Genes encoding the β subunits of tomato were identified by BLASTp searches of the predicted proteome (ITAG release 2.40) for homologs of the seven Arabidopsis β subunits at the SolGenomics website¹. The β2a protein sequence was modeled onto polypeptide H of the structure of the yeast proteasome (2zcy, Groll et al., 2008) using Swiss Model² (Biasini et al., 2014). This β2a model was used in PyMol to replace the β2 in the structure of the yeast proteasome. Only the surface of one ring of β subunits was visualized and the various parts and residues were colored using PyMol. Further annotations were added using CorelDRAW. Transcript levels were extracted from published RNA sequencing experiments on different organs (The Tomato Genome Consortium, 2012).

Tomato (Solanum lycopersicum L. cv “Rio Fuego”) plants were germinated at 26°C for 3 days in the dark, and the seedlings were subsequently transferred to perlite for 2 weeks. Plants were grown hydroponically in a controlled environment in a greenhouse (300 μmol m^-2 s^-1 photon flux density with 12/12 light/dark photoperiod, 25°C, and 55–60% relative humidity) for 3 weeks (Poór et al., 2011). Tomato plants were treated with 0-, 100-, and 250 mM NaCl in the nutrient solution [2 mM Ca(NO₃)₂, 1 mM MgSO₄, 0.5 mM KCl, 0.5 mM, KH₂PO₄, 0.5 mM Na₂HPO₄, 0.001 mM MnSO₄, 0.005 mM ZnSO₄, 0.0001 mM (NH₄)₆Mo₇O₂₄, 0.01 mM H₃BO₄, 0.02 mM Fe(III)-EDTA]. Samples were made in at 9 a.m. and samples were taken in triplicate at 1, 6, and 24 h after salt exposure.

Fluorescein diacetate (FDA; Sigma–Aldrich, St. Louis, MO, USA) was used to determine cell viability according to Gémes et al. (2011). Root tip segments were stained for 10 min at room temperature in the dark with 10 mM FDA dissolved in 3 ml 10 mM 2-(N-morpholino)ethanesulfonic acid (MES) potassium chloride (KCl) buffer (pH 6.15). After staining, the samples were washed two times in 10 min with MES/KCl buffer (pH 6.15). Fluorescence intensity was detected with Zeiss Axiovert 200 M type fluorescent microscope (Carl Zeiss Inc., Jena, Germany) equipped with an 5X objective. Digital photographs were taken from the samples with a high-resolution digital camera (Axiocam HR, HQ CCD camera; Carl Zeiss Inc., Jena, Germany) using a filter set 10 (excitation 450–495 nm, emission 515–565 nm) or filter set 20HE (excitation: 535–585 nm, emission: 600–655 nm). The fluorescence emission (pixel intensity) was measured on digital images with AXIOVISION REL. 4.8 software (Carl Zeiss Inc., Munich, Germany).

Labeled and precipitated proteins were resuspended in UTC buffer (8 M urea, 2 M thiourea, 4% (w/v) CHAPS, 1 g AG 501-X8 Resin) containing 1% (v/v) ampholyte and 65 mM DTT. Samples were isoelectrically focused on 7 cm immobilized pH gradient (IPG) 3–10 pH strips (BioRad–ReadyStrip^TM IPG Strips) using BioRad PROTEAN i12 IEF system with the following focusing conditions: 12 h passive rehydration; 250 V, 15 min, rapid ramp; 4000 V, 1 h, slow ramp; 4000 V, 30000 Vhr, rapid ramp; 500 V hold. After focusing, IPG strips were equilibrated in IEF Equilibration buffer [6 M urea, 5% SDS (w/v), 30% glycerol (v/v)] containing 1 % (w/v) DTT, then in IEF Equilibration buffer containing 2.5% iodoacetamide (w/v). The second dimension electrophoresis was run on a 15% SDS gel. Gels were imaged using a Typhoon 9400 Imager (GE Healthcare) using excitation and emission wavelengths of 532/580 nm. Images were quantified using ImageJ 1.48V by multiplication of the fluorescence intensity and the area of each of the spots (n = 3).

