An evolutionarily distinct chaperone promotes 20S proteasome α-ring assembly in plants

Selective protein degradation is crucial for almost every aspect of cellular physiology. While the degradation of mis-translated, mis-folded, damaged or otherwise malfunctioning proteins is essential for maintaining intracellular proteostasis (Marshall and Vierstra, 2019; Pohl and Dikic, 2019), the timely removal of regulatory proteins is required for most, if not all, cellular processes, including the cell cycle, signal transduction, neuronal signaling, transcription, DNA repair and translation (Ahuja et al., 2017; Brinkmann et al., 2015; Ramachandran et al., 2018). Consequently, disruptions in selective proteolysis often lead to an array of pathologies, including hypersensitivity to proteotoxic stress, cancer, neurodegeneration, immune-related disorders and accelerated ageing (Higuchi-Sanabria et al., 2018; Pilla et al., 2017; Rape, 2018).

A main proteolytic route in eukaryotes is the ubiquitin–proteasome system (UPS) (Finley et al., 2012; Marshall and Vierstra, 2019). Here, substrates are targeted by the selective attachment of polyubiquitin chains with appropriate topologies by a myriad of substrate-specific E3 ubiquitin–protein ligases (Zheng and Shabek, 2017). Breakdown of the ubiquitylated proteins is then directed by the 26S proteasome, which recognizes the polyubiquitin, unfolds the modified protein, releases the ubiquitin moieties for re-use, and cleaves the linear polypeptide into short fragments using relatively non-specific peptidase activities (Finley and Prado, 2020; Greene et al., 2020; Marshall and Vierstra, 2019).

The 26S proteasome is composed of two stable and functionally distinct subcomplexes; the 20S core protease (CP) that provides the peptidase activities, capped at one or both ends by the 19S regulatory particle (RP) that captures and prepares appropriate substrates for breakdown (Bhattacharyya et al., 2014; Greene et al., 2020). The CP assumes a barrel shape generated by four stacked heptameric rings that form upon the association of seven α- and β-subunits in a C₂-symmetric αββα configuration. Whereas eukaryotes employ structurally related but distinct α- and β-subunit isoforms to generate the barrel, their archaeal and bacterial progenitors often use single α- and β-subunit proteins to assemble homo-heptameric CP rings (Gille et al., 2003; Majumder and Baumeister, 2019). Upon assembly, a central chamber is created at the interface between the β-rings; in eukaryotes the β₁, β₂, and β₅ subunits provide six catalytic sites within this lumen (Arendt and Hochstrasser, 1997; Heinemeyer et al., 1997), whereas in archaea and bacteria, all 14 β-subunits are presumably active (Majumder et al., 2019; Seemüller et al., 1995).

Atop each β-subunit ring sits the α-subunit ring, which creates two narrow axial pores that are gated by a short N-terminal sequence extending from several α-subunits (Groll et al., 2000; Köhler et al., 2001; Rabl et al., 2008). Gate opening is normally triggered by docking of the CP to various proteasome regulators (Majumder and Baumeister, 2019), the most prominent being the RP. These regulators, or their subunits, often possess an extended C-terminal HbYX or HbF motif (Hb, hydrophobic residue; Y, tyrosine; F, phenylalanine; and X, any amino acid) that inserts into a shallow pocket formed at the interface between adjacent α-subunits (Rabl et al., 2008; Smith et al., 2007; Tian et al., 2011). Through this distinctive and stable architecture, the CP acts as a self-compartmentalized protease that degrades polypeptides that are deliberately imported into the β-ring chamber.

Assembly of a complete 26S proteasome from its 32 or more distinct constituents is a highly complicated process that requires a suite of dedicated chaperones and maturation factors for precise construction of both the CP and RP (Howell et al., 2017; Marshall and Vierstra, 2019; Rousseau and Bertolotti, 2018). As the eukaryotic CP requires the orderly recruitment of the α- and β-subunits, its correct formation is particularly challenging to maintenance of a healthy proteasome pool devoid of dysfunctional products. CP assembly typically begins with construction of the individual α-rings, followed by β-subunit incorporation (Hirano et al., 2008; Li et al., 2007), though α-ring-independent assembly pathways have also been reported (Hammack et al., 2020; Panfair et al., 2015). Creation of the eukaryotic α-ring is controlled by a hetero-dimeric chaperone termed Pba1–Pba2 (also known as Pba1–Add66) in the budding yeast Saccharomyces cerevisiae and PAC1–PAC2 in mammals (also known as PSMG1–PSMG2) that provides a template onto which several α-subunits correctly associate, namely α₅, α₆ and α₇ (Hirano et al., 2005; Kock et al., 2015; Stadtmueller et al., 2012; Wani et al., 2015). Binding of Pba1 and Pba2 to the cognate α-subunit pairs is stabilized through interactions between their C-terminal HbYX motifs (HbF in mammalian PAC2) and pockets at the α₅–α₆ and α₆–α₇ interfaces, respectively, thus placing the α₅–α₆–α₇ trimer in correct register (Kusmierczyk et al., 2011; Wu et al., 2018). Although still viable, yeast lacking Pba1 or Pba2 accumulate immature CP species containing structurally unstable α-rings, from which α₅ and α₆ readily dissociate (Wani et al., 2015), whereas comparable mammalian mutants accumulate fewer α-rings (Hirano et al., 2005).

How the α₁, α₂, α₃ and α₄ subunits then integrate is unclear, but the hetero-dimeric Pba3–Pba4 chaperone in yeast (also known as Irc25–Poc4) and the PAC3–PAC4 chaperone in mammals (also termed PSMG3–PSMG4) appear critical for correct stoichiometry of the α₃ and α₄ subunits (Kusmierczyk et al., 2008; Padmanabhan et al., 2016; Takagi et al., 2014), and integration of the α₁ subunit appears rate limiting (Howell et al., 2019). The Pba1–Pba2 and Pba3–Pba4 pairs also prevent premature association of CP assembly intermediates with the RP or other activating factors, in some cases by outcompeting HbYX-containing RP subunits such as Rpt5 for CP binding until the CP rings mature (Kock et al., 2015; Stadtmueller et al., 2012; Wani et al., 2015).

Upon completion, the α-ring provides a platform for β-ring assembly, the orderly recruitment of which requires the maturation factor Ump1 (POMP in humans) (Gemperline et al., 2019; Ramos et al., 1998) and the N-terminal propeptides present in several β-subunits (Chen and Hochstrasser, 1996; Hirano et al., 2008; Li et al., 2007). A transient 15S ‘half-proteasome’ emerges, which ultimately dimerizes to generate the mature 20S CP, possibly with the help of the capping factor PA200 (also known as PSME4 in humans; Blm10 in yeast) which also harbors a HbYX motif (Guan et al., 2020). Assembly of the mature CP is rate limited by final integration of the β₇ subunit. This association promotes half-proteasome dimerization followed by auto-catalytic removal of the N-terminal propeptides in β₁, β₂ and β₅, which then exposes their catalytic N-terminal threonine residues.

Although our understanding of CP assembly in yeast and mammals is reasonably well developed, the pathway(s) for constructing the plant CP remain unclear. Our prior proteomic analyses of Arabidopsis proteasomes led to the discovery of likely plant orthologs of yeast Pba1–Pba4 and Ump1, strongly suggesting that CP assembly follows the same path (Gemperline et al., 2019; Wang et al., 2019). Surprisingly, whereas yeast Δump1 mutants are viable, Arabidopsis mutants null for its two functional orthologs are lethal due to defects in gametogenesis and/or seed development, implying that subtle differences in CP assembly exist. Arabidopsis mutants missing the dominant UMP1a isoform (At1g67250) are viable, but they are hypersensitive to proteotoxic stress and have compromised 26S proteasome assembly, consistent with UMP1 having CP maturation activity (Gemperline et al., 2019).

Besides the above chaperones, our mass spectrometric (MS) analyses also identified a fifth possible chaperone associated with the Arabidopsis CP, whose amino acid sequence is distantly related to PBAC1 and PBAC2, but includes the signature HbYX motif (Gemperline et al., 2019). Here we show that this protein, designated as PROTEASOME BIOGENESIS-ASSOCIATED CHAPERONE 5 (PBAC5; At3g07640), assembles with PBAC1 and PBAC2 to form a trimeric complex that can replace the Pba1–Pba2 dimer in yeast α-ring assembly. Arabidopsis pbac5 mutants, like those missing PBAC1 and PBAC2, have CP assembly defects, elevated levels of ubiquitylated proteins and a hypersensitivity to proteotoxic stress, consistent with proteasome insufficiency. Remarkably, while PBAC5 orthologs could be found throughout the plant kingdom, they were also evident in a scattered array of fungal, metazoan and oomycete species. Taken together, the observations reported in this study extend our appreciation of proteasome assembly in plants, and for the first time describe a fifth chaperone required for efficient CP construction.

Our prior MS characterizations of Arabidopsis 26S proteasomes, affinity-purified through FLAG-tagged versions of the CP subunit PAG1 (α₇) or the RP subunit RPT4, identified all predicted core 26S subunits, plus an array of interacting proteins that included homologs of most yeast and mammalian CP and RP assembly chaperones (Book et al., 2010; Gemperline et al., 2019). One notable exception was an uncharacterized protein (PBAC5) of 241 amino acids found at high levels in preparations enriched in the CP alone, but not in the RP or the entire 26S particle. Because PBAC5 was most abundant in samples generated from seedlings treated with the proteasome inhibitor MG132, which also encourages proteasome synthesis (Gemperline et al., 2019), we speculated that it participates in CP assembly and not in the functions of the mature CP or 26S holocomplex.

