The endocytic TPLATE complex internalizes ubiquitinated plasma membrane cargo

Endocytosis controls the perception of stimuli by modulating protein abundance at the plasma membrane. In plants, clathrin-mediated endocytosis is the most prominent internalization pathway and relies on two multimeric adaptor complexes, the AP-2 and the TPLATE complex (TPC). Ubiquitination is a well-established modification triggering endocytosis of cargo proteins, but how this modification is recognized to initiate the endocytic event remains elusive. Here we show that TASH3, one of the large subunits of TPC, recognizes ubiquitinated cargo at the plasma membrane via its SH3 domain-containing appendage. TASH3 lacking this evolutionary specific appendage modification allows TPC formation but the plants show severely reduced endocytic densities, which correlates with reduced endocytic flux. Moreover, comparative plasma membrane proteomics identified differential accumulation of multiple ubiquitinated cargo proteins for which we confirm altered trafficking. Our findings position TPC as a key player for ubiquitinated cargo internalization, allowing future identification of target proteins under specific stress conditions. Ubiquitination triggers endocytosis of proteins from the plasma membrane. The TASH3 subunit of the TPLATE complex recognizes ubiquitinated proteins via its SH3 domain-containing appendage and initiates their internalization from the plasma membrane.

prolonged heat-shock 35 . TASH3 is one of the two large core TPC subunits whose role and function remain unexplored. In contrast to all large adaptor subunits of the heterotetrameric adaptor complexes, including TPLATE, which all carry a bilobal appendage domain composed of an amino-terminal sandwich domain connected to a C-terminal platform domain, TASH3 has evolutionarily acquired an SH3 domain 36 . Similar to other TPC mutants 4,6,37 , available transfer DNA (T-DNA) insertion lines in TASH3, tash3-1 and tash3-2¸ harbouring the T-DNA at the beginning or the middle part of the gene, respectively, caused gametophytic lethality. Both mutants exhibited more than 40% of fluorescein diacetate (FDA)-negative pollen and backcross experiments failed to transfer the T-DNA to the next generation (Fig. 1a-c and Extended Data Fig. 1a).
In contrast, nosh allowed transfer of the T-DNA via the pollen (Fig. 1a-c and Extended Data Fig. 1a) and the isolation of homozygous mutant plants. The T-DNA in nosh is located C-terminally, in the linker before the SH3 domain (Extended Data Fig. 1b). The nosh seedlings exhibited significantly reduced root and hypocotyl length compared to the wild type ( Fig. 1d-g). Adult nosh plants exhibited smaller rosette leaves, delayed flowering time and early senescence (Extended Data Fig. 1c,d). Uptake of the styryl dye N-(3-triethylammoniumpropyl)-4-(6-(4-(diethylamino) phenyl) hexatrienyl) pyridinium dibromide (FM4-64), a proxy for endocytic flux, revealed a significant reduction in internalization in nosh compared to wild type (Fig. 1h,i). Introducing full-length TASH3-GFP (green fluorescent protein) into nosh restored FM4-64 to wild-type levels for two independent complemented mutant lines (Fig. 1h,i), linking the T-DNA insertion with the reduced endocytic capacity. Given the gametophytic lethality of the other null alleles, nosh, therefore, probably produces a truncated variant of TASH3 lacking the C-terminal SH3 domain. Lack of this SH3 domain is not essential for plant life but causes developmental delay and affects CME capacity.

Overall CME dynamics are delayed in nosh
To properly assess CME in nosh, we crossed TPLATE-GFP, CLATHRIN LIGHT CHAIN 1 (CLC1)-GFP and DYNAMIN-RELATED PROTEIN 1 A (DRP1a)-RFP (red fluorescent protein) into nosh and evaluated their dynamic behaviour at the PM. All three endocytic markers revealed significantly lower densities of endocytic foci in nosh compared to control etiolated hypocotyl cells (Fig. 2a-f). Kymograph analysis revealed significantly prolonged life times of endocytic events at PM in nosh compared to control cells ( Fig. 2g-l). This was most pronounced for TPLATE and CLC1 ( Fig. 2g-j) and to a lesser extent for DRP1a (Fig. 2k,l).
The cytokinetic syntaxin KNOLLE is specifically degraded by CME following cell plate maturation 38,39 . Time-lapse imaging of growing root tips monitored the time that KNOLLE-GFP was present in the dividing cells. The presence of KNOLLE-GFP at the fully formed cell plate in nosh was strongly prolonged compared to wild type (Fig. 2m,n). The lower densities and prolonged life times of endocytic foci in nosh therefore correlate with a significant delay in the removal and degradation of endocytic cargo. motifs with internalization of PM cargo via AP-2 in Arabidopsis 10 . Tyrosine motifs have been implicated in endocytosis of BRASSINOSTEROID INSENSITIVE 1 (BRI1), BORON TRANSPORTER 1 (BOR1) or PIN-FORMED 1 (PIN1) [10][11][12][13] . However, there are no examples in which tyrosine motifs were shown to be essential for internalization 14,15 . Recently, the double NPF motif of Secretory Carrier Membrane Protein 5 (SCAMP5) was reported to be recognized by the AtEH1/Pan1 TPC subunit and shown to be important for its internalization 16 .
Besides linear motifs, reversible post-translational modifications also function in endosomal trafficking of transmembrane proteins. Phosphorylation of NOD26-LIKE INTRINSIC PROTEIN 5;1 (NIP5;1), LYSIN MOTIF-CONTAINING RECEPTOR-LIKE KINASE 5 (LYK5) or FLAGELLIN-SENSITIVE 2 (FLS2) was shown to be a prerequisite for their internalization [17][18][19] . Within the last couple of years, ubiquitin became known as an essential mark for PM protein internalization from the cell surface and consequent degradation in the vacuole 20,21 . Mono-, di-or K63-linked poly-ubiquitin chains, covalently linked to the lysine residues of transmembrane proteins, drive the endosomal sorting of PM cargoes to the trans-Golgi network 22 . There, quality control enables vacuolar sorting of ubiquitinated PM proteins, while de-ubiquitinated cargo recycles back to the PM 21,23,24 . A plethora of transmembrane proteins such as IRON-REGULATED TRANSPORTER 1 (IRT1), FLS2, LYK5, BRI1, PIN2 or BOR1 have been shown to undergo ubiquitin-dependent internalization 14,[25][26][27][28][29][30][31] . Among the endocytic machinery, SH3P2, a member of the family of SH3 DOMAIN-CONTAINING PROTEIN (SH3P) and TARGET OF MYB1-LIKE (TOL) proteins, of which some are located at the PM and interact with TPC, bind ubiquitin 23,[32][33][34] . Mechanistic insight as to how ubiquitinated cargo is recognized at the PM and initiates CME remains, however, elusive.
Here we identified nosh (no SH3), a viable TPC mutant that produces a truncated variant of the large TASH3 subunit lacking the carboxy-terminal SH3 domain. This truncation does not prevent complex formation, yet it severely compromises endocytosis. The SH3 domain of TASH3 binds ubiquitin, which consequently leads to an accumulation of ubiquitinated PM proteins in nosh, a feature which is less apparent in other endocytic mutants. Comparative PM proteomics identified multiple potential PM cargo proteins that differentially accumulate in nosh, including the TRANSMEMBRANE KINASE 1 (TMK1), for which we confirmed reduced internalization in nosh. Taken together, we provide evidence that TPC functions in recognizing and internalizing ubiquitinated cargo proteins via its evolutionary unique SH3-containing appendage domain on TASH3.

