Cytoplasmic Transport Machinery of the SPF27 Homologue Num1 in Ustilago maydis

In the phytopathogenic basidiomycete Ustilago maydis, the Num1 protein has a pivotal function in hyphal morphogenesis. Num1 functions as a core component of the spliceosome-associated Prp19/CDC5 complex (NTC). The interaction of Num1 with the kinesin motor Kin1 suggests a connection between a component of the splicing machinery and cytoplasmic trafficking processes. Previously it was shown that Num1 localizes predominantly in the nucleus; however, due to the diffraction-limited spatial resolution of conventional optical microscopy, it was not possible to attribute the localization to specific structures within the cytoplasm. We have now employed super-resolution localization microscopy to visualize Num1 in the cytoplasm by fusing it to a tandem dimeric Eos fluorescent protein (tdEosFP). The Num1 protein is localized within the cytoplasm with an enhanced density in the vicinity of microtubules. Num1 movement is found predominantly close to the nucleus. Movement is dependent on its interaction partner Kin1, but independent of Kin3. Our results provide strong evidence that, in addition to its involvement in splicing in the nucleus, Num1 has an additional functional role in the cytosol connected to the Kin1 motor protein.


Establishment of a strongly fluorescent EosFP for super-resolution microscopy in U. maydis.
The interaction with the motor protein Kin1 observed in the yeast two-hybrid screen implicates a cytoplasmic localization of Num1. Previously, we detected cytoplasmic signals from a Num1:3GFP fusion protein with conventional fluorescence microscopy, however, fluorescent signals were too weak to assign them to specific cellular structures. In this study, we aimed at using PALM to benefit from its single-molecule sensitivity and capability of localizing emitters with nanoscale resolution. To this end, we constructed fusion proteins of Num1 with EosFP, which is a tetrameric protein in its natural form. Because oligomerization is often detrimental in imaging applications requiring functional fusion constructs, dimeric and monomeric EosFP variants have been generated by protein engineering 34 . For example, the change of the amino acid 158 from threonine to arginine (T158R) causes a splitting of the EosFP tetramer into dimers 31 . Other amino acids have been replaced to enhance stability or reduce the oligomerization tendency 31,35 . We have started our efforts with the frequently used photo-convertible fluorescent protein mEos2, a monomeric variant of EosFP that expresses well at 37 °C 36 . The mEos2 open reading frame was dicodon optimized for expression in U. maydis 37 and fused to the 3′ end of the num1 gene. The construct was integrated into the num1 locus of strain AB31 by homologous recombination to express the num1:mEos2 fusion gene under the native promoter of num1 in its natural context. The U. maydis strain AB31 (a1 P crg1 :bE1/P crg1 :bW2) harbors a set of compatible bE1 and bW2 genes under control of the arabinose-responsive P crg1 promoter 38 . In glucose-containing media, AB31 grows yeast-like, but upon arabinose-induced expression of bE1/bW2, the strain switches to polarized growth and forms long filaments reminiscent to those formed after fusion of compatible sporidia 38,39 .
The fluorescence signal of AB31 cells expressing Num1:mEos2 was monitored by conventional widefield fluorescence microscopy. EosFPs were photoconverted using a filter set with maximum transmission at 365 nm; fluorescence of the red and green species was measured in 20 s intervals. Only weak fluorescence in the green channel was detected from the nucleus; in contrast to our data obtained with a Num1:3GFP fusion protein 14 , cytoplasmic signals were not visible (Fig. 1a, upper panel). Conversion to the red EosFP species was not detectable within a time frame of 120 s. Thus, mEos2 does not appear to be feasible for PALM microscopy in U. maydis.
SCIENTIfIC RepORTS | (2018) 8:3611 | DOI: 10.1038/s41598-018-21628-y One option to increase the signal is to express the fusion protein by means of a strong promoter. However, as overexpression might lead to artificial localization, we aimed to increase the sensitivity of the EosFP probe. To this end, we constructed a tandem dimeric EosFP fusion protein optimized for use in U. maydis. We introduced the T158R mutation that splits the tetramer into dimers into a dicodon optimized EosFP open reading frame, and tethered two copies of the gene via a linker sequence encoding 12 amino acids 31 . We did not introduce a mutation to increase the thermostability because the dimeric as well as the monomeric EosFP proteins both express well at the optimal growth temperature of U. maydis (28 °C) 34 . In analogy to mEos2, the tandem dimeric EosFP (tdEosFP) open reading frame was fused in frame to the 3′ end of the num1 gene and introduced into the num1 locus of AB31.
