Direct attenuation of Arabidopsis ERECTA signalling by a pair of U-box E3 ligases

Plants sense a myriad of signals through cell-surface receptors to coordinate their development and environmental response. The Arabidopsis ERECTA receptor kinase regulates diverse developmental processes via perceiving multiple EPIDERMAL PATTERNING FACTOR (EPF)/EPF-LIKE peptide ligands. How the activated ERECTA protein is turned over is unknown. Here we identify two closely related plant U-box ubiquitin E3 ligases, PUB30 and PUB31, as key attenuators of ERECTA signalling for two developmental processes: inflorescence/pedicel growth and stomatal development. Loss-of-function pub30 pub31 mutant plants exhibit extreme inflorescence/pedicel elongation and reduced stomatal numbers owing to excessive ERECTA protein accumulation. Ligand activation of ERECTA leads to phosphorylation of PUB30/31 via BRI1-ASSOCIATED KINASE1 (BAK1), which acts as a coreceptor kinase and a scaffold to promote PUB30/31 to associate with and ubiquitinate ERECTA for eventual degradation. Our work highlights PUB30 and PUB31 as integral components of the ERECTA regulatory circuit that ensure optimal signalling outputs, thereby defining the role for PUB proteins in developmental signalling.

Plants sense a myriad of signals through cell-surface receptors to coordinate their development and environmental response. The Arabidopsis ERECTA receptor kinase regulates diverse developmental processes via perceiving multiple EPIDERMAL PATTERNING FACTOR (EPF)/EPF-LIKE peptide ligands. How the activated ERECTA protein is turned over is unknown. Here we identify two closely related plant U-box ubiquitin E3 ligases, PUB30 and PUB31, as key attenuators of ERECTA signalling for two developmental processes: inflorescence/pedicel growth and stomatal development. Loss-of-function pub30 pub31 mutant plants exhibit extreme inflorescence/ pedicel elongation and reduced stomatal numbers owing to excessive ERECTA protein accumulation. Ligand activation of ERECTA leads to phosphorylation of PUB30/31 via BRI1-ASSOCIATED KINASE1 (BAK1), which acts as a coreceptor kinase and a scaffold to promote PUB30/31 to associate with and ubiquitinate ERECTA for eventual degradation. Our work highlights PUB30 and PUB31 as integral components of the ERECTA regulatory circuit that ensure optimal signalling outputs, thereby defining the role for PUB proteins in developmental signalling.
The development of multicellular organisms relies on coordinated cell proliferation and differentiation in response to external cues. Plants use a battery of membrane-bound cell-surface receptors with an intracellular kinase domain, collectively known as receptor-like kinases (RLKs), to sense and transduce external signals to adjust cellular responses 1 . Among them, those with an extracellular leucine-rich repeat (LRR) domain, LRR-RLKs, comprise the largest subfamily in plants with more than 200 members in Arabidopsis 1 . The LRR-RLKs play critical roles in development, hormone perception, interkingdom communication and immunity. Those with known ligands are referred to as LRR-RKs hereafter 2,3 . Well-studied LRR-RKs include the brassinosteroid receptor BRASSINOSTEROID INSENSITIVE1 (BRI1) 4 and the immune receptor FLAGELLIN SENSING2 (FLS2) 5,6 , among others 2,3 .
To address the genetic relationships of PUB30/31 with ERECTA, we next generated the triple mutant of pub30 pub31 er-105 (erecta null allele). The pedicel length of the triple mutant phenocopied that of er-105 ( Fig. 1a-c), indicating that the erecta mutation is epistatic to pub30 pub31. To further examine the underlying cellular basis of the pub30 pub31 defects and its relationship with erecta, we analysed longitudinal sections of mature pedicels (Fig. 1d). It has been shown that short erecta pedicels accompany reduced cell proliferation and compensatory cell growth in the cortex layer 14,39 . In contrast to erecta, cortex cells in the pub30 pub31 pedicels are small and highly organized (Fig. 1d, asterisks). Quantitative analysis detects no statistical difference in the cortex cell areas of the WT and pub30 pub31 pedicels (Fig. 1e), indicating that the extremely elongated pedicel phenotype of pub30 pub31 is due to excessive cell proliferation, but not cell expansion (Fig. 1d,e). The pub30 pub31 er-105 pedicels exhibit large, expanded cortex cells that are statistically indistinguishable from er-105 (Fig. 1d,e). Thus, the erecta mutation is epistatic to not only the overall pedicel length, but also the underlying cortex cell proliferation phenotype of pub30 pub31. Combined, our results suggest that PUB30 and PUB31 function as negative regulators of ERECTA-mediated inflorescence and pedicel growth.

