Exploiting the endogenous yeast nuclear proteome to identify short linear motifs in vivo

Summary

Peptide-domain interactions mediated by short linear motifs (SLiMs) play crucial roles in cellular biology. The simplicity of SLiMs poses challenges in their computational identification. Existing high-throughput methods for discovering SLiMs lack cellular context as they are typically performed in vitro. We developed a functional selection method using yeast to identify peptides that interact with the endogenous yeast nuclear proteome. Remarkably, peptides selected for in yeast also mediated nuclear import in human cells. Notably, the identified peptides did not resemble classical nuclear localization sequences. This platform has the potential to identify and investigate motifs that interact with the nuclear proteome of yeast and human and to aid in the identification and understanding of alternative protein nuclear import mechanisms.

Graphical abstract

Highlights

• •We develop pNIA2, a method to screen libraries for peptides mediating nuclear import
• •pNIA2 can be used to study non-classical nuclear import, such as piggybacking
• •Peptides identified with pNIA2 direct nuclear import in a mammalian system
• •Highly similar peptide sequences in the human proteome are identified

Motivation

Methodologies for identifying short linear motifs have several limitations. Typically, these approaches are limited by scale, making testing of bait and prey libraries in a “many-to-many” fashion impractical. These limitations are further compounded by their in vitro nature and therefore lack functional context. To address these limitations, we have exploited the endogenously expressed yeast nuclear proteome as prey to screen libraries of genetically encoded peptides in an in vivo setting to identify peptide-mediated interactions. This approach capitalizes on the observation that many nuclear proteins lack an identifiable nuclear localization sequence (NLS) and are co-transported into the nucleus via interaction with proteins containing a true NLS.

Tessier et al. develop a yeast-based system (pNIA2) to functionally screen peptide libraries for their ability to mediate nuclear import via protein interactions. Screening a random peptide library with pNIA2 identifies motifs that function in a mammalian system and don’t resemble the well-characterized classical nuclear localization signal.

Introduction

Peptide-domain interactions mediated by short linear motifs (SLiMs) are the most frequently observed protein-protein interaction (PPI) module within intrinsically disordered protein regions (IDPRs) and are integral to a diverse range of cellular processes.¹ Their simplistic nature has made them an important contributor to the dynamic rewiring and evolution of PPI networks.² Moreover, their fundamental role has made this form of PPI an “Achilles heel,” exposing protein interaction networks to hijacking by pathogens via motif mimicry.^3,4,5

Importantly, SLiMs can make highly specific interactions using 10 amino acids or less.⁶ Typically, only three or four of these residues substantially contribute to binding specificity and affinity.⁷ These attributes make bioinformatic detection of SLiMs problematic.⁸ Integrating additional features, such as evolutionary conservation and/or common protein interactions, has aided in their discovery.⁹

It’s estimated that up to 100,000 SLiMs and thousands of SLiM classes exist within the human proteome.^10,11 The eukaryotic linear motif database, which catalogs experimentally verified SLiMs, currently contains approximately 4,000 literature-curated motifs and only a few hundred SLiM classes.¹² The large discrepancy between the number of identified and predicted SLiMs highlights the need for additional approaches to aid in their discovery.

Multiple approaches for experimentally identifying SLiMs have been used to varying degrees of success.^13,14,15 These include protein and peptide arrays, yeast-two-hybrid (Y2H) screens, and yeast or phage display approaches.¹⁶ Though certainly high throughput, these approaches are often limited by their scale of implementation. The necessity to express or purify proteins limits testing bait and prey to a one- or few-vs.-many level of throughput. More importantly, their in vitro nature means PPIs are identified in the absence of cellular context.

To overcome these limitations, we designed a “many-vs.-many” approach to identify SLiMs. We exploit protein nuclear import via non-classical pathways or “piggybacking,” which indirectly allows proteins lacking nuclear localization sequences (NLSs) to traffic to the nucleus by interacting with endogenous nuclear-targeted proteins.^17,18,19 This allowed us to screen a large number of peptides against endogenous nuclear proteins within a cellular context. This approach was not easily achievable in a mammalian system but was readily implemented in the yeast S. cerevisiae. While systems have been developed for screening peptides in a mammalian system,^20,21 the genetic tractability, lack of specialized equipment, ease, and flexibility of constructing diverse peptide libraries make yeast ideal for this approach. Despite the evolutionary distance between yeast and mammals, protein nuclear import is highly conserved.^22,23 Moreover, most protein domains identified in yeast are found within the human proteome, allowing a domain-centric approach.²⁴

Since this system exploits protein nuclear import, we designed this system to limit detection of SLiMs resembling classical NLSs (cNLSs), which contain multiple basic residues recognized by importin-α (Imp-α).^25,26 Upon binding, the cNLS:Imp-α complex is subsequently trafficked to the nucleus via Imp-β1.²⁷ This pathway is assumed to handle the majority of protein nuclear import; however, many Imp-α interactors in yeast, mice, and humans have no discernable cNLS.^18,25,28 These observations raise fundamental questions regarding alternative non-classical mechanisms of protein nuclear import, such as piggybacking, and the extent to which they exist.^18,28,29

Using a yeast-based approach to exploit protein nuclear import, we performed a small-scale proof-of-principle experiment and identified a number of peptides that share no resemblance to cNLSs. Intriguingly, some of these motifs interacted with human Imp-α and mediated nuclear localization in mammalian cells. Overall, this approach is economical and allows for rapid identification of putative SLiMs in yeast that also function in an orthologous human system.