PALM microscopy reveals cytoplasmic localization of Num1. As the novel tdEosFP performed well with respect to both fluorescent signal intensity as well as photoconversion, we next used PALM microscopy to localize Num1 EosFP fusion proteins in AB31 sporidia expressing Num1:tdEosFP; for comparison, we also studied AB31 expressing Num1:mEos2. PALM images, rendered from 1000 successive camera frames each, were acquired with 561 nm laser excitation and an additional weak 405 nm laser irradiation for continuous green-to-red photoconversion of the fluorescent proteins; the emission was filtered with a 607/70 band-pass filter. A photon number threshold of 100 was set for collecting fluorescent events in the red channel. As expected, both strains showed maximum fluorescence from the nucleus. However, fluorescence signals were also detectable within the cytoplasm (Fig. 1c), indicating a dispersed cytoplasmic localization of Num1. Both the number of detected fluorescent proteins as well as their intensities were much greater in AB31 sporidia expressing Num1:tdEosFP than in those expressing Num1:mEos2. The histograms given in Fig. 1d display the number of localization events as a function of the registered photons per frame for each event. For Num1:tdEosFP, the average number of photons from 965 events was 236; for Num1:mEos2, it was only 198 from 192 events. These data underscore that the tdEosFP fusion construct is significantly superior to the one with mEos2 for super-resolution imaging of proteins.
The localization precision of a fluorophore scales approximately with the square root of the number of photons collected during exposure. Effects such as fluorophore movement or a slight defocus may lead to lower photon numbers collected from an individual molecule and, therefore, less accurate localization. Such events can be excluded by setting a proper photon threshold; events below this threshold will simply be discarded. Taking a high photon number threshold in image reconstruction leads to a smaller number of events, but selective inclusion of very precisely localized events. Consequently, the image quality improves as long as the number of localization events is still sufficiently high. In Fig. 2, we compare a localization microscopy data set rendered with photon number thresholds of 100 ( Fig. 2a) to 300 (Fig. 2b). Under these more stringent imaging conditions, the Num1:tdEosFP protein is still clearly dispersed throughout the entire cytoplasm.
Both the interaction with the microtubule-associated motor protein Kin1 as well as the finding that deletion of num1 affects the movement of EEs, which are transported via a microtubule-dependent machinery 14 , suggest that Num1 might be associated with either microtubules and/or EEs. To address a possible connection between Num1 and the microtubule cytoskeleton, we expressed a GFP:Tub1 fusion protein 40 ectopically in AB31 num1:tdEosFP to visualize microtubules. Cells were sequentially imaged in green (GFP:tub1) and red (num1:tdEosFP) channels, and images were subsequently merged (see Materials and Methods).
In AB31 sporidia, Num1:tdEosFP shows partial colocalization with GFP-labeled microtubules; however, tdE-osFP signals are also visible in regions distinct from microtubules (Fig. 3a). We next questioned whether Num1 might localize predominantly in the vicinity of microtubules, and calculated the density of Num1:tdEosFP (1) in the microtubule region and (2) the cytoplasmic region of the entire cell. Based on microtubule images in the green channel, lines were drawn by hand to trace the microtubules, and the microtubule region was defined as those pixels that were closer than 300 nm (12 pixels with 25 nm pixel size) to the center of the lines (Fig. 3b, middle panel). The ±300 nm tolerance was chosen to account for possible movements of microtubules during imaging in the red channel (1000 frames, 50 s). As a control, the density of Num1:tdEosFP localization events was also calculated for the whole cytoplasmic region, i.e., the entire cell without the nuclear region ( Fig. 3b, right panel). Averaged over nine cells, we obtained 6.5 ± 0.8 µm −2 for the microtubule region, which is significantly larger than the value of 4.5 ± 1.1 µm 2 for the cytoplasmic region (Student's t-test, p < 0.0005) (Fig. 3c). This quantitative analysis supports the notion that Num1 is partially co-localized with microtubules.
