Phytocystatin 6 is a context‐dependent, tight‐binding inhibitor of Arabidopsis thaliana legumain isoform β

SUMMARY Plant legumains are crucial for processing seed storage proteins and are critical regulators of plant programmed cell death. Although research on legumains boosted recently, little is known about their activity regulation. In our study, we used pull‐down experiments to identify AtCYT6 as a natural inhibitor of legumain isoform β (AtLEGβ) in Arabidopsis thaliana. Biochemical analysis revealed that AtCYT6 inhibits both AtLEGβ and papain‐like cysteine proteases through two separate cystatin domains. The N‐terminal domain inhibits papain‐like proteases, while the C‐terminal domain inhibits AtLEGβ. Furthermore, we showed that AtCYT6 interacts with legumain in a substrate‐like manner, facilitated by a conserved asparagine residue in its reactive center loop. Complex formation was additionally stabilized by charged exosite interactions, contributing to pH‐dependent inhibition. Processing of AtCYT6 by AtLEGβ suggests a context‐specific regulatory mechanism with implications for plant physiology, development, and programmed cell death. These findings enhance our understanding of AtLEGβ regulation and its broader physiological significance.


INTRODUCTION
Arabidopsis thaliana expresses four isoforms of the cysteine protease legumain (C13 family, EC 3.4.22.34)denoted as AtLEGa, -b, -c, and -d.Due to their organ-specific roles, they are assorted as vegetative-(AtLEGc and AtLEGa) and seed-type legumains (AtLEGb and AtLEGd) (Gruis et al., 2002;Kinoshita et al., 1995).Together they are accountable for a myriad of defense and developmental responses such as the physiological and stress-induced programmed cell death (PCD), seed papain-like cysteine proteases (PLCP, clan C1) storage mobilization and processing of vacuolar proteins, the latter leading to their synonymous naming as vacuolar processing enzymes (VPEs) (Cheng et al., 2019;Hatsugai et al., 2004;Kinoshita et al., 1999;Kuroyanagi et al., 2005;Li et al., 2012;Nakaune et al., 2005;Rojo et al., 2003Rojo et al., , 2004)).AtLEGb is the most relevant isoform for seed storage processing and mobilization.Naturally occurring A. thaliana specimens with dysfunctional atlegb genes showed aberrant seed storage processing (Shimada et al., 2003).Furthermore, AtLEGbnull mutants display impaired pollen fertility and tapetal cell degradation (Cheng et al., 2020).AtLEGb is synthesized as an inactive proenzyme (proAtLEGb) consisting of a caspase-like catalytic domain and a death-domain-like Cterminal prodomain.Activation proceeds via the autocatalytic, pH-dependent removal of the prodomain.Activated AtLEGb specifically cleaves after asparagine and to a lesser extent aspartate, residues.It is stable and proteolytically active at acidic pH and displays transpeptidase and ligase activity at near-neutral pHs; over time, it undergoes irreversible denaturation at pH >6.0 (Dall et al., 2020;Dall & Brandstetter, 2012;Zauner, Elsasser, et al., 2018).Given its versatile activities, tight regulation is required and may be conferred in vivo by cystatins.
Cystatins are reversible competitive inhibitors of PLCPs, which are well characterized in mammals but less Ó 2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd.This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.so in plants.Human cystatins are subdivided into type I, II, and III cystatins.Type I cystatins are intracellular, un-glycosylated small proteins lacking disulfide bonds, in contrast to type II cystatins which are secreted, typically glycosylated and disulfide-linked.Single-domain type II cystatins harbor a legumain inhibitory site, distinct from the PLCP site.Type III cystatins are multidomain cystatins with or without inhibitory activity encoded within their individual domains (Chaudhuri, 2010;Turk & Bode, 1991).Phytocystatins (phycys), the cystatins from plants, can be classified according to similar criteria as in mammals, except for type II phycys.Type II phycys are two-domain inhibitors that target PLCPs and legumains via two distinct cystatin-like domains (Martinez et al., 2007).While the Nterminal domain encodes PLCP inhibition, legumain inhibition is encoded in the C-terminal domain.The cystatin superfamily shares a highly conserved fold consisting of a five-stranded antiparallel b-sheet wrapped around a central a-helix (Bode et al., 1988).Cystatins bind to PLCPs in the so-called elephant-trunk binding mode, where the strictly conserved and critical inhibition determinants consist of an N-terminal glycine, the Q-X-V-X-G motif on the L1-loop, and the P-W motif on the L3 loop (Bode et al., 1988;Machleidt et al., 1989).By contrast, the binding mode of legumain and type II cystatins has been reported by Dall et al. as a substrate-like interaction between human cystatin E (hCE) and human legumain (hLEG) where the P1-Asn on the reactive center loop (RCL) of hCE inserts into the S1 pocket of hLEG, thus blocking the access of the substrate.The interaction with the enzyme is strengthened by the legumain exosite loop (LEL), which in hCE is long and predominantly hydrophobic (Dall et al., 2015).Furthermore, complex formation is tightly regulated by a pH-dependent equilibrium of RCL cleavage at acidic pH and re-ligation at near neutral pH.In spite of a relatively broad knowledge of the interaction of hLEG and human cystatins, there is little data available on the interplay of plant legumains and their endogenous regulators, for example, phytocystatins.We hypothesized that A. thaliana phytocystatins regulate AtLEGb activity and were able to characterize the type II phytocystatin 6 (AtCYT6) as a potent substrate-like inhibitor of the enzyme.The finding presented here on the interaction of AtCYT6 with AtLEGb contributes to our understanding of AtLEGb activity in vivo and to the identification of structural determinants of type II phytocystatins important for inhibition.

RESULTS
Phytocystatin 6 is a candidate inhibitor of Arabidopsis thaliana legumain b (AtLEGb) To study how the activity of AtLEGb is regulated in Arabidopsis thaliana we were looking out for potential inhibitors of its enzymatic activities.In mammals, legumain activity is inhibited by the endogenous family II cystatins, including cystatin C, M/E, and F (Alvarez-Fernandez et al., 1999).Similarly, type II phycys were shown in previous studies to inhibit legumain from rice (Christoff et al., 2016), barley (Julian et al., 2013), and humans (Martinez et al., 2007).Till now, however, not much is known about phycys in A. thaliana.In 2005, Martinez et al. identified seven phytocystatin candidate genes in A. thaliana (AtCYT1-7) through a comparative phylogenetic analysis (Martinez et al., 2005).Four of them, AtCYT1-3 and AtCYT6, have in the meantime been confirmed on protein level (Belenghi et al., 2003;Guo et al., 2013;Hwang et al., 2009Hwang et al., , 2010;;Zhang et al., 2008).To find out if and which of A. thaliana phytocystatin candidate genes harbor the characteristic tandem two-domain architecture, we used AlphaFold to predict their protein structures (Jumper et al., 2021;Varadi et al., 2022).Intriguingly, although AtCYT6 and AtCYT7 were expected to harbor a two-domain architecture based on their primary sequence (Figure S1), only AtCYT6 resulted in an Alpha-Fold model that contained two folded domains with cystatin-like architecture (Figure 1a; Figure S2).AtCYT7, as well as AtCYT1-5, contained only one fully folded cystatin domain at the N-terminal end.In the model, the characteristic carboxy-terminal extension of type II phycys, although present in the sequence of AtCYT7, was primarily constituted of disordered regions (not shown).Overall, the AtCYT6 AlphaFold model had a good per-residue confidence score, with 70% of all residues having a predicted local distance difference test (pLDDT) value higher than 70 (Figure S2a,c).This translated in confident (90 > pLDDT > 70) or very high confident predictions (pLDDT >90).The exceptions were the N-terminal region up to residue Gly39 (including the signal peptide) and the interdomain linker (Ala128-Glu144) connecting the two predicted cystatin domains.The individual N-and Cterminal domains (NTD and CTD) both displayed an a-helix wrapped by an anti-parallel b-sheet, the characteristic tertiary structure of members of the cystatin family (Bode et al., 1988).The superposition of modeled NTD and CTD yielded a Ca backbone RMSD of 2.20 A (Figure 1b).Interestingly, superposition and sequence alignments indicated a potential legumain reactive center loop (RCL) harboring an asparagine as putative P1-residue, as seen in the legumain inhibitory human cystatin M/E and C, in both the NTD and the CTD (Figure 1b,c) (Dall et al., 2015).Furthermore, sequence analysis also uncovered an L/VGG motif on the NTD, which was previously linked to PLCP inhibition in other cystatins.Based on these findings, we hypothesized that AtCYT6 is a type II phytocystatin capable of inhibiting both A. thaliana legumains and PLCPs.

