Determinants of N-degron recognition by PRT1

of variants mutated in the ZZ domain, as well as reciprocal pull-downs from plant cells under lower expression conditions, might reveal further regulatory mechanisms acting on PRT1 in vivo.

In summary, the results obtained here show that PRT1 acts as an E3 ligase that might be subject to posttranslational regulation. This implies that timely PRT1 action might be required under specific conditions in the plant, and hyperactivity is prevented by cellular mechanisms.

et al., 2006; Xia et al., 2008; Choi et al., 2010; Matta-Camacho et al., 2010; Wadas et al., 2016). When sequentially replacing the amino acids of the nsP4 sequence that interacted with PRT1 in the WT version, it was found that the first six amino acids were most important for PRT1 binding (Fig. 24A).

Next to the bulky hydrophobic N-terminal residue, these included amino acids with polar side chains (Ser-4, Thr-5) which might be required for the mobility of the peptide, in accordance with what was acknowledged before as a requirement for a functional N-degron (Bachmair and Varshavsky, 1989).

In line with this, the sequences of EIN2 and BB which showed very good PRT1 binding, were likewise rich in Gly and Ser residues, consistent with the assumption that the combination of these residues will increase flexibility of a peptide linker sequence (Van Rosmalen et al., 2017).

Interestingly, another hydrophobic residue (Phe-3) was furthermore found within the first six amino acids of nsP4 that proved to be important for PRT1 binding, and especially the well-binding BB sequence was also rich in aromatic residues. This indicates that the substrate binding pocket of PRT1 might be relatively hydrophobic and large. In the case of E. coli ClpS, it was proposed that a deep hydrophobic pocket mediates selection of bacterial N-degrons (starting Phe, Leu, Trp, or Tyr), potentially involving an induced-fit mechanism (Wang et al., 2008a; Schuenemann et al., 2009; Kwon et al., 2018). A similar mechanism might also apply to PRT1 substrate binding. Alternatively, the PRT1 binding pocket itself could be rather small, and hydrophobic residues might be required to properly position the substrate N-terminus on the PRT1 surface. Of interest in this context, when comparing the ZZ domains of p62 and PRT1, Kwon et al. (2018) noted the presence of additional hydrophobic residues in the PRT1 ZZ domain when compared to the binding pocket of p62. Investigation of the PRT1 crystal structure in complex with a substrate peptide would help to test wether such hydrophobic residues of PRT1 are specifically involved in positioning PRT1-substrates in the binding pocket.

Notably, throughout the peptide sequences tested for PRT1 interaction, the three best binding sequences, K2 and the neo-N-termini of EIN2 and BB, all exhibited a low net charge at neutral pH, owing to an overall low number of charged residues. This is in contrast to the Δ138 N-terminus of RD21A which, despite of a neutral net charge, contains a rather high number of charged amino acids, and could not bind PRT1 detectably even when Phe was in N-terminal position (Fig. 26B). It is possible that differently charged amino acids within a peptide sequence interfere with PRT1 binding because they impose a rather rigid peptide structure. The preference of PRT1 for N-termini with low net charges was furthermore supported by the observation that replacement of distal Lys residues (Lys-14 and Lys-15) of nsP4 by Ala or Gly strongly promoted pull-down of PRT1 by this peptide (Fig.

24), whereas replacement by another positively charged amino acid, Arg, had only a modest effect (Fig. 24B). Moreover, the positively charged Lys-2 was the only residue among the first six amino acids of a tested substrate peptide, substitution of which did not dampen PRT1 binding. The preference of low net-charged N-termini by PRT1 differs from E. coli ClpS substrate binding, where a preference for N-termini with positive net charge was reported (Erbse et al., 2006).

Finally, despite of the apparent preference for low net charges, all three PRT1 binders (K2, EIN2, BB) contained a protonatable residue in penultimate position. Interestingly, this was accompanied by the occurrence of a negatively charged residue in more downstream position, namely Glu-7 and Asp-9 in EIN2 and BB, respectively, although this co-occurrence of positive and negative charges is also likely to decrease structural flexibility. Consistent with the presence of negatively charged residues in peptides that bound well to PRT1, replacement of Asp-6 in nsP4 also weakened the interaction (Fig.

24A). Potentially, this negative charge compensates for the positive charge in position 2 to restore neutral net charge.