To address the localization of Num1 in filamentous cells, we induced the expression of bE1 and bW2 in AB31 for 6 h in arabinose-containing medium. The resulting hyphae do not show any obvious phenotype with respect to morphology, septum formation or nuclear distribution, as described previously for num1 deletion strains 14 , indicating that the Num1:tdEosFP fusion protein is functional. Similar to our observations in sporidia, the main signal of Num1:tdEosFP resides within the nucleus, visible as a stretched, elongated structure in U. maydis hyphae 14 . Close to the nucleus, Num1:tdEosFP shows linear distributions (Fig. 3d); similar to the observations in sporidia, additional signals are distributed over the entire cytoplasm and are not restricted only to microtubules.

Localization and movement of Num1 depends on the integrity of the microtubule cytoskeleton.
To answer the question whether the Num1 protein moves within the cytoplasm, kymographs were generated along a straight line tracing the long axis of the cells (see Materials and Methods). As shown in Fig. 4a (upper panel), the Num1:tdEosFP protein showed bi-directional movement; proteins travel from the nucleus towards the sporidial tips and back towards the nucleus.
The influence of the microtubule cytoskeleton on the observed Num1 movement in the cytoplasm was analyzed by treatment of cells with the microtubule-destabilizing drug benomyl 41 . Widefield images revealed that the signal of Num1:tdEosFP in the cytoplasm was significantly reduced in the presence of 40 µM benomyl (Fig. 4a, lower panel). For quantification, 500 continuous frames were analyzed for cytoplasmic Num1:tdEosFP signals with the photon number threshold set to 100. 461 ± 47 events (N = 8) were detected for the control cells; for benomyl treated cells, the number of events decreased to 171 ± 45 events (N = 8) (Fig. 4b).
The kymographs of benomyl treated sporidia (Fig. 4a, lower panel) show markedly reduced transport traces of Num1:tdEosFP in the cytoplasm. Some signals in benomyl treated cells show horizontal lines, indicating that Num1 molecules are immobilized as a result of the disintegrated microtubule cytoskeleton. Thus, the integrity of the microtubule cytoskeleton is required for both cytoplasmic localization as well as cytoplasmic movement of Num1.

Kinesin 1 but not Kinesin 3 is involved in Num1 transport. The requirement for an intact microtubule
cytoskeleton suggests that cytoplasmic localization and/or mobility might depend on microtubule-dependent motor proteins, which would be in line with the previously observed interaction of Num1 with Kin1 14 . Accordingly, we assessed the involvement of the two microtubule-associated motor proteins Kin1 and Kin3 on Num1 trafficking. kin1 and kin3 were replaced in AB31 num1:tdEosFP with a nourseothricin-resistance marker gene by homologous recombination. In the resulting deletion strains AB31 num1:tdEosFP∆kin1 and AB31 num1:tdEosFP∆kin3, movement of Num1:tdEosFP was analyzed in hyphae via kymographs. We focused on hyphae because both the phenotype of kin1 or kin3 as well as of num1 deletions is more prominent in the filament. Deletion of either kin1 or kin3 leads to an altered hyphal morphology in AB31; induction of bE1/bW2 in AB31 wildtype cells results in long filamentous hyphae that originate at one cell pole of the sporidium; in ∆kin1 and ∆kin3 deletion strains, the filament is initiated on both cell poles (bipolar), and the hyphae are much shorter 10 . The deletion also affects shape and localization of the nucleus: in AB31 hyphae, the nucleus migrates into the filament and has a long, stretched appearance, while in ∆kin1 and ∆kin3 strains nuclei remain in the sporidia and are round-shaped.
Although Num1 appears distributed within the entire hypha of AB31 wildtype strains (Figs 3d and 5a), we observe most movement of Num1:tdEosFP in the vicinity (8-10 µm) of the nucleus (Fig. 5a). Deletion of kin3 does not alter movement of Num1; similar to the situation in AB31 wildtype hyphae, movement is observed around the nucleus (Fig. 5b). In contrast, cytoplasmic movement is nearly completely abolished AB31∆kin1 filaments: the Num1 molecules are apparently trapped close to the nucleus and cannot move further into the cytoplasm (Fig. 5c).
