The Nma1 protein promotes long distance transport mediated by early endosomes in Ustilago maydis

Early endosomes (EEs) are part of the endocytic transport pathway and resemble the earliest class of transport vesicles between the internalization of extracellular material, their cellular distribution or vacuolar degradation. In filamentous fungi, EEs fulfill important functions in long distance transport of cargoes as mRNAs, ribosomes, and peroxisomes. Formation and maturation of early endosomes is controlled by the specific membrane‐bound Rab‐GTPase Rab5 and tethering complexes as CORVET (class C core vacuole/endosome tethering). In the basidiomycete Ustilago maydis, Rab5a is the prominent GTPase to recruit CORVET to EEs; in rab5a deletion strains, this function is maintained by the second EE‐associated GTPase Rab5b. The tethering‐ and core‐subunits of CORVET are essential, buttressing a central role for EE transport in U. maydis. The function of EEs in long distance transport is supported by the Nma1 protein that interacts with the Vps3 subunit of CORVET. The interaction stabilizes the binding of Vps3 to the CORVET core complex that is recruited to Rab5a via Vps8. Deletion of nma1 leads to a significantly reduced number of EEs, and an increased conversion rate of EEs to late endosomes. Thus, Nma1 modulates the lifespan of EEs to ensure their availability for the various long distance transport processes.

In contrast to yeast, however, where cells deleted for all three isoforms are still viable (Singer-Krüger et al., 1994), double deletion of rabA and rabB is lethal in A. nidulans, arguing for an at least partially overlapping function of the two Rab5 isoforms (Abenza et al., 2010).
Both Rab5 and Rab7 are effectors for multi-subunit tethering complexes that are required for membrane fusion and endosome maturation, namely the CORVET complex (class C core vacuole/endosome tethering) for Rab5 on EEs and the HOPS complex (homotypic fusion and protein sorting) for Rab7 on late endosomes.
HOPS and CORVET complexes are conserved in function from unicellular eukaryotes to mammals (Perini et al., 2014), although they may interact with different GTPases or interaction partners. In S. cerevisiae, CORVET is an effector primarily of Vps21 (Peplowska et al., 2007), and also for A. nidulans it has been shown that CORVET interacts preferentially with RabB, and only to lesser extend with RabA (Abenza et al., 2010). CORVET is required for homotypic fusion and maturation of EEs, but also involved in transport processes between endosomes and vacuoles and in sorting of transported cargoes Peplowska et al., 2007;Perini et al., 2014;Rink et al., 2005).

HOPS is required for homotypic vacuole-vacuole fusion events, or
fusion of autophagosomes or Golgi-derived vesicles with the vacuole (Liang et al., 2008;Stroupe et al., 2006;Yogosawa et al., 2005Yogosawa et al., , 2006. Both tethering complexes are related and share four subunits as core components, in yeast termed Vps11p, Vps16p, Vps18p, and Vps33p. The two CORVET-specific subunits Vps3p and Vps8p bind to Rab5 proteins, while specificity of HOPS to Rab7 is mediated by the two HOPS-specific subunits Vps41p and Vps39p (reviewed in Solinger & Spang, 2013).
The HOPS and CORVET-specific subunits are located at opposing poles related to the central positioned core subunits of the complex (Figure 1a), which allows to contact their cognate Rab-GTPases on membranes of individual endosomes as a first step for membrane fusion Bröcker et al., 2012;Plemel et al., 2011). Vps33p, one of the core subunits in both complexes, is a member of the Sec1/Munc18 (S/M) protein family, which promote SNARE (soluble N-ethylmaleimide-sensitive fusion protein attachment protein receptors) complex formation. SNAREs are membranebound protein complexes that mediate the fusion of two membranes that harbor compatible SNARE complexes. For HOPS it has been shown that it is required for assembly of SNAREs to induce membrane fusions (Torng et al., 2020;Zick & Wickner, 2016), and, based on the structural similarity, the same is postulated for CORVET.
It was proposed that, as part of endosomal maturation progress,  (Peplowska et al., 2007).
In contrast to S. cerevisiae, where individual components of the endocytic pathway can be deleted with only minor effects on cell growth or morphology (Singer-Krüger et al., 1994), filamentous fungi-like A. nidulans and U. maydis depend on Rab5a for polarized growth (Abenza et al., 2010;Bielska, Higuchi, et al., 2014). The elongated cells of filamentous fungi may impose additional requirements for EEs with respect to EE-mediated long distance transport within the elongated cell.
During the past decades, the basidiomycete U. maydis has developed to an excellent model organism for cell biological studies (summarized in Steinberg & Perez-Martin, 2008). The fungus has a dimorphic lifestyle; the switch from budding growth to filaments is controlled by the b-mating type locus that encodes a pair of homeodomain transcription factors, termed bE and bW (reviewed in Brefort et al., 2009). Strains in which the bE and bW genes are expressed under nutritional-controlled promoters allow the observation of both budding cells and filaments (Brachmann et al., 2001). Endosomal transport has been extensively studied in U. maydis and has revealed novel F I G U R E 1 The class C core vacuole/endosome tethering (CORVET) subunits Vps3 and Vps8 colocalize in Ustilago maydis with early endosomes (EEs). (a) Model of the CORVET complex. CORVET binds with its specific subunits Vps3 and Vps8 to the activated form of Rab5a (GTP-bound) located on EEs to initiate endosomal fusion, indicated by arrows (Modified after Figures from . Movement of EEs is mediated by molecular motors as kinesin and dynein along the microtubule cytoskeleton. (b) Vps3 (left, mKate2, magenta) and Vps8-(right, 3xmCherry, magenta) show predominantly colocalization (white) to Rab5a (GFP, green) on EEs. (c) Kymograph showing that Vps3 (left, mKate2, magenta) and Vps8 (right, 3xmCherry, magenta) are transported with Rab5a (GFP, green) on EEs. (d) CORVET subunits Vps3 (GFP, green) and Vps8 (3xmCherry, magenta) colocalize partially (left panel, white signal) and show jointly bidirectional movement (kymograph, right panel). Kymographs: horizontal scale 10 µm, vertical scale 10 s; Microscope pictures: Scale 10 µm functions for EEs in transport of cargoes as mRNAs, ribosomes, and peroxisomes and their impact on formation and maintenance of filamentous cells (Becht et al., 2006;Guimaraes et al., 2015;Higuchi et al., 2014;Olgeiser et al., 2019;Zander et al., 2016).
We have now studied the impact of Rab5 GTPases and the CORVET complex on function and maturation of EEs in U. maydis.
In particular, we have identified a novel protein, Nma1, that interacts with CORVET and modulates the lifespan of EEs to ensure their availability for the various long distance transport processes.