PUB30/31 inhibit ERECTA pathway on stomatal development
It is well known that ERECTA family LRR-RKs enforce stomatal patterning 9 . Among the three members, ERECTA plays a major role in restricting the initiation of stomatal cell lineages 9,19 . To dissect the genetic relationship between ERECTA and PUB30/31 in stomatal development, we first analysed the cotyledon epidermal phenotype (Fig. 1f,g). The pub30 and pub31 single mutants showed a slightly reduced stomatal index (number of stomata/(number of stomata + non-stomatal epidermal cells) × 100) compared with WT. The stomatal index was reduced notably in the pub30 pub31 double mutant (Fig. 1f,g). Again, transgenic pub30 pub31 plants expressing proPUB30::PUB30 and proPUB31:PUB31 fully rescued the stomatal phenotype of pub30 pub31 (Extended Data Fig. 1f,g), indicating that PUB30 and PUB31 redundantly promote stomatal development.
After receptor activation, the strength of cellular signalling must be promptly downregulated to avoid excessive or untimely signal outputs. Thus, the mechanism of signal downregulation is an integral part of receptor signalling. Studies on FLS2 and BRI1 have highlighted the role of receptor ubiquitination (ubiquitylation) for signal attenuation; interestingly, both ligand-activated FLS2 and BRI1 are ubiquitinated by the two identical plant U-box ubiquitin E3 ligases, PUB12 and PUB13, albeit in a slightly different manner 30,31 . Whereas PUB12/13 target several additional receptor kinases, ERECTA is not ubiquitinated by PUB12/13 (ref. 31). The U-box domain, which was originally identified from Ub Fusion Degradation 2 (UFD2) in yeast, mediates interaction with the ubiquitin-conjugating enzyme [32][33][34][35] . The first reported PUB protein is ARM Repeat Containing1 (ARC1), which interacts with the kinase domain of Brassica S-locus receptor kinases 36 . The PUB proteins constitute a family of more than 60 members in Arabidopsis, some of which are involved in a variety of environmental responses 37 . However, with the exception of a handful of members, their in vivo targets and functions remain unknown. The well-studied PUB proteins function in immunity and/or the stress response 37 , and whether PUB proteins attenuate RLKs regulating development remains an open question. For example, although PUB12/13 target BRI1, neither pub12, pub13 nor the pub12 pub13 double mutant exhibits any growth phenotypes nor do they enhance the elongated rosette phenotype of BRI1 overexpressors 31 . Thus, it is unclear whether PUB12/13 play a major role in brassinosteroid-mediated growth and developmental processes. A genetic study placed the possible role of PUB4 in CLAVATA signalling in the meristem 38 . However, the identity of PUB4 targets remains unknown 38 .
Here, we report two paralogous PUB proteins, PUB30 and PUB31, as key attenuators of ERECTA signaling transduction pathways for both inflorescence/pedicel growth and stomatal development. The pub30 pub31 double mutant plants exhibit characteristic inflorescence with extreme pedicel elongation and a reduction in stomatal development. The erecta mutation is epistatic to pub30 pub31, indicating that PUB30 and PUB31 are redundantly required to downregulate ERECTA activity. We demonstrate that perception of EPF2 and EPFL6 peptides by ERECTA leads to the phosphorylation of PUB30/31 by BAK1 and stronger associations of PUB30/31 with ERECTA and BAK1. In this system, BAK1 acts both as a coreceptor kinase and a scaffold that recruits PUB30/31 to directly ubiquitinate ERECTA, but not BAK1 itself, for eventual degradation and signal attenuation. Our work reveals the modes of action and functions of a pair of PUB proteins in two ERECTA-mediated developmental processes and further suggests a broader view of how plant receptor kinases are attenuated upon signal activation.

PUB30/31 negatively regulate ERECTA-mediated plant growth
Loss-of-function erecta mutant plants exhibit characteristic compact inflorescence and short pedicels (Fig. 1a-c) 7,8,14,39,40 . We hypothesized that potential negative regulators of ERECTA may confer the opposite phenotype-that is, extreme elongation of inflorescence and pedicels. With this in mind, we systematically surveyed the transfer DNA (T-DNA) insertion lines of PUB family genes. This led to the identification of PUB30 and PUB31 null mutant alleles (Methods and Extended Data Fig. 1a-c). Whereas single mutants of pub30 and pub31 do not show obvious growth phenotypes, the pub30 pub31 double mutant produces elongated inflorescence with extremely long pedicels (Fig. 1a-c). Introduction of wild-type (WT) PUB30 or PUB31 coding sequences driven by their native promoters (proPUB30::PUB30 and proPUB31::PUB31) into Article https://doi.org/10.1038/s41477-022-01303-x juxtamembrane domain, kinase domain and the C-terminal tail, was fused to the DNA-binding domain and used as bait. PUB30/31 are predicted cytoplasmic proteins, which contain a U-box domain, an ARM repeat and a linker domain in between (Extended Data Fig. 2b). Full-length PUB30 and PUB31 proteins were fused to the activation domain (AD). As shown in Fig. 2a, ERECTA_CD interacts with PUB30/31. Next, we confirmed the direct interaction of ERECTA_CD and PUB30/31 by in vitro pull-down assays using purified recombinant ERECTA_CD and full-length PUB30 and PUB31 proteins (Extended Data Fig. 2c,d).
Finally, to quantitatively characterize the kinetics of protein-protein interactions between PUB30/31 and ERECTA_CD, we performed biolayer interferometry (BLI) assays (Methods). PUB30 and PUB31 bind with ERECTA_CD at a micromolar affinity (Fig. 2b,c). The observed weak affinity might be attributed to the transient and dynamic interactions of ERECTA and PUB30/31.
To examine the in vivo association of ERECTA with PUB30 and PUB31 in Arabidopsis, we further performed co-immunoprecipitation (co-IP) analyses using transgenic plants carrying epitope-tagged ERECTA  . P values are indicated in the graph. NS, not significant. f, Confocal microscopy of 10-day-old abaxial cotyledon epidermis of WT, pub30, pub31 and pub30 pub31. Scale bar, 25 μm. g, Quantitative analysis. Stomatal index of the cotyledon abaxial epidermis from 10-day-old seedlings of respective genotypes (n = 8). Welch's two-sample two-tailed unpaired t-test was performed for pairwise comparisons with the WT. P values are indicated in the graph. h, Confocal microscopy of 6-day-old abaxial cotyledon epidermis of WT, pub30 pub31, er-105 and pub30 pub31 er-105. Scale bar, 25 μm. i, Quantitative analysis. Stomata + meristemoid index of the cotyledon abaxial epidermis from 6-day-old seedlings of respective genotypes (n = 12). Welch's two-sample two-tailed unpaired t-test was performed for the pairwise comparisons of WT versus pub30 pub31 as well as er-105 versus pub30 pub31 er-105. P values are indicated in the graph.
To address the in vivo role of PUB30/31 in regulating ERECTA, we next compared the in vivo ubiquitination status of ERECTA in erecta null mutant, er-105, complemented with proERECTA::ERECTA-FLAG (hereafter referred to as 'WT') and erecta pub30 pub31 triple mutant complemented with proERECTA::ERECTA-FLAG ('pub30 pub31') seedlings 44 (Fig. 3a,b). The relative signal intensity ratio between ubiquitinated ERECTA detected by anti-ubiquitin antibody and immunoprecipitated ERECTA detected by anti-FLAG antibody, indicates that, in the absence of PUB30/31, ERECTA is less ubiquitinated in vivo (Fig. 3a,b). Poly-ubiquitinated proteins can be destined for degradation via the ubiquitin/26S proteosome pathway or the endocytic/vacuolar degradation pathway [45][46][47][48][49] . We subsequently examined whether PUB30/31 regulate ERECTA stability in vivo. Higher accumulation of ERECTA proteins (ERECTA-FLAG) was detected in pub30 pub31 mutant background compared with the WT (Fig. 3c,d). By contrast, ERECTA transcript levels were not significantly different between WT and pub30 pub31 seedlings (Extended Data Fig. 3a), indicating that the effects of PUB30/31 on ERECTA accumulation is likely posttranslational. To test whether endocytic/vacuolar degradation pathways are involved in the PUB30/31-mediated ERECTA degradation, we subsequently treated the seedlings with Tyrphostin A23 (Tyr A23, an inhibitor of clathrin-mediated endocytosis) 50 and Concanamycin A (Con A, a vacuolar ATPase inhibitor) 51 . As shown in Fig. 3c and Extended Data Fig. 3b-d, treatments of both Tyr A23 and Con A resulted in the significant increase in ERECTA protein accumulation in the 'WT' seedlings, suggesting that endocytosis followed by degradation in vacuole may be involved in the ERECTA degradation.
Next, to address whether the ubiquitination activity of PUB30/31 is essential for their function as regulators of ERECTA-mediated processes, we introduced amino acid substitutions into PUB30/31 sequences that replace the conserved E2-binding tryptophan residue to alanine (PUB30 W97A or PUB31 W93A ) within their U-box motif (Extended Data Fig. 2e). In vitro auto-ubiquitination assays showed that these mutations (W97A in PUB30 or W93A in PUB31) diminished the ubiquitination activity of PUB30 and PUB31 (Extended Data Fig. 3e). Subsequently, these E2-binding defective PUBs were expressed by their native promoters in the pub30 pub31 mutant. Neither transgenic proPUB30::PUB30 W97A nor proPUB31::PUB31 W93A was able to rescue the pedicel growth phenotype or stomatal phenotype (stomatal index) of pub30 pub31 ( Fig. 3f-i). Therefore, the E3 ligase activity of PUB30/31 is indeed required for proper pedicel elongation and stomatal development.
Finally, to address whether PUB30/31 target ligand-activated ERECTA, we examined whether EPF2/EPFL6 peptide perception induces the degradation of ERECTA. To this end, we performed peptide treatment experiments using bioactive as well as inactive (heat denatured) EPF2/EPFL6 peptides. As shown in Extended Data Fig. 3f,g, treatment of bioactive EPF2 or EPFL6, but not their inactive peptides, significantly decreased the accumulation of ERECTA protein in the 'WT' seedlings. By contrast, treatment of either active or inactive EPF2/EPFL6 peptides conferred much less change for the ERECTA protein abundance in the pub30 pub31 mutant backgrounds. Based on these findings, we conclude that PUB30 and PUB31 downregulate the accumulation of ligand-stimulated ERECTA proteins via ubiquitination.