A TASH3 mutant, lacking the SH3 domain, is viable
Knockout mutants of any TPC subunit identified so far exhibit male gametophytic lethality, which is inhibiting the use of standard loss-of-function analysis 4,6 . Only two viable TPC subunit mutants were reported to date: twd40-2-3, which reduces the amount of full-length protein present 7 , and WDXM2, a TPLATE isoform which reduces complex stability at room temperature and abolishes TPC function upon Insertion sites of T-DNA alleles identified in this study (tash3-1, tash3-2 and nosh) are indicated by arrows. The structural domains of the protein (body, linker and SH3 domain) are depicted above the gene model. b, Pollen viability assay of tash3-1, tash3-2 and nosh. Pollen was stained with FDA and visualized by confocal microscopy. Scale bar, 20 µm. c, Quantification of the percentage viable and dead pollen from tash3-1 (n = 143), tash3-2 (n = 41) and nosh (n = 43) based on FDA staining. d, Representative images of 5-day-old Col-0 and nosh seedlings light grown on half-MS media without sucrose. Scale bar, 1 cm. e, Box plot graph of 5-day-old Col-0 and nosh seedlings light grown on half-MS media without sucrose, showing that nosh mutant exhibits slower root growth. f, Representative images of 4-day-old Col-0 and nosh seedlings dark grown on half-MS media without sucrose. Scale bar, 1 cm. g, Box plot graph of 4-day-old seedlings of Col-0 and nosh grown on half-MS media without sucrose in the dark, showing that nosh mutants exhibit shorter etiolated hypocotyl length. h, Representative single confocal slices of FM4-64 stained root cells as a proxy for endocytic flux evaluation of Col-0, nosh and two independent complemented mutant lines (TASH3-GFP 1 and 2 in nosh). Scale bar, 10 µm. i, Box plot graph of the intracellular to PM intensity of FM4-64 in Col-0, nosh and two independent complemented nosh lines. Complementation with TASH3-GFP restores the endocytic capacity of the nosh mutant. e,g,i, The top and bottom lines of the box plot represent the 25th and 75th percentiles, the centre line is the median and whiskers are the full data range. Red asterisks mark the outliers in the box plots. Numbers of quantified cells are indicated at the bottom of each graph. The indicated P values were calculated using the two-sided Wilcoxon-signed rank test by comparing mutants to wild type. Article https://doi.org/10.1038/s41477-022-01280-1 Introducing full-length TASH3-GFP into nosh complemented the deviating density and life times, indicating that the observed effects are caused by truncating TASH3 (Extended Data Fig. 2a-d). The foci density and endocytic life time in the nosh complemented lines, visualized using TASH3-GFP, were analogous to those observed in the complemented tash3-1 mutant expressing TASH3-GFP and were similar to those obtained using TPLATE-GFP in the complemented tplate mutant (Extended Data Fig. 2a-d). TASH3-GFP and TPLATE-GFP can therefore be interchangeably used as markers to visualize TPC dynamics and full-length TASH3-GFP complements the endocytic defects in nosh.
Col-0 nosh Despite the changes in endocytic density and dynamics, TPLATE-GFP localized preferentially to the PM in the nosh tplate double mutant, with no markable differences compared to the control (TPLATE-GFP in tplate) (Fig. 3a,b). We evaluated TPC assembly in nosh, by affinity purification coupled with mass spectrometry (AP-MS) using TPLATE as bait, similarly to what we did for the destabilized TPLATE isoform WDXM2 (ref. 35 ). We observed significant enrichment of AtEH1/Pan1 and AtEH2/ Pan1 proteins alongside clathrin light-and heavy-chain proteins when we compared the complemented tplate mutant line to the nosh tplate double mutant ( Fig. 3c and Supplementary Data). However, relative to the bait, the average intensity of shared peptides for all other TPC subunits was similar between nosh tplate compared to the original TPLATE-GFP complemented line (Extended Data Fig. 3a). These results imply that in nosh, a stable hexameric TPC forms but that the AtEH/Pan1 proteins associate less strongly with the complex and fail to copurify, similar to what was observed for the C-terminal truncation of TML 4 . To validate if the nosh mutation affects the localization of AtEH/Pan1, we performed immunolocalization using a self-raised antibody against the C terminus of AtEH1/Pan1 (Extended Data Fig. 3b-d). Similarly to the control, AtEH1/Pan1 localized at PM in nosh (Fig. 3d). The coverage of TASH3 peptides identified by MS in complemented tplate mutants versus the nosh tplate double mutant allowed us to characterize NOSH. Whereas peptides covered almost the entire length of TASH3, in the tplate background, peptides from the C-terminal part of TASH3 were absent in the nosh tplate double mutant line (Extended Data Fig. 4a). This confirmed our prediction that a truncated TASH3 protein without the SH3 domain is being produced in nosh. Taken together, we confirmed that NOSH lacks the SH3 domain and that it is integrated into a TPC where the interactions between the hexameric complex and the AtEH/Pan1 proteins are weakened.
Loose association of AtEH/Pan1 subunits with the hexameric TPC resembles the TSET complex from Dictyostelium 5 . Moreover, the partial functionality of TPC lacking an SH3 domain is expected given that Dictyostelium TASH3 (TSAUCER) also lacks this domain (Extended Data Fig. 4b). Phylogenetic analysis of TASH3 from different eukaryotic species revealed that this domain firstly appeared in a common ancestor of Chlorophytes and Streptophytes, while in the other eukaryotic supergroups like Excavata, Apusozoa or Amoebozoa, only TSAUCER without SH3 domain is present (Extended Data Fig. 4c). Interestingly, in the TASH3 sequences of some Chlorophytes we identified TASH3 homologues possessing two SH3 domains (Extended Data Fig. 4c), pointing to potential divergence and specialized functions of these domains. To unravel why Archaeplastida did evolve a TPC containing an SH3 domain, we examined its function.

The SH3 domain of TASH3 binds ubiquitin
On the basis of the available literature, SH3 domains carry out a variety of activities and interactions. Among all, two well-characterized roles are recognizing proline-rich regions (PRR, defined as PXXP) 40,41 and binding ubiquitin molecules 42,43 . TPC containing NOSH has reduced association with the AtEH/Pan1 subunits and these contain several PRR (Extended Data Fig. 5a). We therefore tested interactions between AtEH/Pan1 subunits and different TASH3 truncated variants: TASH3 full-length, the body part of TASH3 and the linker and SH3 domain of TASH3. Overexpression of AtEH/Pan1 in Nicotiana benthamiana can be used to perform proteinprotein interaction analysis via partitioning 16,37 . In contrast to single infiltrations (Extended Data Fig. 5b-d), both full-length TASH3-GFP and TASH3_body-GFP partitioned together with AtEH1/Pan1-mCherry or AtEH2/Pan1-mCherry whereas mCherry-TASH3_linker_SH3 did not (Extended Data Fig. 5e-l). This strongly suggests that the SH3 domain of TASH3 has no binding capacity for the AtEH/Pan1 subunits. Thus, the detachment of AtEH/Pan1 subunits from the hexameric TPC observed in nosh is not caused by the absence of the SH3 domain.
Besides recognizing PRR, some SH3 domains also bind ubiquitin 42 . To assess, if the SH3 domain of TASH3 binds ubiquitin, we structurally aligned our SH3 domain with that of yeast Sla1p, for which the ubiquitin-binding residues have been experimentally determined 42 . We focused on residues forming the binding interface (defined as Sla1p residues within a distance of 0.3 nm from the ubiquitin molecule). Despite the F1194Y mutation in TASH3 SH3, the other ubiquitin-interacting residues (Y1145, E1154, W1172 and P1191) of Sla1p SH3 and TASH3 SH3 are well conserved. Moreover, the tyrosine in the TASH3 SH3 domain might mediate the interaction analogously as the phenylalanine in Sla1p (Fig. 4a). Further, we performed in silico docking and folding of the SH3 domain and a single UBQ10 molecule from Arabidopsis. Three different algorithms for protein-protein docking and one for protein folding, positioned the ubiquitin molecule and the SH3 domain almost identically (Fig. 4b). As a control, we performed docking of the TASH3 SH3 domain with ATG8a, which has a ubiquitin-like structure and size 44 . Using two different algorithms, docking of ATG8a did not result in a single orientation/binding with any of them (Extended Data Fig. 6). These results corroborate the specificity of the docking approach and indicate a preference of the TASH3 SH3 domain towards ubiquitin.
To show direct ubiquitin-binding, we heterologously expressed and purified the SH3 domain of TASH3 in Escherichia coli (Extended Data Fig. 7a-j). Co-immunoprecipitation assays with the ubiquitin moiety from total cell extracts of Arabidopsis seedlings showed that the SH3 domain can bind ubiquitinated proteins (Extended Data Fig.  7k). Further, we analysed the specificity of the SH3 domain to different ubiquitin molecules that are present in plants: mono-ubiquitin, K11-, K48-and K63-linked ubiquitin chains using co-immunoprecipitation. In contrast to empty beads, the SH3 domain clearly bound K11-, K48-and K63-linked tetra-ubiquitin chains (Fig. 4c), while it did not bind mono-ubiquitin (Fig. 4c). These results suggest a preference for poly-ubiquitin but, in vitro, no clear specificity of the TASH3 SH3 domain to differently linked chains could be determined.