Num1 does not co-localize with early endosomes. As the deletion of Num1 affects the movement of EEs 14 , we addressed a potential co-localization of Num1 with these vesicles by expressing a GFP:Rab5a fusion protein as a specific marker of EEs 42 in AB31 num1:tdEosFP. Localization microscopy requires a large number of camera frames; thus, the technique is not fast enough to image Rab5a movement in living cells. Therefore, we fixed samples with formaldehyde to study co-localization. GFP:Rab5a labeled EEs are distributed throughout the cytoplasm (Fig. 6); the frequently observed linear array of their localization is in accordance with their microtubule association. However, we did not find co-localization between the GFP:Rab5a and Num1:tdEosFP signals, indicating that Num1 does not localize to EEs.

Discussion
In our previous study 14 we failed to localize the Num1 protein to specific structures within the cytoplasm due to its low abundance. We therefore employed PALM to improve both single-molecule sensitivity and resolution of Num1 fusion proteins. Frequently, the monomeric mEos2 protein is used as a fluorescent probe for PALM, in U. maydis, however, mEos2 showed only a weak fluorescence emission that limited the use for PALM. To improve EosFP as fluorescent probe for U. maydis, we constructed a pseudo-monomeric EosFP tandem construct consisting of two protomers connected via a flexible linker engineered so as to favor dimerization. The usefulness of such a tandem dimer construct has been shown previously in HEK293 cells 31,34 . Our data show that tdEosFP is superior to the monomeric mEos2 with respect to sensitivity and brightness: in a direct comparison with a mEos2   the phenotypic alterations accompanying a non-functional Num1 protein, such as altered hyphal morphology or changed motility of EEs. Thus, tdEosFP appears to be a very feasible tool for super-resolution localization microscopy in U. maydis.
Num1 has been shown to function in the spliceosome-associated NTC-complex as an SPF27 homologue, and, in accordance with this function, the predominant localization of Num1 is within the nuclear compartment. In our previous studies 14 , we observed an interaction of Num1 with several cytoplasmic proteins including the microtubule-dependent motor protein Kin1, which suggested an additional localization of Num1 within the cytoplasm. Faint cytoplasmic emission was observed from Num1:GFP fusion proteins; however, precise localization of the weak signals was not possible. PALM microscopy with the newly established tdEosFP construct now clearly supports both cytoplasmic localization of Num1 as well as Kin1-dependent movement of Num1. Deletion of the num1 gene has been shown to affect the transport of EEs. However, we can exclude that the observed motility of Num1 depends on EEs because (1) we did not observe co-localization of Num1:tdEosFP with the endosomal marker Rab5a, and (2) motility of Num1 in hyphae is affected by deletion of kin1 but not kin3. Kin3 is the motor protein required for the plus-end directed transport of EEs on microtubules 4,5 . Thus, if Num1 movement would result from its association with EEs, we would expect an impact of the kin3 deletion also on Num1 motility. Kin1 is indirectly involved in EE motility via the retrograde transport of dynein; Kin1 activity leads to an accumulation of dynein at a dynein loading zone in the hyphal tip 6 . Since deletion of kin1 or num1 has similar phenotypes with respect to dynein mislocalization 6,14 , we have to assume that Num1 affects the motility of EEs more likely by disturbing the kinesin1/dynein transport machinery. In Saccharomyces cerevisiae, the transport of dynein via the kinesin motor Kip2 is facilitated by Pac1/LIS1, which acts as a processivity factor for kinesin to suppress hindrance by dynein moving in the opposite direction 43 . Off-loading of dynein is mediated by yeast Num1p by relieving Pac1/LIS1-mediated inhibition. Of note, the yeast Num1p protein shares only its name with the U. maydis Num1 protein, but shows no homologies 44 . Similar to yeast Num1p, the U. maydis Num1 protein could also function to establish or maintain the dynamics of a Kin1/Dyn complex. In hyphae, the minus-ends of microtubules accumulate around the nucleus; thus, one would expect the function of a protein that "reverses" the direction of a Kin1/Dyn complex within this region. This would be in accordance with our observation that (a) Num1 is predominantly localized in the vicinity of microtubules within this region (Fig. 3d,b) Num1 motility is mostly observed within this region (Fig. 5a).