Results

Construction and validation of a nuclear import selection system in yeast

Previously, our group published a modified yeast-based system that uses a recombinant protein, pNIA, to test peptides or proteins for their ability to direct nuclear import of a chimeric transcription factor containing the LexA DNA binding domain (LexA_DBD).^30,31 This protein is optimized to work in the L40 yeast strain, which harbors a genomic LexA-responsive β-galactosidase reporter and HIS3 selectable marker. This system provides a simple and easily quantitated measure of NLS function using β-galactosidase assays.

Here, we adapted this system to screen a randomly generated library of peptides. Using the highly efficient homologous recombination process in yeast, we can perform a genetic selection in the L40 yeast strain. This new system, which we have named pNIA2, uses a similar modular protein design to the original platform (Figure 1A). To test peptides using pNIA2, we first identified a suitable location for their expression. Using the disorder prediction program IUPred,³² we identified a configuration that creates an area of predicted disorder located within the linker separating the LexA_DBD and Gal4 activation domain (Gal4_AD). Additionally, for peptides containing a stop codon, this arrangement limits false-positive identifications, as only the LexA_DBD will be translated.

Figure 1.

Schematic of pNIA2 organization, optimization, and validation

(A) Organization of pNIA2 and accompanying IUPred protein disorder prediction. Gray indicates linker region separating LexA and Gal4_AD where peptides are integrated.

(B) Optimization of pNIA2 activity in high- (2μ) and low- (CEN) copy yeast expression vectors and the addition of a maltose-binding protein (MBP) stuffer.

(C) Sensitivity of pNIA2 for measuring nuclear import of NLSs tested in suboptimal positions. C-terminal and internal positions were tested individually for statistical significance.

(D and E) Validation of pNIA2 testing well-recognized NLS SLiMs (D) and non-NLS SLiMs (E) derived from the eukaryotic linear motif database (ELM: AIR2, ELME000385; PAP2, ELME000387). Nuclear import activity was determined using β-galactosidase activity using at least three replicates and represented as mean ± SD (B)–(E).

To optimize the sensitivity of pNIA2 for detecting nuclear import, we tested high-copy (2μ) and low-copy (CEN; centromere sequence) yeast expression vectors and the presence or absence of a maltose-binding protein (MBP) “stuffer” (Figure 1B). MBP is used to increase the overall size of pNIA2 (∼80 kDa) above the passive diffusion limit of the nuclear pore complex.³³ When expressed from a high-copy 2μ plasmid, pNIA2 gave substantial background activity and provided marginal sensitivity upon the addition of MBP and the simian virus 40 (SV40) large T antigen (TAg) cNLS.³⁴ When expressed from a low-copy CEN plasmid, MBP reduced background to virtually zero, presumably due to pNIA2 restriction to the cytoplasm. Upon the addition of the TAg cNLS, activity increased ∼150-fold. Together, these features contribute to a wide dynamic range, allowing detection of putative SLiMs with variable nuclear targeting potential.

While SLiMs are generally regarded as modular, some motifs are restricted to particular protein contexts.³⁵ To determine if the disordered linker region can identify NLSs in a suboptimal context, we measured nuclear import using NLSs that originate from either a terminal or internal position (Figure 1C). As expected, a C-terminally derived cNLS from the adenovirus E1A protein³⁶ functioned optimally on the C terminus of pNIA2. Likewise, the internally derived SV40 TAg cNLS functioned best when expressed internally within the linker region. Only in the internal position did both NLSs function, indicating this location is sensitive enough for detecting motifs in a suboptimal context.

In determining the specificity of pNIA2, we tested a variety of NLSs that included the Xenopus nucleoplasmin bipartite cNLS (NP)³⁷ and human c-Myc monopartite NLS³⁸ and the human immunodeficiency virus 1 (HIV-1) Tat non-classical NLS.³⁹ When expressed on pNIA2, each NLS demonstrated robust nuclear import activity (Figure 1D). Conversely, when testing non-NLS SLiMs, nuclear import was not observed (Figure 1E). As protein shape can influence nuclear transport,⁴⁰ we determined if this had an effect on pNIA2 using a flexible SGSG linker. This had no observable influence on nuclear import activity (Figure 1E). Thus, pNIA2 can distinguish between bona fide NLSs and non-NLSs with high specificity and sensitivity.

pNIA2 selects for nuclear import and can measure piggybacking

Next, we determined if our system selectively favors growth of yeast expressing pNIA2-NLS fusions and whether or not there is enough sensitivity to directly detect piggybacking mechanisms. Serially passaging L40 yeast co-transformed with pNIA2 and pNIA2-TAg cNLS demonstrated a selective growth advantage over pNIA2 alone. This was evident based on increased bulk nuclear import activity following each passage under selective growth conditions (Figure 2A).

Figure 2.

Outline of peptide selection strategy and identification of putative motifs

(A) Serial passage of L40 yeast transformed with either pNIA2, pNIA2 and pNIA2-TAg NLS, or pNIA2-TAg NLS. Mean ± SD, n = 2.

(B) Sensitivity of pNIA2 to measure nuclear import through piggybacking. pNIA2 harboring a TAg NLS or MYND domain-binding SLiM (PxLxP) derived from the adenovirus C5 E1A protein was co-expressed with a nuclear-targeted MYND domain, and nuclear import activity was measured by β-galactosidase assay. Mean ± SD, n = 3.