Ubiquitinated PM proteins accumulate in nosh
Ubiquitin is an anchoring molecule, which marks PM cargo for internalization 45 . The lack of an SH3 domain in TPC containing NOSH could prevent the internalization of ubiquitinated proteins at PM. To investigate  this, we extracted PM fractions from wild type and several other viable endocytic mutants with documented defects in CME 7,35,46 . Western blot analysis of PM fractions from various endocytic mutants clearly revealed an increased amount of ubiquitin signal in the PM fraction of nosh compared to the wild-type control (Col-0) but also compared to twd40-2-3, WDXM2 and ap2m-2 mutants (Fig. 5a,b). We evaluated the  purity of the extracted PM fraction by probing all fractions with specific antibodies. We detected a clear enrichment of Aquaporin PIP2;7 (PIP2;7, a PM marker), while we could not detect the presence of cytosolic ascorbate peroxidase (cAPX, a cytosol marker) or cytochrome c (CytC, a mitochondrial marker) in the PM fraction (Extended Data Fig. 8).
To identify which PM proteins accumulate in nosh, we extracted the PM fraction from dark-grown Col-0 and nosh seedlings and performed MS analysis. We chose the dark treatment as it was previously shown to induce ubiquitination and degradation of proteins 47 . From all identified and significantly enriched proteins in nosh compared to Col-0, almost 37% of them were annotated as PM proteins ( Fig. 5c and Supplementary Data). Further, comparing our significantly enriched PM proteins with published databases of ubiquitinated proteins 48-54 , we identified 38% of enriched PM proteins previously shown to be ubiquitinated (Fig. 5c). These results suggest that deletion of the SH3 domain of TASH3 causes accumulation of ubiquitinated PM proteins and that this is not a feature that correlates with an overall reduction of endocytic capacity.

TMK1 protein exhibits altered dynamics in nosh
To validate the obtained data from our MS analysis, we decided to focus on the receptor protein kinase TMK1, one of the PM proteins that was significantly overaccumulated in the PM fraction of nosh compared to the wild-type control Col-0 (Supplementary Data). We chose TMK1, as it was previously shown to be ubiquitinated in etiolated seedlings 52 . We crossed pTMK1::TMK1-GFP (in tmk1-1) expressing plants with nosh and analysed the localization of TMK1-GFP in the nosh tmk1-1 double mutant background. The ratio between the endosomal and PM signal for TMK1-GFP in the nosh tmk1-1 double mutant shifted towards PM compared to the complemented tmk1-1 mutant (Fig. 5d,e) and the vacuolar accumulation of GFP, visualized after prolonged dark treatment, was significantly reduced in nosh tmk1-1 compared to the tmk1-1 mutant ( Fig. 5f,g). Both the increased PM and reduced vacuolar accumulation support diminished internalization of TMK1 in nosh tmk1-1, confirming our PM proteomics results.
Further, we evaluated whether TMK1 was differentially post-translationally modified by ubiquitination in nosh. Immunoprecipitated TMK1-GFP from the tmk1-1 complemented mutant and the nosh tmk1-1 double mutant, probed with the general P4D1 anti-ubiquitin antibody, visualized substantially more ubiquitinated TMK1 in nosh tmk1-1 compared to the control (Fig. 5h, white arrowheads; lower bands on the ubiquitin blot probably represent ubiquitinated degradation products of TMK1-GFP). These results confirmed that the absence of the SH3 domain of TASH3 reduces internalization and causes accumulation of ubiquitinated cargo at PM.
To evaluate whether nosh is more affected in ubiquitinated cargo internalization than in general endocytic flux, we combined prolonged dark treatment with FM4-64 staining. We measured the Ratio of the intracellular to plasma membrane signal Di erence (TPL-GFP (nosh/tplate) versus TPL-GFP (tplate)) n = 159 n = 163

