Mutation of UPL3 affects plant development
In order to address more defined roles of UPL3, we systematically analyze the phenotype of upl3 lines, two knockout lines (upl3-1, upl3-3), two overexpressing lines (oeUPL3-24, oeUPL3-39), and two complementation lines (comUPL3-1, comUPL3-2) were produced and confirmed at the transcriptional and protein levels (Figure1A-B; Supplementary Figure S1). The predicted size of the entire protein is ~250 kDa appeared in oeUPL3-24/-39 lines, along with an additional 100 kDa band (possible alternative splicing isoform or proteolysis statue) is yet to be confirmed by sequencing. Plants of the upl3 lines exhibited apparently downward-curled leaves and a two-week delay in bolting and leaf senescence, compared to the wildtype (WT) plant (Figure 1C, G, H). In contrast, the overexpressing UPL3 lines displayed premature leaf aging and one-week earlier bolting (Figure 1C, G, H). Complementation by the full-length UPL3 rescued the upl3 mutants’ phenotypes (Figure 1C). Consistently, chlorophyll content (Figure 1D), photosystem II fluorescent activity (Fv/Fm) (Figure 1E) and green leaves/yellow leaves ratio (Figure 1F) increased significantly in the upl3 lines and decreased in the oeUPL3 line, compared to the WT. Thus, UPL3 functions in organ development and aging by accelerating cell senescence, under natural developmental conditions.
Loss of UPL3 induces global ubiquitin enrichment in 6-week-old upl3 plants
UPL3 was highly expressed in senescent leaf (after 6 weeks) (Winter et al., 2007; Supplementary Fig. S2), therefore, we performed a label-free mass spectrometry (MS)-based analysis of protein ubiquitination using K-epsilon-GG remnant antibody enrichment approach using 6-week-old upl3 and WT plants (Figure 2A; Supplementary Fig. S3). Identified proteins based on their tandem mass spectra matching against the UniProt Arabidopsis thaliana Columbia database (current total of 39211 reads) using the MaxQuant software, are listed in Supplementary Dataset 1. Label-free quantification (LFQ), with a false discovery rate (FDR) adjusted to < 1% and a minimum score for modified peptides set as > 40, resulted in a set of over 1,310 potential ubiquitinated targets. This was further refined to a subset of 1155 targets by at least two post-translational modifications (PTMs) (Figure 2B; Supplementary Data Set 2).
To identify the alteration of ubiquitin conjugates associated with the upl3 mutation, a fold-change greater than 1.2 or less than 1/1.2 was used to filter conjugate targets in the library, whose ubiquitination was up-regulated or down-regulated. All the differentially ubiquitin conjugates (DUCs) data in upl3/WT were shown in Figure 2B and Supplementary dataset S2. Among them, ubiquitination of 545 sites (356 proteins) was found to be up-regulated, and ubiquitination of 198 sites (189 proteins) was down-regulated in the upl3, compared to WT plant (Figure 2B). These results were verified by global ubiquitination immunodetection using an anti-ubiquitin antibody, in which deletion of UPL3 led to enhanced signals of global levels of protein ubiquitination, whereas loss of its homolog UPL5 did not, as a control (Figure 2C, Miao and Zentgraf, 2010). Overexpression of UPL3 retained comparable global levels of protein ubiquitination with the WT (Figure 2C). Indeed, more proteins were found in the up-regulated category than in the down-regulated set, in term of modified protein sites (Figure 2D). Specifically, 188 conjugates had enhanced abundance over WT by greater than 1.5-fold, while the levels of only 79 conjugates were reduced by 1/1.5 or less (Figure 2D). Fold-change levels against the statistic P value of individual ubiquitinated site were plotted to give more details on the dataset (Figure 2E). Of them was the fact that ubiquitin-conjugating enzyme 35 (UBC35) and ubiquitin-activating enzyme 1 (UBA1) were among the targets with significantly enhanced ubiquitination levels (Figure 2E), which appeared in the dataset of ubiquitome (Miller et al., 2010), and RPN, a known target of UPL3 (Furniss et al., 2018), was appeared in dataset of ubiquitome with upregulated ubiquitination level, indicating that the quality of ubiquitome dataset was properly sound. Together, this surprising rise in protein ubiquitination caused by loss of UPL3 perhaps means that UPL3 has somehow function on deubiquitinating pathway or UPL3 promotes deubiquitinases (DUBs) function.