(C) Workflow of the pNIA2 genetic selection in L40 yeast. (I) Comparison of a completely degenerate NNN codon to an NHS codon. (II) Oligonucleotides composed of 10xNHS codons and encoded SGSG flanking regions are recombined into pNIA2 backbone through homologous recombination. Recombined pNIA2 plasmid is selected using media lacking leucine (non-selective conditions). (III) Nuclear localization of pNIA2 results in expression of genomically integrated HIS3 and lacZ. Only yeast with nuclear-localized pNIA2 survive selection on media lacking leucine and histidine (selective growth condition). (IV) Yeast are assayed for nuclear import activity by β-galactosidase activity before pNIA2 plasmids are isolated and sequenced. Plasmids are retransformed to determine nuclear import activity under non-selective conditions.

(D) Nuclear import activity comparing primary (dark green, n = 1) and secondary (light green, n = 3) screens. Primary and secondary screens were performed under selective and non-selective growth conditions, respectively. Motifs with statistically significant activity are indicated in bold. NC, secondary screen not completed.

(E) Squelching assay testing peptides identified on pNIA2 using the W3031-A yeast strain. Nuclear Gal4_DBD competes with endogenous Gal4, reducing expression from a Gal4 reporter.

(F) Nuclear import activity of peptides using the squelching assay. Green bars indicate peptides that function on both pNIA2 and Gal4_DBD. Mean ± SD, n = 3.

See also Figure S2.

To evaluate the sensitivity of pNIA2 to detect piggybacking mechanisms, we co-expressed a pNIA2 fusion harboring the human BS69 MYND domain binding peptide, derived from the adenovirus E1A protein, and a nuclear-targeted BS69-MYND domain (residues 427–602).^41,42 Expression of a nuclear-targeted MYND domain increases nuclear import activity nearly 3-fold, demonstrating pNIA2 is sensitive enough to detect motif-mediated nuclear import via piggybacking (Figure 2B).

Identification of putative non-classical SLiMs

Having established that pNIA2 can measure peptide-mediated nuclear import, we screened a randomly generated pool of genetically expressed peptides. A randomly synthesized library of DNA oligonucleotides was transformed with linearized pNIA2 into L40 yeast to undergo homologous recombination and subsequent selection based on nuclear import (Figure 2C).

Briefly, only yeast containing pNIA2 repaired by homologous recombination with a peptide library oligonucleotide will survive selection on media lacking leucine. By design, pNIA2 is restricted to the cytoplasm. Therefore, only yeast harboring a pNIA2 recombinant that directs its nuclear import will be able to activate transcription of the genomic LexA-responsive HIS3 reporter.

A defining feature of our experimental design was to limit the identification of well-described cNLSs. We utilized oligonucleotides synthesized with 10 contiguous NHS codons (where N = A/C/G/T, H = A/C/T, and S = G/C) and flanking SGSG linkers to promote peptide accessibility. Ten codons were chosen to reflect the typical SLiM core binding region size. Although the NHS codon prevents incorporation of certain amino acids, it reduces the occurrence of stop codons, preventing unwanted truncations. More importantly, the occurrence of basic amino acids is nearly eliminated, driving selection for motifs that do not resemble a cNLS.

The theoretical library size using randomly generated 10-mer peptides is enormous. To estimate how many peptides could be screened per experiment, we performed a set of transformations along with serial dilutions to calculate colony forming units (CFUs). Here, CFUs represent a reliable measure of diversity, as is routinely done with other display technologies like phage display.⁴³ Transformation of 2 × 10⁶ L40 yeast, yielded roughly 65,000 CFUs/μg of DNA, corresponding to a transformation efficiency between 3% and 4% (Figure S1A). This is on par with what has been reported for construction of antibody libraries in yeast.⁴⁴ Furthermore, library diversity increased proportionally with the amount of DNA transformed (Figure S1B).

Using our system outlined in Figure 2C, we performed a small-scale, proof-of-principle screen where we transformed approximately 2 × 10⁷ L40 yeast. Based on the transformation efficiency, this represented a library size of 600,000–800,000 peptides. We selected 34 transformants (named nuclear motif [NM]-1 through -34) for further analysis (Figures 2D, S2A, and S2B).

Many of the peptide sequences identified contained no Arg or Lys codons, and those few that did shared no resemblance to a cNLS, validating our NHS codon design. Secondary evaluation of nuclear import activity under non-selective conditions (Leu⁻His⁺) showed a consistent reduction in activity, likely due to the selective growth conditions (Leu⁻His⁻ media) of the initial screen. This step demonstrated many of the peptides in the initial selection were false positives. This secondary screen allowed us to rule these out and identify motifs with statistically significant nuclear import activity.

A hallmark feature of many SLiMs is their modular nature, allowing them to retain their function in different protein contexts.⁸ To evaluate this with peptides identified using pNIA2, we measured nuclear import using an orthologous yeast-based system. Using the galactose-inducible W303-1A yeast strain, we measured nuclear import via the ability of a Gal4_DBD-peptide fusion to quench endogenous Gal4-mediated reporter gene expression (Figures 2E, 2F, and S2C). Several motifs that showed statistically significant activity when expressed on pNIA2 also functioned within this orthologous system, indicating they may function in a modular fashion. Unexpectedly, several motifs that did not show statistically significant activity on pNIA2 appeared to function when fused to the Gal4_DBD. These false negatives are possibly due to pNIA2’s sensitivity for detecting NLS activity in a suboptimal context, suggesting that some peptides may not pass a statistical threshold when expressed within pNIA2 but may demonstrate greater activity as a C-terminal fusion. Indeed, the inverse also appeared to be true, as observed with NM-34, and these may represent motifs where their internal location is critical for function.