Position-specific iterative basic local alignment search tool (PSI-BLAST) searches identified obvious orthologs of Arabidopsis PBAC5 throughout the plant kingdom, with 80–90% amino acid sequence identity in close relatives such as Arabidopsis lyrata and Brassica rapa, but just 30–35% identity in monocots such as rice and maize, and 20% identity in the moss Physcomitrella patens (Fig. S1A). By contrast, similar searches failed to find human or yeast relatives. Surprisingly, comparisons of hidden Markov models (HMM) constructed from different protein families (Söding, 2005) revealed a significant similarity (P<0.01) between the PBAC5 family and the PBAC1 and PBAC2 families (25% and 20% identity, respectively, in Arabidopsis), which by phylogenetic analyses clustered outside the plant PBAC5 family (Fig. 1A; Fig. S1B). This relationship was further supported by the presence of a C-terminal Leu-Tyr-Gly sequence analogous to the HbYX motif present in Pba1/PAC1/PBAC1, Pba2/PBAC2 and other CP α-ring-interacting proteins (Fig. 1B; Fig. S1A; Guan et al., 2020; Kusmierczyk et al., 2011; Rabl et al., 2008; Tian et al., 2011). Given this sequence homology and affinity for the CP, we predicted that PBAC5 participates in CP α-ring assembly.

Under the assumption that proteasome assembly chaperones should be co-ordinately expressed to meet proteolytic demand, we attempted to support this relationship between Arabidopsis PBAC5, PBAC1 and PBAC2 by gene co-expression analysis based on the transcript patterns available within the Transcriptome Variation Analysis database (www.travadb.org/; Klepikova et al., 2016). Unfortunately, a common theme was not detected, even among chaperones that should work together (e.g. PBAC1–PBAC2 and PBAC3–PBAC4; Fig. S1C). PBAC5 expression was highest in seeds, shoots, seedling meristems and, in contrast to the other CP chaperones, in senescing silique pods (Fig. S1C).

To confirm the interaction between PBAC5 and proteasomes, and to identify the subcomplex to which it binds, we generated a transgenic Arabidopsis line overexpressing HA-tagged PBAC5 for co-immunoprecipitation studies with seedlings treated with or without MG132 (Gemperline et al., 2019). Here, the fully functional HA–PBAC5 transgene was introduced into the pbac5-3 null mutant (see below) and expressed with the tag appended in-frame to the N-terminus of PBAC5, to avoid interference with the HbYX motif. Immunoblotting the immunoprecipitates from untreated seedlings easily detected HA–PBAC5 in association with the CP α-ring, based on co-enrichment with the CP α-subunits PAC1 (α₃) and PAG1 (α₇). However, no association was seen between HA–PBAC5 and the PBA1 (β₁) or PBF1 (β₆) subunits from the β-ring, or any of the seven RP subunits examined, despite their easy detection in the input samples (Fig. 2A).

Notably, the interactions between HA–PBAC5 and PAC1 (α₃) and PAG1 (α₇) were enhanced upon pretreating Arabidopsis with MG132 (Fig. 2A). We also detected unprocessed PBA1 (β₁) in these HA–PBAC5 immunoprecipitates, which not only confirmed the efficacy of the MG132 on inhibiting proteasome maturation (Book et al., 2010; Chen and Hochstrasser, 1996), but also demonstrated that PBAC5 preferentially associated with CP assembly intermediates. As compared to the inputs, which contained both unprocessed and mature forms of PBA1, only the unprocessed form was evident in the HA–PBAC5 immunoprecipitates (Fig. 2A).

Based on the observations that yeast and mammalian Pba1/PAC1 and Pba2/PAC2 assemble into hetero-dimers (Hirano et al., 2005; Le Tallec et al., 2007; Stadtmueller et al., 2012), and our observations that Arabidopsis PBAC5 co-enriched with PBAC1 and PBAC2 in affinity-purified proteasomes (Gemperline et al., 2019), we predicted that PBAC5 interacts with PBAC1 and/or PBAC2. As a first test, we assayed each combination using yeast two-hybrid (Y2H) assays. As shown in Fig. 2C, interactions were clearly evident for the PBAC1–PBAC2 and PBAC3–PBAC4 pairs, consistent with the known dimerization of their yeast and mammalian counterparts (Hirano et al., 2006; Kusmierczyk et al., 2008; Le Tallec et al., 2007; Takagi et al., 2014; Yashiroda et al., 2008). By contrast, PBAC5 showed a clear interaction only with PBAC1, and not with any of the other PBAC chaperones.

We next tested the interactions in planta with bimolecular fluorescence complementation (BiFC) assays using tobacco (Nicotiana benthamiana) leaf cells, which expressed the chaperones as N-terminal fusions with either the N- or C-terminal halves of yellow fluorescent protein (YFP). Again, reconstitution of cellular fluorescence was seen when the PBAC1–PBAC2 and PBAC3–PBAC4 pairs were co-expressed, but a similar signal was evident only when PBAC5 was co-expressed with PBAC1 (Fig. 2B; Fig. S2). Taken together, our data implies that Arabidopsis PBAC5 forms a hetero-trimeric complex with the proteasome assembly chaperones PBAC1 and PBAC2 in a PBAC5–PBAC1–PBAC2 orientation, and that this complex primarily interacts with the CP α-ring.

To help confirm that the PBAC5–PBAC1–PBAC2 complex is involved in proteasome assembly, we attempted to rescue yeast mutants missing Pba1 and Pba2 with their potential Arabidopsis orthologs. By themselves, neither the Δpba1 nor the Δpba2 strains (nor the Δpba1 Δpba2 double mutant) show obvious growth defects (Kusmierczyk et al., 2011; Li et al., 2007). However, these mutants become sensitive to proteotoxic stress when proteasome function is further compromised by secondary mutations, such as the doa5-1 allele impairing the CP α₅ subunit Doa5 (also known as Pup2; Chen and Hochstrasser, 1996), which can be seen by suppressed colony growth at low concentrations of the amino acid analog canavanine (Kusmierczyk et al., 2011). Although the doa5-1 Δpba1 and doa5-1 Δpba2 mutants were hypersensitive to canavanine, re-introduction of functional Pba1 or Pba2 from yeast, respectively, rescued this growth suppression (Fig. 3A).

When we introduced Arabidopsis PBAC1 or PBAC2 into the corresponding doa5-1 Δpba1 and doa5-1 Δpba2 backgrounds in various combinations, rescue of the canavanine hypersensitivity phenotype was not evident (Fig. 3A). Neither was PBAC5 effective, initially suggesting that the Arabidopsis chaperones are sufficiently distinct to prevent cross complementation, as was the case with UMP1 (Gemperline et al., 2019). However, when all three Arabidopsis chaperones were co-expressed in doa5-1 Δpba1 Δpba2 cells, canavanine hypersensitivity was nearly abolished (Fig. 3A). This partial rescue implied that a hetero-trimeric PBAC5–PBAC1–PBAC2 complex exists that can functionally replace the yeast Pba1–Pba2 heterodimer.

Considering that the HbYX motif in PBAC5 might be of similar importance as those predicted for PBAC1 and PBAC2, we next tested the need for these sequences in complementing doa5-1 Δpba1 Δpba2 cells, using mutants in which the tyrosine (in PBAC1 and PBAC5) or phenylalanine (in PBAC2) residues were replaced with alanine. Whereas the wild-type (WT) versions of yeast Pba1 and Pba2 rescued the canavanine hypersensitive phenotype, mutations impacting either HbYX sequence blocked full complementation, again demonstrating that this motif in both Pba1 and Pba2 is required for full chaperone function in yeast (Fig. 3B; Kusmierczyk et al., 2011). However, when the three Arabidopsis chaperones were co-expressed, we unexpectedly found that the PBAC5–PBAC1–PBAC2 complex could still partially rescue the canavanine hypersensitivity of the doa5-1 Δpba1 Δpba2 cells if only a single HbYX or HbF motif was altered, but not if two out of the three motifs were altered (Fig. 3B). Thus, in both yeast and Arabidopsis, at least two HbYX or HbF motifs in the Pba1–Pba2 and PBAC5–PBAC1–PBAC2 chaperone complexes are needed.

To support assembly of a hetero-trimeric PBAC5–PBAC1–PBAC2 complex with PBAC1 acting as the central tether, we exploited the complemented yeast strains described above to perform paired co-immunoprecipitation assays. As shown in Fig. 3C, the trimer could only be immunoprecipitated from cells expressing FLAG–PBAC1 along with both myc–PBAC2 and HA–PBAC5, and when the FLAG–PBAC1 gene was omitted, no co-immunoprecipitation of myc–PBAC2 and HA–PBAC5 was observed. For further confirmation and to avoid potential complications provided by other yeast proteasome components or chaperones, E. coli-expressed Arabidopsis PBAC1, PBAC2 and PBAC5 bearing N-terminal glutathione S-transferase (GST), 6His and HA tags, respectively, were affinity purified via their corresponding tags (Fig. 2D) and mixed in various combinations for in vitro pulldown assays using anti-HA antibody or nickel-nitrilotriacetic acid (Ni-NTA) chelate beads. Again, only when GST–PBAC1 was present could we simultaneously enrich for both 6His–PBAC2 and HA–PBAC5 (Fig. 2E).