Bands were excised by hand and treated with trypsin as described elsewhere (Shevchenko et al., 2006). Tryptic digests were desalted on home-made C18 StageTips as described (Rappsilber et al., 2007). After elution from the StageTips samples were dried using a vacuum concentrator (Eppendorf) and the peptides taken up in 10 μL 0.1 % formic acid solution. LC-MS/MS experiments were performed on an Orbitrap Elite instrument (Thermo, Michalski et al., 2012) that was coupled to an EASY-nLC 1000 liquid chromatography (LC) system (Thermo). The LC was operated in the two-column mode. The home-made fused silica column equipped with a glass fiber frit (Maiolica et al., 2005) was packed with Reprosil-Pur 120 C18-AQ 3 μm resin (Dr. Maisch) and connected to the analytical column via an UHPLC union (Upchurch; UH-432). The analytical column was a fused silica capillary (75 μm × 25 cm) with integrated PicoFrit emitter (New Objective) packed in-house with Reprosil-Pur 120 C18-AQ 3 μm resin (Dr. Maisch). The analytical column was attached to a nanospray flex ion source (Thermo). The LC was equipped with two mobile phases: solvent A (0.1% formic acid, FA, in UPLC grade water) and solvent B (0.1% FA in acetonitrile, ACN). Peptides were delivered to the pre-column via the integrated autosampler at a flow rate of 2–3 μl/min in 100% solvent A. Peptides were subsequently separated on the analytical column by running a 70 min gradient of solvents A and B (start with 7% B; gradient 7–35% B for 60 min; gradient 35–100% B for 5 min and 100% B for 5 min) at a flow rate of 300 nl/min. The mass spectrometer was operated using Xcalibur software (version 2.2 SP1.48) and was set in the positive ion mode. Precursor ion scanning was performed in the Orbitrap analyzer (FTMS) in the scan range of m/z 300–1,500 and at a resolution of 120,000 with the internal lock mass option turned on (lock mass was 445.120025 m/z, polysiloxane; Olsen et al., 2005). Product ion spectra were recorded in a data dependent fashion in the ion trap (ITMS) in a variable scan range and at a rapid scan rate. The ionization potential (spray voltage) was set to 1.6–2.0 kV. Peptides were analyzed using a repeating cycle consisting of a full precursor ion scan (1.0 × 10⁶ ions) followed by 15 product ion scans (1.0 × 10⁴ ions) where peptides are isolated based on their intensity in the full survey scan (threshold of 500 counts) for tandem mass spectrum (MS2) generation that permits peptide sequencing and identification. CID collision energy was set to 35% for the generation of MS2 spectra. During MS2 data acquisition dynamic ion exclusion was set to 120 s with a maximum list of excluded ions consisting of 500 members and a repeat count of one. Ion injection time prediction, preview mode for the FTMS, monoisotopic precursor selection and charge state screening were enabled. Only charge states bigger than 1 were considered for fragmentation.

The recorded RAW files were processed in ProteomeDiscoverer 1.4 (PD14, Thermo). MS2 spectra were extracted using the Spectrum Selector node. Precursor selection was set to “use MS1 precursor.” The mass range was set between 350 and 5,000 Da with a minimum peak count of 1. Mass analyzer was set to “any” and MS order to “MS2.” Activation type was set to “is CID” and Scan type was defined as “full” with ionization source set to “is nanospray.” Selected spectra were submitted to the in house MASCOT server [version 2.4.1 (Perkins et al., 1999)] using the PD14 MASCOT node.

Tandem mass spectrum spectra data were searched against the tomato_ITAG.fasta database⁴ (version 2.3; 34725 entries). All searches included a contaminants database (as implemented in MASCOT and MaxQuant, 263 sequences). The contaminants database contains known MS contaminants and was included to estimate the level of contamination. Mascot and Andromeda searches allowed for oxidation of methionine residues (16 Da) and a static modification on cysteine (57 Da, alkylation with iodoacetamide). Enzyme specificity was set to Trypsin/P. The instrument type in MASCOT searches was set to ESI-TRAP and the mass tolerance was set to ± 10 ppm for precursor mass and ± 0.35 Da for product ion masses. MS2 spectra matches were then evaluated using the peptide validation node of PD14 with the standard settings [search against decoy database, target false discovery rate (FDR, strict): 0.01 and target FDR (released): 0.05]. The reported results were further filtered. On peptide level only peptides with a minimum confidence ‘medium’ were reported and on protein level only proteins with a minimum of at least two peptide hits were reported.