AtCYT6 is an interaction partner of AtLEGb
To find out whether AtCYT6 was indeed an inhibitor of AtLEGb in planta, pull-down assays were performed.For that purpose, extracts of A. thaliana leaves of a VPE 0 strain lacking the expression of all four legumain isoforms were used.The extracts were spiked with recombinant inactive proAtLEGb or active AtLEGb, both harboring an N-terminal His 6 -tag, which was used to recover the proteins via magnetic Ni 2+ -beads.In control experiments, the extracts were spiked with buffer instead of AtLEGb.Differential triplex labeling of the proAtLEGb, AtLEGb, and buffer control samples followed by mass spectrometry analysis confirmed that AtLEGb was significantly increased in both proteasetreated samples relative to the buffer control (Figure 1d).Importantly, AtCYT6 showed the highest log2-fold change in the sample spiked with AtLEGb but was not co-purified with proAtLEGb, suggesting that the binding of AtCYT6 to AtLEGb requires an interaction surface that is available in AtLEGb but not in proAtLEGb, for example, the active site.This finding corroborated AtCYT6 as an interaction partner and likely physiologic inhibitor of AtLEGb.

The CTD of AtCYT6 inhibits AtLEGb
To further confirm that AtCYT6 was indeed binding to and inhibiting AtLEGb, we recombinantly expressed and purified to homogeneity the full-length AtCYT6 lacking the N-terminal signal peptide (A35-D235 and AtCYT6-FL), the N-terminal domain alone (A35-D127 and AtCYT6-NTD), and the C-terminal domain alone (E142-D235 and AtCYT6-CTD).
The identity of the recombinant proteins was confirmed by mass spectrometry (Table S1).As a first assessment of legumain interaction sites, we carried out co-migration assays.Specifically, we co-incubated active AtLEGb with AtCYT6-FL, AtCYT6-NTD, or AtCYT6-CTD separately and subjected the samples to size exclusion chromatography (SEC) experiments.The AtCYT6-FL co-migrated with AtLEGb, reiterating its binding properties to AtLEGb seen in the pull-down assay (Figure 2a).When we tested the AtCYT6-NTD and AtCYT6-CTD individually, co-migration with the enzyme was only observed with the AtCYT6-CTD (Figure 2b,c).Subsequently, we tested whether the AtCYT6 constructs were indeed inhibitors of legumain activity.
To that end, we co-incubated active AtLEGb with the individual constructs and monitored the turnover of the legumain-specific AAN-AMC substrate.In line with our SEC experiments, we found that AtCYT6-FL and AtCYT6-CTD inhibited the proteolytic activity of AtLEGb, but not the AtCYT6-NTD (Figure 3a).In the next step, we compared the affinities of AtCYT6-FL and AtCYT6-CTD by determining their inhibition constants (K i ).The constructs displayed similar, sub-nanomolar affinities towards AtLEGb (K i AtCYT6-FL = 0.29 AE 0.05 nM; K i AtCYT6-CTD = 0.61 AE 0.23 nM) (Table 1; Figure S3a,b).Withal, a second variant of the C-terminal domain (A128-D235 and AtCYT6-CTD long ) carrying the interdomain linker was produced and inhibited the enzyme with similar K i as AtCYT6-FL and AtCYT6-CTD (K i AtCYT6-CTDlong = 0.68 AE 0.20 nM).Based on these results we concluded that AtCYT6 was a high-affinity physiologic inhibitor of AtLEGb, and that inhibition was mediated by its C-terminal domain.
AtCYT6-NTD inhibits papain-like cysteine proteases (PLCPs) Since cystatins are well-established inhibitors of PLCPs, in the next step we tested if our recombinant AtCYT6 constructs would also be PLCP inhibitors.The L/VGG motif at the N-terminus of inhibitory cystatins is critical for the inhibition of PLCPs by other cystatins (Stubbs et al., 1990).The absence of such a motif in the AtCYT6-CTD and its presence at the AtCYT6-NTD suggested that the NTD harbored a PLCP-inhibitory activity (Figure 1c).To confirm this hypothesis, papain was used as a model for the PLCP family.The turnover of its FR-AMC substrate was monitored upon incubation of papain with AtCYT6-FL, AtCYT6-NTD, or AtCYT6-CTD.Indeed, AtCYT6-FL and AtCYT6-NTD    1; Figure S3c,d).In accordance with our hypotheses, the N-and C-terminal domains of AtCYT6 exhibited distinct inhibitory properties towards specific cysteine proteases.

AtLEGb selectively binds to AtCYT6-CTD in vivo
To further validate the selectivity of AtCYT6-CTD as legumain inhibitory domain in vivo, we performed coimmunoprecipitation assays (CoIPs) using A. thaliana seed extracts and polyclonal rabbit antibodies raised against AtCYT6-NTD or AtCYT6-CTD.To precipitate endogenous AtCYT6, we immobilized anti-AtCYT6-NTD or anti-AtCYT6-CTD antibodies on protein G-coupled sepharose beads.Subsequently, we loaded the soluble fractions of seed extracts prepared at pH 5.5, followed by washing and elution of captured proteins.Control samples contained protein G-coupled sepharose beads only.After digestion of the samples with trypsin, they were analyzed by label-free mass spectrometry (Figure 4a).The abundance of proteins identified in each sample was calculated as log 2 ratios between label-free quantification intensities of test samples and control beads.Importantly, both antibodies, anti-AtCYT6-NTD, and anti-AtCYT6-CTD, yielded cystatin six-derived peptides as the most abundant species (positive control).Interestingly, only anti-AtCYT6-CTD but not anti-AtCYT6-NTD antibodies co-precipitated endogenous AtLEGb with endogenous AtCYT6.This experiment confirmed that AtCYT6 is a physiologic inhibitor of AtLEGb in A. thaliana seeds and that also in vivo their interaction is mediated by the AtCYT6-CTD.Furthermore, we suggest that AtCYT6 was at least partially processed in our CoIP samples, for example, in the inter-domain linker, and that therefore the anti-AtCYT6-NTD antibodies were not able to co-elute AtLEGb.This suggestion was supported by western blot experiments of the elutions of the CoIPs, which showed no or little protein at the expected height of AtCYT6-FL (Figure S4).Instead, we observed bands corresponding in size to AtCYT6-NTD or AtCYT6-CTD, respectively.Importantly, no other cystatin isoform besides cystatin 6 was co-purified in the CoIPs, confirming the specificity of the antibodies used.To further confirm the observed in vivo selectivity, we repeated the CoIP experiments using beads where recombinant AtCYT6-NTD or AtCYT6-CTD were loaded to the immobilized anti-AtCYT6-NTD or anti-AtCYT6-CTD antibodies respectively.Importantly, AtLEGb was recovered at 20 times higher intensity using anti-AtCYT6-CTD loaded with recombinant AtCYT6-CTD than with anti-AtCYT6-NTD loaded with recombinant AtCYT6-NTD (Figure 4b).Further analysis of our CoIP experiments additionally uncovered specific interaction partners of AtCYT6-NTD, like the PLCPs SAG12, RD19D and a cysteine proteinase superfamily protein (Q9LNC1) with so far unknown function (Figure 4).All of them were specifically observed when the anti-AtCYT6-NTD antibody was used as a bait, with or without AtCYT6-NTD.This suggested that AtCYT6-NTD specifically inhibits PLCPs also in vivo and implies that AtCYT6 may be a physiological regulator of SAG12 and RD19D activities.