To support their roles in Arabidopsis CP assembly, we examined the functions of PBAC1, PBAC2 and PBAC5 genetically using a collection of T-DNA insertion mutants, whose T-DNA positions were confirmed by genomic DNA sequencing around the insertion sites (Fig. 4A; Fig. S1B). Analyses by reverse transcription (RT)-PCR failed to detect corresponding full-length mRNAs in several of the mutants when homozygous, including the sole pbac1 allele (pbac1-1), the two pbac2 alleles (pbac2-1 and pbac2-2) and one pbac5 allele (pbac5-3), suggesting that these lines represent strong, if not null, alleles (Fig. 4B,C). Two additional pbac5 alleles (pbac5-1 and pbac5-2) were identified that had reduced transcript abundance; reverse transcription quantitative PCR (RT-qPCR) assays revealed that both had ∼25% of the mRNA expression found in WT (Fig. 4B,C). Interestingly, mRNA levels for PBAC5 were slightly elevated in the pbac1 and pbac2 null mutants compared to levels in WT (Fig. 4C), possibly indicating that the proteasome-stress regulon controlled by the NAM/ATAF1/CUC2-53 (NAC53) and NAC78 transcription factor pair was activated (Gladman et al., 2016), which would be consistent with a block in proteasome assembly (see below).

Under standard growth conditions, plants homozygous for each mutation appeared phenotypically normal and produced viable seeds, indicating that none of the three chaperones are essential to Arabidopsis by themselves. With a focus on PBAC5, we next tested whether the pbac5-3 mutant seedlings were susceptible to proteotoxic stress, as were mutants in the CP maturation factor UMP1a (Gemperline et al., 2019). We first measured the growth response of the seedlings when exposed to various concentrations of MG132 or canavanine, using the response of double mutant seedlings lacking NAC53 and NAC78 as a positive control (Gladman et al., 2016). WT seedlings grew reasonably well on low doses of MG132 or canavanine beginning at germination (25 or 5 μM, respectively), but stalled at higher concentrations (50 μM MG132 or 25 μM canavanine), whereas growth of the nac53-1 nac78-1 seedlings was strongly inhibited by even low drug doses (Fig. 4D,E). The pbac5-3 seedlings also displayed a strong hypersensitivity to both treatments, though not to the same extent as nac53-1 nac78-1 seedlings, indicating that this chaperone is important but not critical for proteostasis (Fig. 4D,E).

To confirm that this proteotoxic-stress phenotype was linked to the absence of PBAC5, we tested the sensitivity of homozygous pbac5-3 seedlings complemented with a transgene expressing either WT HA–PBAC5 or an HA–PBAC5(ΔHbYX) variant in which the HbYX motif was deleted. For this purpose, transgenic lines were selected that expressed similar levels of the HA–PBAC5 and HA–PBAC5(ΔHbYX) loci relative to WT PBAC5, as judged by RT-qPCR analysis (Fig. 4F), and that accumulated similar levels of the HA–PBAC5 and HA–PBAC5(ΔHbYX) proteins, as determined using immunoblot analysis of seedling extracts with anti-HA antibodies (Fig. 4G). The HA–PBAC5 transgene successfully rescued the MG132 and canavanine hypersensitivity of pbac5-3 seedlings, whereas the HA–PBAC5(ΔHbYX) transgene did not (Fig. 4D,E), thus firmly connecting the observed phenotypes to PBAC5, and demonstrating that the HbYX motif is essential for its activity in planta.

In addition to the proteotoxic stress hypersensitivity, pbac5-3 seedlings displayed other proteasome-related defects. As measured with the fluorogenic succinyl-LLVY-AMC substrate (Kisselev and Goldberg, 2005), pbac5-3 and HA–PBAC5(ΔHbYX) pbac5-3 seedlings had ∼55% less CP activity as compared to WT or HA–PBAC5 pbac5-3 seedlings, consistent with compromised CP assembly (Fig. 4H). The pbac5-3 seedlings also activated the proteasome-stress regulon, as evidenced by the 1.5- to 2-fold increase in transcripts encoding an array of CP and RP subunits, as determined by RT-qPCR (Fig. S3B). Furthermore, pbac5-3 seedlings hyper-accumulated ubiquitin conjugates, as predicted to occur if proteasome levels became insufficient (Fig. S3A). Notably, both activation of the proteasome-stress regulon and the increase in ubiquitin conjugates were suppressed by complementing the pbac5-3 mutant with the HA–PBAC5 transgene, but not with the HA–PBAC5(ΔHbYX) transgene (Fig. S3A,B). The pbac5-1 and pbac5-2 knockdown mutants also displayed similar aberrant phenotypes, but at levels intermediate between those of WT and the pbac5-3 null mutant, consistent with attenuated but not abolished expression of PBAC5 (Fig. 4C; Figs S4A–C, S5A).

Expecting that PBAC1 and PBAC2 work together with PBAC5 in CP assembly, we predicted that mutations eliminating PBAC1 or PBAC2 would similarly impact proteotoxic stress sensitivity, proteasome activity, proteasome-stress regulon activation and ubiquitin conjugate levels, as compared to the pbac5-3 mutant. Besides testing the pbac1-1, pbac2-1 and pbac2-2 single null mutants (Fig. 5A), we also examined higher-order mutant combinations, including the homozygous pbac1-1 pbac2-1 pbac5-3 triple mutant generated by introgression. Surprisingly, even the triple mutant was viable and grew indistinguishably to WT under standard conditions, again showing that none of the three chaperones are essential to Arabidopsis. However, analysis of the entire mutant collection with respect to proteasome function revealed a similar severity for a number of responses, suggesting that the three proteins work collectively in proteasome assembly. As examples, the pbac1-1 and pbac2-1 single homozygous mutants, either alone or in combination with pbac5-3 as double or triple mutants, displayed a similar hypersensitivity to MG132 and canavanine as the pbac5-3 mutant alone (Fig. 5A,B). These single, double and triple mutant combinations also had equal activation of the proteasome-stress regulon, as judged by increased levels of mRNAs encoding CP and RP subunits of the 26S proteasome (Fig. S5A,B), and a similar hyper-accumulation of ubiquitin conjugates (Fig. 5D). All the mutants also had near equal reductions in proteasome activity, based on the succinyl-LLVY-AMC substrate (Fig. 5C).

We previously showed that Arabidopsis mutants with attenuated expression of the CP maturation factor UMP1 are hypersensitive to salt and drought stress (Gemperline et al., 2019). Here, we tested whether our panel of pbac1, pbac2 and pbac5 mutants were similarly compromised, using the pbe1-2 mutant affecting the CP subunit PBE1 (β₅) as a positive control (Han et al., 2019). Growth of the single and double mutants, and the pbac1-1 pbac2-1 pbac5-3 triple mutant, was strongly dampened in the presence of NaCl, mannitol or the drought-responsive hormone abscisic acid (ABA) when used at concentrations that only mildly suppressed growth of WT Arabidopsis, although the effect was less pronounced than for pbe1-2 seedlings (Fig. S6A–D). For the pbac5-3 line, these sensitivities were rescued by the HA–PBAC5 transgene but not by HA–PBAC5(ΔHbYX), again confirming the importance of the HbYX sequence to PBAC5 activity (Fig. S7A,B).

To definitively assess the importance of the PBAC5–PBAC1–PBAC2 trimer to 26S proteasome assembly, we monitored protease integrity using glycerol gradient fractionation. Here, total seedling extracts from WT, the single pbac1-1, pbac2-1 and pbac5-3 mutants, the triple mutant, and the pbac5-3 mutant complemented with the HA–PBAC5 or HA–PBAC5(ΔHbYX) transgenes were analyzed, with the inclusion of ATP at all steps to stabilize the holo-proteasome. Consistent with previous results (Book et al., 2010; Gemperline et al., 2019; Lee et al., 2011), a single 26S species containing both the CP and RP was detected in WT samples by immunoblotting the fractions with a panel of proteasome subunit antibodies (Fig. 6). By contrast, samples from pbac1-1, pbac2-1 and pbac5-3 single mutant seedlings, and the pbac1-1 pbac2-1 pbac5-3 triple mutant, amassed not only the 26S species but also an RP particle of intermediate sedimentation and substantial amounts of CP α- and β-subunits at the top of the gradient, but little to no free CP (Fig. 6; Fig. S4D), all consistent with CP assembly defects (Gemperline et al., 2019; Hirano et al., 2005; Lee et al., 2011).

The sedimentation positions of the CP subunits likely reflected the accumulation of free proteins or partially assembled CP α-rings with or without β-subunits, as was previously seen for mutants missing UMP1 (Gemperline et al., 2019), whereas the positions of the RP subunits were consistent with fully assembled free RP (Gemperline et al., 2019; Lee et al., 2011). Notably, the assembly defects were no worse in the pbac1-1 pbac2-1 pbac5-3 triple mutant as compared to the individual single mutants, again supporting the notion that the three chaperones work together (Fig. 6). These assembly defects were absent in pbac5-3 plants expressing WT HA–PBAC5, but remained in plants expressing HA–PBAC5(ΔHbYX), again indicating that the HbYX motif in PBAC5 is critical for efficient CP assembly (Fig. 6). Interestingly, the HA–PBAC5 protein was found in the top four fractions of the gradient, consistent with an association with CP assembly intermediates, which we assume are in low abundance relative to the fully assembled CP. In contrast, the HA–PBAC5(ΔHbYX) protein was almost exclusively in the top fraction, suggesting that it was free protein precluded from associating with α-subunits without its HbYX motif (Fig. 6).