| The CORVET complex in U. maydis consists of conserved components
The components of the CORVET complex are conserved from yeasts to mammals (Perini et al., 2014), and also in U. maydis all components were identified in the genome based on their high similarity to both the S. cerevisiae and human homologs (Table S1). The composition of CORVET in U. maydis was investigated by co-immunoprecipitation/ mass spectroscopy (Co-IP/MS). To this end, we expressed C-terminally tagged Vps3-3xHA or Vps8-3xMyc under control of the native promoter within the respective endogenous genomic locus in strain AB31. This strain allows controlled switching from budding to filamentous growth via arabinose-induced expression of the bE/bW heterodimer (P crg1 :bE1, P crg1 :bW2) (Bottin et al., 1996;Brachmann et al., 2001). When incubated under repressing conditions in glucose medium, AB31 cells multiply by budding, similar to wildtype haploid cells (sporidia). When shifted to arabinose medium, the induction of the bE1 and bW2 genes leads to the formation of the bE1/bW2 heterodimeric transcription factor that initiates filamentous growth and pathogenic development. The arabinose-induced AB31 filaments resemble the wildtype hyphae formed after fusion of compatible cells: the cells show polarized growth, but only the tip cell is filled with cytoplasm, while most of the hypha consists of empty compartments that are separated from the tip by a retraction septum (Freitag et al., 2011). In axenic culture, cells of the filaments do not divide; further development is linked to the infection of the host plant.

| Rab5a and the CORVET subunits Vps3 and Vps8 colocalize on EEs
To visualize the colocalization of CORVET to EEs in U. maydis, we constructed AB31 derivatives harboring the Rab5a protein fused N-terminally to the green fluorescent protein (GFP), in combination with either Vps3 or Vps8 fused C-terminally to the fluorescent proteins 3xmCherry or mKate2. In all cases, the gene fusions were generated via homologous recombination at the native gene locus to ensure expression of the fusion gene under the native promoter. Movement of EEs was monitored in filamentous cells of AB31 gfp-rab5a vps3-mKate2 and AB31 gfp-rab5a vps8-3xmCherry. Both Vps3-mKate2 and Vps8-3xmCherry colocalized with GFP-Rab5a on EEs (Figure 1b,c); for both proteins, only few signals were detected independent from the localization of Rab5a. In line with this, Vps8-3xmCherry and Vps3-GFP also colocalized on EEs (Figure 1d). Kymographs revealed that the colocalization of GFP-Rab5a with either Vps3-GFP or Vps8-3xmCherry occurred both during acropetal as well as basipetal movement of EEs. Our data corroborate the general scheme that the CORVET complex in U. maydis cells is associated with motile EEs.