The TASH3 SH3 domain recognizes ubiquitinated cargo
Degradation of KNOLLE from the cell plate in nosh is delayed compared to the control (Fig. 2m,n). To correlate the delay in KNOLLE degradation with the inability of TPC containing NOSH to recognize ubiquitinated cargo, we checked whether KNOLLE is ubiquitinated. We probed immunoprecipitated KNOLLE-GFP with a general anti-ubiquitin (P4D1) antibody and observed a specific band representing ubiquitinated KNOLLE-GFP (Fig. 6a). Delayed degradation and ubiquitination of KNOLLE are therefore in agreement with the inability of TPC in nosh to properly recognize ubiquitinated PM cargo.
To investigate if TPC exclusively interacts with ubiquitinated cargo, we compared pBRI1::BRI1-mCitrine with the ubiquitin-dead pBRI1::BRI1-25KR-mCitrine line, which is functional, yet mutated in 25 lysine residues 55 . TPLATE copurified similarly with BRI1-mCitrine and Docking algorithms HDOCK AlphaFold2 ClusPro ZDOCK a b Empty beads HIS- Silver stain  conserved amino acid residues (dark purple). Ubiquitin is shown in the ribbon representation. The different colours depicted for Ubiquitin 10 represent the best-scoring model from the respective modelling algorithms. c, Coimmunoprecipitation assay using the purified recombinant HIS-TEV-SH3 domain from TASH3 or the empty beads control with different forms of ubiquitin (monoubiquitin, K11-linked tetra-ubiquitin, K48-linked tetra-ubiquitin and K63-linked tetra-ubiquitin). The assay was visualized via SDS-PAGE and subsequent silver staining. Compared to the controls, the SH3 domain binds more K11-, K48-and K63-linked tetra-ubiquitin. I, input; FT, flow through; B, beads fraction. Below the blot, the stain-free gel loading control is shown to verify that equal amounts of extracted PM proteins were loaded. b, Box plot graph quantifying the normalized intensity of ubiquitinated PM proteins of the different endocytic mutants. c, SUBA depicted localization of the proteins enriched in the PM fraction of dark-grown nosh seedlings compared to Col-0. d, Representative single-slice confocal images of TMK1-GFP in tmk1-1 and nosh tmk1-1. e, Box plot graph of the intracellular to PM intensity of TMK1-GFP signal in tmk1-1 and nosh tmk1-1 mutant. f, Representative single-slice confocal images of TMK1-GFP in the tmk1-1 and the nosh tmk1-1 double mutant after 12 h of dark treatment. White arrowheads mark GFP signal from TMK1 degradation in the vacuole. g, Box plot graph of the intracellular to PM signal intensity of TMK1-GFP in tmk1-1 and nosh tmk1-1. h, Detection of ubiquitination levels of TMK1-GFP in the tmk1-1 and nosh tmk1-1 backgrounds and quantification of band intensities (normalized to TMK1-GFP in tmk1-1) of three independent repetitions (GFP-band marked by black arrowheads and the higher MW band representing the ubiquitinated full-length TMK1-GFP marked by white arrowheads). i, Representative singleslice confocal images of TMK1-GFP in tmk1-1 and in the nosh tmk1-1 mutant after 12 h of dark treatment and 30 min of FM4-64 treatment. j, Paired slope graph of the intracellular to PM signal intensity of TMK1-GFP and FM4-64 measured from the same cells in tmk1-1 and nosh tmk1-1. Letters represent a two-sided mixed linear model statistic used to determine the difference between GFP/FM4-64 signal in the tmk1-1 and in the nosh tmk1-1 background. Numbers of quantified cells are indicated at the bottom of each graph. The top and bottom lines of box plots represent the 25th and 75th percentiles, the centre line is the median and whiskers are the full data range. Red asterisks mark the outliers. For graphs in b,e and g the indicated P values were calculated using the two-sided Wilcoxonsigned rank test by comparing mutant to control. No adjustment for multiple comparisons was performed. Scale bar, 10 µm. Article https://doi.org/10.1038/s41477-022-01280-1 BRI1-25KR-mCitrine (Fig. 6b), indicating that ubiquitin is not the sole mechanism linking TPC to PM cargo. Next, we co-immunoprecipitated pBRI1::BRI1-mCitrine and pBRI1::BRI1-25KR-mCitrine line in the presence of recombinant HIS-TEV-SH3. The TASH3 SH3 domain specifically copurified with BRI1-mCitrine (Fig. 6c), providing direct evidence that the TASH3 SH3 domain functions in ubiquitinated cargo recognition. Ratio of the intracellular to plasma membrane signal   57 . It is tempting to speculate that the reduced density of endocytic foci in nosh represents a failure to recognize a distinct subset of cargo proteins due to the absence of the SH3 domain. The delayed CME observed in nosh correlates with changes in TPC composition. AP-MS analysis revealed formation of the hexameric complex, while both AtEH/Pan1 subunits associated less strongly with the hexameric core in nosh compared to the control. Immunolocalization revealed no visible difference in AtEH1/Pan1 PM localization in nosh compared to wild type. Together with the observation that AtEH1/ Pan1 remains at the PM upon temperature-dependent inactivation of TPLATE 35 , our results indicate that AtEH1/Pan1 proteins do not need other TPC subunits to localize at the PM and they probably do so via their anionic phospholipid and protein binding EH domains 16 . Moreover, relative to the bait, we identified similar levels of all hexameric core subunits, indicating that comparable amounts of TPC are present in Col-0 and nosh. This differs from our previous results with WDXM2 as bait, where an imbalance between the amount of bait protein versus the other core TPC subunits strongly suggested a reduced amount of TPC units 35 .
TPC organization in nosh therefore resembles the situation in the slime mould Dictyostelium discoideum, which contains a similar, yet hexameric complex, called TSET, lacking AtEH/Pan1-like subunits 4,5 . Our previous work indicated the µHD of TML as a major factor for stabilizing the interactions between the hexameric core and the AtEH/ Pan1 subunits. Dictyostelium TSET does not contain a µHD, which can explain the absence of robust interactions with AtEH/Pan1 4,5,36 . Interactions between TASH3 and AtEH/Pan1 are independent from the SH3 domain, indicating that the reason why AtEH/Pan1 detaches from TPC containing NOSH is caused by another mechanism. This could include structural destabilization by NOSH or even the absence of an unknown stabilizing factor that connects to TPC via a binding motif in the linker region that is partially truncated in nosh.
Even though the SH3 domain of TASH3 does not have an obvious function to stabilize the different TPC subunits, it obviously has an important, albeit not essential, function that is underlined by all the observed phenotypes. The SH3-containing appendage domain is a modification which is absent in any other large core subunit from other heterotetrameric adaptor complexes containing coat family members. The SH3 domain first appeared as a part of TASH3 in Archaeplastida. Moreover, several species of Chlorophytes possess TASH3 isoforms containing two SH3 domains. This might suggest that SH3 domains underwent specialization for different cellular functions. Establishment of a singular SH3 domain in TASH3 throughout evolution indicates the divergence of these domains and potentially roles for independent proteins possessing SH3 domains. Compared to animal genomes possessing more than hundred SH3 domain-containing proteins, there are only five such proteins identified in Arabidopsis: three members of the SH3P family, TASH3 and an SH3 domain-containing protein (AT4G39020). Expression of the last one has not been confirmed by any recent single cell dataset, indicating that it probably has no major function [58][59][60] .
Ubiquitination is involved in internalization and subsequent degradation of PM proteins in plants 21 . Different types of ubiquitin attachments can induce different fates for the cargo proteins. Mono-ubiquitination of IRT1 triggers its internalization from PM 27 , while degradation of BOR1, IRT1, BRI1 or PIN2 requires K63-linked poly-ubiquitination 26,61-64 . Degradation of ubiquitinated PM cargo is negatively regulated by de-ubiquitinating proteins, as shown recently for BRI1, thus negatively modulating its vacuolar targeting 65 . However, the evidence of ubiquitinated cargo recognition at PM by endocytic machinery is still missing. In the animal field, a subfamily of clathrin-associated sorting proteins (CLASPs) epsin 1, epidermal growth factor pathway substrate clone 15 (EPS15) and Eps15-related protein (EPS15-R), cargo-specific monomeric endocytic adaptor proteins, contain ubiquitin-interacting motifs that have been implicated in the recognition of ubiquitinated cargo [66][67][68][69][70][71]  proteins, however, do not contain these conserved ubiquitin-binding domains, suggesting that ubiquitinated proteins need to be recognized before internalization by other mechanisms.
Here, we suggest that this role can be performed by the SH3 domain of TASH3. We show that it binds different ubiquitin-linked chains and PM fractions of nosh contain the highest levels of ubiquitinated proteins compared to other endocytic mutants. Moreover, the SH3 domain of TASH3 specifically copurifies with BRI1 and not with the ubiquitin-dead BRI1-25KR isoform. Together with the in vitro data, this strongly favours the SH3 domain of TASH3 as a poly-ubiquitin-binding domain, recognizing cargo at the PM. However, it is probably not the only one. Proteins associated with the endocytic machinery such as TARGET OF MYB1 (TOM1)-LIKE (TOL) or SH3P2 are also capable of binding ubiquitin 23,32,33 . A role of SH3P in recognizing ubiquitinated cargo would be in agreement with the recent finding that nosh enhanced the plant developmental defects of a triple sh3p mutant, indicating that these proteins have a non-redundant role in plant development 72 . In mammalian cells, WD40-repeat β-propellers also bind ubiquitin 73 , which suggests that two TWD40 subunits of TPC might potentially also play a role in this. Another group of ubiquitin-recognizing proteins is a family of AMSH-like proteins, which can bind K63-and K48-linked poly-ubiquitin chains. AMSH proteins were shown to participate in autophagic degradation as well as vacuolar degradation of endocytic cargo [74][75][76][77] . AMSH3 protein can also associate with SH3P2, which was shown to also participate in autophagic degradation 23 . Recently, TPC and especially AtEH/Pan1 subunits have been shown to participate in autophagosome formation at endoplastic reticulum-PM contact sites 37 , therefore the role of the TASH3 SH3 domain in autophagic degradation cannot be completely excluded.
In conclusion, the addition of an SH3 domain to the large TPC subunit TASH3 represents a unique evolutionary plant synapomorphy. Our study reveals that this adaptation modulates endocytosis in plants by recognizing poly-ubiquitinated cargo at the PM destined for degradation. As such, it provides the first mechanistic insight, to our knowledge, into adaptor-complex-dependent ubiquitin-mediated cargo recognition leading to endocytic initiation in plants.