Coreceptor of ERECTA, BAK1, interacts with PUB30 and PUB31
EPF/EPFL ligands trigger the active receptor complex formation of ERECTA and its coreceptor BAK1 (ref. 20). We thus sought to decipher the regulatory relationships between BAK1 and PUB30/31. First, we asked whether BAK1 could directly interact with PUB30/31. As show in Fig. 4a, the cytosolic domain of BAK1 fused with DNA-binding domain (BD-BAK1_CD) interacts with PUB30/31 (AD-PUB30/31) in the Y2H assays. Further domain analysis in yeast implies that the domains of PUB30/31 encompassing the linker region mediates the association with BAK1 and ERECTA and that the U-box domains are dispensable for the interaction (Extended Data Fig. 4). The in vitro pull-down assays confirmed the interaction between recombinant PUB30/31 (GST-PUB30/31) with BAK1_CD (MBP-BAK1_CD) (Extended Data Fig. 5a,b). To quantitatively characterize the interaction property of BAK1_CD with PUB30/31, we further performed the BLI assays (Fig. 4b,c). Compared with ERECTA_CD (Fig. 2b, c), BAK1_CD exhibited stronger physical interaction in the order of magnitude with PUB30 and PUB31 (Fig. 4b,c). We next performed co-IP analyses to investigate the in vivo association of BAK1 with PUB30 and PUB31 in Arabidopsis. Just like the in vivo interaction of ERECTA with PUB30/31 (Fig. 2d,e), BAK1 was weakly detected in the absence of the peptide treatment. Upon EPF2 or EPFL6 peptide incubation, however, BAK1 strongly associated with PUB30 and PUB31 ( Fig. 4d and Extended Data Fig. 5c).
It has been reported that EPF/EPFL signal perceived by ERECTA-BAK1/SERKs is subsequently transduced via BSK1/2 and YODA MAPK cascade 10,52 . To address the extent to which PUB30/31 associate with the ERECTA signalling components, we expanded our proteinprotein interaction assays. As shown in Extended Data Fig. 5d, no interaction of PUB30/31 with BSK1/2 as well as YODA was detected by Y2H. Combined, our results demonstrate that the coreceptor of ERECTA, BAK1, also interacts with PUB30/31 in the EPF/EPFL ligand-stimulated manner and suggest that the regulation of PUB30/31 activity likely occurs at the level of the active receptor complex but not further downstream components.
To further identify the exact residue(s) of PUB30/31 phosphorylated by BAK1, we performed liquid chromatography tandem mass spectrometry analysis after an in vitro phosphorylation reaction using MBP-BAK1_CD as kinase and GST-PUB30 as substrate. The threonine 155 (T155) residue, located in the linker domain of PUB30, which serves as interaction domain with BAK1 and ERECTA, was identified as a phosphosite (Extended Data Figs. 4 and 6b, and Supplementary Table  1). The T155 in PUB30 is conserved in PUB31 as threonine 151 (T151) (Extended Data Fig. 2e). These threonine residues were replaced by alanines (PUB30 T155A or PUB31 T151A ) to confirm that they are the major phosphosites. Indeed, GST-PUB30 T155A and GST-PUB31 T151A were less phosphorylated by BAK1_CD in vitro (Extended Data Fig. 6c,d). Furthermore, we performed an in vivo phosphorylation assay to confirm the Article https://doi.org/10.1038/s41477-022-01303-x importance of these phosphorylation sites. After incubation together with BAK1 and treatment with EPFL6, the phosphorylation level of WT PUB30/31 was much higher than that of phosphonull versions of PUB30/31 (Extended Data Fig. 6e). These results further support that single amino acid residues in the linker domain of PUB30/31 are the major in vitro phosphosites by BAK1. . Asterisks indicate statistical significance using two-tailed paired Student's t-test (P = 0.0012). e, PUB30 or PUB31 mediates ERECTA ubiquitination in vivo. Arabidopsis protoplasts were cotransfected with ERECTA-HA, FLAG-UBQ and a control vector or PUB30-MYC or PUB31-MYC. Five micromolar EPFL6 was used for treatment for 1 h. After immunoprecipitation using anti-FLAG beads, the ubiquitinated ERECTA (Ubn-ERECTA) was probed with anti-HA antibody. The total ubiquitinated proteins were probed by anti-FLAG antibody and PUB30 or PUB31 proteins were probed by anti-MYC antibody. The inputs of ERECTA were probed with anti-HA antibody. f, Representative pedicels of mature siliques of pub30 pub31, proPUB30::PUB30 W97A ; pub30 pub31, proPUB31::PUB31 W93A ; pub30 pub31 and WT plants. Scale bar, 1 cm. g, Confocal microscopy of 8-day-old abaxial cotyledon epidermis of pub30 pub31, proPUB30::PUB30 W97A ; pub30 pub31, proPUB31::PUB31 W93A ; pub30 pub31 and WT plants. Scale bar, 25 μm. h, Morphometric analysis of pedicel length from each genotype. Mature pedicels (n = 15) from 6-week-old plants were measured. One-way ANOVA followed by Tukey's HSD test was performed and classified their phenotypes into categories (a and b). For P values see Extended Dataset 1. i, Quantitative analysis. Stomatal index of the cotyledon abaxial epidermis from 8-day-old seedlings of respective genotypes (n = 7). One-way ANOVA followed by Tukey's HSD test was performed and classified their phenotypes into categories (a and b). For P values see Supplementary Data. Article https://doi.org/10.1038/s41477-022-01303-x