| Both Rab5a and Rab5b can recruit CORVET to EEs
When compared with S. cerevisiae and A. nidulans, Rab5a from U. maydis is closest to Vps21p (S.c.) and RabA (A.n.), while Rab5b (U.m.) clusters with Ypt52 (S.c.) and RabB (A.n.) (Figures S1, S2 and Table S2). Despite the higher similarity of RabA from A. nidulans to Vps21p from yeast, RabB appears to be the functional ortholog of Vps21p, based on their prominent role for endosomal function (Abenza et al., 2009(Abenza et al., , 2010. Similar to Vps21, RabB, and only to a lesser extend RabA, was shown to interact with CORVET (Abenza et al., 2010;Peplowska et al., 2007). Another argument for Vps21p and RabB as orthologs is that both proteins lack a long insertion present in Ypt52 (S.c.) and RabA (A.n.). Interestingly, this extension is present in U. maydis Rab5a, but absent in Rab5b ( Figure S1).
It has been described previously that deletion of rab5a in U. maydis abolishes movement of EEs and impedes polar growth in filaments (Bielska, Higuchi, et al., 2014). However, in contrast to these earlier observations, we still observed residual movement of EEs in AB31 filaments deleted for rab5a, as indicated by movement of a GFPfusion to Yup1, a t-SNARE localizing on EEs and vacuoles (Wedlich-Söldner et al., 2000;Figure 2a,b). We also observed a cytokinesis defect in AB31∆rab5a sporidia, leading to chains of unseparated cells ( Figure S3a), although growth rate of AB31∆rab5a sporidia was not altered ( Figure S3c).
Despite the lack of the amino acid extension in Rab5b, which would suggest that the protein could be a functional ortholog of A. nidulans RabB and yeast Vps21p, the deletion of rab5b in AB31 did not result in detectable alterations with respect to cell morphology and growth rates ( Figure S3). Furthermore, endosomal movement was not changed in AB31∆rab5b, as indicated by movement of GFP-Rab5a (AB31∆rab5b gfp-rab5a) and Yup1-GFP (AB31∆rab5b yup1-gfp; Figure S4).
We next attempted to generate strains deleted for both rab5a and rab5b. However, we never obtained transformants in efforts to delete rab5b in AB31∆rab5a or to delete rab5a in the AB31 ∆rab5b background, which indicates that the double deletion might be lethal. To confirm this F I G U R E 2 Early endosomes (EEs) show reduced, but not abolished movement in rab5a deletion strains. (a) In wildtype AB31 background, the t-SNARE Yup1 (GFP, green) shows evenly distribution through filaments with bidirectional movement as well as static signals. (b) Deletion of rab5a in AB31 results in accumulations of static Yup1 signals at cell poles of bipolar growing filaments (previously shown by Bielska, Higuchi, et al., 2014). Only in regions close to the tip of the cells bidirectional movement can be observed. (c) Rab5a (GFP, green) and Rab5b (mCherry, magenta) show colocalization (white) (upper panel) and bidirectional co-motility (lower panel). (d) Deletion of rab5a reduces, but does not abolish colocalization (white) (upper panel) or joint movement (lower panel) of Vps3 (GFP, green) and Vps8 (3xmCherry, magenta). (e) In wildtype AB31 hyphae (upper panel) Vps3 (GFP, green) and Rab5b (mCherry, magenta) show partial colocalization (white) and co-motility. Deletion of rab5a (lower panel) reduces endosomal movement, but Vps3 and Rab5b still colocalize (white) on motile EEs. Kymographs: horizontal scale 10 µm, vertical scale 10 s hypothesis, we deleted rab5a in the haploid wildtype strain FB1 (a1b1) and rab5b in the compatible strain FB2 (a2b2). We inoculated maize plant seedlings with a mixture of FB1∆rab5a::nat R and FB2∆rab5b::gent R strains, and harvested diploid spores from tumor material. During germination of spores, meiosis occurs to generate haploid sporidia. For unlinked genes, an equal distribution of the possible recombination products is expected. However, sporidia that were resistant to both Geneticin (gent R ) and Nourseothricin (nat R ) were not identified, indicative that cells with a double deletion of rab5a and rab5b are not viable (Table S3).
In addition to Rab5a and Rab5b, we have identified Rab5c as an additional protein with similarity to Rab5-GTPases. Rab5c shows 43.6% and 42.9% identity to Rab5a and Rab5b, respectively. In S. cerevisiae, it is most similar to Ypt53, (46.9%), a gene duplication of Vps21 (Klöpper et al., 2012), and in A. nidulans to RabX, a protein with unknown function (Pinar & Penalva, 2021; Table S2). In addition to conserved domains present in the Rab5 protein family, the Rab5c protein shows an N-terminal extension of 340 amino acids and an internal insertion of 112 amino acids, both with unknown function (Figures S1 and S2).
Deletion of rab5c did not cause any obvious alteration of the phenotype with respect to cell morphology or movement of EEs; also, the combined deletion with rab5a or rab5b did not alter the phenotype of strains harboring single rab5a or rab5b deletions ( Figure S5). The finding that in cells deleted for rab5a movement of EEs is severely reduced, but still observable, argues that Rab5a is the prominent Rab5-GTPase required for the function of EEs. As the double deletion of ∆rab5a∆rab5b is lethal, the residual movement (and activity) in ∆rab5a strains has to be attributed to Rab5b that can partially overtake the function of Rab5a.
Rab5c apparently has no influence on endosomal function or movement. In accordance with the partially redundant function of Rab5a and Rab5b, both proteins were found to colocalize on 50%-60% of the EEs in strain AB31 gfp-rab5a mCherry-rab5b (N = 40 independent hyphae; Figure 2c). For Rab5c (AB31 P nar1 :gfp-rab5c vps8-3xmCherry), no specific localization to EEs was observable ( Figure S6). We next examined the Rab5-dependent recruitment of the CORVET complex to EEs. Deletion of rab5b (AB31∆rab5b vps3-gfp vps8-3xmCherry) did not alter the localization of the CORVET subunits Vps3-GFP or Vps8-3xmCherry to EEs. In accordance with the central function of Rab5a, the rab5a-deletion in strain AB31∆rab5a vps3-gfp vps8-3xmCherry led to a severe reduction in EE motility ( Figure 2d). However, Vps3 and Vps8 colocalized on the residual moving EE, emphasizing that Rab5b can complement partially the function of Rab5a in recruiting the CORVET complex to EE (Figure 2d). This is in line with the finding that Vps3 and Rab5b show colocalization on motile EEs in the rab5a deletion strain AB31∆rab5a mCherry-rab5b vps3-gfp (Figure 2e), demonstrating that Rab5b is sufficient to recruit CORVET to EEs in the absence of Rab5a.