Peptides identified with pNIA2 localize to the nucleus and interact with nuclear import machinery in human cells

Given the high degree of conservation between yeast and mammalian proteomes, we next determined if these peptides function in mammalian cell lines. For this, we chose two motifs, NM-9 and -34. First, we confirmed their nuclear import activity in yeast was not attributed to some unique ability to transactivate (Figure S3A). When expressed as EGFP fusions, both NM-9 and -34 relocalized EGFP to the nucleus of transfected HeLa and HT-1080 cells (Figures 3A and S3B). Nuclear localization of EGFP was more pronounced in HeLa cells and also sensitive to ivermectin, an inhibitor of Imp-α-mediated nuclear import.⁴⁵

Figure 3.

Motifs identified using pNIA2 confer nuclear localization and interact with human Imp-α

(A) Confocal microscopy of transfected HeLa cells expressing EGFP fusions of NM-9 and -34 with or without ivermectin (IVM). Scale bars represent 50 μm.

(B and C) Co-immunoprecipitation experiments in HT-1080 cells transfected with FLAG-tagged Imp-α1 isoforms and EGFP-motif fusions.

(D) Top BLAST results for NM-3 and NM-28 against the human proteome and their alignment. TET3- and PER1-aligned regions correspond to residues 456–462 and 608–616, respectively. Highly similar amino acids indicated with “+”.

(E) Co-immunoprecipitation experiments in HT-1080 cells transfected with EGFP fusions of negative control motifs and Imp-α1 or -α5. Rev-NES is a nuclear export signal from the HIV-1 Rev protein. NM-4 is a false-positive hit from the initial pNIA2 screen.

See also Figures S3 and S4 and Data S1, S2, S3, and S4.

To further investigate the weaker nuclear localization observed in HT-1080 cells, we performed biochemical fractionation of HT-1080 nuclei and quantitative western blotting. Comparison of nuclear lysates demonstrated a statistically significant increase in nuclear EGFP mediated by NM-9 but not NM-34 (Figures S3C, S4A, and S4B). While not quantitative, western blotting using increasing amounts of nuclear lysate from HT-1080 cells expressing EGFP-NM34 demonstrated a substantial increase in nuclear localization compared to EGFP (Figures S4C and S4D).

Despite NM-9 and -34 bearing no resemblance to a classical NLS, their sensitivity to ivermectin suggested involvement of the classical nuclear import pathway. Therefore, we investigated potential protein interactions with isoforms α1, α3, and α5 as representative human members of the three Imp-α superfamilies.⁴⁶ Co-immunoprecipitation experiments between EGFP-NM9 or -NM34 and FLAG-Imp-α confirmed interactions with the classical nuclear import receptors (Figure 3B). NM-9 appeared specific to Imp-α5, while NM-34 displayed dual specificity for Imp-α1 and -α5. Neither motif bound Imp-α3 (Figure S4E). We tested three additional motifs that were selected for on pNIA2 and detected an interaction with Imp-α1 for two of those motifs, NM-3 and NM-28, which showed sequence similarity with human TET3 and PER1, respectively (Figures 3C and 3D; α3 and α5 were not tested; see Data S1 and S2). Importantly, two negative controls, the HIV-1 Rev nuclear export sequence (NES)⁴⁷ and NM-4, which had no observed nuclear import activity in yeast (Figures 2D and 2F), failed to interact with Imp-α1, -α3, and -α5, demonstrating the specificity of pNIA2 for bona fide interactors (Figure 3E).

An intrinsically disordered region within human DDX10 interacts with importin alpha

Motifs identified in the pNIA2 screen were randomly synthesized and have no direct link to cellular proteins. To determine if either NM-9 or -34 share any sequence homology to human proteins, we performed BLAST⁴⁸ searches against the human proteome. While NM-34 shared no homology to any human proteins, NM-9 showed striking resemblance to a region within human DDX10, a nuclear DEAD box helicase (Figure 4A, see supplemental Data S3 and S4). Since SLiMs are almost exclusively found within IDPRs, we used MobiDB to confirm this region is disordered (Figure 4B).⁴⁹ Moreover, MobiDB predicted this region to contain a linear interacting peptide. As an additional measure of protein disorder, we consulted Phosphosite Plus to look for evidence of post-translational modifications (PTMs), as these are most often found within IDPRs.^50,51 The presence of nearby ubiquitylation (K522) and phosphorylation (S539) sites within the flanking region of the core motif provided further evidence this region is likely to represent an IDPR. Upon testing the entire disordered region (DDX10-IDR; 516 to 545) using pNIA2, we observed nearly identical nuclear import activity as NM-9 (Figure 4C).

Figure 4.

An intrinsically disordered region within human DDX10 interacts with Imp-α

(A and B) The indicated region within human DDX10 (527–536, IDRcore) shares sequence similarity with NM-9 (A) and falls within a larger region of predicted disorder (516–545, IDR) (B). Intrinsic disorder prediction was determined using MobiDB.

(C) DDX10-IDR nuclear import activity in L40 yeast was tested using pNIA2. Mean ± SD, n = 3.

(D and E) Co-immunoprecipitation experiments in HT-1080 cells transfected with FLAG-tagged Imp-α isoforms and EGFP fusions of DDX10-IDR and -IDRcore (D) or -IDR with the core sequence scrambled (SCRcore) (E).

(F) Transfected HT-1080 cells expressing EGFP fusions of DDX10-IDR and -IDRcore were evaluated for co-immunoprecipitation with endogenous Imp-β1 (kpnb1).