Differential ubiquitomic enrichment reveals that metabolic enzymes are among the processes significantly affected by UPL3
GO term enrichment showed that the molecular functions of assembled ubiquitin conjugates were related to protein-protein interactions (47%), catalytic activity (34%), transport activity (7%), structural molecule activity (5%), and other (7%). However, for the differentially ubiquitin conjugates (DUCs) of upl3 relative to WT, the protein-protein interaction function enrichment was increased to 49% (Figure 3A), implying a major involvement in molecular interaction for the UPL3-regulated targets. Further assignment to biological processes showed that the DUCs included mainly proteins involved in the response to metal ion stresses and enzymes related to carbon metabolism and nucleotide metabolism (Supplementary Figure S4). Hence, it is plausible that UPL3 maintains the ubiquitination status of enzymes and regulatory factors to fine-tune cellular metabolism.
This notion was further validated via KEGG pathway analysis (Figure 3B; Supplementary Fig S4). A large subset of the DUCs were enzymes related to biosynthesis of secondary metabolites, carbon fixation, carbon metabolism, and amino acid metabolism. The ubiquitinated forms of the Calvin-Benson enzymes, such as ribulose-1,5-P2-carboxylase (RuBisCO), phosphoglycerate kinase (PGK), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and the aldoketo reductase family members (ALDOs), as well as the CAM enzymes malate dehydrogenases MDH1 and MAEB homolog, were enriched in the upl3 plants (Figure 3C). Carbon metabolism was also represented by a reduction in ubiquitin conjugates of phosphoenolpyruvate carboxylase (PPC2) and hexokinase-1 (HXK1) in the upl3 background. On the other hand, a decline in ubiquitin conjugates was found in a small subset of cysteine and methionine metabolism-related enzymes, such as 5-methyltetrahydropteroyl-triglutamate, homocysteine methyltransferase 2 (MS2), S-adenosylmethionine synthase 1 (SAM1), cysteine synthase 1 (OASA1), and methionine aminotransferase (BCAT4), SNARE-like superfamily protein (YKT61), vesicle-associated membrane protein 714 (VAMP714) in the upl3 plants (Figure 3D, Supplementary Dataset S2).
Our Arabidopsis ubiquitome proteins can be classified as cytoplasmic (36%), nuclear (20%), chloroplast (20%), membrane (14%) and others (10%). Among these, the percent ratio of the UPL3-influenced ubiquitin conjugates was higher in the cytoplasmic and the nuclear sections (Figure 3E). According to fold-change categories of DUCs, most of the proteins in Q1, experiencing large drop in ubiquitination (0 < fold change < 1/1.5) were chloroplast and chromatin components; most of those in Q2 (1/1.5 < fold change < 1/1.2) were vacuole and endomembrane components. In contrast, most of the enriched ubiquitin conjugates in Q3 (1.2 < fold change < 1.5) were components of the stromal and vesicle membrane; and the proteins with the most significantly up-regulated ubiquitination, Q4 (fold change > 1.5) were predicted to be cytoplasmic (Supplementary Figure S5). The Cytoscape protein-protein interaction network of the 545 UPL3-influenced DUCs generated distinctly clustered interaction nodes (Supplementary Fig S6) (Shannon, 2003). The UPL3-influenced conjugates were dispersed throughout the network, suggesting that UPL3 was likely involved in broad control of stress response (Supplementary Fig S6A). The network also revealed a relatively significant enrichment in response to stimulation (abiotic or biotic) of the translation machinery and protein metabolism (Supplementary Fig S6B). However, UPL3 fused GFP transformed transiently in tobacco leaves showed that UPL3 was clearly localized in the nucleus (Figure 3F). Therefore, nuclear proteins such as those functioning in chromatin remodeling, chromosome regulation, and transcriptional complexes are likely primary UPL3-dependent targets.