As motility and localization of Num1 depend on both the integrity of the cytoskeleton and on Kin1, we have to infer a microtubule-associated localization of Num1. Although we did not detect a clear co-localization with GFP:Tub1, the density of Num1:tdEosFP appears enhanced around microtubules, perhaps due to a weak interaction with microtubule-bound Kin1. Indeed, the interaction between the two proteins appears to be rather weak, as indicated by the observation that co-purification of Num1 and Kin1 is only possible after crosslinking 14 . Such a weak interaction would be in line with a transient function of Num1 in controlling the dynamics of a Kin1/Dyn complex on microtubules.
In ∆kin1 deletion strains, localization and motility of Num1 are restricted to the nucleus and to a region very close to the nucleus, suggesting that Kin1 is intimately associated with the transfer of the protein from the nucleus to the cytoplasm. Kin1 has been described to interact with the nuclear pore complex 45 , It is well possible that the initial interaction of Num1 and Kin1 appear at this stage, and that Num1 stays attached close to the nuclear envelope without Kin1.
Num1 functions as a component of the splicing-associated NTC 14 . Within U. maydis hyphae, mRNA is tethered to EEs mediated by the RNA-binding protein Rrm4 3 . The functional connection between nuclear export and loading on EEs of mRNA molecules is currently not yet understood. Similar to what has been described for proteins of the exon junction complex and for the Drosophila oskar mRNA reviewed in 46 , one could imagine that Num1 remains attached to mRNA molecules after splicing. An interaction of Kin1 with Num1 at the nuclear pore complex could then be involved in loading of mRNAs to EEs, which would be in agreement with the observed movement of Num1 close to the nucleus. The distinct localization in ∆kin1 strains could be attributed to the tethering of Num1 to the nuclear pore complex, the minimal movement reflecting the dynamic movement of the nuclear envelope (or the nuclear pore complexes).
With the fusion of an optimized tdEosFP to Num1, we have shown the feasibility of this construct for PALM microscopy in U. maydis by proving the cytoplasmic localization and the Kin1 dependent motility of a spliceosome-associated protein. In general, localization-based super-resolution microscopy using pseudo-monomeric EosFP tandem fusion proteins appears as a promising approach for cell biology in U. maydis and other fungal systems.

Materials and Methods
DNA procedures. Molecular methods followed established protocols 47 . U. maydis transformation procedures were performed as previously described 48 . PCR-generated linear DNA was used for transformation of U. maydis 49 . Homologous integration of all constructs was verified by gel blot analysis.
SCIENTIfIC RepORTS | (2018) 8:3611 | DOI:10.1038/s41598-018-21628-y Plasmid constructions. For construction of tandem dimeric EosFP, we followed the strategy described in 31 . The open reading frame (ORF) for eos was di-codon optimized 37 for U. maydis based on the EosFP amino acid sequence (GenBank accession: AAV54099.1) using an online tool developed by Florian Finkernagel, IMT, Marburg, Germany (http://dicodon-optimization.appspot.com) and synthesized by Eurofins Genomics (Ebersberg, Germany). The synthesized EosFP ORF harbors a mutation (corresponding to T158R) to prevent tetramerization; nucleotides corresponding to amino acid positions 220 and 108 were altered by silent mutagenesis to delete internal BsaI and NcoI sites that would interfere with subsequent cloning. The first copy of EosFP (T158R) was PCR amplified with di-codon optimized EosFP DNA as template using primers P1 and P2 (see Suppl. Table 1) to add terminal 5′ SfiI and 3′ BsaI sites and to delete the STOP codon of the EosFP ORF. The second copy of EosFP (T158R) was generated by PCR amplification using the same template by primer pair P3 and P4 (see Suppl. Table 1); P3 adds a BsaI site and a 36 nt linker sequence corresponding to the amino acid sequence GHGTGSTGSGSS 5′ to the tdEosFP ORF; P4 adds a AscI site 3′ of the STOP codon.
The type IIS restriction enzyme BsaI cleaves outside of its recognition sequence (in primers P2 and P3), creating four base flanking overhangs that were customized to allow the direct assembly of the two tdEos fragments. The fusion via BsaI resulted in a tandem arrangement of two EosFP monomers linked by the sequence GHGTGSTGSGSS (tdEosFP). The sequence of the tdEosFP protein is given in Suppl. Fig. 1. The SfiI/AscI fragment harbouring the tdEosFP ORF was subsequently exchanged with the SfiI/AscI fragment harboring the eGFP gene in plasmid pUMa317 50 . In the resulting plasmid pBL7, the tdEosFP ORF is flanked at the 3′ end by the nos terminator, followed by a hygromycin-resistance cassette (hyg R ) to allow selection in U. maydis (Suppl. Fig. 2). The entire tdEosFP/hyg R fragment is bordered by two SfiI sites that allow fusion of DNA-fragments for homologous recombination in U. maydis in a single step 49,50 .