Molecular cloning
Primers used for constructing all the constructs in this manuscript are present in the Supplementary Data. To yield the expression construct for full-length TASH3-GFP, entry clones of TASH3 without a stop codon in pDONR221 (ref. 4 ), pDONRP4-P1R-Histone3p 78 and pDONRP2-P3R-GFP 79 were combined with pB7m34GW 79 in a triple gateway reaction.
To create the TASH3_body-GFP construct, TASH3_body without stop codon was amplified from the full-length TASH3 entry clone in pDONR221 using AtTASH3_Body_221_Fw and AtTASH3_ Body_221_Rv primers, cloned in pDONR221 and combined with pDONRP4-P1R-Histone3p 78 , pDONRP2-P3R-GFP 79 and pB7m34GW 79 in a triple gateway reaction.
The SH3 domain of TASH3 was amplified from TASH3 without a stop codon in pDONR221 (ref. 4 ) using SH3domain_Fwd and SH3do-main_Rv primers. The polymerase chain reaction (PCR) fragment was cloned into the pET22b plasmid (Novogen) by restriction digestion (Nde/Xho). The final construct has an N-terminal HIS-tag followed by a TEV protease cleavage site and contains amino acids 1,136-1,198 of TASH3.

destination vector in a Golden Gate reaction.
To yield the vector for antibody production, the expression plasmid peAtEH1-Pan1c-term for heterologous expression of the C-terminal part of AtEH1/Pan1 with a 6× HIS-tag in E. coli was generated via restriction digest-mediated (NdeI/BamHI) cloning into the pLT32 expression vector 84 . The pDONR221-AtEH1_no_stop 4 plasmid was used as a template for PCR with primers EH_atb_Fw and EH_atb_Rv.
For backcross experiments, heterozygous mutant plants of tash3-1, tash3-2 and nosh were used as male to cross with Col-0 as female. The transfer of the T-DNA, requiring the functionality of the complementing fusion construct, was analysed by genotyping PCR on F 1 plants.
Homozygous tplate mutant plants carrying pLat52::TPLATE-GFP were crossed with nosh. N. benthamiana plants were grown in a greenhouse under long-day conditions (6:00 -22:00 h light, 100 photosynthetically active radiation, 21 °C) in soil (Saniflo Osmocote pro NPK: 16-11-10 + magnesium and trace elements). Transient expression was performed by leaf infiltration according to ref. 92 . An optical density of 0.5 of Agrobacterium strains was used for all constructs during co-expression. Transiently transformed N. benthamiana epidermal leaf cells were imaged 2-3 d after infiltration.

Root and hypocotyl growth
Arabidopsis seedlings were grown on half-strength Murashige and Skoog (MS) medium without sucrose. Plates with seeds were stratified for 48 h at 4 °C and then placed at 21 °C in continuous light. For root growth analysis, seedlings were grown in continuous light for 5 d. At 2 d after transfer to the light, seeds that did not germinate or that grew into the agar were marked and excluded from further analysis. Root length measurements were carried out with the Fiji 93 software package. For hypocotyl analysis, following stratification, plates were transferred to continuous light for at least 3 h and afterwards covered in aluminium foil. Plates were kept covered for another 4 d. Hypocotyl length measurements were carried out with the Fiji 93 software package.

Phylogenic analysis
To identify TASH3 homologues, predicted proteins of selected genomes from the Joint Genome Institute database (https://genome.jgi.doe.gov/ portal/) were searched using the BLASTP algorithm 94 with Arabidopsis Article https://doi.org/10.1038/s41477-022-01280-1 TASH3 as input sequence. TSAUCER sequences were obtained from ref. 5 . The SMART database 95 was used to decipher the presence of the SH3 domain. A multiple sequence alignment was constructed with the MAFFT algorithm in the einsi mode 96 (Supplementary Alignment File, a Jalview compatible multiple alignment file). The phylogenetic analysis was carried out using PhyML v.3.0 (ref. 97 ) with the smart model selection 98 . The phylogenetic tree was visualized using iTOL v.6 (ref. 99 ).

Docking
Structures of the Arabidopsis TASH3_SH3 domain (Uniprot code F4IL68, amino acids 1,137-1,198), the Arabidopsis Ubiquitin10 (Uniprot code Q8H159) and the Arabidopsis ATG8a (Uniprot code Q8LEM4) were downloaded from the AlphaFold2 structure database 100 . Four different modelling algorithms were used to position the TASH3_SH3 domain to either Ubiquitin10 or ATG8a. Namely, we used the Alpha-Fold2 algorithm as implemented in ColabFold 101 , ClusPro 102 , HDOCK 103 and ZDOCK 104 . The best-scoring models were analysed and visualized in the ChimeraX programme 105 . To analyse amino acid conservation, we used the Consurf server 106 .

SH3 domain purification
The pET22b-HIS-TEV-SH3 construct was transformed into BL21(DE3) (no. C2527H, NEB). Cells were grown at 37 °C in LB+ medium and induced by adding 0.4 mM isopropyl-d-thiogalactoside at optical density 0.6 for 5 h. Proteins were extracted using sonication in 20 mM HEPES pH 7.4, 150 mM NaCl, 10 mM imidazole and protease inhibitors (cOmplete Mini Protease Inhibitor Cocktail, Roche). Purification was performed on an ÄKTA purifier (GE Healthcare) system by subsequent purification using immobilized metal affinity chromatography (IMAC) via a HisPrep FF 16/10 column (GE Healthcare) using 20 mM HEPES pH 7.4, 150 mM NaCl, 10 mM imidazole as binding buffer and 20 mM HEPES pH 7.4, 150 mM NaCl, 500 mM imidazole as elution buffer. The first size exclusion chromatography (SEC) step was performed using a HiLoad 16/600 Superdex 75 pg column (GE Healthcare) using 20 mM HEPES pH 7.4, 150 mM NaCl as elution buffer. When no HIS-tag was required, the protein was incubated overnight with 1:40 protein:HIS-TEV protease (own production) 107 at room temperature without shaking. Uncleaved protein and protease were removed via reverse IMAC using a 1 ml HisTrap FF column (GE Healthcare) with the same buffers as in the first IMAC step followed by SEC using a HiLoad 16/600 Superdex 75 pg column (GE Healthcare) with the same buffer as in the first SEC step. The yield was ~0.250 mg l −1 culture after all purification steps. The protein sequence of the SH3 domain along with its native molecular weight was verified by MS analysis. The HIS-HRV3C-GFP control (Ray Owens, OPPF) on an OPINF backbone 16 was produced and purified analogously to the TASH3 SH3 domain.

SH3 ubiquitin-binding assays
For the SH3 ubiquitin-binding assays, 100 µg of HIS-TEV-SH3 and 326 µg of HIS-HRV3C-GFP were covalently coupled via their primary amines to 3 mg Pierce NHS-Activated Agarose (Thermo Scientific). This was done by incubating the recombinant proteins with the beads in purification buffer: 20 mM HEPES pH 7.4, 150 mM NaCl and protease inhibitors (cOmplete Mini Protease Inhibitor Cocktail, Roche) rotating at room temperature for 1 h. For the empty beads control, recombinant protein was omitted. The coupling efficiency of HIS-TEV-SH3 and HIS-HRV3C-GFP was determined via a Coomassie-stained SDS-polyacrylamide gel electrophoresis (SDS-PAGE) gel by comparing input and flow through. The beads were washed three times with purification buffer. Subsequently, free binding spots on the beads were quenched with 1 M Tris-HCl pH 7.4 for 20 min of rotating at room temperature. The beads were washed with binding buffer: 20 mM HEPES pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , 0.05% Tween-20 and then incubated with 10 µg ml −1 of tetra-ubiquitin (K11-, K48-and K63-linked ubiquitin, Biotechne/ R&Dsytems) or 40 µg ml −1 of mono-ubiquitin (Biotechne-R&Dsystems) in binding buffer, rotating at 4 °C for 2 h. The bound fraction was eluted by incubating at 70 °C for 10 min in 1× Laemmli buffer (BioRad) and 1× NuPage reducing agent (Invitrogen). Input, flow through and bound fractions were assessed via SDS-PAGE using a 4-20% SDS-PAGE TGX gel (BioRad) and subsequent silver staining using the Pierce Silver Stain Kit (Thermo Fisher Scientific).
For the SH3 ubiquitinated proteins pulldown from Arabidopsis thaliana Col-0 extracts, 100 µg of SH3 (TEV-cleaved) and 100 µg of HIS-HRV3C-GFP were coupled as described above. Six extracts from A. thaliana Col-0 were prepared by flash-freezing and grinding the material in liquid nitrogen. The material was then incubated in a 1:1 ratio with protein extraction buffer: 50 mM Tris pH 7.6, 150 mM NaCl, 5 mM dithiothreitol, 1 mM phenylmethyl sulfonyl fluoride and protease inhibitors (cOmplete Mini Protease Inhibitor Cocktail, Roche) and allowed to rotate for 30 min at 4 °C. Subsequently, the supernatant was cleared by centrifugation at 20,000g at 4 °C for 20 min. Protein concentrations from the extracts were measured using the Qubit Protein assay kit (Thermo Fisher). Equal concentrations of the extracts were incubated with the beads rotating at 4 °C for 2 h. The beads were washed three times with protein extraction buffer. The bound fraction was eluted by incubating at 70 °C for 10 min in 1× Laemmli buffer (BioRad) and 1× NuPage reducing agent (Invitrogen). Input, flow through and bound fractions from different replicates were pooled and assessed via western blot.