Together, based on the binding studies performed with peptide arrays it can be concluded that a 17-mer N-terminal peptide optimal for PRT1 recognition is characterized by 1) an aromatic N-terminal residue; 2) a protonatable residue in penultimate position; 3) a negatively charged residue in a central position; 4) a high number of Gly and Ser residues providing segmental mobility; and 5) preferentially additional hydrophobic residues throughout the sequence.

4.3.1.1 Distal Lys residues in the N-terminus of a PRT1 substrate

The results obtained here concerning Lys in penultimate position of a PRT1 substrate peptide are rather contradictive. On the one hand, the neo-N-terminus of BB which proofed to be a very good ligand for PRT1 in vitro, bears a Lys residue in position 2, indicating that Lys at this position is advantageous for binding. This is also in agreement with the assumption that a protonatable residue in penultimate position of a peptide substrate increases PRT1 affinity. On the other hand, replacement of Lys-2 by Ala in the nsP4 sequence did not diminish PRT1 binding, or even improved pull-down of the N-recognin slightly (Fig. 24A). It would have to be tested whether Lys in second position is only beneficial in certain sequence contexts, for example when contributing to a neutral net charge as in the case of BB, or if binding of PRT1 to the BB sequence would be even improved upon replacement of Lys-2.

In contrast, Lys in the distal part of an N-terminal substrate sequence (Lys-14 and Lys-15 of nsP4) clearly decreased PRT1 binding compared to Ala in this position (Fig. 24). This was surprising since E3 ligase action depends on the presence of Lys residues for Ub transfer. One explanation for this observation could be the idea that N-recognins act by scanning the substrate sequence for a ubiquitination site, eventually allowing for substrate release (Eldeeb et al., 2018a). Although Ub was not provided in the SPOT assays performed here, it is conceivable that the absence of Lys in a sequence leads to longer dwell times of the E3 at the peptide during a process of sequence scanning, resulting in higher steady-state protein levels at the peptide spot. In disagreement with this idea however, optimal Ub transfer to a substrate by an N-recognin appears to occur considerably more downstream of the substrate’s N-terminus than residue number 15 (Bachmair and Varshavsky, 1989). Hence, a potential process of sequence scanning by PRT1 would likely target more distal parts of the substrate protein, and dwell time effects on 17-mer artificial peptides might be negligible. In contrast, the simpler conclusion for the observation that PRT1 pull-down was augmented upon replacement of distal Lys residues would be that Lys in a downstream peptide sequence decreased PRT1 affinity. In fact, this hypothesis would be directly in line with observations made by (Bachmair and Varshavsky, 1989). Here, replacement of Lys-15 in the e^K N-degron linker sequence by Arg even improved the efficiency of the linker to act as a type II N-degron in yeast. This effect was however position specific, since K17R mutation led to a significant reduction in N-degron efficiency, prompting the authors to propose the requirement of properly spaced Lys residues for degron function (Bachmair and Varshavsky, 1989). Bachmair and Varshavsky (1989) inferred a competition mechanism between the two Lys residues of e^K at position 15 and 17 for a binding site at the N-recognin, and suggested that degron efficiency might be decreased when Lys-15 is occupying this binding site as compared to Lys-17.

These considerations would suggest that decreased affinity, rather than shortened dwell time of PRT1 at the peptide, explains the lower protein signals at the containing compared to the Lys-deleted peptide sequences. However, in the experiment presented in Fig. 24B, PRT1 occupancy was also strongly increased upon replacement of all Lys residues in nsP4 (i.e., Lys-15 and Lys-16), in a manner only partially independent of the positive charge elimination (Fig. 24B). This contradicts the explanation that competition effects between the Lys residues are responsible for a binding effect.

Instead, the result obtained here suggests that 1) positive charge in the distal region of a substrate’s N-terminus interferes with PRT1 binding, and 2) Lys in this region furthermore conflicts with the stability of the association between PRT1 and a peptide ligand by a yet unknown mechanism.