AtCYT6 is a substrate-like inhibitor of AtLEGb
Knowing that the C-terminal domain of AtCYT6 was the actual legumain inhibitor, we were in the next step wondering whether it would exploit a substrate-like mode of interaction, as seen in human legumain-inhibitory cystatins  (Dall et al., 2015).Along that line, based on our model, Asn173 could serve as a P1-residue.To test this hypothesis, we prepared an AtCYT6-CTD-N173A mutant and tested whether it would be inhibiting AtLEGb.Using AAN-AMC as a substrate, we indeed found that the mutant lost its ability to inhibit AtLEGb activity (Figure 5a).This finding confirmed the essential role of Asn173 within the proposed RCL as a mediator of the interaction of AtCYT6-CTD with AtLEGb.Since an asparagine was indeed essential for the interaction, we then hypothesized that Asn173 serves as a P1 residue binding in a productive, substrate-like orientation to the S1 pocket on AtLEGb.To test this hypothesis, we co-incubated AtLEGb with AtCYT6-FL or the AtCYT6-CTD at pH 5.5 in a 1:5 molar ratio (enzyme: inhibitor) and analyzed the reaction for cleavage products using SDS-PAGE and mass spectrometry.Indeed, we found that both AtCYT6 variants were processed by AtLEGb (Figure 5b).Interestingly, cleavage was only observed for less than 20% of the AtCYT6-FL or AtCYT6-CTD proteins, leaving 80% of these proteins intact.This led us to the conclusion that AtCYT6 remained in a stable complex with AtLEGb even though partial AtCYT6 cleavage by AtLEGb was observed.Additional mass spectrometry experiments confirmed that Asn173 was the most prominent cleavage site in both constructs (Table S2; Figure 5c).Although AtCYT6-NTD harbored an asparagine residue (Asn68) at a position structurally equivalent to Asn173 on AtCYT6-CTD, it did not show inhibition of AtLEGb.To test whether AtCYT6-NTD was a substrate to AtLEGb rather than an inhibitor, we co-incubated them at pH 5.5 and similarly analyzed the reaction by SDS-PAGE and mass spectrometry.Interestingly we found that in contrast to the other AtCYT6 constructs, the N-terminal domain was processed completely within less than 10 min of incubation (Figure 5b).Mass spectrometry experiments confirmed that cleavage occurred after the P1-Asn68 residue (Table S1).Additional cleavage events resulted in complete fragmentation of the N-terminal domain by AtLEGb (Figure 5c).These experiments confirmed our hypothesis that AtCYT6-NTD was rather a substrate to AtLEGb than an inhibitor.

The double basic Arg205-Lys208 motif establishes pHdependent exosite interactions
Interestingly, both cystatin-like domains on AtCYT6 harbor an asparagine on the RCL that may be recognized as a substrate by AtLEGb.However, while the N-terminal domain is a substrate, the C-terminal domain additionally functions as an inhibitor.Prompted by this observation, we wondered which feature was turning the C-terminal domain into an inhibitor.To address this question, we prepared models of AtLEGb in complex with AtCYT6-FL, AtCYT6-NTD, or AtCYT6-CTD using AlphaFold-Multimer and compared them to the crystal structure of hLEG in complex with hCE (PDB 4N6O; Figure 6a; Figure S5) (Jumper et al., 2021).Importantly, the modeled structure of AtLEGb was in good agreement with a crystallographic structure of proAtLEGb (PDB 6YSA) which we previously determined (Ca RMSD = 0.38 A; determined with PyMOL).Also, the Ca RMSD between the C-terminal domain modeled in complex with AtLEGb in the presence and absence of the Nterminal domain, that is, AtCYT6-FL and AtCYT6-CTD, was <0.1 A (Figure S5a), which is in accordance with our experimental data supporting similar K i values for both constructs against AtLEGb.The models suggested that the RCL interaction was similar in hCE and AtCYT6-CTD.
The RCL of AtCYT6-CTD established substrate-like interactions to the S3-S2 0 -substrate binding sites on AtLEGb, where Asn173 serves as a P1 residue (Figure 6a,b; Figure S5b).Interestingly, AlphaFold-Multimer positioned AtCYT6-NTD in an 'upside-down' orientation onto the active site of AtLEGb, with Asn48 positioned to the S1pocket (Figure S5c).These models are in qualitative agreement with our experimental data, which showed that only AtCYT6-CTD is an effective inhibitor of AtLEGb, but not AtCYT6-NTD.To force AlphaFold-Multimer to position Asn68 of AtCYT6-NTD to the active site of AtLEGb we replaced all asparagine residues that were identified as cleavage sites by mass spectrometry experiments, except for Asn68, by alanine in the sequence used for modeling.This approach indeed resulted in a model with a similar orientation of AtCYT6-NTD as compared to hCE and AtCYT6-CTD.However, AtCYT6-NTD was tilted relative to AtCYT6-CTD (Figure 6a).Interestingly, in the model both the AtCYT6-NTD and -CTD presented a leucine residue in position P2 0 , which perfectly matches the architecture of the S2 0 -binding site on AtLEGb (Figure S5b) (Dall et al., 2020).
The legumain exosite loop (LEL) of hCE establishes an exosite interaction to hLEG, which turns hCE from a substrate into an inhibitor (Dall et al., 2015).Importantly, such a LEL structure was essentially missing in the AtCYT6 sequence.Instead, the LEL was replaced by a rather short hairpin loop in the AlphaFold model of AtCYT6 (Figure 6a,c).Interestingly, the hairpin loop of AtCYT6-CTD harbored a double basic motif formed by Arg205 and Lys208 which, according to our model, could establish ionic interactions with Glu212 on AtLEGb (Figure 6a,b; Figure S5d).Glu212 is part of the S1 0 -substrate binding site and sits directly next to the catalytic Cys211 residue.Based on this observation, we hypothesized that the Arg205-Lys208 double basic motif might serve as an electrostatic exosite anchor that stabilizes the enzyme-inhibitor complex, functionally resembling the LEL in hCE.To test this hypothesis, we prepared AtCYT6-CTD-R205A and -K208A mutants and tested their interaction with AtLEGb.In line with our hypothesis, the mutants displayed reduced inhibition of AtLEGb protease activity compared to the wild-type inhibitor (K i AtCYT6-CTD-R205A = 3.36 AE 0.70 nM; K i AtCYT6-CTD-K208A = 3.98 AE 0.66 nM) (Table 1).This experiment confirmed that the R205A-K208A double basic motif indeed functions as an important exosite that turns the AtCYT6-CTD into an inhibitor rather than a substrate.Consistent with the identified charged exosite interactions, we found that the inhibition of AtLEGb by AtCYT6-CTD was pH-dependent.Inhibition was most efficient at around neutral pH and lower at very acidic pH (Figure 6c).Moreover, we found that the processing of the different AtCYT6 constructs was also pH-dependent.While the AtCYT6-CTD was rather resistant to cleavage at moderately acidic pH (pH 5.5), it was completely processed at pH 4.0, whereas the AtCYT6-NTD was quickly processed by AtLEGb regardless of the pH (Figure 5b,d).Importantly, the Arg205-Lys208 double basic motif was replaced by a negatively charged Glu97 on the AtCYT6-NTD (Figure 6a,b; Figure S5d).According to our model, Glu97 will result in repulsive interactions with Glu212 on AtLEGb, which is in line with our observation that AtCYT6-NTD was not inhibiting AtLEGb but rather acted as a substrate.To test whether we could turn AtCYT6-NTD into an inhibitor, we introduced an E97R mutation.Interestingly, we found that the AtCYT6-NTD-E97R mutant was not able to significantly impair AtLEGb activity, even at a 10:1 inhibitor:enzyme ratio (Figure 6d).
Sequence comparison of other five type II phytocystatins which had both domains individually tested against papain and different legumains indicated conservation of the properties seen in AtCYT6: (i) an asparagine residue on the RCL of N-and C-terminal domains, (ii) an R-G-X-K basic motif on the LEL of the legumain-inhibitory CTD, and a single Asp or Glu on the LEL of their NTDs (Aceituno-Valenzuela et al., 2018;Christoff et al., 2016;Martinez et al., 2007) (Figure S6).
Altogether, we conclude that complex formation between AtLEGb and AtCYT6-CTD is pH-dependent, and its dissociation constant is modulated by the double basic motif Arg205-Lys208 within the LEL.Our experiments suggest that the LEL is not the only motif that is required for discrimination between inhibitory and non-inhibitory cystatin domains, as neither the R205A and K208A mutations on AtCYT6-CTD abolished legumain inhibition, nor the point mutation E97R elicited significant inhibition of AtCYT6-NTD against AtLEGb.