BAK1 is required for ERECTA-PUB30/31 interaction
We have shown that EPF/EPFL ligand perception intensifies the association of ERECTA as well as BAK1 with PUB30/31, leading to subsequent ubiquitination and degradation of ERECTA by PUB30/31 (Figs. 2-4). These findings suggest that the ligand-activated ERECTA-BAK1 receptor complex recruits and activates PUB30/31. To address this question, we first examined whether BAK1 is required for the interactions of ERECTA and PUB30/31. For this purpose, we performed in vivo co-IP experiments using protoplasts from the serk1 bak1-5 double mutant, which confers a stomatal development phenotype but not severe growth defects 20 . We observed a striking reduction in the association of ERECTA-HA with PUB30-MYC as well as PUB31-MYC in serk1 bak1-5 background (Extended Data Fig. 7a). The results indicate that BAK1 (and SERK1) is required for ERECTA-PUB30/31 interaction.
Next, to address whether the phosphorylation of PUB30/31 at the T151/T155 residues by BAK1 promotes the association PUB30/31 with ERECTA, we performed a series of in vitro and in vivo assays. The in vitro pull-down experiments showed that the phosphomimetic versions of PUB30/31 exhibit stronger interaction with ERECTA_CD than their WT versions (Extended Data Fig. 7b,c). Subsequently, we performed the in vivo co-IP experiments using Arabidopsis protoplasts expressing the epitope-tagged ERECTA and WT, phosphomimetic and phosphonull versions of PUB30/31 (Methods). The association of ERECTA-HA with PUB30 T155A -MYC as well as PUB31 T151A -MYC was markedly reduced compared with the WT versions of PUB30/31-MYC ( Fig. 5a and Extended Data Fig. 7d,e). By contrast, the phosphomimetic mutant PUB30 T155D -MYC and PUB31 T151D -MYC exhibited stronger interactions with ERECTA than the WT PUB30 and PUB31, respectively ( Fig. 5a and Extended Data Fig. 7d,e). Combined, these results indicate that BAK1-mediated phosphorylation of PUB30/31 intensifies their association with ERECTA.

ERECTA ubiquitination-degradation by PUB30/31 require BAK1
To investigate whether phosphorylation of PUB30/31 by BAK1 is required for the ubiquitination of ERECTA by PUB30/31, we first performed in vitro ubiquitination assays of MBP-ERECTA_CD by purified recombinant GST-PUB30/31 as well as the phosphomimetic (PUB30 T155D Article https://doi.org/10.1038/s41477-022-01303-x and PUB31 T151D ) and phosphonull (PUB30 T155A and PUB31 T151A ) versions. The ubiquitination of ERECTA_CD by PUB30 T155D and PUB31 T151D was stronger than that by WT PUB30 and PUB31, respectively (Extended Data Fig. 7f,g). By contrast, the ubiquitination of ERECTA_CD by PUB30 T155A and PUB31 T151A was slightly reduced compared with the WT versions of PUB30/31 (Extended Data Fig. 7f,g). No difference was observed for the auto-ubiquitination of PUB30/31 WT, phosphomimetic and phosphonull versions, as detected by anti-GST immunoblots (Extended Data Fig. 7f,g). Therefore, although we cannot fully exclude the possibility that these phosphomimetics/phosphonull mutations may affect the protein structure, our results indicate that the phosphorylation of PUB30/31 by BAK1 may not have a direct effect on their ubiquitination activity per se. We next performed in vivo ubiquitination assays in Arabidopsis protoplasts upon EPFL6 peptide application, the condition that triggers polyubiquitination of ERECTA (Fig. 3e). As evidenced by the reduced ladder-like smear formation, the phosphonull mutants, PUB30 T155A -MYC and PUB31 T151A -MYC, conferred reduced ubiquitination of ERECTA-HA compared with WT PUB30-MYC and PUB31-MYC, respectively (Fig. 5b). By contrast, the phosphomimetic mutants, PUB30 T155D -MYC and PUB31 T151D -MYC caused increased ubiquitination on ERECTA-HA compared with WT versions of PUB30/31 (Fig. 5b). Combined, these results support that the phosphorylation of PUB30/31 linker domains promotes the ubiquitination of ERECTA.
To unambiguously address whether BAK1 is required for PUB30/31-mediated ubiquitination of ERECTA, we further performed the in vivo ubiquitination assays using Arabidopsis WT and serk1 bak1-5 protoplasts. As evidenced by the reduced ladder-like smear formation, PUB30-MYC and PUB31-MYC conferred reduced ubiquitination of ERECTA-HA in the serk1 bak1-5 background (Fig. 5c). By contrast, the phosphomimetic mutants, PUB30 T155D -MYC and PUB31 T151D -MYC caused comparable ubiquitination of ERECTA-HA regardless of the presence or absence of BAK1 (and SERK1) (Fig. 5c).
Finally, we tested the effects of PUB30/31 phosphorylation by BAK1 on degradation of ERECTA protein. For this purpose, we co-expressed ERECTA and WT, phosphomimetic and phosphonull versions of PUB30/31 in Arabidopsis protoplast and performed cotreatment with EPFL6 and cycloheximide (de novo protein synthesis inhibitor). Upon EPFL6 treatment, phosphomimetic mutants, PUB30 T155D -MYC and PUB31 T151D -MYC caused a distinct decrease in ERECTA-HA protein level (Extended Data Fig. 7h,i). By contrast, phosphonull mutants, PUB30 T155A -MYC and PUB31 T151A -MYC conferred less decrease in ERECTA-HA protein level (Extended Data Fig. 7h,i). Taken together, our results show that the BAK1 phosphorylation of PUB30 and PUB31 at T155 and T151 residues, respectively, facilitates the ubiquitination and subsequent degradation of ERECTA.