4.3.1.2 PRT1 substrates bear a protonatable residue in position 2

All three peptide sequences that were found to interact strongly with PRT1 in SPOT peptide assays (i.e., K2, and the neo-N-termini of EIN2 and BB) contained a protonatable residue in penultimate position. Remarkably, also all of the previously employed artificial substrates used to address PRT1 activity in vivo, i.e. for capturing the prt1-1 allele (Bachmair et al., 1993; Potuschak et al., 1998), as well as for subsequent studies (Stary et al., 2003; Garzón et al., 2007; Faden et al., 2016), harbored a protonatable residue in second position. The experiments obtained from SPOT assay experiments now suggest that the affinity of PRT1 might in fact be strongest towards substrates initiated by aromatic N-termini followed by a basic amino acid. Notably, in a proteomics approach, Majovsky et al. (2014) found that some substrates of the type I branch of the N-end rule were also modestly enriched upon loss of PRT1 function, indicating that PRT1 also slightly contributes to the degradation of these proteins. The assumption that the PRT1 binding pocket prefers peptides with positive charge close to the N-terminus would potentially explain this observation. Compellingly, this finding also implies the possibility that the genetic screen performed by Bachmair et al. (1993) might have actually resulted in the identification of a specialized N-recognin targeting type II N-degrons with a protonatable second residue. This would lead to the conclusion that other substrates with aromatic neo-N-termini might remain unstable upon loss of function of PRT1, and would allow the postulation of the existence of further N-recognins with type II specificity in Arabidopsis. Alternatively, the specificity for basic amino acids in second position might be a common feature among type II N-recognins, given that also several ClpS proteins showed preferential binding to peptides with positively charged amino acid in penultimate position during SPOT assay experiments (Erbse et al., 2006; Stein et al., 2016).

There are three proteins from other organisms than Arabidopsis known to confer binding preferentially to aromatic, rather than branched hydrophobic N-degrons. These are the ZZ domain of mammalian p62 (Cha-Molstad et al., 2017; Kwon et al., 2018), the ClpS2 N-end rule substrate adaptor of Agrobacterium tumefaciens, and ClpS1 from Synechococcus elongatus (Stein et al., 2016).

Comparison of PRT1 and p62 substrate binding is worthwhile given that these two are the only known N-recognins engaging a ZZ domain for substrate interaction. Kwon et al. (2018) investigated the crystal structure of the ZZ domain of p62 in complex with different N-degrons. Thereby they also noted structural homology of the p62 ZZ domain with the one from PRT1, including the conservation of an Asp residue responsible for binding of the substrate’s α-amino group (Kwon et al., 2018). Given that p62 has the strongest affinity towards Arg-initiated degrons which is a type I residue (Kwon et

al., 2018), it would be also conceivable that preference of PRT1 towards substrates with a positive residue in second position reflects an evolutionarily old feature of the ZZ domain.

Interestingly, many α-proteobacteria and also cyanobacteria contain an additional ClpS protein, and the variants from A. tumefaciens and S. elongatus were shown to be specific for aromatic N-degrons (Stein et al., 2016), as is PRT1. Moreover, like PRT1, all ClpS proteins investigated so far exhibit a preference for positively charged residues in position 2 (Erbse et al., 2006; Stein et al., 2016). In contrast, the N-terminus of the known natural substrate of p62, BiP, is of highly acidic nature (Cha-Molstad et al., 2015). Hence, although sequentially different, the PRT1 binding pocket might actually have specialized to function similarly to the bacterial ClpS domain (Erbse et al., 2006; Stein et al., 2016). While E. coli ClpS does not discriminate between aromatic amino acids and Leu in N-terminal position, the examples of A. tumefaciens ClpS2 and S. elongatus ClpS1 show that sequence variations of ClpS can principally lead to specificity for aromatic amino acids. It would be interesting to learn if such specialized ClpS proteins show structural analogy to PRT1, and if PRT1 and these ClpS proteins have evolved in parallel to perform similar substrate binding reactions.

With the criteria for PRT1-binding determined above, it is possible to predict whether a neo-N-terminus, once proteolytically exposed, principally qualifies as a substrate for PRT1. The uncertainty of in-planta generated N-terminal sequences however renders the search for potential substrate candidates difficult. While a growing number of studies provides information on the steady-state N-terminal proteome of Arabidopsis (Tsiatsiani et al., 2013; Rowland et al., 2015; Li et al., 2017b;

Willems et al., 2017), data on the unstable N-terminome is still limited, and to date has mostly addressed the type I substrate branch of the Arabidopsis N-end rule (Majovsky et al., 2014; Zhang et al., 2015b, 2018b). The analysis of proteomes upon proteasome inhibition using methods suitable for the identification of endogenously created N-termini, such as COFRADIC and TAILS, could help in the future to provide the information needed for the identification of plant N-end rule substrates outside of the oxygen- and NO-dependent pathway.