The AtCYT6-CTD has a pH-stabilizing effect on AtLEGb
The catalytic domains of legumains are known to undergo irreversible denaturation at near-neutral pH.Some interaction partners can stabilize the catalytic domain at such pHs, as is the case for hCE (Dall et al., 2015).Our model established a similar contact surface for AtLEGb and AtCYT6-CTD to that of hCE and hLEG (Figure 6a).Therefore, we hypothesized that a similar pH-protective role would be encoded in AtCYT6.To test this hypothesis, we performed nanoDSF experiments, which monitored the changes in the intrinsic fluorescence of aromatic residues upon thermal unfolding.The ratio of fluorescence measured at 330 nm (absorption peak when aromatic residues are buried) and at 350 nm (absorption peak of solventexposed aromatic residues) was plotted against temperature, and the inflection temperatures (T i ) of the samples were compared.As seen in Figure 7(a), at pH 6.5 AtLEGb alone displayed a T i of 49.6°C, whereas when incubated with AtCYT6-FL the T i shifted to 58.9°C.Similarly, incubation of AtLEGb with AtCYT6-CTD resulted in an increase of T i to 57.0°C.Taken together, complex formation resulted in thermal stabilization of AtLEGb by approx.8.5°C.Importantly, when AtCYT6-NTD was incubated with AtLEGb, no significant shift in T i was detected (Figure S7).Similarly, the P1 AtCYT6-CTD-N173A mutant was also not able to elicit the same stabilizing effect over AtLEGb (Figure 7b; Figure S7b).However, the AtCYT6-CTD-K208A mutant showed a stabilization capacity comparable to its wild-type counterpart.This further confirmed that the stabilizing effect was directly linked to complex formation.

Cystatin 6 is also a potent inhibitor of AtLEGc
Once the structural requirements for inhibition of AtLEGb by AtCYT6 were revealed, we wondered whether the inhibition would be specific to certain AtLEG isoforms.Considering that the catalytic domains of AtLEGb and AtLEGc are highly conserved and share critical features found to be important for inhibition by AtCYT6, for example, the P1-Asn specificity and the Glu212 residue neighboring the catalytic cysteine, we hypothesized that AtLEGc would also be inhibited by AtCYT6.Indeed, AtCYT6-FL and AtCYT6-CTD displayed a binding affinity (K i ) of 1.21 AE 0.23 nM and 0.86 AE 0.22 nM, respectively, towards AtLEGc (Table 2; Figure S8a,b), suggesting a physiological role for inhibition of both isoforms by AtCYT6.To test if the inhibition would be specific against plant legumains, we assessed the inhibition of human legumain by the different AtCYT6 constructs.Interestingly, we found significantly lower binding affinities for AtCYT6-FL and AtCYT6-CTD (K i AtCYT6-FL = 106.4AE 11.9 nM; K i AtCYT6-CTD = 61.4AE 7.5 nM), indicating intra-species specificity of AtCYT6 (Table 2; Figure S8d,e).The reduction in inhibition of human legumain might be explained by the missing S2 0 substrate specificity pocket.Importantly, the N-terminal domain did not inhibit either of the enzymes (Figure S8c,f).

DISCUSSION
Based on our findings we suggest a bivalent binding mechanism of AtCYT6 to AtLEGb which is established by two major binding sites.Binding site 1 mediates substrate-like binding via the P1-Asn173 residue on the RCL.Binding site 2 establishes electrostatic interactions via the double basic motif Arg205-Lys208 and thereby steers (pre-)complex formation (Figure 8).Substrate-like binding inhibitors are not rarely substrates themselvesslow degradation may follow binding to the enzyme.Indeed, we observed that cleavage of AtCYT6-CTD upon incubation with AtLEGb increased over time at pH 5.5 (Figure 5b).We conclude that the inhibition of AtLEGb by the AtCYT6-CTD happens at the expense of inhibitor degradation in a competitive, substrate-like manner.The presence of an asparagine residue in the corresponding RCL of non-inhibitory AtCYT6-NTD (Asn68) suggests that the P1-Asn is a prerequisite rather than a determining factor for AtLEG inhibition by phytocystatins.
AtCYT6 is a high-affinity inhibitor of legumains from A. thaliana.Despite the high conservation between the catalytic domains of AtLEGb and AtLEGc (more than 70% identity and 80% similarity), a fourfold difference in K i was observed for the AtCYT6-FL.This difference could be explained by substitution on the non-prime side of the enzymes: a tyrosine in AtLEGb (Tyr240) is substituted by a tryptophane (Trp249) in AtLEGc, which results in a tighter S3 pocket to accommodate the bulky P3-Arg171.On the other hand, AtCYT6-FL features a 50-fold lower binding affinity to hLEG than it does for the A. thaliana legumains.In this case, the missing S2 0 pocket in hLEG may be responsible for the reduction in affinity.Therefore, the P2 0 -Leu will bind with higher affinity to the A. thaliana legumains (Dall et al., 2021).
Our co-immunoprecipitation assays confirmed complex formation between AtCYT6-CTD and AtLEGb in vivo (Figure 4a,b).Notably, AtLEGb was also recovered, although to a low amount, when anti-AtCYT6-NTD antibodies loaded with AtCYT6-NTD were used as a bait.This finding is in agreement with our observation that AtCYT6-NTD is recognized as a substrate by AtLEGb rather than an inhibitor.Importantly, proteins like the PLCPs SAG12 and RD19 were exclusively co-eluted with anti-AtCYT6-NTD antibodies with and without recombinant AtCYT6-NTD.SAG12 is a senescence marker implicated in nitrogen content and yield of seeds of A. thaliana, whose expression is upregulated in cvpe knockout plants (Cheng et al., 2019;James et al., 2018;Otegui et al., 2005).
We hypothesized that the LEL of the cystatins is a structural feature that discriminates inhibitory from non-   inhibitory cystatin domains.However, introducing a supposedly 'inhibitory' AtCYT6-NTD-E97R mutation did not convert the non-inhibitory cystatin domain into a functional inhibitor.Our models also suggest that AtCYT6-NTD binds in a structurally different, tilted way to AtLEGb.As a result, the AtCYT6-NTD LEL adopts a sterically different position, which can likely explain why the E97R mutation was not able to reconstitute inhibition on AtCYT6-NTD.Our models furthermore show that the orientations of Arg205 and Lys208 allow a strong electrostatic interaction with Glu212 from AtLEGba neighbor residue to the catalytic cysteine Cys211.This double basic motif finds a resemblance in another human cystatin.Arg96 and Arg119 of the crystallographic structure of cystatin C (hCC, PDB 3GAX) align nicely with and are oriented as Lys208 and Arg205 of our modeled AtCYT6-CTD (Figure S9).Importantly, the interaction between hCC and hLEG is pH-sensitive; it is disrupted at pH 4.0.This exosite interaction is mediated via Glu190 on hLEG, which also sits directly next to the catalytic Cys189 (Dall et al., 2015).This is in good agreement with our experimental data and allows us to postulate that the LEL may be the structural motif that encodes pH dependency to the stability of the enzyme-inhibitor complexes.However, since the R205A and K208A mutants showed a reduction in inhibition but not a complete loss of inhibition, we suggest that RCL cleavage and religation may be in equilibrium at pH 5.5 and higherbut not at pH 4an important aspect for complex stabilization observed previously in hCE (Dall et al., 2018).Human cystatin M/E is able to protect hLEG from quick and irreversible pH-driven denaturation.A pHdependent stability was also observed for the catalytic domain of AtLEGb: its conformational stability is high at pH 5.0 and decreases at pH 4.0 and 7.0 (Dall et al., 2020).Our results on the thermal unfolding (Figure 7a) pose an interesting parallel, as the AtCYT6-CTD was able to significantly increase AtLEGb stability at pH 6.5 and is, therefore, a candidate for legumain stabilization in vivo, for example, outside of the vacuole.
Despite AtCYT6 being an AtLEGb inhibitor, data from literature addressing their colocalization is still missing.The majority of AtCYT6 was shown to be expressed in seeds and siliques, and it is predicted to harbor an Nterminal signal peptide for extracellular secretion, as assigned in Uniprot (Huala et al., 2001;UniProt Consortium, 2023).Our experiments revealed AtCYT6 to be an isoform-unspecific inhibitor displaying a high affinity for AtLEGb and AtLEGc.Along that line, CoIP experiments using immobilized anti-AtCYT6-CTD antibody incubated with A. thaliana leaf extract co-eluted AtLEGc (data not shown).Interestingly, at pH 7.5 AtLEGc was the sole legumain isoform that was co-eluted, whereas at pH 5.5 AtLEGb was also present but to a significantly lesser extent.Together, these findings support the hypothesis that AtCYT6 could have tissue-specific targets.However, it remains unclear in which subcellular compartment and physiological context the inhibitor is to exert its inhibitory activity upon the enzymes.Despite the absence of experimental evidence of the presence of A. thaliana legumains in the extracellular environment, it is plausible to suppose its colocalization with its dedicated inhibitor, that is, type II phytocystatins, in one or multiple physiological contexts.In that regard, the results of our pull-down assays in which AtLEGb was used as bait (Figure 1d) show coprecipitation with a plasmodesmal protein (Uniprot O65449) and could indicate the presence of AtLEGb in the extracellular environment (Fernandez-Calvino et al., 2011).In this case, AtCYT6 could shield the enzyme from pHdriven denaturation upon alkalinization of the extracellular pH during plant-pathogen interaction and abiotic stress (Geilfus, 2017).Additionally, soluble extracellular content and vacuolar enzymes can meet up within the vacuolar system via endocytosis of apoplastic content (Etxeberria et al., 2012;Fan et al., 2015;Shimada et al., 2018).Lastly, the vacuole-mediated programmed cell death mechanisms comprise means by which AtCYT6 and AtLEGb/c may find each other.They consist of a series of events involving the lytic vacuole that culminate in either its collapse and consequent discharge of vacuolar proteases into the or fusion with the cell membrane, which also promotes the mixing of vacuolar and apoplastic content (Hara-Nishimura & Hatsugai, 2011;Hatsugai et al., 2004Hatsugai et al., , 2009;;Kuriyama & Fukuda, 2002).
Our experimental data demonstrate AtCYT6-NTD to be a potent inhibitor of papain, and AtLEGb to be capable of promptly processing it, and its equal competence to process AtCYT6-CTD at acidic pH (Figure 5b,d).It is tempting to speculate that such post-translational processing of AtCYT6 could release the individual N-and C-terminal domains as truncation products, which would consequently allow them to be translocated into different subcellular compartments in vivo.This hypothesis is further strengthened by studies on type III phycys.Upon incubation with multiple endoproteases, functional subdomains with sizes of approximately 10, 22, and 32 kDa were released from the 85 kDa potato multicystatin and retained their inhibitory activity against papain (Walsh & Strickland, 1993).Similarly, tomato leaves expressed a multicystatin of approximately 88 kDa which released functional fragments after short-term proteolytic processing by trypsin and chymotrypsin (Wu & Haard, 2000).
Taken together, our work suggests that the separation of functional subdomains within multi-domain phytocystatins could be a feature conserved throughout the cystatins.Additionally, the processing of AtCYT6 in vitro suggests a regulatory mechanism of its inhibitory activity by itself, as it demonstrates that individual domains can be processed by selected enzymes under differential conditions such as pH.