The UPL3-dependent conjugates are enriched in histone H1/H5 and stress-related protein domains
Scanning protein domains of the DUCs using InterProScan identified several UPL3-dependent (reduced ubiquitination in the upl3 background) protein domains including the histone H1/H5 domain, jacalin-like lectin domain, GST domain, S15/NS1 RNS binding domain, and heavy metal associated domain (Figure 4A blue), whilst the UPL3–influenced ubiquitin conjugates (enhanced ubiquitination in the upl3 background) contained leucine-rich repeat, histone H2A/H2B/H3, ribosomal protein S5, double-stranded RNA binding domain, and histone fold domains (Figure 4A orange). For protein domains of the individual DUC within each fold-change range (Q1, Q2, Q3 and Q4), the results were shown in Supplementary Figure S4 and Supplementary dataset S2. Again, it was noted that protein domains associated with RNA binding, protein translation, and amino acid synthesis related proteins were overrepresented in proteins showing reduced ubiquitination in the upl3 plants, followed by heavy-metal-associated domain, remorin- and jacalin-like lectin domain of abiotic (metal)/biotic stress responsive proteins. In contrast, protein domains associated with protein binding, transport ATPase, ubiquitin carboxyl-terminal hydrolase, and metabolism enzyme were the most significant domains in proteins showing enhanced ubiquitination in the upl3 plants, followed by domains contained in the ubiquitination/26S proteasome system (UPS) regulatory complex (Supplementary Fig S4; Supplementary dataset S2). It suggested that proteins containing heavy-metal-associated domains, GST domains, jacalin-like lectin domains, and histone H1/H5 domains were likely UPL3-dependent ubiquitin conjugates. Consistently, UPL3–influenced ubiquitin conjugates contain histone H2A/H2B/H3, histone fold, and SPK1/BTB/POZ binding domain, most of which have functions related to protein binding activity and gene transcriptional control.
Histone ubiquitination was earlier recognized as a marker of transcriptionally active chromatin, where the ubiquitinated forms of histones H2B/H2A were associated specifically with activated or repressed transcribed genes (Cao et al., 2008; Kralemann et al., 2020). It was also reported that the H1 (or linker) histone played critical functions in determining the accessibility of chromatin DNA to trans-acting factors and mediated chromatin organization in the epigenetic control during developmental and cellular transitions (Kotinski et al., 2017; Rutowicz et al., 2019). In this scenario, UPL3–influenced ubiquitin conjugates in the nucleus are likely mediators of chromatin accessibility and transcriptional processes that control downstream gene expression. To test this hypothesis, we compared transcriptome data of the upl3 and WT plants, which released from Furniss (2018) (Supplementary dataset S3). A total of 1467 differentially expressed genes between the upl3 and WT seedlings were identified (Figure 4B; Supplementary Dataset S3). Of these, genes related to stress responses (drug, hypoxia, oxidative, and UV) were most significantly upregulated, followed by genes associated with cellular glucan metabolic processes, anthocyanin-containing compound biosynthesis, and cellular polysaccharide catabolic processes. A third set of up-regulated genes was involved in protein transport and leaf senescence. In contrast, genes for response to (a)biotic stress (metal ion salt stress, and fungus) and plant aging were most significantly downregulated, followed by genes related to secondary metabolic processes, MAPK signaling pathway, leaf cell death, and phenylpropanoid biosynthesis (Figure 4C). Based on a delaying senescence phenotype of upl3 mutant, we further selected 14 genes coding for transcription factors such as WRKYs, NACs, ZFs, and MYB related to leaf aging (e.g., WRKY53 and WRKY75; Miao and Zentgraf, 2007; Guo et al., 2017) and salicylic acid responsive senescence (e.g., WRKY38, WRKY63, and WRKY51 etc.; Zhang et al., 2017; Krinke et al., 2009) for confirmative RT-qPCR analysis (Figure 4D). The results showed that the expression level of WRKY53, WRKY75, WRKY38, WRKY63 and WRKY51, as well as ZFs was significantly decreased in the upl3 (Figure 4D). These data confirmed the notion that UPL3 had a profound functional involvement in stress responsive cell senescence, developmental cell senescence (aging), and secondary metabolic processes, possibly via ubiquitination of their regulators either directly or indirectly.