Protein expression for antibody production
Protein expression and purification were performed by the VIB protein core facility (https://vib.be/labs/vib-protein-core/research) in two independent runs. One batch was used for rabbit immunization (Extended Data Fig. 3b) and one batch was used for antibody purification from the obtained serum (Extended Data Fig. 3c). Protein expression was induced via isopropyl-d-thiogalactoside; afterwards the cells were pelleted and resuspended in lysis buffer (50 mM HEPES pH 7.4; NaCl; 20 mM imidazole; 1 tablet per 50 ml of cOmplete Mini Protease Inhibitor Cocktail, Roche; 1 mg 100 ml −1 of DNase1) followed by sonication. The sonicated solution was pelleted twice for 30 min at 10,000 and 18,000 r.p.m. at 4 °C to separate the cell lysate from the debris. The lysate was subsequently filtered with a 0.22 µm filter and the protein was purified on an Äkta purifier (GE Healthcare) in three independent steps. First, the protein was purified via IMAC using a HisPrep FF 16/10 GE 20 ml column. After applying the lysate, the column was washed with 20 column volumes (CV) of IMAC buffer (50 mM HEPES pH 7.4; 20 mM NaCl; 20 mM imidazole; 0.1% empigen detergent) and then with 5 CV of IMAC buffer containing 50 mM imidazole. The protein was then eluted with IMAC elution buffer (50 mM HEPES pH 7.4; 20 mM NaCl; 400 mM imidazole). The protein was further purified via ion exchange chromatography using Source Q15 resin. Therefore, the IMAC eluate was diluted 1:100 to guarantee binding to the resin. After washing the column with 20 CV of 20 mM Tris-HCl pH 8.0, the protein was eluted via an NaCl gradient increasing the concentration from 0 M to 1 M over 5 CV. Finally, the protein was purified by SEC. SEC was performed with a HiLoad Superdex 200 26/600 prep grade column using PBS buffer. Coomassie-stained gels of the obtained SEC fractions for each run are depicted in Extended Data Fig. 3b,c. For the batch used for rabbit immunization, the fractions 18-23 were pooled (Extended Data Fig.  3b); for the batch used for antibody purification, the fractions 25-29 were pooled (Extended Data Fig. 3c). Both pools were analysed via HPLC using a Superdex 200 increase 10/300 as a quality control (chromatograms in Extended Data Fig. 3b,c).

Polyclonal AtEH1/Pan1 antibody generation
The Biotem antibody service (https://www.biotem-antibody.com/) was used to generate a polyclonal rabbit antibody against the C terminus of AtEH1/Pan1. Rabbits for immunization were selected on the basis of their blood serum responsiveness to the Arabidopsis proteome. Selected rabbits were injected on days 0, 7, 14 and 34 with purified protein concentrated to 0.724 mg ml −1 . After 42 d, the rabbits were exsanguinated and their sera were tested against the antigen. The antibody of the most responsive serum was purified by Biotem against the C terminus of AtEH1/Pan1. The purification (Biotem) was performed using a second batch of recombinant protein concentrated at 3 mg ml −1 .
To test the purified antibody for specificity, 5-day-old seedlings expressing AtEH1/Pan1-mRuby3 in a Col-0 or ateh1 pan1 mutant background 35 as well as Col-0 control seedlings were shock frozen and ground in liquid nitrogen. Subsequently, heated 1.2× Laemmli buffer was added to the plant material followed by thorough vortexing. After 8 min of incubation at 75 °C, samples were centrifuged twice at 20,000g for 1 min at 4 °C. Samples were separated on a 4-20% Gradient TGX SDS stain-free gel (BioRad) and comparable protein amounts between samples were confirmed via stain-free imaging (Extended Data Fig. 3d). The gel was blotted on a nitrocellulose membrane (BioRad) and developed with the purified polyclonal antibody diluted 1:2,000 in PBS-T containing 5% milk powder. The specificity could be confirmed, as Col-0 and the AtEH1/Pan1-mRuby3 (Col-0) lines show the native AtEH1/Pan1 band roughly at 140 kDa, which is not present in the complemented mutant lines (Extended Data Fig. 3d, marked by black arrowheads). Transgenic lines expressing AtEH1/Pan1-mRuby3 in Col-0 and (ateh1 pan1 1-2 −/− backgrounds also exhibited a band for AtEH1/Pan1-mRuby3 at ~175 kDa (Extended Data Fig. 3d, marked by white arrowheads).

Plasma membrane protein extraction
Sample preparation. Arabidopsis seedlings were grown on half-strength MS medium without sucrose at 21 °C in the dark for 4 d. We chose dark germination to induce ubiquitination and degradation of proteins. The tissue samples were flash-frozen and crushed using a liquid cooled mortar and pestle and the crushed material was used to extract the PM fraction using the Minute Plasma Membrane Protein Isolation and Cell Fractionation Kit (Invent Biotechnologies) on the basis of the manufacturer's manual. For liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis of the PM fractions, samples were prepared on the basis of the Mini Urea protocol, using 90 µg of PM proteins, as described in ref. 108 . For the analysis of the fractions by western blot, samples were prepared on the basis of the manufacturer's manual with a slight modification. The pellet of the total membrane fraction and the pellet of the PM fraction were each washed once in the respective buffers that were used to acquire the pellet during that step (Buffer A for total membrane; Buffer B+ PBS for PM). This was followed by an identical centrifugation step. For western blot analysis, samples were prepared as described in the western blot section.

LC-MS/MS analysis.
Purified peptides were redissolved in 25 µl of loading solvent A (0.1% TFA in water/ACN (98:2, v/v)) and 5 µl was injected for LC-MS/MS analysis on an Ultimate 3000 RSLCnano LC (Thermo Fisher Scientific) in-line connected to a Q Exactive mass spectrometer (Thermo Fisher Scientific). Trapping was performed at 10 µl min −1 for 4 min in loading solvent A on a 20 mm trapping column (made inhouse, 100 µm internal diameter, 5 µm beads, C18 Reprosil-HD, Dr. Maisch) and the sample was loaded on a 200 mm analytical column (made inhouse, 75 µm internal diameter, 1.9 µm beads C18 Reprosil-HD, Dr. Maisch). Peptides were eluted by a nonlinear gradient from 5 to 55% MS solvent B (0.1% FA in water/acetonitrile (2:8, v/v)) over 145 min at a constant flow rate of 300 nl min −1 , reaching 99% MS solvent B after 150 min, followed by a 10 min wash with 99% MS solvent B and re-equilibration with MS solvent A (0.1% FA in water). The column temperature was kept constant at 45 °C in a column oven (Butterfly, Phoenix S&T). The MS was operated in data-dependent mode, automatically switching between MS and MS/MS acquisition for the 16 most abundant ion peaks per MS spectrum. Full-scan MS spectra (375-1,500 m/z) were acquired at a resolution of 60,000 in the orbitrap analyser after accumulation to a target value of 3E6. The 16 most intense ions above a threshold value of 2.0 × 10 4 were isolated for fragmentation at a normalized collision energy of 28% after filling the trap at a target value of 1 × 10 5 for maximum 50 ms. MS/MS spectra (200-2,000 m/z) were acquired at a resolution of 15,000 in the orbitrap analyser.