Construct design
Expression constructs of full-length Arabidopsis thaliana cystatin 6 (AtCYT6-FL) were designed based on the sequence deposited in Uniprot under the accession code Q8HOX6-2.A full-length cDNA sequence lacking the N-terminal signal peptide (M1-M34) and harboring 5' NdeI and 3' XhoI restriction enzyme cleavage sites was purchased from Eurofins Genomics (Ebersberg, Germany).The expression construct was subcloned into the pET-22b(+) expression vector using the same restriction enzymes, resulting in a sequence encoding a C-terminal His 6 -tag.Additionally, the expression construct was also subcloned into the pET-15b vector, to obtain an N-terminal His 6 -tag.For that, a PCR reaction was performed using the primers TATAGGTACCGCTTTAGTCGGAGGTG TTGGCG (forward, KpnI recognition site) and GGTGGGATCCT-CAGTCATGGTGTTGCTCCGCGTGG (reverse, BamHI).The PCR product was purified, digested by the respective restriction enzymes, and ligated into the pET-15b vector.Furthermore, constructs composed of the individual C-terminal domain (AtCYT6-CTD) and the C-terminal domain harboring the interdomain linker (AtCYT6-CTD long ) were prepared using a similar protocol.For the AtCYT6-CTD CAAGGGTACCGAACATGAATCTGGATGGAGGG (forward, KpnI) and the full-length reverse primer were used, and for the AtCYT6-CTD long , a TGCCGGTACC GCCCCTGCTATCACTTCC TCCG (forward, KpnI) forward primer was used together with the reverse primer of the full-length construct.As the expression of the AtCYT6-NTD construct in the pET15b vector was inviable, we cloned the sequence into vector pET-28 and it included, as a result, a C-terminal His 6 -tag.A PCR reaction with the forward primer AGTCCCATGGCTTTAGTCGGAGGTGTTGGCG (NcoI) and reverse primer AGCTCTCGAGATCGGAGGAAGTGATAGCAGGGG C (XhoI) was done and followed by purification, enzymatic digestion, and ligation into pET28.AtCYT6 point mutants were prepared using round-the-horn site-directed mutagenesis.The forward primer GCGTCC TTGTTCCCTTATGAACTTCTGGAGGTTGTGC and reverse primer AGACCTCTGCTGAATGGTCTTGACAGCCTGCTCAGC were used to obtain the P1-Asn mutant N173A.The LEL mutation R205A in AtC YT6-CTD was obtained with the forward GCGGGAGAAAAAGAG-GAAAAGTTCAAGGTGGAAGTTCAC and reverse CTTCAACTTC AGTAGCATGTTGTATTTTGCAGCCTCGC primers.The K208A mutation was introduced in AtCYT6-CTD with the forward GC GGAGGAAAAGTTCAAGGTGGAAGTTCACAAGAACC and reverse TTCTCCCCTCTTCAACTTCAGTAGCATGTTGTATTTTGC primers, and E97R was introduced in AtCYT6-NTD with CGCGCAGGACA-GAAGAAGCTATACGAAGC (forward) and CAGAATCTCCAGAGT CAGGTGATGC (reverse) primers.

Expression
The final expression vectors were transformed into the E. coli BL21(DE3) strain for expression.Expression flasks of 2 L containing 500 ml of LB medium (Carl Roth, Karlsruhe, Germany) and ampicillin (for constructs cloned into pET15b) or kanamycin (for constructs cloned into pET28) at 100 lg ml À1 concentration were inoculated with 25 ml of preculture and incubated at 37°C and 240 rpm until an OD 600 of approximately 1.0 was reached.Expression was induced upon the addition of IPTG (Isopropyl b-D-1-thiogalactopyranoside) at a final concentration of 1 mM.After induction, the cultures were kept overnight at 25°C and 240 rpm.The cells were harvested and either lysed for immediate use or frozen as pellets at À20°C.

Purification
Cell pellets were resuspended in 25 ml of lysis buffer (Tris-HCl 100 mM pH 7.5, NaCl 500 mM) and lysed by sonication at 40% power, 4 times for 45 sec (Bandelin Sonopuls, Berlin, Germany) in the presence of EDTA-free, EASY-pack protease inhibitor cocktail tablets (Roche, Basel, Switzerland) according to instructions of the manufacturer.The lysates were centrifuged at 17 500 g for 1 h at 4°C.The supernatant was incubated for 30 min at 4°C with 5 ml of Ni 2+ -beads (Qiagen Ni-NTA Superflow) equilibrated in lysis buffer.After removal of the flow-through, the beads were washed with two column volumes of buffer (Tris-HCl 100 mM pH 7.5, NaCl 500 mM) containing 5, 10, and 20 mM imidazole, in that order.The elution of the protein of interest was done in a buffer composed of 100 mM Tris-HCl pH 7.5, 500 mM NaCl, and 250 mM imidazole in three steps of one column volume each with 15 min incubation.The eluate was concentrated in Amicon Ultra centrifugal filter units (MWCO: 3 kDa; Cytiva, Dreieich, Germany).Following affinity chromatography, the proteins were further purified by size exclusion chromatography using an € AKTA FPLC system equipped with an S200 10/300 GL or S75 10/300 GL column equilibrated in a buffer composed of 20 mM Tris-HCl pH 7.5, 50 mM NaCl, and 2 mM DTT (if indicated).