We note that RPN5, but not the other Arabidopsis RPN subunits tested, also consistently sedimented with a distinct species, which was slightly smaller than free RP (Fig. 6). The nature of this particle is unknown but could reflect RPN5 also integrating into the structurally related COP9 signalosome complex, as seen in yeast (Yu et al., 2011).

To better understand how PBAC5 interacts with proteasomes, we identified, using MS, the collection of Arabidopsis proteins that bind HA–PBAC5. MS analysis of the anti-HA antibody eluants prepared from HA–PBAC5 pbac5-3 seedlings in the absence of MG132 not only detected PBAC5 but also several CP α-subunits and other CP assembly chaperones that were enriched as compared to their levels in eluants from WT seedlings (Fig. 7). The most significantly enriched were the α-subunits PAD1 (α₄), PAE1 and PAE2 (both α₅), and the chaperones PBAC1 and UMP1b. The enrichment of PBAC1 was consistent with our demonstrations that PBAC1 and PBAC5 interact (Figs 2B,C,E and 3C), whereas those for PAD1, PAE1 and PAE2 were consistent with PBAC5 specifically binding to and aiding in the association of these two α-subunits, which are adjacent in the mature CP.

Also moderately enriched were the α-subunits PAA2 (α₁), PAC1 (α₃) and PAG1 (α₇), plus the PBAC2 and PBAC4 chaperones (Fig. 7), suggesting that these five CP α-subunits, four assembly chaperones and UMP1 represent an early CP α-ring assembly module (Fig. 7). Two isoforms of α₂ (PBA1 and PBA2) and α₆ (PAF1 and PAF2), and the PBAC3 and UMP1a chaperones, were also identified, but only at amounts well below the level of significance, and with little enrichment compared to their levels in the eluants from WT (Fig. 7). We also note that, besides PBB2 (β₂) and PBF1 (β₆), no other β-subunits were detected, nor was the PA200 regulator evident, again consistent with these HA–PBAC5 enriched species reflecting early assembly intermediates consisting of just α-subunits and their assembly chaperones and other factors.

Treating Arabidopsis with MG132 leads to an acceleration of 26S proteasome synthesis, mainly through an activation of the proteasome-stress regulon that controls expression of most, if not all, RP and CP subunits, and associated assembly chaperones and maturation factors (Gemperline et al., 2019; Gladman et al., 2016). MS analysis of MG132-treated WT or HA–PBAC5 pbac5-3 detected a significantly increased number of these factors, now including all five PBAC chaperones, UMP1a, UMP1b, and even PA200 (Fig. 7). Isoforms for all seven α-subunits and six of the seven β-subunits [the exception being PBG1 (β₇)] were also evident, along with a collection of RP subunits with lower significance (Fig. 7). Because β₇ is the last subunit to be integrated into the complex prior to formation of CP half-barrels, its absence is consistent with the immunoprecipitations being enriched in 15S intermediates. Taken together, the MS analysis supported the association of PBAC5 with an α-ring intermediate also containing PBAC1, with the enrichment of PAD1 (α₄), PAE1 and PAE2 (both α₅) confirming direct binding to this α-ring pair.

Although PBAC5 was clearly found to be present throughout the plant kingdom (Fig. 1A), no obvious homologs were detected in humans or yeast, leading to the notion that PBAC5 is a plant-specific chaperone. To help support this possibility, we performed profile HMM searches of the GenBank DNA sequence database (https://www.ncbi.nlm.nih.gov/genbank/), which identified 181 possible PBAC5 orthologs (Table S1). As predicted, a majority were plant sequences, with all taxa appearing to possess a PBAC5 homolog distinct from PBAC1 and PBAC2, including monocots, eudicots, gymnosperms and seedless plants, plus several green and brown algal species (Chlamydomonas spp., Ectocarpus spp., Gonium spp., Micromonas spp., and Volvox spp.; Fig. 8A). Surprisingly, 41 of the sequences were from non-plant genomes. Although many of these matches were of borderline significance initially, they became significant after creating specific HMM searches using the plant family to facilitate HMM-to-HMM comparisons.

Non-plant species with potential PBAC5 orthologs included a number of oomycetes (e.g. Hyaloperonospora spp., Phytophthora spp., Pythium spp., Saprolegnia spp., Thraustotheca spp.), and a few fungi and metazoans, although the distributions were sporadic. For example, all fungal sequences containing PBAC5 were from basal lineages (e.g. Chytridomycota, Mucoromycota, Neocallimastigomycota and Zoopagomycota) but, even within these lineages, only very few species encoded a likely ortholog (Fig. 8A; Table S1). The single celled protist Sphaeroforma arctica possesses a PBAC5-type sequence, but was the only holozoan representative. Remarkably, the few metazoan PBAC5 sequences identified were from disparate organisms that included the fish-like lancelet Branchiostoma floridae, the Japanese sea scallop Mizuhopecten yessoensis, the slime sponge Oscarella carmela, the sea snail Pomacea canaliculata, brachiopods of the Lingula genus and flatworms of the Macrostomum genus (Fig. 8A; Fig. S8).

Subsequent searches using the entire PBAC5 family also revealed significant matches to the families encompassing the archaeal and bacterial Pba chaperones that assemble the homo-heptameric CP α-ring (Kusmierczyk et al., 2011), thus helping establish that PBAC5 is a member of the Pba/PAC/PBAC superfamily. Because all of the PBAC5-containing species also possessed conventional PBAC1 and PBAC2 members, a likely scenario is that PBAC5 split from the Pba1–Pba2/PAC1–PAC2 branches early in evolution, and was later lost in most lineages, possibly accompanied by a gain of function for the Pba1–Pba2/PAC–PAC2 dimer that resulted in PBAC5 becoming dispensable.

26S proteasomes are large proteolytic machines with a sophisticated architecture built from multiple heteromeric rings (Bhattacharyya et al., 2014; Greene et al., 2020; Marshall and Vierstra, 2019), which in eukaryotes must be assembled in correct register from evolutionarily related subunits (Gille et al., 2003). It is therefore unsurprising that proteasome assembly is a complicated, multistep process driven by the sequential action of various extrinsic assembly chaperones and maturation factors (Howell et al., 2017; Marshall and Vierstra, 2019; Rousseau and Bertolotti, 2018). The initial step in CP assembly is particularly challenging, requiring formation of a hetero-heptameric ring of related α-subunits, which is promoted by two hetero-dimeric chaperones (Pba1–Pba2 and Pba3–Pba4 in yeast) that help ensure correct subunit positioning and stoichiometry (Hirano et al., 2005, 2006; Kock et al., 2015; Kusmierczyk et al., 2008, 2011; Le Tallec et al., 2007; Stadtmueller et al., 2012; Wani et al., 2015; Wu et al., 2018; Yashiroda et al., 2008). Additionally, at least in yeast, α₁ subunit integration is rate limiting during assembly (Howell et al., 2019), hence chaperones are required to prevent premature association of incomplete proteasome intermediates. Upon α-ring formation, the Ump1 maturation factor then aids consecutive docking of the β-subunits onto the α-ring template (Burri et al., 2000; Li et al., 2007; Ramos et al., 1998).

Despite extensive knowledge obtained from yeast and mammalian studies, the low sequence identities of assembly chaperones across kingdoms have challenged understanding of proteasome assembly in plants (Gemperline et al., 2019; Wang et al., 2019). Recently, we overcame this hurdle through the MS analysis of Arabidopsis proteasomes, which discovered a suite of CP- and RP-specific interacting proteins, several of which are confirmed homologs of the yeast and mammalian Pba/PAC CP α-ring assembly chaperones and UMP1, despite low sequence homology (Gemperline et al., 2019). Here, we demonstrate that the CP-interacting protein PBAC5 is an additional assembly chaperone in Arabidopsis that, like its relatives PBAC1 and PBAC2, contains the signature C-terminal HbYX motif that promotes CP binding. Through various interaction studies, we showed that PBAC5 assembles into an ordered PBAC5–PBAC1–PBAC2 chaperone hetero-trimer, which then binds to a subcomplex of the CP, comprising almost exclusively the α-ring subunits α₄, α₅, α₆, α₇ and possibly α₃, which might represent the initial step in plant α-ring assembly.

As predicted, Arabidopsis pbac5 mutants displayed numerous phenotypes consistent with CP construction defects. Included were: (1) the accumulation of CP assembly intermediates apparently stalled in α-subunit integration; (2) reduced levels of enzymatically active proteasomes; (3) a hypersensitivity to proteotoxic stress induced by proteasome inhibition or the amino acid analog canavanine; (4) a hyper-accumulation of ubiquitin conjugates, presumably caused by proteasome insufficiency; and (5) an activation of the proteasome-stress regulon that attempts to overcome this insufficiency by promoting new proteasome synthesis. Also observed were hypersensitivities of Arabidopsis seedlings to salt and osmotic stress, as was the case for mutants missing the UMP1a chaperone or carrying the pbe1-2 allele compromising the β₅ subunit PBE1 (Gemperline et al., 2019; Han et al., 2019). In all cases, these aberrant pbac5 phenotypes could be rescued by re-introducing WT HA–PBAC5, but not by a version of PBAC5 lacking the HbYX motif, thus confirming the importance of this sequence to PBAC5 activity. Subsequent genetic analysis of the pbac1, pbac2 and pbac5 mutants, alone or in combination, revealed that all three chaperones collectively promote α-ring assembly, with their absence generating the same phenotypic severities described above.