See also Figure S4 and Data S3.

Similar to NM-9 and -34, we determined if DDX10-IDR can interact with human Imp-α and whether or not the core region (DDX10-IDRcore) was necessary and sufficient for that interaction. Co-immunoprecipitation experiments using human HT-1080 cells expressing DDX10-IDR and -IDRcore as EGFP fusions, along with FLAG-Imp-α1, -α3, or -α5, demonstrated that both regions of DDX10 interact with Imp-α1 and -α5 but not -α3 (Figures 4D and S4F). Unlike NM-9, DDX10-IDR was able to interact with Imp-α1. These differences may be due to the minor sequence variations between them. While DDX10-IDR and -IDRcore show a similar interaction with Imp-α5, the Imp-α1 interaction with DDX10-IDRcore appears stronger compared to DDX10-IDR.

To determine if DDX10-IDRcore is necessary for binding Imp-α1 and -α5, we performed a similar set of co-immunoprecipitation experiments where we scrambled the IDRcore sequence (SCRcore) within the context of full-length DDX10-IDR. While this appeared to have a marginal effect on binding Imp-α1, the interaction with Imp-α5 was completely lost (Figure 4E). Other than valine at position 6 of SCRcore, all other amino acids differed from the non-scrambled core sequence. Notably, positions 2 and 10 share similar amino properties and may contribute to the weak binding observed with SCRcore.

Since these interactions were tested in a transfection-based system, we determined if DDX10-IDR and -IDRcore can interact with the endogenous classical nuclear import pathway in the absence of exogenously transfected Imp-α. Rather than testing interactions for each endogenous Imp-α isoform, we instead tested for binding to endogenous Imp-β1, which recognizes all Imp-α isoforms. Co-immunoprecipitations from transfected HT-1080 cells expressing DDX10-IDR and -IDRcore as EGFP fusions demonstrated that both regions interacted with endogenous Imp-β1 (Figure 4F). These results provided additional evidence that this region from DDX10 can interact with the mammalian endogenous classical nuclear import pathway, and these observations are not an artifact of overexpression.

Discussion

There is a substantial discrepancy between the number of validated and predicted SLiMs. This is especially apparent with respect to NLSs, where a large number of nuclear proteins lack an identifiable NLS.¹⁸ High-throughput methods for identifying SLiMs have been used successfully; however, they fail to identify SLiMs in a functional context. Additionally, these approaches often only allow testing a large number of baits against a limited number of preys. Here, the pNIA2 system exploits the endogenously expressed yeast nuclear proteome as prey, making this an in vivo SLiM discovery approach that can theoretically test a library of peptides against all nuclear proteins. Other yeast-based approaches, such as Y2H, rely on expression of bait and prey proteins without capitalizing on the endogenous proteome. In fact, competition with the endogenous yeast proteome may actually impede PPIs, which may explain the suspiciously low number of Imp-α interactors identified in the Human Reference Interactome database.⁵²

To date, the cNLS remains the best described NLS. While non-cNLSs have been documented, their sequence attributes remain poorly defined or they require a folded domain for binding.⁵³ Moreover, many non-cNLSs that bind members of the Imp-β family are still rich in basic amino acids like the cNLS.⁵³ Using pNIA2, we could measure nuclear import of known NLSs with remarkably high specificity and sensitivity and screen a randomly synthesized library of genetically encoded peptides to identify motifs bearing no resemblance to cNLSs. These attributes make pNIA2 ideal for studying underappreciated mechanisms of nuclear import, such as piggybacking, or possibly help refine our knowledge of non-cNLSs. Indeed, we demonstrated that pNIA2 can piggyback into the nucleus using a MYND domain-binding peptide. Furthermore, peptides NM-9 and -34 showed a common interaction with human Imp-α5, which belongs to the α3 superfamily and shares the highest similarity to the sole yeast Imp-α, Srp1 (Kap60).⁴⁶ Therefore, it’s possible these peptides were selected for based on an ability to piggyback on Srp1 cargos, hence why they both interact with Imp-α5.

Nuclear localization of peptide fusions with EGFP was less pronounced in HT-1080 cells than HeLa cells. This may be related to Imp-β1 expression level, a rate-limiting step for efficient nuclear import.⁵⁴ While Imp-β1 protein levels have not been directly compared between cell lines, the Protein Atlas reports reduced Imp-β1 normalized transcript levels in HT-1080 vs. HeLa cells.⁵⁵ Alternatively, bulk nuclear import has been associated with nuclear size.⁵⁶ HeLa cells are routinely used for microscopy due to their large nuclei, which may be indicative of enhanced bulk nuclear import compared to HT-1080 cells. Together, these factors may explain the EGFP subcellular localization differences between HT-1080 and HeLa cells. Intriguingly, nuclear import mediated by these peptides was sensitive to ivermectin, which broadly inhibits the classical nuclear import pathway.^57,58 These findings suggest these peptides are utilizing a non-classical import method that involves piggybacking on the classical nuclear import machinery.

Given the number of potential sequences in a library spanning 10 random amino acids, it would be impossible to achieve complete coverage with pNIA2. However, library transformation efficiency was in line with expectations for a yeast-based library. Therefore, complete determination of smaller targeted libraries remains feasible. For example, a tiled array of 10⁶ peptides representing the entire human disordered proteome has been screened using phage display.⁵⁹ This approach could be adapted to pNIA2 and could enable the study of proteomes from additional eukaryotic species.

