expressing cells imaged under identical conditions were then used as inputs to quantify absolute signal intensities. By comparing the intensity distributions of the NPCs spots and the VP2 particles, we computed a copy number of 13.8 ± 4 for Nup84 per NPC. This result is fully consistent with and confirms our measurements based on single NPC intensities (Fig. 2C).

Altogether, our results demonstrate that on average no more than 16 copies of Nup84 are present in the yeast NPC. Based on our relative abundance measurements, most Nups within the NPC of budding yeast, therefore, exist in 16 copies. This finding is consistent with molecular mass estimates for the yeast NPC, which is significantly smaller compared with its vertebrate counterpart (24, 25). Furthermore, this suggests that there are considerable differences in the architectural organization of the yeast and vertebrate NPCs. From an evolutionary perspective, this implies that a macromolecular machine with conserved function and constructed from conserved subunits can adopt fundamentally distinct structural layouts in different species.

Altering Nup Expression Leads to NPCs with Different Stoichiometries.

Our automated imaging pipeline can also be applied to mutant yeast strains in which the composition of the NPC is altered. This paves the way for investigating a wide range of questions regarding the NPC architecture that would have otherwise been difficult to address. For example, in recently proposed models of the human NPC (22, 23), Nup155 is an abundant Nup present in 48 copies playing a critical role in linking the Y complex to the NPC core inner ring and nuclear membrane. However, despite this central structural role, in yeast, its function is split between two nonessential and functionally redundant paralogous genes NUP170 and NUP157 that have separated due to whole-genome duplication ∼100 Mya (41⇓–43). Of note, each of these Nups is present in only ∼16 copies in the yeast NPC (Table 1). To study this further, we utilized NuRIM to investigate whether Nup157 can replace Nup170 when the NUP170 gene is deleted. We hypothesized that if substitution was to occur, we should see an increase in the intensity due to additional copies of Nup157 now at the pore. However, in NUP157-yEGFP cells, the deletion of NUP170 did not cause any increase in Nup157-yEGFP intensity (Fig. 3 A and B), indicating that sites that are normally occupied by Nup170 remain empty in its absence. To rule out the possibility that protein expression is a limiting factor for substitution, we constructed a NUP157-yEGFP nup170Δ strain that contained an extra copy of NUP157-yEGFP inserted into the URA3 locus, thus doubling the production of this protein. This, however, also did not cause any significant increase in its abundance at the NPC. In contrast and as expected, the reexpression of NUP170 by insertion of a NUP170-yEGFP gene into the URA3 locus produced twice as bright NPCs (Fig. 3 A and B). We conclude that Nup157-yEGFP is not able to substitute for the loss of Nup170, indicating that yeast NPCs with only ∼16 copies of Nup157 are fully functional.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Analysis of Nup stoichiometry in yeast mutants. (A, C, and E) Representative images illustrating Nup-yEGFP localization in actively growing yeast cells with specified genotypes. The corresponding NE intensities quantified using NuRIM are shown in B, D, and F, respectively. Mean ± SD. n, number of image frames analyzed for B, D, and F. (Scale bars: 2 μm.) (G) Representative intensity histograms and Gaussian fits obtained for single NPCs in NUP170-yEGFP and NUP170-yEGFP nup170Δ URA3::NUP170-yEGFP strains.

Gene duplications supply the raw material for the emergence of new functions through the forces of mutation and natural selection (44). We therefore also wanted to test the inverse scenario, that is, whether Nup170 could substitute for the loss of Nup157 at the NPC. As in the mirror strain, we did not observe any increase in abundance of Nup170-yEGFP upon deletion of NUP157 (Fig. 3 C and D). However, in this case, the introduction of an additional copy of NUP170-yEGFP led to a twofold increase in Nup170 abundance at the NPC as measured by NuRIM (Fig. 3 C and D). This intensity increase was further confirmed by brightness measurements of individual NPCs, and nup157Δ cells expressing two copies of Nup170-yEGFP exhibited a 2.07 ± 0.26-fold higher NPC intensity compared with cells expressing a single Nup170-yEGFP copy (Fig. 3G). This reveals an interesting functional asymmetry between these two paralogs and shows that gene expression levels can affect the NPC stoichiometry producing an NPC with ∼32 copies of Nup170.

Motivated by these findings, we also investigated another paralog pair, comprised of Nup188 and Nup192. These two paralogs presumably originated from an ancient gene duplication event—about 800 Mya (45). Since the gene encoding for Nup192 is essential, we restricted our analysis to the deletion of NUP188. Similar to the results with Nup170 and Nup157, no intensity increase was observed in nup188Δ cells in the presence of endogeneous levels of Nup192-yEGFP (Fig. 3 E and F). However, once an extra copy of NUP192-yEGFP was introduced, an increase in NPC abundance of more than 50% was detected (Fig. 3 E and F). This suggests that Nup192 is capable of occupying the empty Nup188 sites, but it is unable to do so entirely. Taken together, these results identify intriguing redundancies between Nup paralogs and reveal a remarkable compositional plasticity of the NPC.