MS data analysis. Data analysis was performed by MaxQuant
(v.1.6.10.43) using the Andromeda search engine with default settings, including a false discovery rate (FDR) set at 1% on both the peptide and protein level. Spectra were searched against the Araport11plus database, consisting of the Araport11_genes.2016.06.pep.fasta downloaded from arabidopsis.org, extended with sequences of all types of Article https://doi.org/10.1038/s41477-022-01280-1 possible contaminants in proteomics experiments in general. These contaminants include the cRAP protein sequences, a list of proteins commonly found in proteomics experiments, which are present either by accident or by unavoidable contamination of protein samples (The Global Proteome Machine, http://www.thegpm.org/crap/). In addition, commonly used tag sequences and typical contaminants, such as sequences derived from the resins or the proteases used, were added. The Araport11plus database contains in total 49,057 sequence entries. The mass tolerance for precursor and fragment ions was set to 4.5 and 20 ppm, respectively, including matching between runs. Enzyme specificity was set as C-terminal to arginine and lysine (trypsin), also allowing cleavage at arginine/lysine-proline bonds with a maximum of two missed cleavages. Fixed modification was set to carbamidomethylation of cysteines. Variable modifications were set to oxidation of methionine residues and acetylation of protein N termini. Proteins were quantified by the MaxLFQ algorithm integrated in the MaxQuant software. Only proteins with at least one unique peptide were retained for identification.
Differential analysis was performed with the Perseus software (v.1.6.7.0) after loading the proteingroups.txt file from MaxQuant, with reverse and contaminant hits removed. Label-free quantitation (LFQ) intensities were log 2 transformed and biological replicate samples were grouped together. Proteins with less than two valid values in at least one group were removed and missing values were imputed from a normal distribution around the detection limit. On the quantified proteins, a t-test was performed for pairwise comparison of the samples. Correction for multiple testing was done by permutation-based FDR, with thresholds FDR = 0.05, S0 = 0.5 or S0 = 0.1. The result is listed in the Supplementary Data.

GFP pulldown
Sample preparation. Samples were prepared as described in ref. 109 . Briefly, samples were lysed in extraction buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.1% NP-40, 1× cOmplete Mini Protease Inhibitor Cocktail, Roche) and sonicated. After centrifugation (15 min at 4 °C at 18,000 r.p.m. or 39,000g) 100 µl of anti-GFP µBeads (µMACS, Miltenyi) and samples were incubated for 2 h at 4 °C. Afterwards, the extracts were run through µMACS Separator columns, washed twice with extraction buffer, washed twice with ABC buffer (50 mM NH 4 HCO 3 in H 2 O) and eluted with 95 °C preheated ABC buffer. For LC-MS/MS analysis, samples were prepared as described in ref. 108 .

LC-MS/MS analysis.
Peptides were redissolved in 20 µl of loading solvent A (0.1% TFA in water/ACN (98:2, v/v)) of which 5 µl was injected for LC-MS/MS analysis on an Ultimate 3000 RSLCnano LC (Thermo Fisher Scientific) in-line connected to a Q Exactive mass spectrometer (Thermo Fisher Scientific). The peptides were first loaded on a trapping column made inhouse, 100 µm internal diameter × 20 mm, 5 µm beads C18 Reprosil-HD, Dr. Maisch) and after flushing from the trapping column the peptides were separated on a 50 cm µPAC column with C18-endcapped functionality (Pharmafluidics) kept at a constant temperature of 50 °C. Peptides were eluted by a linear gradient from 99% solvent A′ (0.1% formic acid in water) to 55% solvent B′ (0.1% formic acid in water/acetonitrile, 20/80 (v/v)) in 30 min at a flow rate of 300 nl min −1 , followed by a 5 min wash reaching 95% solvent B′. The MS was operated in data-dependent, positive ionization mode, automatically switching between MS and MS/MS acquisition for the five most abundant peaks in a given MS spectrum. One MS1 scan (400-2,000 m/z, AGC target 3 × 106 ions, maximum ion injection time 80 ms), acquired at a resolution of 70,000 (at 200 m/z), was followed by up to five tandem MS scans (resolution 17,500 at 200 m/z) of the most intense ions fulfilling predefined selection criteria (AGC target 5 × 104 ions, maximum ion injection time 80 ms, isolation window 2 m/z, fixed first mass 140 m/z, spectrum data type: centroid, intensity threshold 1.3 × 10 4 , exclusion of unassigned, 1, 5-8 and >8 positively charged precursors, peptide match preferred, exclude isotopes on, dynamic exclusion time 12 s).

MS data analysis.
The raw data were searched with MaxQuant (v.1.6.10.43) as described for the PM proteins. Differential analysis was performed with the Perseus software (v.1.6.15.0) with largely the same workflow as described for the PM proteins. Except here, proteins with less than three valid values in at least one group were removed and thresholds FDR = 0.01 and S0 = 1 were applied. The results are shown in the volcano plot in Fig. 3c and listed in the Supplementary Data.

Live-cell imaging
FM4-64 imaging. Before imaging, whole 5-day-old seedlings were incubated with 2 µM FM4-64 (Invitrogen) solution in half-strength MS liquid medium without sucrose at room temperature for 30 min.
For the FM4-64 uptake in Fig. 1, confocal images were taken using a FluoView FV1000 (Olympus) confocal, equipped with a ×60 water-corrected objective (numerical aperture (NA) = 1.2). Fluorescence was imaged in a single channel setting with 559 nm excitation light and emission fluorescence was captured in the frame-scanning mode via a 570 to 670 nm band pass emission window.
For the FM4-64 uptake in Fig. 5, FM4-64 was visualized on a Leica SP8X confocal with settings as outlined below.
FDA staining. FDA staining was performed on mature pollen stained on a glass slide in a solution containing 50 µM FDA in 10% sucrose for 5 min in the dark. Afterwards, confocal images were taken using a FluoView FV1000 confocal microscope (Olympus), equipped with a ×60 water-immersion corrected objective (NA = 1.2). Fluorescence was imaged in a single channel setting with 488 nm excitation light and emission fluorescence was captured in the frame-scanning mode via a 515-565 nm band pass emission window.
Protein localization in Arabidopsis roots. Arabidopsis root images for PM signal-quantification (TPLATE and TMK1) were obtained using a Leica SP8X confocal microscope equipped with a White Light Laser and using the LASX software package (Leica). Images were acquired on Hybrid detectors (HyD, gating 0.3-10 ns) using bidirectional imaging with a ×40 water-immersion corrected objective (NA = 1.10), frame or line signal averaging and with a ×3 digital zoom. A single excitation (488 nm laser) line and emission windows ranging between 500 and 550 nm for GFP; and 600-740 nm for propidium iodide (PI; Invitrogen) were used. The PI served to generate the mask for quantification of the cytoplasm to PM ratio in Figs. 3a,b and 5d-g. For the combined imaging of GFP and FM4-64 in Fig. 5, the GFP settings from above were combined with 561 nm laser excitation and a 570-630 nm detection window for FM4-64 in a line sequential imaging mode.
Root tracking. Arabidopsis seedlings expressing the pKNOLLE::KNOLLE-GFP marker were prepared as described in ref. 111 . In brief, 3-day-old seedlings were transferred into an inhouse-made two-part chambered coverglass (adapted from https://cellgrowth-lab. weebly.com/3d-prints.html), covered with a slice of half-MS agar and moved back to the growth chamber for 1 h to recover. A Zeiss LSM 900 KMAT confocal microscope equipped with a Zeiss Plan-Apochromat Article https://doi.org/10.1038/s41477-022-01280-1 ×20/0.8 dry objective, 488 nm diode laser and a GaASP-PMT detector (detection window of 490-565 nm) was used to acquire the images. For several roots, 3 µm step-size Z-stacks were taken over 120 timepoints with a time interval of 720 s with 8 bit depth, 1,024 × 1,024 pixels, pixel time of 0.52 µs, bidirectional scanning and four times line averaging. A tip track Matlab script 111 was used for tracking the root tips. Afterwards, assembly of the obtained videos was performed using the provided Fiji/Image script described in ref. 111 .