Both Imp-α and Imp-β family members are highly conserved from yeast to plants and mammals. The number of Imp-β family members does not differ substantially between yeast and higher-level eukaryotes.⁶⁰ While not perfect, these similarities make a yeast-based system, like pNIA2, suitable for evaluating non-classical Imp-β-mediated import. In contrast, Imp-α family members have expanded over evolutionary time, with a single member in yeast and up to seven and nine in human and Arabidopsis, respectively.⁶¹ Despite these differences, extensive analysis of peptides, which bind to Imp-α′s from yeast, plant, and human, generally shows highly similar sequences.²⁶ Overall, the high degree of conservation among Imp-α and -β family members suggests that pNIA2 can be used to study nuclear import across eukaryotic species.

Intriguingly, we identified no mammalian proteins with sequence similarity to NM-34, which could represent a synthetic protein interaction motif. However, NM-9 shared high sequence similarity to a region of predicted disorder within the human DDX10 protein. Scrambling the core region within the context of the entire IDPR abrogated binding to human Imp-α5 but not -α1, indicating the core motif is necessary and sufficient for binding Imp-α5. Additionally, motifs NM-3 and -28 interacted with Imp-α1 and showed substantial sequence similarity to the human proteins TET3 and PER1, both of which have nuclear functions.^62,63 Importantly, both regions are within IDPRs and even overlap with MobiDB predicted linear interacting peptides.⁴⁹

While pNIA2 is biased toward identifying peptides related to nuclear import, many intriguing aspects of protein nuclear import, such as piggybacking, remain understudied. Substantial progress has been made beyond the classical nuclear import pathway; however, many seemingly fundamental questions remain unaddressed. While unbiased approaches, such as proteome-wide Y2H screening, have their advantages, these have failed to recapitulate known cargos within the classical nuclear import pathway. For this reason, targeted approaches that focus on specific pathways may provide deeper insight into peptide- or SLiM-mediated interactions.

Limitations of the study

While nuclear import pathways are generally conserved across eukaryotes, this system is biased toward the nuclear import machinery present in yeast. This approach is suitable for addressing the broader principles of nuclear import; however, questions targeted toward specific Imp-α or -β isoforms in higher eukaryotic species may be problematic. Additionally, not all SLiMs conform to 10 amino acids or less, and flanking regions can contribute to binding. In this study, restricting peptides to 10 amino acids will fail to account for these additional SLiM attributes.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-EGFP	Takara	Cat# 632592; RRID: AB_2336883
Mouse monoclonal anti-FLAG (clone M2)	MilliporeSigma	Cat# F1804; RRID: AB_262044
Mouse monoclonal anti-α-Tubulin (clone DM1A)	MilliporeSigma	Cat# T6199; RRID: AB_477583
Mouse monoclonal anti-karyopherin β1 (clone H7)	Santa Cruz	Cat# sc-137016; RRID: AB_2133993
Rabbit polyclonal anti-G6PD	MilliporeSigma	Cat# A9521; RRID: AB_258454
Rat monoclonal anti-HA (clone 3F10)	Roche	Cat #: ROAHAHA; RRID: AB_2687407
Rabbit polyclonal anti-LexA	MilliporeSigma	Cat# 06–719; RRID: AB_310223
Mouse monoclonal anti-FLAG magnetic beads (clone M2)	MilliporeSigma	Cat# M8823; RRID: AB_2637089
Rabbit polyclonal anti-Lamin A/C	Cell Signaling Technology	Cat# 2032; RRID: AB_2136278
Goat anti-rabbit Alexa Fluor 488	ThermoFisher	Cat# A-11008; RRID: AB_143165
Chemicals, peptides, and recombinant proteins
3-Amino-1,2,4-triazole	MilliporeSigma	A8056
Protease inhibitor cocktail	MilliporeSigma	P8340
Ivermectin	MilliporeSigma	I8898
X-tremeGENE™ HP DNA Transfection Reagent	MilliporeSigma	6366236001
ProLong™ Gold Antifade Mountant with DAPI	ThermoFisher	P36935
Synthetic drop-out media	MilliporeSigma	Y2001
Uracil	MilliporeSigma	U1128
Leucine	MilliporeSigma	L8000
Tryptophan	MilliporeSigma	T0254
Histidine	MilliporeSigma	H6034
Experimental models: Cell lines
HT-1080 cells	ATCC	CCL-121
HeLa cells	ATCC	Cat# CCL-2: RRID: CVCL_0030
Experimental models: Organisms/strains
S. cerevisiae: strain L40	ATCC	MYA-3332
S. cerevisiae: strain W303-1A	ATCC	208352
Oligonucleotides
NHS Ultramer: TGAAGGGCTGGCGGTTGGGGTTAT TCGCAACGGCGACTGGCTGGAATTCTCTGGATCA GGTNHS(x10)TCTGGATCAGGTGAATTCAATTTTAAT CAAAGTGGGAATATTGCTGATAGCTCATTGTC	IDT	N/A
Fw NHS Ultramer: TGAAGGGCTGGCGGTT	IDT	N/A
Rv NHS Ultramer: GACAATGAGCTATCAGCAATATTCCC	IDT	N/A
Recombinant DNA
Plasmid: FLAG-Imp-α1 (pcDNA3 backbone)	This work	N/A
Plasmid: FLAG-Imp-α3 (pcDNA3 backbone)	This work	N/A
Plasmid: FLAG-Imp-α5 (pcDNA3 backbone)	This work	N/A
Plasmid: EGFP	This work	N/A
Plasmid: EGFP-TAg NLS	This work	N/A
Plasmid: EGFP-NM-3	This work	N/A
Plasmid: EGFP-NM-4	This work	N/A
Plasmid: EGFP-NM-9	This work	N/A
Plasmid: EGFP-NM-27	This work	N/A
Plasmid: EGFP-NM-28	This work	N/A
Plasmid: EGFP-NM-34	This work	N/A
Plasmid: EGFP-DDX10-IDR	This work	N/A
Plasmid: EGFP-DDX10-IDRcore	This work	N/A
Plasmid: EGFP-DDX10-SCRcore	This work	N/A
Plasmid: EGFP-REV NES	This work	N/A
Plasmid: LexA-Gal4AD (2μm)	Rhee et al.31	N/A
Plasmid: LexA-Gal4AD (CEN)	This work	N/A
Plasmid: LexA-Gal4AD-MBP (2μm)	This work	N/A
Plasmid: LexA-Gal4AD-MBP (CEN)	This work	N/A
Plasmid: LexA-TAg-Gal4AD (2μm)	This work	N/A
Plasmid: LexA-TAg-Gal4AD (CEN)	This work	N/A
Plasmid: pNIA2 and related vectors	This work	N/A
Plasmid: Gal4DBD vectors (pAS1U backbone)	This work	N/A
Plasmid: MYND domain (pJG4-5 backbone)	This work	N/A
Software and algorithms
IUPred	Erdős et al.32	https://iupred2a.elte.hu/
MobiDB	Piovesan et al.49	N/A
BLAST	Altschul et al.48	BLAST+ 2.11.0: October 19, 2020
R Programming Language (R 4.2.0)	R core team64	https://cran.r-project.org/
GraphPad Prism	GraphPad Software Inc	https://www.graphpad.com/
Mimotopes	N/A	www.mimotopes.com
ImageJ (Fiji Distribution)	Schneider et al.65	https://ImageJ.nih.gov/ij