| Recruitment of the CORVET complex to early endosomes is initiated by Vps8
For functional analysis of the CORVET subunits Vps3 and Vps8 and the core-subunit Vps11, we attempted to generate the respective deletion strains. As we were not able to generate haploid ∆vps8 or ∆vps11 strains, we deleted one copy of either vps8 or vps11 in the diploid, solopathogenic strain FBD11 (a1a2b1b2) (Banuett & Herskowitz, 1989). Analysis of the meiotic offspring revealed no haploid ∆vps8 sporidia (N = 60) or ∆vps11 sporidia (N = 178), demonstrating that both vps8 and vps11 are essential genes in U. maydis (Table S3).
However, deletion of vps3 did neither abolish the movement of EEs nor recruitment of the CORVET complex to EEs: in AB31∆vps3 gfp-rab5a vps8-3xmCherry, Rab5a was localized on motile EEs and colocalized with Vps8 ( Figure 3c). Apparently, similar to the situation in S. cerevisiae, the first step in recruitment of CORVET to EEs requires Rab5a and Vps8; Vps3 is then required for recruitment of a second Rab5a to initiate the subsequent steps in endosome maturation and/ or fusion (Pawelec et al., 2010;Peplowska et al., 2007). As Vps8 is essential, we were not able to examine whether Vps3 is able to partially complement the function of Vps8 to recruit CORVET to EEs.
To examine whether deletion of vps3 alters the composition of CORVET, we performed Co-IP/MS with protein extracts from strain AB31∆vps3 vps8-3xmyc compared with the respective wildtype strain. All components of CORVET (with exception of Vps3) co-purified with Vps8. Interestingly, when compared with the Co-IP from wildtype strains, we detected an enrichment of Vps39 in ∆vps3 cells ( Figure 3d). In yeast, during maturation of EEs to late endosomes, Vps3 is substituted by Vps39 in the CORVET-related HOPS-complex, which is specific for late endosomes (Ostrowicz et al., 2010;Plemel et al., 2011). Our findings suggest now that deletion of vps3 alters the specificity of CORVET by substitution of Vps3 with Vps39, which would lead to a hybrid CORVET-HOPS complex. Indeed, we observed GFP-Rab5a in ∆vps3 cells increasingly associated with immobile structures that probably resemble late endosomes (Figures 3e and S7).

Interestingly, in both Vps3 and Vps8 Co-IP/MS experiments, in
addition to the conserved CORVET complex components, a so far uncharacterized protein (Umag_00933) was co-purified with high abundance (Table S1 and Figure 3d). Based on the finding that Umag_00933 co-purified with components of CORVET, we examined the possible association of the protein with EEs in strain AB31 umag_00933-3xmCherry gfp-rab5a. Indeed, Umag_00933 was found to colocalize with Rab5a on motile EEs (Figure 4a). In addition, we observed a colocalization of Umag_00933 with microtubles (AB31 umag_00933-3xgfp P otef :mCherry-tub1; Figure 4b). Also, treatment with Benomyl, a microtubule-destabilizing agent, led to disruption of the fibrillar Umag_00933 localization ( Figure S8). Heterologous expression of Umag_00933-GFP in A. nidulans harboring a mCherry-RabA fusion protein (homologous to Rab5a) revealed that Umag_00933 also localized to fibrillar structures (microtubules). While the EE-marker RabA showed bidirectional motility, we could not detect any movement for Umag_00933, indicating that microtubule-association, but not the localization to motile EEs, is maintained in A. nidulans ( Figure S9).