Notably, although the Arabidopsis PBAC1 and PBAC2 proteins together failed to rescue a yeast mutant lacking the Pba1–Pba2 dimer, the PBAC5–PBAC1–PBAC2 trimer was mostly successful, implying that the contacts provided by Arabidopsis PBAC1 and PBAC2 are alone insufficient for yeast α-ring assembly, and that PBAC5 provides additional contact(s) to overcome this apparent weak affinity. However, this stronger binding does not appear to depend on the HbYX motif in PBAC5, because the PBAC1, PBAC2 and PBAC5(Y240A) combination, in which the central tyrosine in the HbYX motif of PBAC5 was replaced with an alanine, was equally successful as the WT version. Our observations that the HbYX contact in PBAC5 was essential to its activity in Arabidopsis, but not in yeast, argues that additional interactions beyond the HbYX motif participate in α-subunit docking (Stadtmueller et al., 2012).

Our co-immunoprecipitation studies showing that HA–PBAC5 most strongly enriched for the PAD1 (α₄), PAE1 and PAE2 (α₅) subunits, together with the available crystal structure of yeast Pba1–Pba2 associated with the α-ring (Stadtmueller et al., 2012; Yashiroda et al., 2008), allowed us to develop a model for Arabidopsis CP α-ring assembly using the PBAC5–PBAC1–PBAC2 trimer (Fig. 8B). As in yeast and mammalian cells (Kusmierczyk et al., 2011; Stadtmueller et al., 2012), Arabidopsis PBAC1 would contact the shallow pocket at the α₅–α₆ interface through its HbYX motif, while PBAC2 would contact the α₆–α₇ interface through its HbF motif. PBAC5, through its binding to PBAC1, would then be placed directly over a potential pocket at the α₄–α₅ interface, thus helping PBAC5 encourage integration of α₄ alongside α₅ through binding by its HbYX motif and other surfaces. The entire trimer would then correctly register an α-ring tetramer consisting of α₄–α₅–α₆–α₇ (Wu et al., 2018). The dimeric PBAC3–PBAC4 chaperone, either alone or in combination with UMP1, could then help recruit α₂ and α₃, and ultimately α₁ to complete the heptameric α-ring.

Surprisingly, Arabidopsis mutants missing PBAC1, PBAC2 and PBAC5 were viable under standard growth conditions and still generated a pool of assembled 26S complexes, as opposed to the strong ump1 mutations that are gametophytic lethal (Gemperline et al., 2019). One possibility is that the α-rings can self-associate, possibly with the help of only UMP1, PBAC3 and PBAC4. Alternative assembly routes, such as the complex comprising the α₁–α₄ and β₂–β₄ subunits seen in yeast, which can integrate the remaining subunits into half-proteasomes (Hammack et al., 2020; Panfair et al., 2015), might also exist. The diminished levels of the complete 26S particle seen in the chaperone mutants are clearly sufficient to maintain a healthy Arabidopsis proteome under normal growth conditions, but presumably become insufficient under conditions that induce proteotoxic stress, such as proteasome inhibitors, amino acid analogs and osmotic stress. Given that autophagy components are also upregulated upon proteasome stress (Gladman et al., 2016), it is likely that this recycling route is activated to augment proteolytic capacity under these conditions.

Our observations that PBAC5 has been retained in all plant and many oomycete species, but is absent from all but a few fungal and metazoan species, were both remarkable and puzzling. Because all species with PBAC5 orthologs also possess the conventional PBAC1 and PBAC2 chaperones, a possible scenario is that PBAC5 split from the PBAC1/PBAC2 branch early during evolution, and was later lost in most lineages (Albalat and Cañestro, 2016; Guijarro-Clarke et al., 2020). Given that PBAC5 is required for efficient plant proteasome assembly, and that all three subunits of the plant PBAC5–PBAC1–PBAC2 hetero-trimeric complex are needed to replace the Pba1–Pba2 chaperone in yeast, it might be the case that loss of PBAC5 conferred a selective advantage in most species, or that its loss was preceded by a gain of function for the Pba1–Pba2/PAC1–PAC2 dimer that rendered PBAC5 dispensable. Alternatively, we speculate that PBAC5 might be necessary to promote CP assembly in organisms more exposed to challenging environments. Because plants can tolerate remarkable extremes in temperature and other stresses (Dudler, 2013; Üstün et al., 2018), retaining PBAC5 along with PBAC1 and PBAC2 might aid CP assembly during adversity.

In summary, we describe here a novel proteasome CP α-ring assembly chaperone in Arabidopsis that is highly conserved in all plant and many oomycete species, but is absent in most fungi and metazoans. This discovery will aid future biochemical and genetic analyses of proteasome assembly in plants, which given its critical role in stress protection (Xu and Xue, 2019), might provide strategies to mitigate proteotoxic stress in agronomically important species. Given the weak homology seen between the PBAC5 family and those of PBAC1 and PBAC2, it is also intriguing to consider that other proteasome chaperones exist that are sufficiently divergent in sequence, or present at sufficiently low levels, to have thus far evaded detection.

The predicted full-length nucleotide and protein sequences of Arabidopsis thaliana PBAC1, PBAC2 and PBAC5 [obtained from the Arabidopsis Information Resource database version 10.1 (www.arabidopsis.org/)] were used as queries in PSI-BLAST searches for orthologous loci in other plant genomes available in the Joint Genome Initiative's Phytozome database (www.phytozome.net). Progressive alignments of the predicted full-length amino acid sequences were performed using Clustal Omega (www.clustal.org/omega/) with the default settings. Following minor manual editing, the final alignments were displayed using BoxShade version 3.2.3 (https://embnet.vital-it.ch/software/BOX_form.html). Bayesian phylogenetic analyses were then performed with MrBayes version 3.2.2 (Ronquist and Huelsenbeck, 2003) using the General Time Reversible evolutionary model with the mixed amino acid model and γ-distributed rate variation with a proportion of invariable sites, as previously described (Gemperline et al., 2019). The resulting consensus trees were displayed using FigTree version 1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/).

Alternatively, to search for PBAC5 sequences in non-plant species, profile-HMM searches in sequence databases were performed using the generalized profiles method (Bucher et al., 1996) by iterative refinement, using a P-value <0.001 as an inclusion criterion for the next iteration cycle. Comparisons between different sequence families were performed using HHSEARCH (Söding, 2005) to generate HMM-to-HMM comparisons. Multiple sequence alignments were then created using the L-INS-I algorithm in the MAFFT package (Katoh et al., 2002), and phylogenetic trees were generated by the neighbor-joining method (Saitou and Nei, 1987) and displayed as above. Species abbreviations used are: Al, Arabidopsis lyrata; Ath, Arabidopsis thaliana; Atr, Amborella trichopoda; Ar, Anaeromyces robustus; Bd, Brachypodium distachyon; Bf, Branchiostoma floridae; Bl, Bremia lactucae; Bm, Basidiobolus meristosporus; Br, Brassica rapa; Ca, Capsicum annuum; Ce, Chlamydomonas eustigma; Cg, Corynebacterium glutamicum; Cp, Carica papaya; Cr, Coemansia reversa; Csa, Cucumis sativa; Csi, Citrus sinensis; De, Diversispora epigaea; Eg, Eucalyptus grandis; Es, Ectocarpus siliculosus; Gm, Glycine max; Gp, Gonium pectorale; Gt, Guillardia theta; Gu, Globisporangium ultimum; Han, Helianthus annuus; Har, Hyaloperonospora arabidopsidis; Hs, Homo sapiens; Lu, Lingula unguis; Mc, Micromonas commoda; Me, Mortierella elongata; Ml, Macrostomum lignano; Mt, Mycobacterium tuberculosis; My, Mizuhopecten yessoensis; Nc, Neocallimastix californiae; Os, Oryza sativa; Pc, Pomacea canaliculata; Pf, Pyrococcus furiosus; Pi, Phytophthora infestans; Ps, Picia sitchensis; Pp, Physcomitrella patens; Rc, Ricinus communis; Sa, Sphaeroforma arctica; Sb, Sorghum bicolor; Sce, Saccharomyces cerevisiae; Sco, Streptomyces coelicolori; Sd, Saprolegnia diclina; Si, Setaria italica; Sl, Solanum lycopersicum; Tc, Thraustotheca clavata; Vc, Volvox carteri; Zm, Zea mays.

Raw transcript abundance data for the Arabidopsis PBAC family were downloaded from the Transcriptome Variation Analysis database (www.travadb.org/; Klepikova et al., 2016) and plotted in R (www.r-project.org/) using the heatmap.2 function, with centroided hierarchical clustering based on Pearson's correlation used as the distance function. All read counts were normalized to values ranging from 0 to 1 by the ‘median-of-ratios’ method, as in DESeq2 (Love et al., 2014), and then divided by the maximum expression for each gene.

All A. thaliana lines were derived from the Columbia-0 (Col-0) ecotype, except pbac2-1 and pbac2-2, which were derived from the Col-3 ecotype. Details of all T-DNA insertion mutants and stable transgenic lines are provided in Tables S2 and S3. All insertion mutants were confirmed by genomic PCR using 5′ and 3′ gene-specific primers (LP and RP, respectively) in conjunction with an appropriate T-DNA left border-specific primer. Details of all oligonucleotide primers are provided in Table S4. The exact positions of the T-DNA insertions were determined by direct DNA sequencing of appropriate PCR products. Before analysis, all mutants were backcrossed at least three times to the WT Col-0 parent and then selfed to obtain homozygous progeny, using growth on medium containing 50 μg/ml kanamycin, 10 μg/ml BASTA and/or 6 μg/ml sulfadiazine for selection, depending on the source of the mutant.