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and fulfilled by the lead contact Joe Mymryk (jmymryk@uwo.ca).

Materials availability

All reagents generated in this study are available upon request from the lead contact.

Data and code availability

• •All data reported in this paper will be shared by the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

Human cell lines

HT-1080 cells (fibrosarcoma, male) and HeLa cells (adenocarcinoma, female) were grown at 37°C with 5% CO₂ in Dulbecco’s modified Eagle’s medium (DMEM; Multicell Technologies) supplemented with 1% penicillin/streptomycin (Multicell) and 10% fetal bovine serum (FBS; Gibco). Transfection of DNA was done with X-tremeGENE HP (MilliporeSigma) using a 2:1 ratio of transfection reagent to μg of plasmid.

Yeast strains

L40 yeast (MATa leu2 his3 trp1 ade2 GAL4 gal80 LYS2(lexAop)4-HIS3 URA3(lexAop)-lacZ) and W303-1A yeast (MATa leu2-3,112 trp1-1 can1-100 ura3-1 ade2-1 his3-11,15) were grown at 30°C in standard YPD media prior to transformation. Following transformation yeast were grown in selective synthetic drop-out media (MilliporeSigma) supplemented with appropriate amino acids (MilliporeSigma).

Method details

Plasmid cloning

Cloning of motifs into pNIA2 and EGFP-C2 was carried out using self annealed oligos with restriction site compatible sticky ends flanking the desired coding region for a specific peptide/protein sequence. Oligos were individually phosphorylated with polynucleotide kinase and annealed using a thermocycler. For cloning of inserts corresponding to motifs selected with pNIA2 into the pAS1U Gal4_DBD vector, inserts were PCR amplified from isolated pNIA2 plasmid with oligos containing compatible restriction digest sites.

Yeast transformations, plasmid and protein isolation

Yeast were transformed using the standard lithium acetate method⁶⁶ and plated on synthetic dropout media (MilliporeSigma) supplemented with glucose and the appropriate amino acids (MilliporeSigma). Transformants were inoculated into 5 mL of appropriate selection media and grown overnight at 30°C with agitation. Plasmids were isolated from yeast using the “smash and grab” procedure⁶⁷ and protein was isolated for western blotting as described.⁶⁸

β-galactosidase assays

Transformants were inoculated into 5 mL of synthetic drop-out media containing glucose and grown overnight at 30°C. Cultures were diluted down and grown to an OD₆₀₀ of approximately 0.7. One mL of culture was collected via centrifugation and β-galactosidase assays were performed as previously described.⁶⁹ β-galactosidase assays performed in the W303-1A yeast strain were performed identically; however, yeast were grown in overnight cultures supplemented with raffinose instead of glucose. The following morning cultures were diluted 1 in 5 into media containing raffinose supplemented with 1% galactose until cultures reached an OD₆₀₀ of approximately 0.7.

Serial passage of L40 yeast

L40 yeast were transformed with either 500 ng pNIA2, an equal mixture (500 ng each) of pNIA2 and pNIA2-TAg cNLS, or 500 ng pNIA2-TAg cNLS and directly inoculated into 5 mL of synthetic drop-out media lacking leucine supplemented with glucose. Cultures were grown to an OD₆₀₀ of 0.7 and β-galactosidase activity was determined (Passage 1). From Passage 1, 100 μL was inoculated into 5 mL of fresh synthetic drop-out media lacking leucine and histidine supplemented with glucose. Cultures were grown to an OD₆₀₀ of 0.7 and again diluted into fresh synthetic drop-out media lacking leucine. When cultures reached an OD₆₀₀ of 0.7 β-galactosidase activity was determined (Passage 2).