BAK1 is not a substrate of PUB30/31
Our experimental evidence supports the model that BAK1 plays a key role in mediating the downregulation of ligand-activated ERECTA by PUB30/31 to attenuate signal transduction. The obvious and important question is whether BAK1 itself is also a substrate of PUB30/31. To address this question, we first examined whether PUB30/31 regulate the stability of BAK1 in vivo. Abundance of the endogenous BAK1 proteins, as detected by anti-BAK1 antibody, is similar in WT and pub30 pub31 seedlings (Extended Data Fig. 8a), indicating that PUB30/31 do not have discernible effects on BAK1 protein levels. Next, we performed in vitro ubiquitination assays of BAK1. Compared with the positive control ERECTA (MBP-ER_CD), no laddering bands of BAK1 (MBP-BAK1_CD) were detected when co-incubated with PUB30/31 proteins (GST-PUB30 and GST-PUB31), His-UBC8 (E1) and His-UBA1 (E2) (Extended Data Fig. 8b,c), indicating that BAK1 is not ubiquitinated by PUB30/31 in vitro.
Because BAK1 directly phosphorylates PUB30/31 ( Fig. 4 and Extended Data Fig. 6), we further tested whether phosphorylation status of PUB30/31 affects its interaction with BAK1 by in vitro pull-down experiments. As shown in Extended Data Fig. 8d,e, both the WT and phosphomimetic versions of PUB30/31 exhibit similar interaction with BAK1_CD. We further test whether the kinase activity of BAK1 is essential for its interaction with PUB30/31. Again, the in vitro pull-down experiments showed that BAK1_CD and BAK1_CDmut had similar interaction with PUB30/31 (Extended Data Fig. 8f,g). On the basis of these findings, we conclude that BAK1 is unlikely a substrate of PUB30/31 but rather an indispensable kinase and a scaffold to bring PUB30/31 to a proximity of ligand-activated ERECTA.

PUB30/31 biological functions require phosphorylation by BAK1
We have shown that BAK1-mediated phosphorylation of PUB30 and PUB31 is critical for subsequent association with, and ubiquitination and degradation of ERECTA. Do PUB30/31 phosphorylation events affect their in vivo functions in plant development? To evaluate the contribution of PUB30 T155 and PUB31 T151 phosphorylation on their biological functions, we introduced the phosphomimetic and phosphonull versions of PUB30/31 driven by their native promoters into the pub30 pub31 mutant. Transgenic plants expressing proPUB30::PUB30 T155A and proPUB31::PUB31 T151A failed to rescue either pedicel growth or stomatal phenotype of pub30 pub31 (Fig. 5d-g). By contrast, transgenic plants expressing proPUB30::PUB30 T155D and proPUB31::PUB31 T151D fully rescued the mutant phenotypes, both in the context of pedicel growth and stomatal index (Fig. 5d-g). To exclude the possibility that the absence of phenotypic rescues by the phosphonull mutants of PUB30/31 may be attributed to their reduced protein accumulation, we examined the protein expression levels in these transgenic lines. The PUB30 T155A /31 T151A protein and transcript levels are comparable (slightly higher than) with the phosphomimetic or WT versions of PUB30/31 (Extended Data Fig. 9), although they failed to rescue the pub30 pub31 mutant phenotypes. Collectively, our results highlight that the BAK1-mediated phosphorylation of PUB30 and PUB31 is required for their biological functions in regulating inflorescence/pedicel growth and stomatal development.