N. benthamiana transient expression-based imaging.
Imaging the interactions between TASH3 and AtEH/Pan1 in N. benthamiana was performed on a PerkinElmer UltraView spinning-disk system, attached to a Nikon Ti inverted microscope and operated using the Volocity software package (Quorum Technologies). Images were acquired on a 512 × 512 pixel ImagEM CCD camera (Hamamatsu C9100-13) using frame-sequential imaging with a ×60 water-immersion objective (NA = 1.40). Specific excitation and emission was achieved using a 488 nm laser combined with a single band pass filter (500-550 nm) for GFP. RFP was visualized using 561 nm laser excitation and a 410-480/580-650 nm detection window.
Endocytic dynamic imaging. Endocytic dynamics were imaged on the UltraView spinning-disk system (PerkinElmer) described above with similar laser and filter settings and using the Nikon Perfect Focus System (PFSIII) for Z-drift compensation. Images of hypocotyl epidermal cells of 3-day-old etiolated seedlings expressing single fluorescent markers were acquired with a ×100 oil-immersion corrected objective (Plan Apo, NA = 1.45). Single-marker line movies were acquired with an exposure time of 1 s per frame. Movies were acquired with a duration of 2 or 3 min. Specific excitation and emission was achieved using a 488 nm laser combined with a single band pass filter (500-550 nm) for GFP. RFP was visualized using 561 nm laser excitation and a 410-480/580-650 nm dual band pass filter.

Image quantification
FM4-64 uptake assays. Acquired FM4-64 labelled Arabidopsis root confocal images were analysed using the Fiji software package. PM and cytosol regions of interest (ROIs) of individual epidermal cells were outlined using the Select Brush and Freehand selections tool, respectively, and histograms of pixel intensities were generated for the indicated ROIs. Pictures which contained >1% saturated pixels were excluded from the quantifications. Cytoplasm:PM ratios were calculated from average intensities of the top 1% highest intensity pixels on the basis of the histograms.
Quantification of endocytic dynamics. Densities of endocytic foci were measured using the Find Maxima function of the Fiji software package. In a single slice of the obtained videos, an ROI was selected in the middle of the image. The background was subtracted from this ROI and the number of endocytic foci was assessed using the Find Maxima function. We used the pixel size and the area to convert this to spots per µm 2 . For each of the analysed sample sets, at least 12 cells from six different seedlings were analysed.
Life times of individual endocytic events were measured from kymographs generated by the Volocity software package (PerkinElmer). Life times had to be measured manually using the Fiji software package. The cmeAnalysis package, previously described in refs. 112,113 , unfortunately could not be used with our samples due to the high fluorescent background present in nosh. For each of the analysed sample sets, at least 12 cells from six different seedlings were analysed.
Quantification PM versus cytoplasm fluorescence. Quantifications of PM over cytoplasm ratios were performed with a custom-made Fiji-based script (available at https://github.com/pegro-psb/ Cyto-PM-signal-quantification). The script uses the PI staining in the red channel as a mask to allow automatic detection of PM and cytoplasm regions in the GFP channel. GFP and PI images are merged together in Fiji. Cells are manually annotated and stored in the Fiji ROI Manager. The script automatically measures the top 5% mean intensity pixels of the channel 2 (GFP) in the detected PM and cytoplasm ROI of each annotated cell based on the PI signal in channel 1 (PI).
For quantification of GFP versus FM4-64 ratios within one cell, FM4-64 staining was used as a mask to detect PM and cytoplasm regions. GFP and FM4-64 images were merged together in Fiji and cells were manually annotated and stored in the Fiji ROI Manager. For quantification of the FM4-64 signal, acquired FM4-64 images were duplicated in Fiji and for one of them the look-up table was changed to green. FM4-64-green and FM4-64-red images were merged together in Fiji and the ROIs stored in ROI Manager from GFP signal-quantification were also used for FM4-64 internalization. The script automatically measured the top 5% mean intensity pixels of channel 2 (GFP or FM4-64-green) in the detected membrane and the cytoplasm ROI of each annotated cell based on the FM4-64 signal in channel 1 (FM4-64-red).
Partitioning assay quantification. Quantification of the partitioning assays in N. benthamiana between the TASH3 constructs and both AtEH/Pan1 proteins was performed using the MitoTally script 114 . Particle regions of interest were determined on the basis of AtEH1/Pan1 or AtEH2/Pan1 positive foci.

Statistics and reproducibility
All the statistical analyses were performed using RStudio (v.1.2.5033) with R (RStudio, 2015). For all statistical analyses, a Mann-Whitney-Wilcoxon test was used, except for Fig. 5i, where a linear mixed effect model was used to quantify the difference between GFP/FM4-64. Group characterization in Fig. 5 and Extended Data Fig. 5

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
All materials are available from the corresponding authors upon request. All data generated or analysed during this study are included in this article and its Extended Data) and/or in public repositories. The raw mass spectrometry data and MaxQuant result files have been deposited to the ProteomeXchange Consortium via PRIDE (PXD035444). The script for comparative quantification of fluorescent signal at PM versus cytoplasm is available for download (https://github.com/pegro-psb/ Cyto-PM-signal-quantification). Araport11plus database consisting of the Araport11_genes.2016.06.pep.fasta downloaded from arabidopsis. org was used for MS analysis. Source data are provided with this paper. Fig. 2 | Complementation of nosh with full-length TASH3-GFP restores its endocytic defects. (a, b) Representative single-slice spinningdisk images and box plot graphs of endocytic foci densities in epidermal dark-grown hypocotyl cells of TPLATE-GFP (tplate), TASH3-GFP (tash3-1) and two independent TASH3-GFP expressing lines in the nosh background. The densities of endocytic foci in both complemented nosh mutants are similar to the values in TPLATE-GFP (tplate) and to those in the complemented tash3-1 mutant allele (TASH3-GFP in tash3-1). Numbers of quantified cells (2 cells per seedling) are indicated. The top and bottom lines of box plots represent 25th and 75th percentiles, the centre line is the median and whiskers are the full data range. The statistical test used was a two-sided Wilcoxon-signed rank test by comparing mutants to wild type. No adjustment for multiple comparisons was performed. Scale bar = 5 µm. (c, d) Representative kymographs and violin plot graphs of the life-time measurements from the spinning-disk time lapses from panel a. Analogous life-time distributions of endocytic events were observed for TASH3-GFP in nosh as for TPLATE-GFP (tplate) and TASH3-GFP (tash3-1). The number of events analysed for each independent line is indicated at the bottom of each graph. At least 12 movies from 6 seedlings were imaged and analysed for each independent transgenic line. The widest part of the violin plot represents the highest point density, whereas the top and bottom are the maximum and minimum data respectively. Red circles represent the mean and the red line represents the standard deviation. The statistical test used was a two-sided Wilcoxon-signed rank test by comparing mutants to wild type. No adjustment for multiple comparisons was performed. Scale bar = 50 µm. n.s. = not significant.

Corresponding author(s): Daniel Van Damme
Last updated by author(s): Oct 10, 2022 Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

March 2021
Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy All materials are available from the corresponding authors upon request. All data generated or analyzed during this study are included in this published article (and its Extended Data information files) and/or in public repositories. The raw mass spectrometry data and MaxQuant result files have been deposited to the ProteomeXchange Consortium via PRIDE (PXD035444). The script for comparative quantification of fluorescent signal at PM versus cytoplasm is available for download (https://github.com/pegro-psb/Cyto-PM-signal-quantification). Araport11plus database consisting of the Araport11_genes.2016.06.pep.fasta downloaded from arabidopsis.org was used for MS analysis.
Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
No statistical methods were used to predetermine sample sizes. Sample sizes were chosen based on the experiment subjects, estimated based on the analogous studies to the one presented in this manuscript and in the manner to be sufficient to address subsequent statistical analysis. The sample size for FM uptake, cytoplasm/PM fluorescent signal measurements and foci life time at the plasma membrane was used as described in previous studies (Wang et  Data exclusions No data were excluded from our analyses.

Replication
Each experiment was repeated for at least three times, except pollen viability, partitioning assay, endocytic foci analysis in complemented mutant, PM extraction fraction analysis and GFP/FM analysis in TMK1 lines, which were performed twice and SH3 domain binding different ubiquitin, TPC interacting with BRI1/BRI-25KR and SH3 purification, which were performed once. All individual biological replicates were with similar results. Samples for MS analysis were analyzed in triplicate. No replication problems were found in any experiment.
Randomization Seedlings (and Genotypes) were always randomly distributed during growth and treatment. Plants were analyzed from multiple independently grown biological replicates.

Blinding
Blinding experiments are no very relevant to this study. Moreover, the author who conducted the experiments also performed data acquisition and analysis, making the blinding difficult.
Reporting for specific materials, systems and methods We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.