Plasmid pPtefRab5aGn-Cbx R harbors an N-terminal fusion of the eGFP open reading frame to rab5a Rab5a (UMAG_10615; GenBank accession XP_011387349.1) under control of the tef-promoter 52 , and a carboxin cassette for the integration into the ip-locus 53  Strains and growth conditions. Escherichia coli strain TOP10 (Invitrogen, Carlsbad, USA) was used for cloning purposes. Growth conditions and media for cultivation were followed as described previously 47 . U. maydis cells were grown in CM Complete Medium 54 , containing 1% glucose (CM-G) or 1% arabinose (CM-A), respectively, at 28 °C. Solid media contained 2% agar. For microscopic analyses, cells were taken from logarithmically growing liquid cultures in CM-G medium. For the investigation of hyphae, cells were transferred from CM-G to CM-A and induced for 6 h, as described 38 . 200 µl cells were mixed with 200 µl 4% low melting agarose in H 2 0 (45 °C) in 8 well chambers (μ-Slide 8 well, ibidi GmbH, Munich, Germany) and analyzed immediately.
U. maydis strains used in this study were generated as follows. AB31 num1:tdEosFP: To generate a tdEosFP fusion to Num1 (UMAG_01682, GenBank XP_011387660.1) 1 kb fragments corresponding (a) to the num1 ORF (primers P5, P6, Suppl. Table 1) and (b) to the genomic 3′ region of num1 (primers P7, P8, Suppl. Table 1) were generated by PCR, digested with SfiI (added by primers P6 and P7, respectively) and ligated to the SfiI-digested tdEosFP:hyg R cassette from plasmid pBL7. The ligation product was PCR amplified and integrated to the native num1 locus of strain AB31 38 via homologous recombination, following previously described protocols 49 .
Cell imaging. Fluorescence  Carl Zeiss, Jena, Germany) were used for epifluorescence analysis. EosFPs were photoconverted with the DAPI filter set with a wavelength maximum of 365 nm in 20 s intervals. After each interval, red and green emitting species were imaged with red and green filter sets, respectively, with 600 ms exposure time (2) Andor Revolution XD spinning disk confocal laser scanning microscope (BFi OPTiLAS, München, Germany) with an OLYMPUS ApoN60×/1.49 oil immersion objective. Cells were imaged with a 488 nm laser to excite mEos2 or tdEosFP in the green channel and filtered with a 525/50 (peak wavelength/width) band pass filter (Semrock, New York, NY); images were further analyzed with ImageJ.
The fusion proteins Num1:mEos2 and Num1:tdEosFP were photoconverted to their red emitting species by 405 nm laser irradiation with 0-50 W cm −1 laser intensity and simultaneously excited by a 561 nm laser (200-400 W cm −1 ). A 607/70 band-pass filter (Semrock, New York, NY) was used to filter the fluorescence emission, which was detected by an EMCCD camera with 50 ms exposure time for 1000-1500 frames. Molecule localization and image calculations were performed with the custom-written analysis software a-livePALM 30  For images in two color channels, we first excited the cells with a 473 nm laser (100-200 W cm −1 ). Typically, 100 continuous frames were taken to image the emission from GFP:Tub1 and GFP:Rab5, filtered with a 512/25 band pass filter (Semrock, New York, NY). After a delay of typically 3 s to change filters, we imaged Num1:tdEosFP as described above. We used a-livePALM to generate localization microscopy images for the data in the red channel and merged them with images in the green channel created by using ImageJ.
Kymographs were generated with the ImageJ software 56 using a Kymograph Plugin (https://www.embl.de/ eamnet/html/body_kymograph.html) along a straight line tracing the long axis of the whole cell. The line width was set equal to the width of the cell to cover the entire cell, so the intensity was integrated perpendicular to the long axis. All kymographs were generated from image stacks with 500 frames.