Nuclear import selection

L40 yeast were grown to an OD₆₀₀ of 0.7 and 10 OD₆₀₀ units (∼2x10⁷ yeast total) were transformed with 5 μg EcoRI digested pNIA2 and 800 ng of PCR amplified NHS Ultramer (IDT), corresponding to an 8:1 ratio of NHS Ultramer to linearized pNIA2. Here, NHS Ultramer was PCR amplified in 10 low-cycle reactions using the Fw and Rv Ultramer oligos and pooled prior to transformation. Yeast were plated onto synthetic drop-out media lacking leucine and histidine, supplemented with glucose and 10mM 3-Amino-1,2,4-triazole. Transformants that underwent selection were inoculated into synthetic drop-out media lacking leucine and grown overnight at 30°C. The following day β-galactosidase activity was assayed and pNIA2 plasmid was isolated, via smash and grab method, and transformed into DH5α E. coli. Plasmids were subsequently isolated and sequenced via Sanger sequencing to determine encoded peptide sequences. Isolated plasmids were subsequently retransformed back into L40 yeast to measure nuclear import activity from biological replicates and perform statistical analysis.

Peptide sequences that were inputted to Basic Local Alignment Search Tool (version BLAST+ 2.11.0: October 19, 2020) were searched against the human proteome (taxid: 9606) containing non-redundant protein sequences using the Web BLAST implementation. The blastp algorithm was used to identify the most similar sequences by sorting based on query coverage and percent identity.

To estimate library diversity that was covered in this screen, 2x10⁶ L40 yeast were transformed with 100–800 ng of EcoRI digested pNIA2 and an 8:1 ratio of NHS Ultramer. Following transformation using the standard lithium acetate method, yeast were resuspended in 100 μL H₂O and serially diluted onto SD media lacking leucine to determine CFUs. Using CFUs as a proxy for library size, it was estimated that the library covered in this screen was approximately 800,000 peptides.

Co-immunoprecipitation and Western blot analysis in human cell lines

Human HT1080 fibrosarcoma cells were transfected with plasmids expressing FLAG-tagged mammalian importin α1, α3, or α5 and indicated EGFP fusions. For the EGFP-scrambled core (SCRcore) fusion, the core region was scrambled using the online peptide scrambling tool provided by mimotopes (www.mimotopes.com), with the number of scrambled peptides set to one. The following day, cells were lysed with NP40 lysis buffer (50 mM Tris–HCl pH 7.4, 150 mM NaCl, 2 mM EDTA, 1 mM EGTA, 10% glycerol and 0.1% NP-40) supplemented with a protease inhibitor cocktail (MilliporeSigma) and incubated with magnetic FLAG beads (MilliporeSigma) for 2 h at 4°C with agitation, 2% input of protein was kept as a loading control. Samples were separated on NuPage Bis-Tris gels (Life Technologies) and transferred onto PVDF membrane and western blotted with the following antibodies: EGFP (Takara), FLAG (MilliporeSigma), Kpnb1 (Santa Cruz) and Tubulin (MilliporeSigma).

Immunofluorescence confocal microscopy and subcellular fractionation

Human HT-1080 fibrosarcoma cells and HeLa adenocarcinoma cells were mounted on glass cover slips and transfected with the indicated EGFP constructs and fixed using 3.7% paraformaldehyde at room temperature. Cells were permeabilized on ice with 0.2% Triton X-100 for 30 min and blocked with 3% BSA in phosphate-buffered saline (PBS). Samples were incubated with primary antibody (Takara) for 1 h and incubated with anti-rabbit Alexa Fluor 488 secondary antibody (Life Technologies) for 30 min. Cover slips were then mounted onto glass slides using ProLong Gold Antifade Mountant with DAPI (Thermofisher). For treatment of HeLa cells with ivermectin (MilliporeSigma), cells were transfected and media was replaced 16–18 h later containing 25 μM ivermectin. Cells were left to incubate with ivermectin for 1.5 h before being collected for immunofluorescence analysis. Images were taken with a Nikon Eclipse Ti2 confocal microscope under 60× magnification using the Nikon NIS Elements acquisition software.

For biochemical fraction of cytoplasmic and nuclear compartments HT-1080 cells were transfected with the indicated EGFP plasmids. Cells were harvested and processed following the REAP method.^70,71 Nuclear and cytoplasmic fractions were assessed with anti-Lamin A/C and anti-α-tubulin, respectively. These experiments were performed in triplicate for quantitative analysis using ImageJ software (Fiji distribution).⁶⁵ Nuclear localization of EGFP was expressed as the ratio of nuclear to whole cell extract.

Quantification and statistical analysis

All statistical analyses were performed in PRISM GraphPad. One-way ANOVA and Dunnett’s multiple comparison were performed for the following figures: Figures 1C–1E comparing each sample to pNIA2; Figure 3B comparing each sample to EGFP; and Figure 4C comparing each sample to pNIA2-SG. One-way ANOVA and Tukey multiple comparison were performed for the following figures: Figures 1B and 2A. A t test was performed for Figure 2B comparing each sample to it’s respective No MYND control. For Figures 2D and 2F, multiple t tests were performed comparing all motifs against pNIA2-SG (D) or Gal4_DBD (F). To reduce stringency and increase identification of putative motifs we specifically did not account for multiple comparisons in either of these tests. Construction of the dumbbell plot depicted in Figure 2D was done in R using the ggplot2 library.⁶⁴ All data is displayed as mean ± SD. ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗p < 0.001. p < 0.05 was used a threshold for significance in all experiments.

Published: November 9, 2023