Sterilized seeds (obtained via vapor-phase or liquid-phase sterilization) were vernalized at 4°C for 3–4 days and typically germinated on solid GM medium [3.2 g/l Gamborg's B5 basal medium with minimal organics, 1% (w/v) sucrose, 0.05% (w/v) MES, pH 5.7, and 0.7% (w/v) agar] at 21–23°C under a long-day (LD) photoperiod (16 h light/8 h darkness) with a light intensity of 75–100 µmol/m²/sec and a relative humidity of 40–50%. Where indicated, agar plates were supplemented with the indicated concentrations of MG132 [(N-benzyloxycarbonyl)-leucinyl-leucinyl-leucinal], canavanine, NaCl, mannitol or ABA. The nac53-1 nac78-1 and pbe1-2 mutants were used as positive controls for hypersensitivity to proteotoxic and salt or osmotic stress, respectively (Gladman et al., 2016; Han et al., 2019). When required, seedlings were transferred to soil [mixed in a 1:1 ratio with organic Coco Coir planting mixture, supplemented before use with 2 g/l Peters 20-20-20 fertilizer, 80 mg/l Ca(NO₃)₂ and 80 mg/l MgSO₄] after 2–3 weeks, and grown at 21–23°C under an LD photoperiod.

To generate transgenic plants expressing HA-tagged PBAC5, the genomic region of PBAC5 encompassing the full coding sequence plus 39-bp upstream of the ATG start codon and 222-bp downstream of the TAA stop codon, which constitute the 5′- and 3′-UTRs, was PCR-amplified from WT Col-0 genomic DNA and recombined into pDONR221 via the Gateway BP clonase II reaction (Thermo Fisher Scientific). Codons for the HA tag (YPYDVPDYA) were then inserted, or the HbYX motif (LYG) deleted, by site-directed mutagenesis using the QuikChange II site-directed mutagenesis kit (Agilent Technologies). The sequence-confirmed HA–PBAC5 and HA–PBAC5(ΔHbYX) clones were recombined into the pMDC99-UBQ10 vector (Curtis and Grossniklaus, 2003; Suttangkakul et al., 2011) via the Gateway LR clonase II reaction (Thermo Fisher Scientific). The plasmids were introduced into Agrobacterium tumefaciens strain GV3101 and transformed into homozygous pbac5-3 plants by the Agrobacterium-mediated floral-dip method (Clough and Bent, 1998). F₁ plants resistant to 25 μg/ml hygromycin B were selected and, after a self-cross, double homozygous plants were identified in the F₂ generation by PCR genotyping, immunoblot analysis (see below) and segregation of the F₃ generation on hygromycin-containing medium.

WT and UBQ10::HA–PBAC5 seedlings were grown for 6 days in 50 ml liquid GM medium (as above, but without agar) and then transferred to fresh medium containing 50 μM MG132 or an equivalent volume of DMSO and incubated for an additional 16 h, with ∼100 mg of dry seeds used per culture, resulting in ∼5 g of fresh weight tissue. Frozen seedlings were ground to a fine powder at liquid N₂ temperatures, and proteins were extracted on ice for 20 min with 1.5 volumes of extraction buffer [50 mM HEPES, pH 7.5, 150 mM NaCl, 10 mM MgCl₂, 10% (v/v) glycerol, 5 mM dithiothreitol (DTT), 2 mM phenylmethylsulfonyl fluoride (PMSF), 0.1% (v/v) Triton X-100 and 1× plant protease inhibitor cocktail (Sigma Aldrich, P9599)]. Extracts were filtered through two layers of Miracloth (Calbiochem), clarified at 30,000 g for 20 min at 4°C, and the resulting supernatants were immediately applied three times at 4°C over a 12 ml PolyPrep chromatography column containing 100 µl (equal to a 50 µl bead volume) of monoclonal anti-HA agarose antibody beads (Sigma-Aldrich, A2095) pre-equilibrated in extraction buffer. The column was washed three times with wash buffer [50 mM HEPES, pH 7.5, 150 mM NaCl, 10 mM MgCl₂, 10% (v/v) glycerol and 2 mM DTT], and remaining bound proteins were eluted with 300 μl of 200 mM glycine-HCl (pH 2.5) and immediately neutralized with 60 μl of 1 M Tris-HCl (pH 8.0). Samples of the crude extract, flow through (both diluted 1 in 10), third wash step and elution were analyzed by SDS–PAGE followed by staining for total protein with silver or immunoblotting with appropriate antibodies, as previously described (Gemperline et al., 2019; Marshall et al., 2017). Alternatively, elution fractions were subjected to tandem MS (see below).

Frozen Arabidopsis seedlings were homogenized in three volumes of protein extraction buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM DTT, 1 mM PMSF, 50 mM MG132 and 1× plant protease inhibitor cocktail) and clarified by centrifugation at 16,000 g for 5 min at 4°C. The supernatant was then mixed with 0.25 volumes of hot 5× SDS–PAGE sample buffer [200 mM Tris-HCl, pH 6.8, 25% (v/v) glycerol, 2% (w/v) SDS, 5% (v/v) 2-mercaptoethanol and 0.1% (w/v) Bromophenol Blue]. The total protein extracts (or elutions from co-immunoprecipitations) were subjected to SDS–PAGE followed by electrophoretic transfer onto Immobilon-P polyvinylidene difluoride (PVDF) membranes (Millipore) for 16 h at 80 mA. Immunoblots were then performed as previously described (Gemperline et al., 2019). Primary antibodies against RPN1 (Yang et al., 2004), RPN3 (Marshall et al., 2015), RPN5, RPN12a, RPT2, PAC1, PBA1, PBF1 (Smalle et al., 2002), RPN10, ubiquitin (van Nocker et al., 1996), RPT4 (Gemperline et al., 2019) and PAG1 (Book et al., 2010) were as previously described (all used at a 1:3000 dilution). Antibodies against 6His (1:1000), FLAG (1:5000), GST (1:5000), HA (1:1000), histone H3 (1:3000) and myc (1:1000) were purchased from BioLegend (906101), Sigma-Aldrich (F1804), Santa Cruz Biotechnology (SC-138), Sigma-Aldrich (H6908), AbCam (AB1791) and BioLegend (908805), respectively.

Assays for direct protein–protein interactions by Y2H were performed using the ProQuest two-hybrid system (Thermo Fisher Scientific). To clone the PBAC1–PBAC5 coding sequences, total RNA was first extracted from WT Arabidopsis seedlings and converted into cDNA, as described below. Coding sequences amplified by PCR were recombined into pDONR221 via the Gateway BP clonase II reaction. Sequence-confirmed clones were then recombined in-frame with either the Gal4 activation domain or Gal4 DNA-binding domain in the pDEST22 or pDEST32 vectors (Thermo Fisher Scientific), respectively, via the Gateway LR clonase II reaction. Pairwise combinations of coding sequences in pDEST22 and pDEST32 (or the empty vectors as controls) were then co-transformed into yeast strain MaV203 (Table S5). Cells transformed with both plasmids were selected by growth for 2 days at 30°C on synthetic dropout medium lacking leucine and tryptophan. Protein–protein interactions were then identified by growing for 2 days at 30°C on synthetic dropout medium lacking leucine, tryptophan and histidine, and containing 25 mM 3-amino-1,2,4-triazole (3-AT), with at least four individual colonies tested for each interaction pair. To confirm interactions, single colonies were diluted in sterile H₂O to an OD₆₀₀ of 0.1, and 5 μl were spotted onto both types of selective medium and again grown for 2 days at 30°C.

Sequence-confirmed coding sequences cloned into pDONR221, as above, were recombined in-frame with the N- or C-terminal halves of EYFP in the pSITE-N-EYFP-C1 or pSITE-C-EYFP-C1 vectors (ABRC stock numbers CD3-1648 and CD3-1649, respectively) via Gateway LR clonase II reactions, with expression driven by the cauliflower mosaic virus (CaMV) 35S promoter. The resulting plasmids were introduced into A. tumefaciens strain GV3101, overnight cultures of which were resuspended in 5 ml infiltration buffer [10 mM MgCl₂, 10 mM MES, pH 5.7, 100 μM 3′,5′-dimethoxy-4′-hydroxyacetophenone (acetosyringone)], incubated at room temperature for 8–12 h and then used for direct infiltration of 4 to 6-week-old N. benthamiana leaves.

Leaf sections ∼2 mm×2 mm were excised 36–48 h after infiltration and visualized by confocal fluorescence microscopy using a Zeiss 510 Meta confocal laser scanning microscope equipped with 20× or 40× oil objectives (numerical apertures 0.75 and 1.30, respectively). Excitation was performed at 488 nm, and emission was collected between 500 and 550 nm. Images were processed using Elements Viewer (Nikon Imaging Software) and/or Adobe Photoshop CC, before conversion to TIFF files for use in the figures. Within each figure, all images (including negative controls) were captured using identical microscope settings, with the exception of bright-field images, where the channel gain was adjusted to provide uniform exposures between images.