Preparation of Arabidopsis thaliana legumain-b (AtLEGb) and -c (AtLEGc), human legumain and papain
Papain from Carica papaya was purchased from Merck (Darmstadt, Germany).AtLEGb, AtLEGb-C211A, AtLEGc, and human legumain were expressed, purified, and activated based on a protocol previously described (Dall et al., 2022).Briefly, the sequence of corresponding constructs of proLEG were cloned into pLEXSY-sat2.1vectors and transfected into the LEXSY P10 host strain of Leishmania tarentolae cells of the LEXSYcon2.1 expression system (Jena Biosciences, Jena, Germany).The resulting constructs carried N-terminal His 6 -tags and an Nterminal signal sequence for secretion into the supernatant.Cultures were grown in brain heart infusion (BHI) medium (Jena Biosciences) in tissue culture flasks with 5 lg ml À1 of hemin, 50 units ml À1 penicillin, and 50 units ml À1 streptomycin (Pen-Strep, Carl Roth) in the presence of nourseothricin (Jena Bioscience) as selection antibiotic.Expression was carried out for 48 h in the dark at 26°C and 140 rpm.The supernatant containing the proteins of interest was harvested by centrifugation and incubated with Ni 2+ -beads at 4°C.The washing buffer consisted of 50 mM HEPES pH 7.5 and 300 mM NaCl.The elution buffer was the same as the washing buffer with the addition of 250 mM imidazole.After elution, the proteins were concentrated, and PD-10 columns (GE Healthcare, Uppsala, Sweden) were used for buffer exchange with storage buffer 20 mM HEPES pH 7.5 and 50 mM NaCl.At this point, proteins were either stored at À20°C for further use or auto-activated for elimination of the C-terminal prodomain.The pH-driven auto activation was performed with incubation at room temperature at pH 3.5-4.0after which size exclusion chromatography or buffer exchange with PD-10 columns was performed with buffer composed of 20 mM citric acid pH 4.0-5.0 and 50 mM NaCl.
The details described above may vary slightly according to the legumain isoform in question.For a more detailed and precise protocol, please refer to the respective publications (Dall et al., 2020;Dall & Brandstetter, 2012;Zauner, Dall, et al., 2018).Differential scanning fluorimetry 10 ll of AtLEGb (1 mg ml À1 ) in a buffer composed of 20 mM citric acid pH 4.0, 50 mM NaCl, and 0.5 mM MMTSmethyl methanethiosulphonate) were mixed with an equimolar amount of cystatins for 5 min at 21°C, in a final volume of 12 ll.Control reactions were supplemented with buffer instead of cystatins or legumain, respectively.For measurements done at pH 6.5, 1 ll of 1 M MES pH 6.5 buffer was added immediately before the addition of the cystatin.Thermal denaturation curves were obtained using the Nanotemper Tycho NT.6 instrument by monitoring the fluorescence intensity at 330 and 350 nm upon heating the samples from 35°C to 95°C.

Activity assays
Initial velocities were obtained by monitoring the fluorescence increase (in RFU) following the breakdown of synthetic peptidic substrates in an Infinite M200 Plate Reader (Tecan, Gr€ odig, Austria) at 25 °C.For legumain assays, the Z-Ala-Ala-Asn-AMC (Bachem) substrate was used at 50 lM concentration in reaction buffer composed of 20 mM citric acid pH 5.5, 100 mM NaCl, 0.02% Tween 20, and 2 mM DTT).For papain assays, the Z-Phe-Arg-AMC (Bachem, Bubendorf, Switzerland) substrate was used at 50 lM concentration in reaction buffer composed of 20 mM MES pH 6.0, 50 mM NaCl, 0.02% Tween 20, and 2 mM DTT).The enzymes and the inhibitors were added 10-fold concentrated in relation to their desired final concentration in the assay.Control experiments were supplemented with buffer instead of inhibitor.Reactions were started by the addition of the enzyme.The first 100 sec of the reactions were used for the calculation of the initial velocities.For the K i determinations, the final concentration of enzymes was 50 nM, and the inhibitor concentrations were defined by 1:2 serial dilutions starting at 15-fold the enzyme concentration, measured in quadruplicates.To obtain K i values, data points were fitted to the Morrison equation using the GraphPad Prism program (version 5.0, La Jolla, CA, USA).

Analysis of complex formation via co-migration assays
Co-migration experiments were performed in buffer composed of 20 mM MES pH 6.0, and 100 mM NaCl.400 lg of AtLEGb or papain were inhibited with 0.5 mM or 5 mM S-methyl methanethiosulfonate (MMTS; Merck, Darmstadt, Germany), respectively.Subsequently, an equimolar amount of AtCYT6-derived constructs was added, and enzyme and inhibitor were incubated for 1 h on ice.Samples were loaded on a Superdex S75 10/300 GL column (Cytiva) preequilibrated in assay buffer at 21°C.Immediately after elution, MMTS was added at a final concentration of 5 mM.Samples collected were analyzed by SDS-PAGE.Samples containing legumain were supplemented with 1 mM of the covalent Ac-Tyr-Val-Ala-Asp-chloromethyle ketone (YVAD-cmk) inhibitor prior to the addition of the loading buffer.

pH-dependent cleavage
AtLEGb was incubated with AtCYT6-derived constructs at a 1:5 molar ratio in a buffer composed of 100 mM citric acid pH 4.0 or 5.5, 100 mM NaCl, and 2 mM DTT at 21°C.Samples were collected after 1, 5, and 10 min of incubation and transferred to a new tube containing 0.1 ll of YVAD-cmk inhibitor (final concentration: 1 mM).Subsequently, samples were analyzed by SDS-PAGE.

Cleavage analysis by mass spectrometry
AtCYT6-derived constructs were incubated with AtLEGb in a 1:5 molar ratio in a buffer composed of 100 mM citric acid pH 4.0 or 5.5, 100 mM NaCl, and 2 mM DTT at 21°C for 30 min.For mass spectrometric analysis, samples were desalted with C 18 ZipTips (Merck, Darmstadt, Germany), eluted from the tips with 50% acetonitrile in 0.1% formic acid, and directly infused into the mass spectrometer (Q-Exactive; Thermo Fisher Scientific, Waltham, Massachusetts) at a flow rate of 1 ll min À1 .Capillary voltage at the nanospray head was 2 kV.Raw data were processed with Protein Deconvolution 2.0 (Thermo Fisher Scientific, Waltham, Massachusetts).Masses were assigned to the protein sequence with the Protein/Peptide Editor module of BioLynx (part of MassLynx V4.1; Waters, Eschborn, Germany).

Modeling of protein complexes
The model of AtCYT6 was generated using AlphaFold2 via Local-ColabFold (Jumper et al., 2021;Mirdita et al., 2022).The models of the complexes were obtained using AlphaFold Multimer via the Colab notebook (Liu et al., 2023).The sequence used for modeling of AtLEGb comprised residues V48-N329 according to Uniprot accession code Q39044.The models of AtCYT6-NTD and AtCYT6-CTD corresponded to residues A35-D127 and G141-D234, respectively, from the Uniprot entry Q8H0X6.The PyMOL Molecular Graphics System (Schr€ odinger LLC, Portland, United States) was used to illustrate protein models.