Furthermore, 26 proteins were identified as the overlapping genes of DEGs and DUCs, with altered ubiquitinated protein level (fold-change > 1.5) and gene expression level (fold-change > 1.3) by the upl3 mutation (Figure 4B). Among these proteins, four proteins (Figure 4E, fold-change depicted with two orange value) with increased ubiquitin conjugation were up-regulated in gene expression in the upl3 background, including glucomannan 4-beta-mannosyltransferase 9 (CSLA9), inositol-3-phosphate synthase isozyme 1 (MIPS1), calcium-binding protein 16 (CML16), and the Patellin-2 (PATL2) (Figure 4E). Two other proteins (Figure 4E, fold-change depicted with a blue and an orange value) with reduced ubiquitin conjugates but with increased transcript level in the upl3 mutant were cysteine lyase (CORI3) and cinnamyl alcohol dehydrogenase 7 (CAD7), both of which had reported functions in the amino acid metabolic process (Tsuwamoto and Harada, 2011; Tanaka et al., 2018). On the other hand, 9 proteins (Figure 4E, fold-change depicted with an orange and a blue value) showed an enhanced ubiquitin conjugation level but a down-regulated expression level in the upl3 mutant plants. These included calmodulin-like protein10 (CML10), aquaporin (PIP1-5), triacylglycerol lipase-like 1 (TLL1), leucine-rich repeat ser/thr protein kinase (LRR-RLK), the jacalin-related lectin 23 (JAL23), ABCB transporter member 19 (ABCB19), methionine aminotransferase (BCAT4), and the ankyrin repeat-containing protein (BDA1), whose functions were involved in response to (a)biotic stress and in plant development and plant aging in response to light, auxin, JA, or calcium (Debernardi et al., 2014; Wu et al., 2010; Yang et al., 2012 ; Yang et al., 2013 ). The remaining eleven proteins (Figure 4E, fold-change depicted in blue) were those with both reduced ubiquitin conjugation level and down-regulated gene expression in the upl3 mutant. Their functions seemed to be related to glycoside metabolism and transport pathways. From these data, it is obvious that the UPL3-centered molecular network involves both feed-forward and feed-back regulatory pathways and mostly impacts on cellular metabolism, stress responsive cell death and aging.
Identifying the potential ubiquitylated targets in UPL3-bound proteins
To evaluate the direct connection between UPL3 and conjugates, we assessed protein-protein interaction using a GFP-nanotrap-assisted pulldown-MS assay (Supplementary Figure S7), in which the GFP-tagged UPL3 was a bait. To do this, the rosette leaves of 6-week-old stable transgenic plants expressing either the UPL3-GFP or the control GFP driven by the ACTIN3 promoter were used for total protein isolation, purification and immunodetection (Figure 5A-B). The trypsin-digested proteins were subjected to mass spectrometry analysis with high-energy collisional dissociation quantum efficiency mass spectrometry (QE-MS). Tandem mass spectra were searched against UniProt Arabidopsis thaliana Columbia (89247_20181227) database via Mascot2.2 software. A total of 81 putative proteins were identified after subtracting the GFP control resulted from the UPL3-GFP candidate list (Supplementary Dataset S5). With these putative UPL3-interacting patterns, we identified 29 (or 11 under more stringent conditions) overlapping proteins in the ubiquitome DUCs (Figure 5C).
These 29 proteins were clustered in four categories based on the KEGG pathway database. Consistently, proteins involved in the carbon fixation pathway have enhanced ubiquitination in the upl3 mutant, while those in the inositol-1,4,5-trisphosphate-3-kinase (IP3K) signaling pathway have reduced ubiquitination in the upl3 plants versus WT (Figure 5D-E), which included 5 proteins, namely ABCG36, MS2, PPC2, LOS1, and AT3G63160 (Figure 5E, fold-change depicted in blue). 14 other interacting candidates showing upl3-mutation-enhanced ubiquitination in the upl3 background were H2AXb, RPS2B, PHOT1, ERD14, the SWI-SNF chromatin remodeling ATPase BRAHMA (BRM) and SWIS3C, as well as carbon-metabolism-related enzymes (Figure 5E, fold-change depicted in orange). Notably, UBP12, UBP13, and UBP26 are also among the UPL3 interacting candidates (Figure 5E). UBP12 is a ubiquitin hydrolase, with a demonstrated de-ubiquitination activity in vitro and localization both in the cytoplasm and the nucleus (Deracheva et al., 2016; Kralemann et al., 2020).