Sixty microliter of MVB072-labeled sample containing 1 M NaCl, 25 mM MgCl₂ 10 mM DTT, and 12.5X protease inhibitor cocktail (Sigma) were treated with 1 or 5 μl of alkaline phosphatase (Sigma P0114) and incubated at 37°C for 1 h. Samples were analyzed on 16% SDS-PAGE, whereas the remainder of the sample were separated on 10% SDS-PAGE and transferred to polyvinylidene difluoride (PVDF) membrane for detection of phosphorylated MAPK using primary antibody, Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) antibody (CST, #9101) and secondary Goat anti-Rabbit IgG (Thermo, #31466) and visualized using chemiluminescent substrates (SuperSignal West/Pico Chemiluminescent substrates, Thermo scientific).

Nine microliter of MVB072-labelled tomato root extract and Bovine Fetuin (Promega) were treated with 1 μl of 10X glycoprotein denaturing buffer (New England BioLabs) and heated at 95°C for 5 min. The denatured proteins were chilled on ice. Two microliter 10X GlycoBuffer (New England BioLabs), 2 μl 10% NP40 (Promega), and 6 μl H₂O was added to the reaction. The mixture was treated with 1 μl PNGase F (New England BioLabs) or with 1 μl H₂O and incubated at 37°C for 1 h. Samples were analyzed on 16% SDS-PAGE.

Eighteen microliter of MVB072-labelled sample and Bovine Fetuin (Promega) were treated with 2 μl of 10X denaturing solution (Promega) and heated at 95°C for 10 min. The denatured proteins were chilled on ice for 5 min. To the denatured samples 5 μl of 10X Deglycosylation Reaction Buffer (Promega), 5 μl of 10% NP40 (Promega), and 15 μl of water were added. Samples were treated with 5 μl of Protein Deglycosylation Mix (Promega – PNGase F, O-Glycosidase, Neuraminidase, β1–4 Galactosidase, β-N-Acetylglucosaminidase) and incubated at 37°C for 8 h. Samples were analyzed on 16% SDS-PAGE.

To investigate the tomato proteasome, we performed BLAST searches with the Arabidopsis β subunits on the predicted tomato proteome⁵ and identified eight tomato genes encoding β subunits. Phylogenetic analysis of the tomato and Arabidopsis β subunits revealed that tomato genome has one gene for each of the six β subunits (β1, β3, β4, β5, β6, and β7), and two genes encoding β2 (Figure 1A). The two β2 proteins in tomato (β2a and β2b) are more closely related to each other when compared to the two β2 proteins of Arabidopsis (PBB1 and PBB2), consistent with the fact that genome duplication occurred in each lineage, after divergence (The Tomato Genome Consortium, 2012). Tomato, however, must have lost the paralogous copies of each except one proteasome β subunit. Sequence alignment with the Arabidopsis orthologs indicates that each of the eight tomato genes encodes a putative functional subunit, including an N-terminal pro-domain for all subunits and a catalytic Thr for the β1, β2, and β5 subunits (Supplementary Figure S1).

The branch lengths indicate that the two β2 subunits are substantially different (Figure 1A). Indeed, we counted 18 amino acid residues that differ between the mature β2a and β2b subunits (Figure 1B). Most of these amino acid substitutions are biochemically dissimilar. To estimate if these variant residues can affect the proteolytic chamber, we generated a structural model of the tomato β2 protein using the yeast proteasome (2zcy, Groll et al., 2008) as a template. Mapping the residues that vary between β2a and β2b onto the structural model revealed that none of the variant residues are exposed to the proteolytic chamber (Figure 1C). Interestingly, nearly all the variant residues reside on the outer surface and are likely solvent-exposed (Figure 1C).

To determine which of the β subunit-encoding genes are expressed in different tissues, we mined RNAseq datasets for the transcript levels of each of these genes from RNAseq data (The Tomato Genome Consortium, 2012). As expected for subunits that assemble in stoichiometric complexes, transcript levels of each of the β subunit genes are very similar, with the exception of β2a transcripts, which accumulate 5- to 10-fold lower when compared to β2b and the other β subunit-encoding transcripts (Figure 1D). Nevertheless, detection of β2a transcripts suggests that β2a is not a pseudogene, but the transcript levels are low under normal conditions. The ratio between β2a and β2b transcript levels does not significantly change in different tissues (Figure 1D), suggesting that proteasome assembly might be similar for β2a and β2b subunits in different tissues.