| Nma1 (Umag_00933) interacts with CORVET and localizes at EEs and microtubules
Due to the endosome-and microtubule-association of Umag_00933, we termed the protein Nma1 (N-dosome and Microtubule Associated).
In accordance with the association of Nma1 with CORVET (indicated by Co-IP, Table S1 and Figure 3d), Nma1 colocalized and moved together with both Vps3 and Vps8 (AB31 vps3-gfp nma1-3xmCherry and AB31 vps8-3xmCherry nma1-3xgfp; Figure 5a,b). Deletion of rab5a strongly reduced localization of both Vps3 and Nma1 to EEs and resulted in drastically diminished motility of EEs, as previously observed for ∆rab5a deletion strains (Figure 5c). While deletion of vps3 only led to reduced motility of Rab5 or Vps8 labeled EEs (see Figure 3c), we could not F I G U R E 4 Nma1 localizes at early endosomes and microtubules. (a) Rab5a (GFP, green) and Nma1 (3xmCherry, magenta) show colocalization (white, left panel) and bidirectional co-motility (white, right panel) in AB31 strains. Nma1 localizes in addition to static structures. (b) Nma1 (GFP, green, middle panel) and Tub1 (mCherry, magenta, lower panel) show colocalization (white, upper panel) at fibrillar structures through whole AB31 hyphae. See also enlargement on the right (blue box). Kymographs: horizontal scale 10 µm, vertical scale 10 s F I G U R E 3 Deletion of vps3 does not abolish recruitment of class C core vacuole/endosome tethering (CORVET) to early endosomes (EEs). (a) Deletion of vps3 (right) results in cytokinesis defects in sporidia and polarity defects in AB31∆vps3 hyphae, while AB31 wildtype cells (left) separate after each cell division and grow as unipolar filaments. (b) AB31∆vps3 sporidia (right) contain a reduced number of vacuoles (stained with CMAC) when compared with wildtype AB31 (left). Number of vacuoles is significantly (***) reduced from 6.6 ± 1.77 in AB31 to 3.84 ± 1.55 in AB31∆vps3 (arithmetic mean, standard deviation; t-test p < .001, N = 300). (c) Rab5a (GFP, green) and Vps8 (3xmCherry, magenta) show colocalization (white) (upper panel) and bidirectional co-motility (white) (lower panel) in AB31∆vps3 strains. (d) Impact of a vps3 deletion on the composition of CORVET in Ustilago maydis. AB31 vps3-3xHA, AB31 vps8-3xMyc, and AB31 vps8-3xMyc ∆vps3 strains were analyzed by Co-IP/MS experiments in two independent biological replicates. As control, the untagged AB31 wildtype strain and either Anti-HA coupled magnetic beads or Anti-Myc coupled agarose beads were used. MS-data were analyzed with MaxQuant 1.6.0.16 (Tyanova et al., 2016) (https://maxqu ant.org). Given are peptide counts after subtraction of nonspecifically enriched peptides from control samples. Threshold for peptide counts: >5 peptides in at least one of the biological replicates. Values for AB31 vps3-3xHA, AB31 vps8-3xMyc are also given in Table S1. The color range from red (45) to grey (0) indicates the % of cumulative peptide counts. (e) Deletion of vps3 (lower panel) in AB31 results in localization of Rab5a signals (GFP) to bidirectional motile EEs, but also result in enhanced localization at static structures when compared with wildtype AB31 (upper panel). Kymographs: horizontal scale 10 µm, vertical scale 10 s detect any motility of Nma1 in vps3 deletion strains (AB31∆vps3 nma1-3xgfp; Figure 5d). Consistently, in Co-IP/MS experiments with Vps8-3xMyc, deletion of vps3 abolished the co-purification of Nma1 together with conserved CORVET components (Figures 3d and 6a), indicating that Vps3 is required for the interaction of Nma1 with CORVET. A suggested direct interaction between Nma1 and Vps3 was further confirmed by yeast two-hybrid analysis: a 184 aa C-terminal fragment of Nma1 fused to the Gal-AD was tested in combination with either Vps3, Vps11, and Vps18 fused to Gal-DB, and only for the combination of Nma1 and Vps3 activation of the reporter genes was observed (Figure 6b). We also observed a delocalization of Vps8 in ∆nma1 hyphae: in AB31 vps8-3xmCherry∆nma1, Vps8 signals accumulated at the hyphal tip and the septum, and only few motile signals were observed in between (Figure 7c). The motility of Rab5a was not significantly altered in the ∆nma1 strain AB31∆nma1 gfp-rab5a vps3-mKate2.

|
Reminiscent to the cellular distribution of Vps8, we also observed an accumulation of Rab5a at the septum and hyphal tip ( Figure S10).