The UPL3-interacting and target proteins UBP12 and BRM are involved in leaf development and aging
To confirm these putative UPL3 interacting partners, we carried out a yeast two-hybrid assay by selecting 16 candidates. The self-interacting N-terminal fragment (470 aa) of UPL3 containing armadillo repeats was shown interacting with most of the selected candidates except for UPL5 and PHOT1, however, the full-length UPL3 only showed a strong interaction with BRM and UBP12, and very weak interaction with HXK1, PPC2 and UBC35 (Figure 6A-B), but the full-length UPL3 bait was not able to interact with its N-terminus (Figure 6B). The failure to detect many interactions with the full-length bait may be explained by a degradation effect mediated by the ligase activity retained in the yeast cell, although that needs to be further verified. Next, the interaction of UPL3 with UBP12 was confirmed using BD-UBP12 as a bait, which in turn showed a weak interaction with two other UPL3-interactors, namely the BRM and PPC2 (Figure 6C).
To examine whether UPL3 altered the protein level of the bound candidates of UPL3 between upl3 and WT. To the end, we compared the total proteomes of 6-week-old wild-type and the upl3 plants by tandem MS using the precursor ion intensity of the MS1 scans for quantification. Altogether, 3557 Arabidopsis proteins could be reproducibly identified and quantified in both samples by our liquid chromatography-mass spectrometry (LC-MS) regime analyzed in triplicate (Supplementary Fig S8; Supplementary Dataset S4). We searched for the protein level of UBP members and several interested proteins in differentially expressed proteins (DEPs) of proteome dataset from upl3/WT (Supplementary Dataset S4). Several UPL3-bound proteins BRM, UBC35, UBP12, and UPL13, as well as other UBP members such as UBP1C, UBP6, UBP26 were slightly upregulated 1-1.4 folds in the upl3 mutant relative to WT; the rest UPL3-bound proteins HXK1 and PPC2 exhibited a downregulated 1.5-3 folds in the upl3 mutant relative to WT (Figure 6D; Supplementary Dataset S4). Furthermore, when the ubiquitination enrichment was normalized to protein level, the level of ubiquitinated UBP12, UBP13, and UBP26 was not significantly changed between the upl3 and WT, however, the protein level of ubiquitinated BRM, UBC7 and HXK1, PPC2 maintained same up/down altered potential (Figure 6D-E). This suggests that UBP12 and other UBP members are interacting partners but not substrates of UPL3 as E3 ligase.
To examine whether the interaction between UPL3 and UBP12 might contribute to the level of ubiquitinated targets in planta, we first detected BRM and PPC2 ubiquitination in ubp12 and upl3 mutant background using antibodies against poly-ubiquitin, BRM, and PPC2. The total ubiquitination level in the ubp12 or the upl3 was comparable or slightly stronger than in the wildtype, while it declined considerably in the UBP12-overexpressing line (Figure 6E). The protein levels of PPC2 were increased by either upl3 mutation or its overexpression, as well as in ubp12 mutant, they were in contrast diminished by UBP12 overexpression (Figure 6E). It suggested that UBP12 affected PPC2 protein level more than UPL3 did, although UPL3 affected PPC2 both ubiquitination enrichment and protein level. However, UBP12 and UPL3 displayed opposite effect on BRM protein level when their genes were mutated. Loss of UBP12 reduced, but loss of UPL3 increased, the BRM protein accumulation relative to WT, while overexpression of the two genes had the opposite effect to deletion (Figure 6C). We conclude that UBP12 has a major effect on PPC2 protein degradation, while UPL3 and UBP12 antagonistically affect the BRM protein level. These results suggest a complicated molecular interaction network between UPL3 and these proteins, probably in terms of homeostasis in protein ubiquitination.