For recombinant protein expression, the PBAC1, PBAC2 and HA–PBAC5 coding sequences in pDONR221 were recombined into pDEST15, pDEST17 or pDEST14 (Thermo Fisher Scientific), respectively, via the Gateway LR clonase II reaction, resulting in the addition of N-terminal GST, 6His or HA tags, respectively. All proteins were expressed in E. coli strain BL21(DE3) pLysS. The cells were cultured at 37°C in 800 ml LB medium to an OD₆₀₀ of 0.6–0.8, followed by a 4-h induction at 30°C with 1 mM isopropyl-β-D-thiogalactopyranoside. Cells were harvested by centrifugation at 4000 g for 20 mins at 4°C, frozen in liquid nitrogen, and lysed in two rounds of 10 ml BugBuster Master Mix (EMD Millipore), as according to the manufacturer's instructions.

GST-PBAC1 was affinity purified using GST-Bind Resin (EMD Millipore), as according to the manufacturer's instructions with some minor modifications (Marshall et al., 2015). Briefly, E. coli cell lysates were incubated with 2 ml phosphate-buffered saline (PBS)-washed GST-Bind Resin for 1 h at 4°C with continual rotation, then applied to a 30 ml EconoPac chromatography column (Bio-Rad). The beads were washed twice with ice-cold PBS containing 1 M NaCl and once with ice-cold PBS containing 2 M NaCl. Bound proteins were eluted with 10 ml of ice-cold 25 mM MOPS-KOH (pH 7.5) containing 10 mM reduced glutathione (GSH), and the purified proteins were dialyzed overnight at 4°C against 25 mM MOPS-KOH (pH 7.5), before being concentrated 10-fold using an Amicon Ultra-15 centrifugal filter unit with a 10 kDa cut-off limit (Millipore).

6His–PBAC2 was affinity purified using nickel-nitrilotriacetic acid (Ni-NTA) agarose beads (QIAGEN). E. coli cell lysates in BugBuster Master Mix containing 10 mM imidazole were applied three times to 1 ml PBS-washed Ni-NTA beads in a 30 ml EconoPac chromatography column at 4°C. After flow-through, the beads were washed once with NaH₂PO₄ wash buffer (50 mM NaH₂PO₄, pH 7.8, 300 mM NaCl and 20 mM imidazole), once with Tris-40 buffer (50 mM Tris-HCl, pH 7.8, 300 mM NaCl and 40 mM imidazole), and once with Tris-60 buffer (50 mM Tris-HCl, pH 7.8, 300 mM NaCl and 60 mM imidazole). Bound proteins were eluted with 5 ml ice-cold elution buffer (50 mM Tris-HCl, pH 7.8, 100 mM NaCl and 300 mM imidazole), and the purified proteins were dialyzed overnight at 4°C to remove the imidazole.

HA–PBAC5 was affinity purified using EZview red anti-HA affinity gel (Sigma-Aldrich). The E. coli lysates were incubated with 250 μl PBS-washed anti-HA antibody beads (500 μl of a 50% slurry) for 1 h at 4°C with continual rotation, then applied to a 12 ml PolyPrep chromatography column (Bio-Rad). The beads were then washed three times with PBS containing 1 M NaCl. Bound protein was eluted by incubating the beads for 30 min at 4°C with 300 μl of PBS containing 500 ng/μl of the HA peptide, and the eluant was dialyzed overnight at 4°C against PBS.

To assay direct and indirect interactions between PBAC1, PBAC2 and PBAC5, equal amounts (2 μg) of all three proteins (as determined using a Pierce BCA protein assay kit; Thermo Fisher Scientific) were combined in 500 μl of binding buffer [150 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM MgCl₂, 1 mM DTT, 5% (v/v) glycerol and 0.01% (v/v) Triton X-100] and incubated for 2 h at 4°C with continual rotation, before the addition of 100 μl GST-Bind Resin or Ni-NTA agarose beads pre-equilibrated with binding buffer. Samples were incubated for a further 2 h at 4°C with continual rotation, and the beads were then pelleted by centrifugation at 3000 g for 1 min at 4°C. The beads were washed five times with binding buffer, and bound proteins were eluted into 100 μl of 2× SDS–PAGE sample buffer by heating at 95°C for 5 min. Samples were analyzed by immunoblotting with appropriate antibodies (see above).

Unless otherwise stated, all yeast (S. cerevisiae) manipulations were performed according to standard protocols, as described by Marshall et al. (2016). Details of all strains used in this study are given in Table S5. Strain MHY500 (a gift of Mark Hochstrasser, Yale University, New Haven, CT) was used as the WT control. The Δpba1 and Δpba2 strains were obtained from the yeast knockout collection (Dharmacon) and cultured on YPDA medium containing 200 µg/ml Geneticin. All genomic deletions were confirmed by PCR genotyping, using the oligonucleotide primer pairs A+B, A+KanB, C+D, KanC+D, and A+D for each deletion strain (Table S4). The doa5-1 mutation impacting the Doa5/Pup2/α₅ subunit of the CP was described previously (Chen and Hochstrasser, 1996).

For complementation studies, the full-length coding sequences of yeast PBA1 and PBA2 were obtained from BY4741 cells by converting RNA isolated by the RNeasy mini kit (QIAGEN) into cDNA, as described below. Arabidopsis PBAC1, PBAC2 and PBAC5 were amplified from total WT Col-0 cDNA, also generated as described below. The sequences for FLAG, myc or HA tags were incorporated into the appropriate PCR amplification primers. The resulting PCR products were recombined into pDONR221 via the Gateway BP clonase II reaction, and mutations altering the various HbYX and HbF motifs were then introduced by QuikChange II site-directed mutagenesis. The sequence-confirmed clones were recombined into the pAG423-GPD1-ccdB, pAG425-GPD1-ccdB or pAG426-GPD1-ccdB vectors (Addgene, product numbers 14150, 14154, and 14156, respectively) via Gateway LR clonase II reactions. Different plasmid combinations (or the empty vectors as a control) were transformed into the indicated yeast strains using the standard lithium acetate procedure, and transformed cells were cultured on synthetic dropout medium lacking histidine, leucine and/or uracil as required. Cells were then cultured in 15 ml of synthetic dropout medium lacking the auxotrophic markers and, following 16 h overnight growth, diluted to an OD₆₀₀ of 1.0, subjected to a series of 5-fold dilutions, and 5 μl of each dilution was spotted onto solid synthetic dropout medium lacking the auxotrophic markers and containing or lacking 5 μM canavanine. Cells were then grown for 48 h at 30°C prior to imaging.

To assess the in vivo interactions between PBAC1, PBAC2 and PBAC5 in yeast, doa5-1 Δpba1 Δpba2 cells expressing the indicated tagged PBAC protein combinations were grown overnight at 30°C, diluted to an OD₆₀₀ of 0.1 in 15 ml, then grown for an additional 2–3 h until reaching OD₆₀₀ ∼0.5. Cells were harvested by centrifugation at 5000 g for 1 min, resuspended in 1 ml of IPL buffer [50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM MgCl₂ and 10% (v/v) glycerol, with 20 mM N-ethylmaleimide, 2 mM ATP, 2 mM PMSF, 1 mM benzamidine, 10 µg/ml pepstatin A, 1 µg/ml antipain and 1× protease inhibitor cocktail (Sigma-Aldrich, P8215) added just before use], lysed by vigorous vortexing at 4°C in the presence of acid-washed glass beads at a volume of ∼50 µl (with five rounds of vortexing for 30 s, then resting on ice for 30 s), and clarified by centrifugation at 16,000 g for 5 min at 4°C. The supernatant was incubated for 2 h at 4°C with 50 µl of either EZview red anti-HA affinity gel, anti-FLAG M2 affinity gel or EZview red anti-myc affinity gel (Sigma-Aldrich; E6779, A2220 and E6654, respectively) pre-equilibrated in lysis buffer, with continual rotation. The beads were collected by centrifugation at 6000 g for 5 min at 4°C, washed five times with ice-cold IPL buffer, and bound proteins were eluted in 100 µl SDS–PAGE sample buffer by heating at 95°C for 5 min. Samples were analyzed by immunoblotting with antibodies against FLAG, HA and myc, using immunoblotting with anti-histone H3 antibodies as the negative control.

Total RNA was extracted from 50–100 mg of 7-day-old liquid-grown seedlings using the RNeasy plant mini kit (QIAGEN), as according to the manufacturer's instructions. Following quantification with a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific), 1 μg of RNA was treated with DNase I (Thermo Fisher Scientific) and converted into cDNA using the SuperScript III first-strand synthesis system (Thermo Fisher Scientific) and oligo(dT)₂₀ primers. For standard reverse transcriptase PCR, the cDNA was diluted 1 in 30 following first-strand synthesis, and 5 μl was then amplified in a 20 μl reaction volume also containing 10 μl EconoTaq Plus Green master mix (Lucigen), 3 μl sterile H₂O and 1 μl each of 10 mM forward and reverse primers (Table S4).