Salt stress is common in plants and is associated with the release of ROS. Higher salt concentration also triggers PCD (Hasegawa et al., 2000; Poór et al., 2014). To investigate salt stress in tomato roots, plants were treated with sublethal- (100 mM) and lethal (250 mM) concentrations of NaCl (Figure 2A). We studied the early stages of abiotic stress by collecting samples at 1, 6, and 24 h after salt exposure. Root tips were stained with FDA to detect and quantify viable cells. Low FDA staining after 6 h upon treatment with 250 mM NaCl indicates a massive and quick PCD that completes within 24 h (Figure 2B). By contrast, treatment with 100 mM NaCl caused a slower loss of viability, where decreased viability was detected only at 24 h.

Papain-like cysteine Proteases and the proteasome have been implicated in stress and PCD. To examine the activity of both PLCPs and the proteasome, we first tested MV151, which labels both the proteasome and a subset of the PLCPs (Gu et al., 2010). MV151 labeling causes weak signals at 30–40 kDa which represent PLCPs and three stronger signals at 26 kDa that may represent the active proteasome subunits or PLCPs (Figure 3A). Interestingly, we detected a strongly activated band at ∼26 kDa with threefold higher intensity upon 250 mM salt treatment at 6 h (Figure 3A). This extra signal was robustly detected in all biological replicates (Supplementary Figure S2), and statistically significant upon quantification of fluorescence intensities from the three biological replicates (Figure 3B). To determine if this signal is caused by the proteasome or PLCPs, a competition assay was performed using proteasome inhibitor epoxomicin and PLCP inhibitor E-64. Pre-incubation with E-64 suppresses labeling at 30–40 kDa (Figure 3C), confirming that these signals are caused by PLCPs. The 26 kDa signals are not suppressed by E-64 (Figure 3C), indicating that these signals might be from the proteasome. Indeed, pre-incubation with the selective proteasome inhibitor epoxomicin suppresses labeling of all 26 kDa signals (Figure 3C), indicating that the significantly activated band is caused by the proteasome. Thus, although these MV151 labeling experiments did not display differential activities of PLCPs, it did uncover differential proteasome activity profiles.

To confirm differential proteasome activity, we used a newly re-synthesized epoxomicin-based MVB072, which carries both a bodipy tag for fluorescent detection and a biotin tag for affinity purification (Kolodziejek et al., 2011). When compared to MV151, MVB072 is a much more selective proteasome probe without any known off targets, ideal to confirm differential proteasome activity in our samples (Kolodziejek et al., 2011). Importantly, MVB072 labeling displays the same altered activity profile upon salt treatment as MV151 labeling (Figure 4A). Quantification of fluorescence intensities of the various signals demonstrate a highly reproducible increased intensity of the upper signal at 6 h upon 250 mM NaCl treatment over multiple biological replicates, whilst other signals seem to reduce (Figure 4B).

To identify the differentially active catalytic subunits of the proteasome, we separated MVB072-labeled proteomes of roots treated with and without 250 mM NaCl for 6 h by IEF and SDS gel electrophoresis. Over 14 fluorescent spots were robustly detected on 2D gels (Figure 5A). These fluorescent spots do not correlate with abundant proteins detected upon SYPRO Ruby staining of these gels (Supplementary Figure S3). The fluorescence intensity of each of these 14 spots was quantified using ImageJ and plotted in a histogram over three biological replicates (Figure 5B; Supplementary Figure S4). We detect significant increase of the fluorescence intensities of spots #5 and #13 and a significant decrease of fluorescence intensities of spots #8, #10, #11, #13, and #14, whilst fluorescence intensities of spots #7 and #9 were strongly increased upon salt treatment (Figure 5B). These changes in fluorescence intensities occurs in two regions in the 2D gel (boxed in Figure 5A) and correlates with the shift in fluorescent intensity on 1D gels, as illustrated in Figure 5C. At the acidic pI range, five spots (#7–11) showed reduced intensity upon salt stress, whilst one spot (#5) has increased signal intensity (blue box in Figure 5C). At basic pI range, the bottom signal (#14) decreases whilst the top signal (#13) intensifies upon salt stress (red box in Figure 5C).