Pull-down assay with recombinant AtLEGb
For pulldown experiments with recombinant AtLEGb, VPE0 mutant (accession N67918) seed stocks were obtained from the Nottingham Arabidopsis Stock Center (NASC, Nottingham, UK) and grown on soil at short day conditions (9 h light with an intensity of 120 lE m À2 sec À1 at 22°C and 15 h darkness at 18°C, 75% RH) after stratification for 3 days at 4°C.Leaves of 5-week-old plants were harvested and homogenized using a Polytron PT-2500 (Kinematica, Luzern, Switzerland) in extraction buffer (100 mM MES pH 6.0, 100 mM NaCl, 250 mM sucrose, 1% Triton X-100, and HALT protease inhibitor cocktail; ThermoFisher, Dreieich, Germany) followed by centrifugation at 4000 g, 4°C for 5 min.The lysate was then quantified using the BCA assay (ThermoFisher).A total of 2 mg of lysate per condition together with 25 ll of Dynabeads TM for His-Tag isolation (Invitrogen, Camarillo, CA, USA) were incubated with His 6 -AtLEGb or the His 6 -proAtLEGb-C211A dead mutant or lysate alone for an hour at room temperature with rotation.To block its proteolytic activity, His 6 -AtLEGb was pre-inhibited with MMTS.The Dynabeads were then washed four times with 400 ll of wash buffer (100 mM MES pH 6.0, 300 mM NaCl, 1% Triton X-100).Elution was performed in 100 ll of elution buffer (300 mM imidazole, 300 mM NaCl, 100 mM HEPES pH 7.5, 1% Triton X-100).For sample processing prior to mass spectrometry measurement, each eluted fraction was then heated to 56°C for 10 min followed by reduction with 20 mM DTT for a further 30 min under shaking conditions at 1500 rpm and alkylation with 50 mM CAA (chloroacetamide) at room temperature for 30 min in the dark.The reaction was quenched with 50 mM DTT for 30 min and purified with SP3 beads.For each reaction, 5 ll of SP3 bead stock (1:1 mixture of hydrophilic and hydrophobic carboxylated beads, 1 ll per 20 lg of protein) was added and protein binding to the beads was induced with a final concentration of 80% ethanol followed by incubation at room temperature for 30 min with rotation.The beads were then washed twice with 90% acetonitrile before being resuspended in trypsin solution (50 mM HEPES pH 7.5, 5 mM CaCl 2 , 1 lg trypsin) and incubated at 37°C for 16 h and shaking at 1500 rpm.For triplex differential labeling of the peptides, the conditions were labeled with 30 mM light, CH 2 O (lysate alone), medium, CD 2 O (AtLEGb), heavy, 13  (proAtLEGb-C211A), and 15 mM sodium cyanoborohydride at 37°C for 1 h, 1500 rpm.The reaction was then quenched with 100 mM Tris for 30 min.The labeled peptides from the three conditions were pooled 1:1:1 and acidified with final 1% formic acid trifluoroacetic acid before passing through pre-activated C18 StageTips.

AtCYT6 CoIPs
For co-immunoprecipitation assays, we purchased rabbit sera (BioGenes, Berlin, Germany) of animals treated with injections of AtCYT6-NTD or AtCYT6-CTD separately.The antibodies raised against each AtCYT6 domain were purified from the sera using CNBr-activated Sepharose 4 Fast Flow beads (GE Healthcare) with immobilized recombinant AtCYT6-NTD and AtCYT6-CTD according to the manufacturer's protocol.The integrity of the purified antibodies was tested in western blot assays against AtCYT6-NTD and AtCYT6-CTD, and each pool of purified polyclonal antibodies displayed high sensitivity and specificity to its respective antigen after this step, so we moved on to the co-immunoprecipitation assay with plant material.To that end, we used rec-Protein G-Sepharose 4B Conjugate beads (Invitrogen) cross-linked to the polyclonal antibodies with dimethylpimelidate-dihydrochlorid (Thermo Fisher Scientific) to prevent leakage of antibodies to the supernatant in further steps.The empty beads, that is, Sepharose beads conjugated with protein G that were not cross-linked to antibodies, constituted an important control and are referred to as bait 1.The beads that were cross-linked to antibodies are bait 2.
For the co-immunoprecipitation using antibodies loaded with their recombinant antigen as baits, antibodies freshly cross-linked to the beads were further incubated with recombinantly expressed antigens, that is, AtCYT6-NTD or AtCYT6-CTD, in buffer composed of 50 mM HEPES pH 7.5 and 100 mM NaCl for 1 h at 4°C.The unbound antigens were washed away with multiple centrifugations (at 4000 g and 4°C) with elimination of the supernatant and resuspension in buffer composed of 50 mM HEPES PH 7.5 and 100 mM NaCl 100 mM until the UV signal at 280 nm was stable and null.This corresponds to the bait 3.
The preparation of the plant material followed with the maceration of 100 mg of A. thaliana seeds in liquid nitrogen until a fine powder was obtained.Immediately after grinding, 0.6 ml of incubation buffer [100 mM citric acid PH 5.5 and 250 mM NaCl with inhibitor cocktail (Roche, Rotkreuz, Switzerland) dilution according to the instructions of the manufacturer] was added and thoroughly mixed for 10 min at 4°C.The suspension was centrifuged three times at 13 000 g for 15 min at 4°C with change of tubes in between.The resulting supernatant was split into three tubes with the same volume of material, and to each tube, either bait 1, 2, or 3 was added and incubated at 4°C with gentle and constant rotation for 1 h.After that, the beads were washed thoroughly with incubation buffer in multiple cycles of centrifugation and discard of supernatant until the UV signal at 280 nm of the supernatant was stable and null.The co-precipitated proteins were eluted with 19 Laemmli/SDS sample buffer without DTT at 37°C for 15 min in a shaker at 750 rpm.Then, we performed an in-gel digestion with trypsin.The eluted samples from coimmunoprecipitation assays with bait 1, 2, and 3 were loaded onto an acrylamide gel for SDS-PAGE and allowed to run 1-2 cm into the separation gel.All the area supposedly containing proteins was cut out, reduced to pieces of approximately 1 mm 9 1 mm, and collected in low binding tubes.The gel pieces of each tube were washed with buffer A (50 mM ammonium bicarbonate) and then with buffer B (25 mM ammonium bicarbonate in 50% acetonitrile).The proteins trapped within the gel pieces were reduced with reduction buffer (buffer A with 20 mM DTT) for 30 min at 56°C, alkylated with alkylation buffer (buffer A with 10 mM iodoacetamide) for 30 min at room temperature in the dark, and then washed with buffer A and B. The drying of the gel pieces started with shrinking by the addition of 100% acetonitrile and finished with complete drying in a SpeedVac at 50°C until the gel pieces were fully dried.For trypsin digestion, the gel pieces were hydrated again with a solution of 12.5 ng ll À1 trypsin in digestion buffer (25 mM ammonium bicarbonate, 5% acetonitrile, and 5 mM CaCl 2 ) and incubated overnight at 37°C.The supernatants containing the peptides were transferred to new tubes and 100% acetonitrile was added to the gel pieces and incubated for 15 min at room temperature.The acetonitrile supernatant was combined with the supernatant previously collected and the resulting sample was dried in a SpeedVac at 50°C until a minimum volume was reached.The volume of the samples was adjusted to 25-50 ll and pH was adjusted to <3 with 2.5% formic acid before desalting the samples with C18 double-layer stage tips (Empore Octadecyl C18, Supelco, Bellefonte, PA, USA).

Western blots of CoIP samples
The samples collected in the CoIP experiment were further analyzed via Western blotting using rabbit anti-AtCYT6-NTD or -CTD antisera respectively.Specifically, 2 ll of the elutions were subjected to SDS-PAGE and transferred to an Amersham Protran 0.45 lm nitrocellulose blotting membrane (GE Healthcare) using a Trans-Blot semi-dry transfer cell (Bio-Rad, Vienna, Austria).10 ng of recombinant AtCYT6-FL, -NTD, and -CTD were loaded as controls.The membranes were washed three times with TBST and subsequently blocked with 5% milk powder dissolved in TBST.Incubation with the respective rabbit antisera was done in the dark at 22°C for 1 h in a 1:40 dilution with TBST.Subsequently, membranes were washed with TBST and incubated with goat anti-rabbit-IgG coupled to horseradish peroxidase (Santa Cruz Biotechnology, Dallas, TX, USA) in a 1:12 000 dilution in TBST for 1 h.Chemiluminescence signals were detected using the Image Lab software (version 6.0.1;Bio-Rad) after the addition of Amersham ECL Prime Western Blotting detection reagent (GE Healthcare).