| DISCUSS ION
For the function and maturation of EEs, tethering and fusion of membrane compartments is an essential process. Tethering is regulated via multi protein complexes such as the CORVET complex, which interacts specifically with Rab5 GTPases (reviewed in Ungermann & Kümmel, 2019). Similar to S. cerevisiae and A. nidulans (Abenza et al., 2010;Singer-Krüger et al., 1994) one of the Rab5 proteins, Rab5a, plays a more prominent role with respect to function of EEs (Figures 2 and S3). The isoforms of the Ypt5 family of At least some of the functions of Rab5a (with respect to endosomal movement) can be overtaken by Rab5b (Figure 2d). The redundant function of the two Rab5 GTPases is further supported by the lethal phenotype of the double deletion in U. maydis, and, similarly, in A. nidulans (Abenza et al., 2010). In S. cerevisiae, the double deletion of vps21/ypt52 (equivalent to Rab5a/Rab5b) is viable (Singer-Krüger et al., 1994), which can be explained by the nonessential role of endocytosis in yeasts. Rab5c, the third Rab5-GTPase present in the U. maydis genome, has obviously no function related to endosomal movement ( Figure S5). The overall structure of Rab5c with the conserved Rab5 domain covering only a third of the length of the protein ( Figures S1 and S2) may indicate a function independent from the other two Rab5 GTPases in U. maydis.
The organization of the CORVET complex in U. maydis with the six subunits Vps3, Vps8, Vps11, Vps16, Vps18, and Vps33 (Figure 3d) follows the general scheme described for various other organisms, including the well-studied components from S. cerevisiae and mammalian cells (Peplowska et al., 2007;Perini et al., 2014). Exemptions from this conserved organization are spare; for example, for D. melanogaster a "mini-CORVET" is described, which lacks the subunit Vps3 (Lörincz et al., 2016).
Identical to the known mechanisms in yeast for the assembly of the CORVET complex (Pawelec et al., 2010), the CORVET subunit Vps8 is sufficient to recruit the complex to EEs via Rab5, as evi- ∆vps3 strains display a phenotype similar to ∆rab5a strains with severe defects in polar growth, cytokinesis, and movement of EEs ( Figure 3). In yeast, the function of Vps3 is to connect the CORVET complex bound to Rab5a via Vps8 of one EE to Rab5a attached to a second EE to initiate fusion/maturation Pawelec et al., 2010;Peplowska et al., 2007). Apparently, in U. maydis the presence of CORVET on EEs (via Vps8) appears to be more important than its function in vesicle fusion (via Vps3).
This may reflect a prominent function of CORVET with respect to the long distance transport function of EEs. For both A. nidulans and U. maydis, it has been shown that the conserved fused toes (FTS)-Hook-Hook interacting protein (FHIP) complex (FHF) serves as an adapter between EEs and the dynein and kinesin motor proteins instrumental for microtuble-dependent transport of EEs. All three components of the complex are critical for movement of EEs Yao et al., 2014;Zhang et al., 2014).
In A. nidulans, it has been shown that the Hook-EE interaction is mediated by FHIP (Yao et al., 2014), and for the human FHF complex it has been described that it interacts with VPS 16 and VPS18, conserved component of HOPS and CORVET (Xu et al., 2008). In mammalian cells, Rab5 has also been described to interact directly with the FHF complex Guo et al., 2016).