Discussion
In this study, we identified a pair of U-box E3 ligases, PUB30 and PUB31, as conserved negative regulators of ERECTA. Our genetic, molecular and biochemical analyses place PUB30/31 as integral components of the regulatory circuit that fine-tunes the ERECTA signalling outputs (Fig. 6). Upon EPF/EPFL ligand perception by ERECTA, ERECTA forms an active receptor complex with BAK1, which directly phosphorylates PUB30/31. This triggers PUB30/31 to ubiquitinate ERECTA, but not BAK1, for degradation (Fig. 6). Following activation, the ERECTA-BAK1 complex relays the signal through BSKs, the MAPK cascade and then the downstream factors to modulate developmental outcomes 23,52,54 . We propose that negative regulation of ERECTA signalling by PUB30/31 ensures robust yet appropriate signalling strengths upon the ligand perception (Fig. 6).
Both ERECTA-mediated inflorescence/pedicel growth and stomatal development are negatively regulated by PUB30 and PUB31 in a largely redundant manner (Figs. 1, 6 and Extended Data Fig. 1). The mutations that abolish the E2-binding (PUB30 W97A and PUB31 W93A ) as well as the BAK1-mediated phosphosites (PUB30 T155A and PUB31 T151A ) uniformly failed to rescue both extremely elongated pedicel and reduced stomatal index phenotypes of the pub30 pub31 plants (Figs.  3 and 5). Thus, whereas each EPF/EPFL peptide ligand elicits a unique developmental response via ERECTA, once the ligands are perceived, subsequent processes of signal activation and attenuation are likely conserved. This appears to be the case for the EPF2-mediated inhibition of stomatal development (Fig. 6a) and the EPFL6-mediated elongation of inflorescence and pedicels (Fig. 6b). Nonetheless, the ERECTA receptor complex harbours intricate unequal redundancy: ERECTA-LIKE1 (ERL1) and ERL2, two paralogous receptors, synergistically function with ERECTA and form active receptor complexes with Article https://doi.org/10.1038/s41477-022-01303-x SERK1, SERK2, SERK3/BAK1 and SERK4 (refs. 8, 9 , 20). In addition to inflorescence/pedicel and stomatal phenotypes, pub30 pub31 mutant exhibits seedling hyperplasia (Extended Data Fig. 10), suggesting that PUB30/31 may restrict cotyledon and leaf size. Exploring the molecular and biochemical basis of such a phenotype may reveal the roles of these additional paralogues of EPF/EPFL, ERECTA family and BAK1/ SERKs family members.
PUB30/31 ubiquitinate the cytoplasmic domain of ERECTA (Extended Data Fig. 2f,g). This explains the previous finding that a truncated ERECTA protein lacking the entire cytoplasmic domain (ERECTAΔK) accumulates at high levels, thereby causing dominant-negative effects 39 . ERECTAΔK is unable to transduce signals (owing to the lack of the kinase domain) or be turned over via PUB30/31-mediated ubiquitination. A more recent study has shown that ERL1, after activation by the EPF1 peptide, undergoes rapid internalization via multivesicular bodies/late endosomes to vacuolar degradation 55 . Again, truncated ERL1 lacking the entire cytoplasmic domain (ERL1ΔK) is stably accumulated at the plasma membrane, irrespective of the ligand perception 55 . It is worth noting that endocytosis of BRI1 relies on the PUB12/13-mediated ubiquitination 31 . Because we found that the inhibitors of clathrin-mediated endocytosis and vacuolar trafficking/degradation inhibit PUB30/31-mediated ERECTA degradation ( Fig. 3c and Extended Data Fig. 3b-d), it is likely that ERECTA turnover is coupled with endocytosis.
Our work broadens the roles of PUB proteins as a regulator of receptor kinase signalling and highlights the similarities and differences in their exact mode of actions. We found that BAK1 phosphorylates PUB30/31 at their linker domain (T155 and T151, respectively; Fig. 4). This echoes the idea of phosphorylation as a key activation mechanism of PUB proteins by LRR-RKs and other signalling kinases.
For example, MPK3 interacts with and phosphorylates another U-box E3 ligase PUB22 (ref. 56); one of which two phosphosites, T88, lies in the linker domain. BRI1 phosphorylates PUB13 at S344, which also falls in the linker domain 31 . This phosphorylation event subsequently facilitates the turnover of BRI1. Most importantly, we found that BAK1 is required for the association of ERECTA and PUB30/31 as well as further ubiquitination of ERECTA by PUB30/31. Whereas BAK1 phosphorylates and activates PUB30/31, BAK1 itself is not ubiquitinated or subjected to degradation by PUB30/31 (Figs. 4, 5 and Extended Data Fig. 8). The action of BAK1 as a kinase and a scaffold is consistent with its higher binding affinity with PUB30/31 (Fig. 4b,c) and further emphasizes its role as a 'universal' coreceptor, allowing the activation and tuning of ligand-perceiving receptors (for example, ERECTA) while maintaining its own homeostasis (Fig. 6). Notably, this scaffold mechanism may also apply to other PUB proteins and their interacting kinases. For example, it has been shown that PUB1, a negative regulator of Medicago symbiosis, as well as PUB22 associates with and phosphorylated by signalling kinases, LYK3 (a lysin motif RLK) and MPK3, respectively 56,57 . In either case, however, LYK3 and MPK3 are not ubiquitinated by their interacting PUB proteins 56,57 . Therefore, they may also serve as scaffolds to recruit other components in the signalling pathway to be ubiquitinated by the PUB proteins.
The exact steps of phosphorylation-ubiquitination events differ among the known kinase-PUB signalling modules. For instance, PUB13 ubiquitinates FLS2 upon phosphorylation of PUB13 by BAK1, which in turn strengthens the interaction of PUB13 and FLS2 (ref. 30). By contrast, PUB13 ubiquitinates BRI1 in a BAK1-independent manner 31 . Whereas PUB13 marks elicitor-activated FLS2, PUB25/26 specifically target non-activated BOTRYTIS-INDUCED KINASE1 (BIK1), a receptor-like cytoplasmic kinase, for degradation 58 Article https://doi.org/10.1038/s41477-022-01303-x regulatory mechanism of the EPF/EPFL-ERECTA-BAK1-PUB30/31 circuit resembles that of flg22-FLS2-BAK1-PUB12/13 but differs from the ubiquitination of BRI1 and BIK1 by PUB12/13 and PUB25/26, respectively. Our study revealed the unique aspects of PUB30/31 actions. Besides the phosphorylation-ubiquitination events among the kinase-PUB signalling modules, the phosphorylation status of E3s may also affect protein stability. For instance, phosphorylation of PUB22 by MPK3 stabilizes PUB22 by inhibiting its oligomerization and auto-ubiquitination 56 . By contrast, we found that the phosphorylation of PUB30/31 by BAK1 does not discernibly change their auto-ubiquitination (Extended Data Fig. 7f,g). In addition, the protein levels of phosphonull and phosphomimetic PUB30/31 are consistent with the transcript levels (Extended Data Fig. 9). Thus, our results suggest that the phosphorylation by BAK1 may not be acting as a degron-like signal on PUB30/31. It would be important to note that, structurally, PUB30/31 belong to a different subclade from PUB12/13, and such structural differences may be contributing to the unique mode of PUB30/31 actions. For example, unlike PUB12/13, PU30/31 lack the U-box N-terminal Domain (UND) 37,60 . Furthermore, based on the AlphaFold Structural Database 61 , the PUB30/31 linker region is predicted to adopt a single alpha helix, which is not as flexible as the inker domain in other PUB proteins. This structural feature may account for the low affinity and processivity in vitro of PUB30/31 to ubiquitinate ERECTA. Future biophysical and structural analyses may decipher the specificity and uniqueness of distinct class of PUB proteins in regulating receptor kinase signalling.
Whereas PUB30/31 together regulate ERECTA-mediated developmental processes, PUB30, but not PUB31, has been shown to mediate salt stress tolerance via interacting with and ubiquitinating BRI KINASE INHIBITOR 62 . This raises an important question of what mechanisms govern the target specificities among the PUB proteins. In this regard, it is worth mentioning that PUB12/13 also interact with and ubiquitinate ABA-insensitive 1 (ABI1), an abscisic acid (ABA) coreceptor, which belongs to a protein phosphatase 2C family regulating ABA signalling and drought response 63 . However, PUB12/13 do not ubiquitinate ERECTA, which is structurally resembling to their bona fide target FLS2 rather than ABI 31 . Searching for the interactors and resolving the structural basis of association with otherwise unrelated targets will shed light on the versatile roles of PUB proteins in signal transduction pathways underpinning development and environmental responses.