Mass spectrometry analysis of pulldown and CoIP samples
AtLEGb pulldown assays were analyzed on a two-column nano-HPLC setup (Ultimate 3000 nano-RSLC; Thermo Fisher Scientific, Waltham, Massachusetts) equipped with Acclaim PepMap 100 C18 columns (ID 75 lm, particle size 3 lm columns, trap column of 2 cm length, analytical column of 50 cm length, Thermo) operated at 60°C.For AtCYT6-CoIP samples, the same system was equipped with a lPAC reverse phase trap and analytical (50 cm) columns (Thermo) operated at a column temperature of 40°C.Peptides were eluted with a binary gradient from 5% to 32.5% B for 80 min (A: H2O + 0.1% FA, B: ACN + 0.1% FA) and a total runtime of 2 h per sample.Spectra were acquired in data-dependent data acquisition mode using a high-resolution Q-TOF mass spectrometer (Impact II, Bruker, Bremen, Germany) as described (Soh et al., 2020).

Mass spectrometry data analysis
Tandem mass spectra acquired for AtLEGb pulldown assays were analyzed with the MaxQuant software package (Tyanova et al., 2016), v1.6.0.16, with  N-terminal acetylation as variable modifications and triplex stable isotope labeling of peptide N-termini and Lys residues with light formaldehyde and sodium cyanoborohydride (+28.0313Da), deuterated formaldehyde ( 12 CD 2 O) and sodium cyanoborohydride (+32.0564Da) and 13 CD 2 O formaldehyde and sodium cyanoborodeuteride (+36.0756Da).Standard settings for Bruker QTOF data were used with standard settings for protein identification including PSM and protein FDR <0.01.AtCYT6-CoIP data was analyzed with MaxQuant v.2.0.3.0 using essentially the same settings and database, except that no labeling was applied and LFQ was enabled with minimal LFQ count set to 1 for label-free quantification.

ACCESSION NUMBERS
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (Deutsch et al., 2023) via the PRIDE (Perez-Riverol et al., 2022) partner repository with the dataset identifiers PXD042260 for the AtLEGb pulldowns and PXD042261 for the AtCYS6 CoIP data.S1.Mass spectrometry analysis (intact mass) of AtCYT6derived constructs expressed in E. coli.Table S2.Cleavage sites (P1 residues) of AtLEGb within AtCYT6-NTD and AtCYT6-CTD identified by mass spectrometry analysis.

Figure 1 .
Figure 1.AtCYT6 is a two-domain phytocystatin that interacts with AtLEGb.(a) AlphaFold model of full-length AtCYT6 depicting the N-terminal signal peptide (gray), the N-terminal domain (orange), the inter-domain linker, and the C-terminal domain (blue).(b) Structure alignment of the AlphaFold models for the N-(orange, AtCYT6-NTD) and C-terminal (blue, AtCYT6-CTD) domains.P1-Asn residues are shown in sticks; reactive center loops (RCL) are highlighted in yellow; legumain exosite loops are highlighted in gray; the binding site for PLCPs in the N-terminal domain is indicated by dashed lines.(c) Sequence alignment of hCE (black), AtCYT6-NTD (orange), and AtCYT6-CTD (blue) generated with Aline.Secondary structure elements are indicated, the P1-Asn residues are highlighted in green, and PLCPs inhibition motifs are in gray.(d) Correlation of pull-down data of AtLEGb/beads and proAtLEGb/beads.Extracts of A. thaliana leaves of a VPE 0 strain were spiked with recombinant His 6 -tagged (pro)AtLEGb and co-precipitated via Ni 2+ -affinity chromatography.AtCYT6 and AtLEGb-derived peptides displayed the highest log 2 -increase relative to control beads.

Figure 2 .
Figure 2. The interaction of AtCYT6 with AtLEGb is mediated by the C-terminal domain (AtCYT6-CTD).AtLEGb was inhibited with methyl-methanethiosulfonate (MMTS), incubated with (a) alkylated AtCYT6-FL, (b) AtCYT6-NTD or (c) AtCYT6-CTD at pH 5.5 and complex formation was analyzed by size exclusion chromatography.Indicated fractions were analyzed by SDS-PAGE.Control reactions contained only AtLEGb or the respective AtCYT6 construct.

Figure 3 .
Figure 3. AtCYT6 is a potent inhibitor of AtLEGb and papain.(a) The turnover of the AAN-AMC substrate by AtLEGb was assayed at pH 5.5 in the absence (light gray) and presence (dark gray) of AtCYT6-FL, AtCYT6-NTD or AtCYT6-CTD.(b) Turnover of the FR-AMC substrate by papain was measured at pH 6.5.

Figure 4 .
Figure 4. Co-immunoprecipitation assays cross-confirmed interaction of AtLEGb and AtCYT6 in vivo.(a) Polyclonal anti-AtCYT6-NTD (Anti-AtCYT6-NTD) or anti-AtCYT6-CTD antibodies were cross-linked to protein G-beads and incubated with seed extract at pH 5.5.Protein abundances are shown as log 2 of the ratio of label-free quantification intensities of test samples and control beads (null values were substituted by 1 to enable the calculations).(b) Same as (a) but antibodies were loaded with the respective recombinantly expressed antigen before incubation with the seed extract.AtCYT6-derived peptides were the most abundant in both experiments.AtLEGb was co-eluted in the presence of the anti-AtCYT6-CTD antibody and with antibodies loaded with recombinant AtCYT6-NTD or AtCYT6-CTD.Ó 2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2023), 116, 1681-1695

Figure 6 .
Figure 6.The interaction of AtCYT6 with AtLEGb critically depends on the RCL and the LEL.(a) Proposed binding model of AtCYT6-NTD (orange) and AtCYT6-CTD (blue) to AtLEGb (green) in comparison to the crystallographic structure (PDB 4N6O) of the complex between hLEG (wheat) and hCE (pink).The reactive center loops (RCLs) are indicated by a black arrow, and the position of the P1-Asn residues are labeled with a star (LEL, legumain exosite loop).Models were generated with AlphaFold Multimer.(b) Model depicting the relative positions of Asn68/173 (black star) on the RCLs and Glu97 (light purple circle) and Arg205 and Lys208 (dark purple circles) on the LELs of AtCYT6-NTD or -CTD respectively.(c) Turnover of the AAN-AMC substrate by AtLEGb at indicated pH values.The inhibition of AtLEGb by AtCYT6-CTD is pH dependent.(d) AtCYT6-NTD and the AtCYT6-NTD-E97R mutant did not inhibit AtLEGb at pH 5.5.

Ó
2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2023), 116, 1681-1695

Figure 7 .
Figure 7. AtCYT6 has a stabilizing effect on AtLEGb.The thermal stability of AtLEGb was measured by nanoDSF (differential scanning fluorimetry) measurements.Unfolding transitions (inflection point; T i ) are indicated by solid circles.(a) AtLEGb was inhibited with MMTS and incubated at pH 6.5 in the absence (black) and presence of AtCYT6-FL (dark gray) or AtCYT6-CTD (light gray).(b) Same as (a) but after incubation with AtCYT6-CTD-N173A (dark gray) or AtCYT6-CTD-K208A (light gray).For control measurements of AtCYT6 constructs alone, see Figure S6.

Figure 8 .
Figure 8. AtCYT6 is a substrate-like inhibitor of AtLEGb.(a) At near neutral pH, the interaction of AtCYT6-CTD to AtLEGb is stabilized by ionic interactions of the legumain exosite loop (LEL) to Glu202 on AtLEGb.(b) At acidic pH (<4.5) the complex is destabilized because of protonation of Glu202.Notably, AtCYT6-NTD functions as a substrate rather than an inhibitor at all pH conditions.

Ó
2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2023), 116, 1681-1695 the Uniprot Arabidopsis thaliana reference proteome database (release 2019_03) with appended sequence of recombinant proAtLEGb-C211A and list of common contaminants included in MaxQuant.Searches considered tryptic digest with up to two missed cleavages, Met oxidation and protein Ó 2023 The Authors.The Plant Journal published by Society for Experimental Biology and John Wiley & Sons Ltd., The Plant Journal, (2023), 116, 1681-1695

Figure S6 .
Figure S6.Sequence alignment of five type II phycys with demonstrated inhibitory activity against PLCPs and legumains.Figure S7.Thermal stability of different AtCYT6 constructs.

Table 1 K
i values of different AtCYT6 constructs against AtLEGb and papain (n.i.: no inhibition)