F I G U R E 6
Association of Nma1 with CORVET depends on Vps3. (a) Deletion of vps3 abolishes interaction with CORVET. Co-IP was performed with Nma-3xHA as bait with proteins extracted from strains AB31 nma1-3xHA vps8-3xMyc, AB31∆vps3 nma1-3xHA vps8-3xMyc, and AB31 vps8-3xMyc (control with untagged bait protein). Western blot was probed with antibodies specific for-Nma1-3xHA (128 kDa) Vps8-3xMyc (248 kDa) or actin (42 kDa). Vps8-3xMyc (marker protein for CORVET) co-purifies with Nma1-3xHA in wildtype background, while in the vps3 deletion strain Vps8-3xMyc is not co-purified. (b) Nma1 interacts directly with Vps3 in the yeast two-hybrid system. Vps3, Vps11, and Vps18 were expressed from plasmid pGBKT7 (bait) in combination with a N-terminally truncated version of Nma1 (Nma1-C, corresponding to 184 amino acids of the C-terminus of Nma1) expressed from pGADT7 (prey). Transformants were spotted in tenfold dilution steps on control plates (SD-LW) and on plates (SD-LWHA) indicative for protein-protein interaction (Vps3 and Nma1-C). The combination of pGBKT7-Vps3 and the empty pGADT7 vector serves as a control for self-activation by Vps3. No interaction was detectable in strains expressing Nma1-C and Vps11 or Nma1-C and Vps18. Cells transformed with pGBKT7-p53 and pGADT7-T (interaction between p53 and T-antigen) or pGBKT7-Lam and pGADT7-T serve as positive and negative controls In fungi, it appears well feasible that CORVET is required as a connection between FHF and Rab5 to facilitate movement of EEs.
The EE-mediated transport of cargoes as mRNA/polysomes, lipid droplets, peroxisomes, or the endoplasmic reticulum (Baumann et al., 2012;Guimaraes et al., 2015;Lin et al., 2016;Marsalek et al., 2017;Olgeiser et al., 2019;Salogiannis et al., 2021) appears to be more critical in U. maydis than in A. nidulans, where deletion of hookA leads to nonmotile EEs and leads to colonies only slightly more compact than those of the wildtype strain (Zhang et al., 2014), while deletion of hok1 in U. maydis leads to severe growth defects .
Interestingly, deletion of vps3 alters the composition of CORVET in U. maydis: Vps3 was found to be exchanged with Vps39 (Figure 3d), a subunit of the HOPS complex that interacts with Vps11, a coresubunit present in both HOPS and CORVET (Plemel et al., 2011). For mammalian cells, it has been shown that Vps11 has a similar binding affinity for Vps3 and Vps39 (van der Kant et al., 2015); apparently, the unoccupied binding site of Vps11 for Vps3 in ∆vps3 strains is likely to be adopted by Vps39. In S. cerevisiae, such intermediate complexes consisting of Vps8 and Vps39 (i-HOPS) are found during the conversion from CORVET (EEs) to HOPS (late endosomes), and also in A.
nidulans there is genetic evidence supporting the physiological function of such mixed complexes (Lopez-Berges et al., 2017). As Vps39 interacts with Rab7, the GTPase specific for late endosomes, one could expect that the "i-HOPS" complex is able to interact with both early and late endosomes. Indeed, in ∆vps3 strains, Vps8 and Rab5a As emphasized previously, the contribution to long distance transport is an important expansion in the range of functions for EEs in filamentous fungi. However, despite this "novel" function, the basic components of the endocytic machinery appear to be conserved between yeasts and filamentous fungi. We have now identified the Nma1 protein with impact on fusion and maturation of EEs, by that modulating the lifespan of EEs as a prerequisite for long distance transport processes.
In ∆nma1 strains, the number of EEs is significantly reduced  Figure S10). For both A. nidulans and mammalian cells, it has been described that maturation of EEs to late endosomes is coupled to dynein-mediated transport processes directed inwards from the cell poles to the center (Abenza et al., 2012;Jordens et al., 2001).
How can Nma1 affect the number of EEs and the conversion rate to late endosomes? We have shown that Nma1 interacts with the Vps3 protein of CORVET (Figures 3d and 6a,b). Vps3 and Vps8 contact Rab5a-GTPases on membranes of individual endosomes as a first step for membrane fusion. The interaction of Nma1 with Vps3 could now prevent this "bridging" by the CORVET complex, by that decreasing the fusion rate of microtubule-associated EEs and enhancing the time span of EEs. In addition, Nma1 could directly affect the conversion rate from EE to late endosomes: one of the initial steps is the recruitment of the GEF for Rab7 by Rab5 to EEs, leading to an increase of GTP-bound Rab7 in the EE membrane and sequentially to the substitution of Rab5 with Rab7 (Langemeyer et al., 2020). It is conceivable that recruitment of the Rab7-GEF by Rab5a is reduced when Rab5a is "occupied" by the Nma1-blocked CORVET complex, which would further decrease the conversion rate to late endosomes. The finding that Nma1 is associated with microtubles would direct the regulatory effect of Nma1 mostly to the EE population moving on microtuble tracks, that is, exactly the EEs involved in long distance transport for which an extended lifespan would be beneficial. The primary function of Nma1 is to stabilize the population of EEs as a prerequisite for EE-mediated transport processes.
The interaction of Nma1 with Vps3 leads also to a stabilization of the binding of Vps3 to the CORVET core complex (Figure 6d,e). In the context that depletion of Vps3 leads to the formation of an intermediate "i-HOPS" complex, the stabilization of the Vps3-CORVET interaction may also contribute to a decreased conversion rate of EEs to late endosomes.
In yeasts as S. cerevisiae, the function of EEs is mostly restricted to endocytic processes, delivering extracellular cargoes to vacuoles or recycling compartments; in accordance with the spherical cell shape, long distance transport processes play a minor role (reviewed in Mellman, 1996).
In filamentous fungi, EEs adopt a prominent role for long distance transport, which, in comparison with yeast EEs, requires a prolonged life span.
In U. maydis, this process is sustained via Nma1 by modulation endosomal maturation via its influence on the CORVET complex.

| Strains and growth conditions
Escherichia coli strain TOP10 (Invitrogen, Thermofischer, Carlsbad, CA, USA) was used for cloning. Growth conditions and media for the cultivation of E. coli followed the protocols described before (Sambrook et al., 1989). Saccharomyces cerevisiae strain AH109 (Clontech/ Takara Bio, Mountain View, CA, USA) was used for yeast two-hybrid interaction studies. S. cerevisiae cells were grown in minimal medium (SD) supplemented with the amino acids dropout-mix needed for selection, as described in the Clontech/Takara Bio Matchmaker TM GAL4 Two-Hybrid System 3 Manual (https://www.takar abio.com). U. maydis cells were grown in YEPSL (Brachmann et al., 2001), CM (complete medium) supplemented with 1% glucose (CM-Glc) or 1% arabinose (CM-Ara), respectively (Holliday, 1974) at 28°C. Solid media contained 2% agar. The induction of hyphal growth in AB31 derivatives was done as previously described in Brachmann et al. (2001). A. nidulans was grown in MM (minimal medium) (Hill & Kafer, 2001). All U. maydis and A. nidulans strains used in this study are listed in Tables S4 and S5, respectively. Growth rate was measured with OD 600 from three biological replicates and technical duplicates hourly over a time span of 12 hr.