Plant materials and growth conditions
The Arabidopsis accession Columbia (Col) was used as the wild type.

Quantitative PCR with reverse transcription analysis
RNA extraction, cDNA synthesis and quantitative PCR with reverse transcription (RT-qPCR) were performed as previously described 65 . For a list of primers, see Extended Data Table 3.

Histological analysis and microscopy
For histological analysis, mature pedicels were fixed, dehydrated and embedded in polymethacryl resin Technovit 7100 (Heraeus Kulzer) as described previously 39 . Tissue sections were prepared using a Leica RM-6145 microtome, and tissue sections were stained with 0.1% toluidine blue (Sigma) in 0.1 M NaPO 4 buffer (pH 7.0) and observed under Olympus BX40 light microscope. Confocal microscope images were taken using either a Zeiss LSM700 operated by Zen2009 (Zeiss) as described previously 19 or a Leica SP5-WLL operated by LAS AF (Leica). Cell peripheries of seedlings were visualized with propidium iodide (Molecular Probes). Images were taken with excitation at 514 nm and emission at 518-600 nm for YFP, and excitation at 619 nm and emission at 642 nm for propidium iodide. For Lecia SP5 confocal microscope, a HyD detector was used. The confocal images were false coloured, and brightness/contrast were uniformly adjusted using Photoshop 2021 (Adobe). The Z-stack projection images were taken at an interval of 0.99 μm, covering the thickness of the entire cotyledon.

Quantitative analysis and statistics
For analysis of epidermis, abaxial cotyledons from 10-, 6-or 8-d-old seedlings of respected genotypes were subjected to propidium iodide staining and confocal microscopy. The central regions overlying the distal vascular loop were imaged and numbers of epidermal cells, stomata and their cluster size were quantified. Pedicel lengths were measured using ImageJ. Statistical analysis was performed using R v.4.1.0 operated under R-Studio v.1.4.1717 (https://www.rstudio. com), and graphs were generated using the R ggplot2 package. For all boxplots, each box represents interquartile range, in which the top, middle and bottom lines indicate the 75, 50 and 25 percentiles, respectively, and the bar represents maximum to minimal values. Each dot represents a jittered individual data point. Data points that go beyond the bar are outliers. Violin plots (Fig. 1e) show the data and probability density distribution with individual data points overlaid as jittered dots. For all analyses, Welche's unpaired t-test or Student's t-test were performed for a pairwise comparison, and one-way analysis Article https://doi.org/10.1038/s41477-022-01303-x of variance (ANOVA) followed by Tukey's HSD test were performed for comparison of groups. For individual sample size (n) and P values, see the corresponding figures or figure legends.

Yeast two-hybrid assay
Bait and prey constructs were cotransformed into the yeast strain AH109 using the yeast transformation kit (Frozen-EZ Yeast Transformation II Kit, Zymo Research). The resulting transformants with appropriate positive and negative controls were spotted on SD (−Leu, −Trp) plates to check for growth in the absence of selection. Transformants were then spotted on SD (−Trp, −Leu, −His) selection media containing 1 mM 3-amino-1,2,4-triazole (3-AT, A8056, Sigma). The positive interactors were then scored based on the stringency of the selection.
For pull-down assay of PUB30/31 phosphomimetic and WT versions with ER_CD, ~15 μg of GST-PUB30T155D or GST-PUB30 or GST-PUB31T151D or GST-PUB31 or GST proteins were incubated with ~15 μg of MBP-ERECTA_CD protein in 900 μl of pull-down buffer. For pull-down assay of PUB30/31 with BAK1_CD and BAK1_CD mut, ~15 μg of GST-PUB30 or GST-PUB31 proteins were incubated with ~15 μg of MBP-BAK1_CD or MBP-BAK1_CD mut or MBP proteins in 900 μl of pull-down buffer. For these pull-down assays, 30 μl of MBP beads (amylose resin; New England Biolabs, catalogue no. E8021S) were incubated with each reaction mixture with gentle shaking at 4 °C for ~1 h.
After the reaction, beads were washed three times and heated for 5 min in a 95 °C metal bath. The immunoprecipitated proteins were separated by SDS-PAGE and detected by anti-GST (1:5,000; GenScript, catalogue no. A00865-200) and anti-MBP (1:10,000; New England Biolabs, catalogue no. E8032) antibodies, respectively. As a secondary antibody, sheep anti-mouse IgG horseradish peroxidase-linked antibody (GE Healthcare, catalogue no. NA931) was used at a dilution of 1:50,000. The protein blots were visualized as described in the previous section.
For in vivo ubiquitination assays, Arabidopsis protoplasts were cotransfected with FLAG-tagged ubiquitin (FLAG-UBQ), HA-tagged ERECTA and together with a control vector or MYC-tagged PUB30 or PUB31 (WT or various phosphor-mutated) and incubated for 8 h followed by treatment with 5 μM EPFL6 for 1 h in the presence of 2 μM MG132 (Sigma, catalogue no. M7449). The ubiquitinated ERECTA was detected with anti-HA (1:1,000; Abcam, catalogue no. ab18181) immunoblotting after immunoprecipitation with anti-FLAG (Abcam, catalogue no. ab205606) antibody. The total ubiquitinated proteins were detected by anti-FLAG (1:5,000; Sigma, catalogue no. F-3165) and anti-MYC (1:1,000; Abcam, catalogue no. ab32) as primary antibodies. As secondary antibody, sheep anti-mouse IgG horseradish peroxidase-linked antibody (GE Healthcare, catalogue no. NA931) was used at a dilution of 1:50,000. The protein blots were visualized as described in the previous section.
The reactions were incubated at 30 °C for 30 minutes and stopped with the SDS sample buffer. After SDS-PAGE, gels were dried, exposed to a GE multipurpose standard screen (catalogue no. 63-0034-87) for 18 h and imaged using a GE Typhoon FLA 9000 gel imager.