We further sought insights from phenotypic analysis of the upl3-1, ubp12, and brm-em mutant plants. The results showed that oeUPL3-24, ubp12, and brm-em plants shared a common phenotype of early bolting and flowering, but only a double mutant of ubp12 ubp13 (with 91% sequence identity at the amino acid level) caused the curled-leaves and premature leaf aging phenotype found in the brm-em plants (Figure 7A-B; Supplementary Fig S9). In contrast, upl3 plants displayed a late-senescence phenotype with delayed flowering-time (more than one week), the effect of UPL3 on plant development contrasted those of both UBP12 and BRM (Figure 7C-D), thereby, a mild curled-leaves phenotype was also observed in aging leaves of the upl3 plants at a later stage of development (Figure 1A, 7A). It is consistent with previous reports of phenotype of ubp12 and brm mutants (Cui et al., 2013; Park et al., 2019; Xu et al., 2016; Li et al., 2016; Archacki et al., 2017).
Furthermore, the starch content of the mutant plants was monitored to check whether the opposite phenotypes observed in upl3 and ubp12 were related to carbon metabolism. The upl3 mutant showed a decreased starch accumulation compared with WT, while complementation or overexpression of UPL3 restored the starch content to the WT level (Figure 7E). In contrast, the ubp12 mutant displayed significantly stronger starch accumulation than that in WT, while effect of UBP12 overexpression was similar to that of the upl3 mutant, with a much lower starch content in the plants (Figure 7E). Therefore, starch accumulation is positively correlated to the early flowering phenotypes associated with mutations in the two antagonistic ubiquitination pathway genes.
Analysis of ubiquitin footprints (K sites).
The H89R substitution in the tagged ubiquitinated assay used here enables the detection of ubiquitination sites (“footprints”) and identified a consensus ubiquitin attachment sequence (Xu et al, 2010). By scanning all generated datasets, we identified 2778 ubiquitinated sites in total (Supplementary dataset S1). Among these, 2359 sites were quantified, 1641 of the 2359 sites were in the upl3 plants, 414 sites were differentially displayed relative to WT plant with a cut-off log2FC of 1.5.
We identified 110 ubiquitinated modification sites on 77 differentially ubiquitinated proteins in upl3/WT (log2FC > 2) (Supplementary dataset S2). Motif analysis around the modified lysine using MEME identified a consensus ubiquitin attachment sequence in 44 of the 110 sites (Figure 8A) that strongly matched the c-K-x-E/D/G ubiquitination motif (where c and x represent a hydrophobic and any amino acid, respectively), which was a prevalent motif in yeast and animal ubiquitinated targets (Xu et al. 2010). However, the remaining 66 sites (60%) were unrelated to this motif, indicating that noncanonical sites were also common. In addition, referring to the GPS-SUMO algorism (Zhao et al., 2014), one or more copies of this consensus sequence were detected in ubiquitinated targets accounting for 74%, 66%, and 59% of the three enriched ubiquitination categories, the UPL3-dependent, abundant, and total ubiquitinated proteins, respectively. Among the 44 sites with a consensus sequence, out of the 110 sites, 14 out of the 21 mapped attachment sites on 18 UPL3-upregulated targets belonged to the canonical c-K-x-E/D/G motif, with the remainder had alternative sequences (Supplementary Dataset S6), including GAPC1, GAPC2,CASA1,NADP-ME2, RuBisCO, AAT1, FBAB, PGK3, MDH1, GLO1, and SHM4 (Figure 8B). Specifically, the ubiquitinated sites identified here for the UPL3-dependent targets, namely ALDH, OASA1, HXK1, and PPC2 were within a non-canonical x-A-K-x- motif or x-K-A-x motif (Figure 8B), due to HXK1 at K77 was previously reported to be ubiquitinated at a site of non-canonical linkages in animal cell (Huang and Li, 2018). Thus, our omics results confirm the assembly of poly-ubiquitin chains in plants. By scanning our ubiquitome datasets for ubiquitination sites using footprints containing ubiquitin remnants after trypsin cleavage, modifications by SUMO1 at K23 and K42 were detected in addition to polyubiquitin linked via K48 linkages. This provides further evidence for SUMOylating of some ubiquitylated proteins in Arabidopsis (Miller et al., 2010).