| DNA procedures
Molecular techniques followed established protocols described in Sambrook et al. (1989). DNA from Agarose gels was extracted following the protocol of Vogelstein and Gillespie (1979).
Constructs for gene fusions were subcloned in pCR2.1 TOPO (Invitrogen, ThermoFischer Scientific, Carlsbad, CA, USA) and sequenced before transformation. Constructs for genome modification are listed (U. maydis: Table S7; A. nidulans: Table S8). Sequences of oligonucleotides used for PCR are listed in Table S9. Homologous integration of all constructs was verified by Southern blot analysis (Southern, 1975). Genomic DNA from U. maydis was prepared as described in Zhou et al. (2018). DNA from A. nidulans was isolated according to Timberlake and Marshall (1989) and transformations were carried out as described in Yelton et al. (1984).

| Co-immunoprecipitation, western blot, and LC/MS analysis
Co-immunoprecipitation was carried out in vivo with cultures of U. maydis AB31 (Brachmann et al., 2001) derivates, which were grown in 150 ml CM-Glc to an OD 600 of 0.2 at 28°C on a rotary shaker. Filamentous growth was induced for 7 hr in CM-Ara (Bottin et al., 1996;Brachmann et al., 2001). Cells were washed once with PBS buffer, resuspended in 1 ml PBS, supplemented with "Complete" proteinase inhibitor cocktail (Roche Life Sciences, Penzberg, Germany), frozen in liquid nitrogen, and homogenized in a Retsch Mill MM200 (Retsch, Haan, Germany) for 10 min at 30 Hz. Proteins were separated by SDS-polyacrylamid-gelelectrophorese (PAGE) (Laemmli, 1970)  18 min 2 mA −1 per cm 2 membrane) (Towbin et al., 1979). Proteins were detected using the ECL system, following the protocol pro- For LC/MS analysis, protein samples were trypsin-digested with an in-gel digestion following the protocol in Shevchenko et al. (1996).

| Microscopy, image processing, and quantitative analysis
For microscopic analyses, logarithmically growing U. maydis AB31 (Brachmann et al., 2001) derivatives were grown in 10 ml CM-Glc to an OD 600 of 0.5 at 28°C on a rotary shaker. For the induction of hyphal growth, cells were shifted to CM-Ara for 15-18 hr.
For microscopy of A. nidulans germlings and young hyphae, MM on cover slips was inoculated with a small amount of spores and incubated for 12-18 hr at 37°C. For microscopy with the LSM 800/LSM 900 invers microscopes (Carl Zeiss, Jena, Germany), cells were mixed 1:1 with 4% low gelling agarose (Sigma Agarose Typ VII: Low Gelling Temperature) and transferred to a microscopy chamber (µ-Slide 8 Well Glass Bottom, ibidi GmbH, Gräfelfing, Germany). Endosome motility was recorded in image sequences with 1 min duration, taken with an exposure time of 100 ms. The resulting movies were "bleach corrected" (histogram matching) and converted into kymographs (Plugin kymograph builder) using Fiji ImageJ software (Schindelin et al., 2012(Schindelin et al., , 2015Schneider et al., 2012). The scale of kymographs was calculated with "set scale:" (distance: 9.7674 distance in pixels, known distance 1.00, pixel aspect ratio 1.0, unit micron; time: 10 distance in pixels, known distance 1.00, pixel aspect ratio 1.0, unit s). Quantification of vesicles (number and average size) was performed with 8-bit converted microscopic images of 100 ms exposure time in Fiji ImageJ software (Schindelin et al., 2012(Schindelin et al., , 2015Schneider et al., 2012). Reduction of background signal with "median filter" (process-filters-median, 1 pixel radius) and subtraction of background (process-subtract background, with values rolling = 5 pixels, sliding paraboloid). To sharpen borders of objects laces were reinforced (image-adjustauto local threshold with parameters Bernsen, radius = 15, parameter 1 = 0, parameter 2 = 0, white objects on black background).
Images for quantification of colocalization were processed prior to measurement as follows: Reduction of background signal with "median filter," subtraction of background, and reinforcement of object borders were performed as described above for vesicle quantification. Measurement of signal overlaps with JACoP plugin (Manders1&Manders2 coefficient). The M1&M2 coefficient is summing the intensities of pixels from one channel and dividing the sum by its integrated density. A pixel from the GFP channel is considered as colocalized if it has a nonzero intensity counterpart in the mCherry channel. The M1&M2 coefficients give an estimate of the amount of colocalizing signal from a signal over another, without making any assumption on the stoichiometry it may adopt (Bolte & Cordelieres, 2006).
All image processing, including adjustment of brightness, contrast, and gamma-values was performed with the AxioVision and ZEN software (Carl Zeiss, Jena, Germany), respectively.

ACK N OWLED G M ENTS
Ka.S. and J.K. like to thank Matteo Jurca for providing plasmid pMF5-9g, Lukas Baumann for help with Y2H experiments, Rabea Suhrborg for constructions of the U. maydis strains URS3 and URS6, and Kai Heimel for his helpful comments and suggestions for the manuscript.
Open Access funding enabled and organized by Projekt DEAL.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that supports the findings of this study are available in the supplementary material of this article. Additional, data and material are available from the corresponding author upon reasonable request.