Mass spectrometry and identification of the phosphosites
For the in vitro phosphorylation reaction, 1 μg of MBP-BAK1_CD protein was incubated with 10 μg of GST-PUB30 in 30 μl of kinase reaction buffer at 30 °C for 3 h (with gentle shaking). After the reaction, the SDS loading buffer was used to stop the kinase reactions. Samples were separated by 10% SDS-PAGE. The gels were fixed for 60 min in 50% methanol + 7% acetic acid, rinsed thoroughly with Milli Q water and stained with GelCode Blue Stain Reagent (ThermoFisher Scientific, catalogue no. 24590). Target bands for GST-PUB30 are cut off from the electrophoresis gel and digested using chymotrypsin (Sigma) at 37 °C overnight. To analyse the chymotryptic peptides, nano-flow reverse-phase liquid chromatography and tandem mass spectrometry was performed using a Q Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer (ThermoFisher Scientific) as described previously 66 . Subsequently, peptide identification was performed by searching the Arabidopsis thaliana reference genome (downloaded from https:// www.uniprot.org) using the SEQUEST (ThermoFisher Scientific) search engine. The parameter of dynamic modifications with phosphorylation filter was added for the identification of phosphopeptides. Details of the peptide spectrum match with phosphorylated residues were validated manually and annotated in Thermo Proteome Discoverer v.2.4.

In vivo phosphorylation assay
Arabidopsis protoplasts were cotransfected with HA-tagged BAK1 (BAK1-HA), and MYC-tagged WT and phosphonull versions of PUB30 or PUB31 and incubated for 8 h followed by treatment with 5 μM EPFL6 for 1 h in the presence of 2 μM MG132 (M7449, Sigma).

Peptide and inhibitor treatment and immunoblot assays
The proERECTA::ERECTA-FLAG er-105 and proERECTA::ERECTA-FLAG pub30 pub31 er-105 seedlings were grown vertically at 22 °C on half-strength Murashige and Skoog medium for 7 d. Thereafter, different treatments were further performed. For the endocytosis inhibitor treatment, seedlings were incubated with or without 50 μM Tyrphostin A23 (Tyr A23) (Sigma, catalogue no. T7165) for 30 min. For the vacuolar ATPase inhibitor treatment, seedlings were incubated with or without 1 μM Con A (Abcam, catalogue no. ab144227) for 30 min. For peptides EPF2/EPFL6 treatment, seedlings were incubated with or without 5 μM peptides for 180 min. Inactive (boiled at 70 °C for 30 min before use) EPF2/EPFL6 were used as negative controls.
Total protein extracts were separated on a 10% SDS-polyacrylamide gel and detected by immunoblot analysis with anti-FLAG (1:5,000; Sigma, catalogue no. F-3165) and anti-actin (1:2,000; Abcam, catalogue no. ab230169) as primary antibodies. As secondary antibody, goat anti-mouse IgG H&L (HRP) (Abcam, catalogue no. ab205719) was used at a dilution of 1:5,000. Protein blots were visualized as described in the previous section.

Biolayer interferometry
The binding affinities of the ERECTA_CD with GST-tagged PUB30 and PUB31 were measured using the Octet Red96 system (ForteBio, Pall Life Sciences) following the manufacturer's protocols. The optical probes coated with anti-GST were first loaded with 2,000 nM GST-PUB30 or PUB31 before kinetic binding analyses. The experiment was performed in 96-well plates maintained at 30 °C. Each well was loaded with 200 μl of reaction volume, and the binding buffer used in these experiments contained 1× PBS (pH 7.2) supplemented with 0.02% Tween 20. The concentrations of the ERECTA_CD as the analyte in the binding buffer were 20,000, 10,000, 5,000, 2,500, 1,250, 625 and 312.5 nM. Similarly, for the binding of BAK1-CD with GST-PUB30/31, the optical probes coated with anti-GST were first loaded with 1,000 nM GST-PUB30 or PUB31 before kinetic binding analyses. The concentrations of the BAK1_CD as the analyte in the binding buffer were 5,000, 2,500, 1,250, 625, 312.5, 156.3 and 78.2 nM. All preformed complexes remained stable as suggested by the constant signal during the washing step after loading. There was no binding of the analytes to the unloaded probes as shown by the control wells. Binding kinetics to all seven concentrations of the analytes were measured simultaneously using default parameters on the instrument. The data were analysed using the Octet data analysis software. The association and dissociation curves were fit with the 1:1 homogeneous ligand model. The k obs (observed rate constant) values were used to calculate dissociation constant (K d ), with steady-state analysis of the direct binding.

ERECTA protein stability in protoplasts
To determine ERECTA protein stability, protoplasts cotransfected with ERECTA-HA and phosphor-mutant and WT versions of PUB30/31-MYC were treated with 50 μM cycloheximide (Sigma, catalogue no. C4859) in the presence or absence of 5 μM EPFL6 for 3 h. Total proteins were separated on SDS-PAGE gels and transferred to a PVDF membrane (Millipore) for immunoblot analysis. ERECTA protein and the input PUB30 or PUB31 proteins were detected with anti-HA antibody (1:1,000; Abcam, catalogue no. ab18181) and anti-MYC antibody (1:1,000; Abcam, catalogue no. ab32) and as primary antibodies, respectively. As secondary antibody, goat anti-mouse IgG H&L (HRP) (Abcam, catalogue no. Article https://doi.org/10.1038/s41477-022-01303-x ab205719) was used at a dilution of 1:5,000. Protein blots were visualized as described in the previous section.

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

Data availability
All generated and analysed data from this study are included in the main figures, Extended Data figures and supplementary information. Source data are provided with this paper. Corresponding author(s): TORII, Keiko U.
Last updated by author(s): Oct 24, 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

Software and code
Policy information about availability of computer code Data collection Image J (version 1.53c), Image Lab (Bio-Rad, version 6.0.1) were used in data collection. The data of biolayer interferometry (BLI) were collected using the Octet data acquisition software 10.0.

Data analysis
Statistical analysis of phenotypes experiments was performed using R ver. 4.1.0 operated under R-Studio ver. 1.4.1717 (https:// www.rstudio.com), and graphs were generated using R ggplot2 package. The data of biolayer interferometry (BLI) were analyzed using the ForteBio data analysis software (Octet BLI Analysis 10.0). Statistical analysis of biochemistry experiments was performed using GraphPad Prism 3.1. The MS/MS spectra were analyzed with SEQUEST (ThermoFisher Scientific). Images were cropped using Adobe Photoshop 2022.
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.

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 generated and analyzed data from this study are included in the main figures, Extended Data figures and supplementary information. Source data are provided with this paper.