Quantitative real-time PCR was performed on three independent biological replicates using a CFX Connect Real-Time System (Bio-Rad) and the LightCycler 480 SYBR Green I master mix (Roche Diagnostics), with three technical replicates for each reaction. In all cases, the amplification factor of the primer pair was experimentally determined to be between 1.90 and 2.10, with the gradient of the standard curve being between −3.59 and −3.10, and the R² value being greater than 0.975. For each individual PCR reaction, 5 µl of cDNA (diluted 1 in 30 following first-strand synthesis) was amplified in a 20 µl reaction volume also containing 10 µl SYBR Green I master mix, 3 µl sterile H₂O and 1 µl each of 10 µM forward and reverse primers. Reaction parameters were 95°C for 5 min; 45 cycles of 95°C for 10 s, 55°C for 10 s and 72°C for 30 s; followed by a melting curve program and cooling at 40°C for 2 min. Fluorescence data were collected at the end of each 72°C extension step and continuously during the melting curve program. The relative transcript abundance of target genes was determined using the comparative threshold cycle method (Pfaffl, 2001) using the ACT2 and PP2A reference genes as internal controls (Czechowski et al., 2005). All data was normalized to the WT Col-0 samples.

Total proteasome activity was assayed as previously described (Marshall et al., 2015), with minor modifications. Seedlings were grown in 5 ml of liquid GM medium at 21–23°C under continuous light for 10 days with gentle shaking (90 rpm). Typically, ∼30 mg of dry seeds was used per culture, resulting in ∼2 g of fresh weight tissue. Frozen tissue was ground to a powder in a mortar and pestle at liquid nitrogen temperatures, and proteins were extracted on ice for 20 min in one volume of extraction buffer [20 mM HEPES-KOH, pH 7.5, 5 mM MgCl₂, 10% (v/v) glycerol, 2 mM ATP, 1 mM DTT, 2 mM PMSF, 6 μM chymostatin and 1× plant protease inhibitor cocktail (all from Sigma-Aldrich)]. Extracts were filtered through two layers of Miracloth, clarified at 30,000 g for 20 min at 4°C, and the supernatant was made 10% (w/v) in PEG 8000 [from an initial stock of 40% (w/v)] and incubated for 30 min at 4°C. The resulting precipitate was collected by centrifugation at 12,000 g for 15 min at 4°C and resuspended in 500 µl of extraction buffer, a step that has been shown to remove the majority of lower molecular weight proteases that might interfere with the assay (Yang et al., 2004). The total protein concentration of each sample was determined using a Pierce BCA protein assay kit, and an equal mass of protein (10 µg) from each sample was assayed for proteasome activity in the presence or absence of 80 µM MG132 (Kisselev and Goldberg, 2005). Protein samples in a volume of 20 µl were incubated for 20 min at 37°C in 1 ml of assay buffer (50 mM Tris-HCl, pH 7.0, 2 mM MgCl₂, 1 mM ATP and 2 mM 2-mercaptoethanol) containing 100 µM of the fluorogenic substrate N-succinyl-leucyl-leucyl-valyl-tyrosyl-7-amino-4-methylcoumarin (succinyl-LLVY-AMC; Sigma-Aldrich). Reactions were quenched by the addition of 1 ml of 80 mM sodium acetate (pH 4.3), and the fluorescence of released AMC was monitored using a TKO 100 fluorometer (Hoefer Scientific Instruments), with an excitation wavelength of 365 nm and an emission wavelength of 460 nm. Three technical replicates were assayed for each sample; the data from three biological replicates were averaged and normalized to the activity observed in WT Col-0 tissue in the absence of MG132.

Glycerol gradient fractionation of 26S proteasome subcomplexes was performed essentially as previously described (Gemperline et al., 2019). Seedlings were grown as above and proteins were extracted in one volume of extraction buffer [20 mM HEPES-KOH, pH 7.5, 5 mM MgCl₂, 10% (v/v) glycerol, 2 mM ATP, 1 mM DTT, 2 mM PMSF, 6 μM chymostatin, 1× plant protease inhibitor cocktail, 10 mM phosphocreatine and 1 mg/ml creatine phosphokinase (all from Sigma-Aldrich)]. Extracts were filtered through two layers of Miracloth, clarified at 30,000 g for 20 min at 4°C, and 1 ml of supernatant was loaded onto an 11 ml, 10–40% (v/v) glycerol density gradient in extraction buffer, made using an Auto-DensiFlow density gradient fractionator (Labconco). Following centrifugation at 100,000 g for 18 h at 4°C, 500 μl fractions were manually collected with a Gilson-type P1000 pipette, and 10 μl of each fraction was subjected to immunoblot analysis as described above.

Samples from three independent anti-HA immunoprecipitation experiments were subjected to tandem MS as previously described (Marshall et al., 2019), with minor modifications. Briefly, eluants were vacuum dried and denatured in 300 μl of 8 M urea, 25 mM (NH₄)HCO₃. Proteins were then reduced with 10 mM DTT at room temperature for 1 h and alkylated in the dark in the presence of 50 mM 2-chloroacetamide at room temperature for a further 1 h. Excess alkylating agent was quenched with 50 mM DTT for 5 min at room temperature, and samples were diluted with 1.2 ml of 25 mM (NH₄)HCO₃ to reduce the urea concentration to below 1.5 M. Proteolytic digestion was initiated by incubating the samples with 1 μg of sequencing-grade modified porcine trypsin (Promega) for 18 h at 37°C. Resulting peptides were vacuum dried to a final volume of ∼300 μl, acidified with 10% (v/v) trifluoroacetic acid (TFA) until the pH was less than 3.0, and desalted and concentrated on a 100 μl Bond Elut OMIX C18 pipette tip (Agilent Technologies), as according to the manufacturer's instructions. Peptides were eluted in 50 μl of 75% (v/v) acetonitrile, 0.1% (v/v) acetic acid, then lyophilized and resuspended in 17 μl 5% (v/v) acetonitrile, 0.1% (v/v) formic acid.

Nano-scale ultra-high-performance liquid chromatographic (UHPLC) separation of tryptic peptides was performed on a Dionex Ultimate 3000 Rapid Separation LC system (Thermo Fisher Scientific) with a C18 analytical column [Acclaim PepMap RSLC C18 column, 2 μm particle size, 100 Å pore size, 75 μm×25 cm (Thermo Fisher Scientific)] by the application of a linear 2 h gradient from 4.0% to 36.0% (v/v) acetonitrile in 0.1% (v/v) formic acid, with the column flow rate set to 250 nl/min. MS analysis of the eluted tryptic peptides was performed online using a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific) possessing a Nanospray Flex ion source (Thermo Fisher Scientific) fitted with a stainless steel nano-bore emitter operated in positive electrospray ionization (ESI) mode at a capillary voltage of 1.9 kV. Data-dependent acquisition of full MS scans within a mass range of 380–1500 m/z was performed at a resolution of 70,000, with the automatic gain control (AGC) target set to 3.0×10⁶ ion intensity and the maximum fill time set to 200 ms. High energy collision-induced dissociation (HCD) fragmentation of the top 15 most intense peaks was performed with a normalized collision energy of 28, an intensity threshold of 4×10⁴ counts and an isolation window of 3.0 m/z, excluding precursors that had unassigned, +1, +7 or +8 charge states. MS/MS scans were acquired at a resolution of 17,500, with an AGC target of 2×10⁵ and a maximum fill time of 100 ms. All peaks were recorded in profile mode. Dynamic exclusion was performed with a repeat count of two and an exclusion duration of 30 s, while the minimum MS ion count for triggering MS/MS was set to 4.0×10³ counts.

The resulting MS/MS spectral files were processed using Proteome Discoverer (version 2.0.0.802; Thermo Fisher Scientific), set up to interrogate the A. thaliana Col-0 proteome file TAIR10_pep_20101214_updated.fasta (available at the Arabidopsis Information Resource; www.arabidopsis.org/). Peptides were assigned using SEQUEST HT (Eng et al., 1994), with search parameters set to assume the digestion enzyme trypsin with a maximum of one missed cleavage, a minimum peptide length of six, precursor mass tolerances of 10 ppm, and fragment mass tolerances of 0.02 Da. Carbamidomethylation of cysteine was specified as a static modification, whereas oxidation of methionine and N-terminal acetylation were specified as dynamic modifications. A target false discovery rate (FDR) was determined using the Protein FDR Validator node, where all proteins with a q-value higher than 0.01 (strict, high confidence) and 0.05 (relaxed, medium confidence) were included in the analysis. Proteins that contained similar peptides, and which could not be differentiated based on the MS/MS analysis alone, were grouped to satisfy the principles of parsimony. Peptide abundances were quantified from the precursor ion intensities available in the MS1 scans, with the relative abundances for the full-length proteins being generated from the averages of three biological replicates, each analysed in duplicate.

Label-free MS1 quantification was performed as previously described (Silva et al., 2006) using Proteome Discoverer with a minimum Quan value threshold set to 0.0001 using unique peptides, and ‘3 Top N’ peptides used for area calculation. The resulting values were log₂-transformed and missing values were imputed assuming a normal distribution, width distribution shrinkage of 0.3 and a downshift of 1.8 standard deviations using the Perseus computational platform (Tyanova et al., 2016). Significant changes in protein abundance were calculated in Perseus by analysis of variance (ANOVA) contrasts (P-value <0.05; FDR <1%; S0=2), allowing 250 permutations. Details of all quantification data are provided in Table S6.

Details of MS-based statistical analyses are provided above. All other datasets were statistically analyzed using one-way ANOVA followed by Tukey's post hoc tests to identify significantly different data points. At least three biological replicates were performed in all cases, unless otherwise indicated in the figure legend.

All accession numbers for genes and proteins used in this study are given in Table S1.

We thank Mark Hochstrasser for sharing yeast strains, and Erin Gemperline, Nicholas P. Gladman, Weiming Hu, and Joseph M. Walker for advice and technical support.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.249862.reviewer-comments.pdf