A total of 28 fluorescent spots were excised from both gels and analyzed by MS. Twenty-seven of these spots contain catalytic subunits of the proteasome (Supplementary Figures S4–S6 and Table S1). In total, four different catalytic subunits were detected: Solyc07g016200.2.1 (β1), Solyc04g024420.2.1 (β2a), Solyc05g013820.2.1 (β2b), and Solyc05g056160.2.1 (β5), each with multiple unique peptides and significant Mascot protein scores (Supplemental Figures S5–S6 and Table S2).

β2a was identified in spots #5, #7, and #10 in the control treatment and in spots #4, #5, and #10 in the salt-treated sample (Figure 5D; Supplementary Figure S6). However, the majority of the MS signal in these spots comes from β2b (Supplementary Figure S6). This indicates that β2a is part of the proteasome, but contributes only a minor fraction, irrespective of the stress condition. These data do not support the hypothesis that the stress proteasome has a different β2a/β2b ratio.

To assign the fluorescence signal to a particular catalytic subunit, we ranked the protein scores in Mascot for each spot and highlighted the four detected catalytic subunits. This analysis shows that we often identified more than one catalytic subunit from several spots (Figure 5D). The identification of multiple subunits per spot might be caused by incomplete separation during IEF or by contamination during gel excision. We assigned the signal to a single catalytic subunit in case a single subunit ranks consistently high in a spot. This way we assigned β1 to spots #1, #2, #3, and #6; β2 to #4 and #5; β5 to #13 and #14, and we found stronger signals for multiple subunits in the remaining spots (#8–12; Figure 5E; Supplementary Figure S6). Taken together, these data indicate that salt stress induces a shift to a higher apparent MW for both labeled β2 and β5, and a shift to lower pI for labeled β5.

To confirm the identity of the shifting of the labeled catalytic subunits, the root extract was labeled with subunit-specific activity-based probes LW124 and MVB127 (Li et al., 2013; Misas-Villamil et al., 2017). LW124 is specific for β1 and has an epoxyketone reactive group and a fluorophore with excitation and emission wavelength at 470 and 530 nm, respectively. By contrast, MVB127 is selective for β5, carries a vinyl-sulfone reactive group and a fluorophore with excitation and emission wavelength at 532 and 580 nm, respectively. As they have different specifies and their excitation and emission wavelength are different, LW124 and MVB127 can be used by co-labeling (Misas-Villamil et al., 2017). Consistent with our earlier observations LW124 labeling does not display significant differences upon salt treatment, with the exception of on extra signal (arrowhead in Figure 6A). MVB127 labeling, however, shows a shift upward, confirming the MW shift of β5 (Figure 6A).

To confirm the composition of the LW124- or MVB127-labeled signals, we pre-incubated the samples with selective inhibitors, N3β1 and N3β5, which target β1 and β5, respectively (Misas-Villamil et al., 2017). N3β1 strongly suppressed labeling of LW124 (β1), whilst N3β5 blocked labeling of MVB127 (β5), respectively, verifying that the top differential MV151- and MVB072 signals contains a β5 subunit (Figure 6A). However, to a lesser extent, N3β1 also suppresses β5 labeling and N3β5 suppresses β1 labeling (Figure 6A), which can be explained by some cross reactivity of the inhibitors, but also by allosteric effects caused by subunit inhibition, including sterical hindrance in the proteolytic chamber.

To confirm the specific labeling and the pI/MW shifts of the catalytic subunits, we analyzed LW124/MVB127 co-labeled samples on 2D IEF-PAGE gels. The labeling pattern confirms the MS data: spots in the acidic range are labeled with LW124, consistent with being β1-derived signals, whereas the spots in the basic pI are labeled with MVB127, consistent with being β5-derived (Figure 6B). We do not detect cross reactivity of LW124 on β5, but there is some cross reactivity of MVB127 on β1 (Figure 6B). Consistent with proceeding data, the MVB127-labeled signals shift increase in MW and shift to more acidic pI upon salt treatment, even though in this case we detect some of this ‘stress β5 isoform’ also in plants treated without salt (Figure 6B). By contrast, LW124-labeled signals are constant in the samples treated with or without salt, with the exception of one additional signal that correlates to the signal detected on 1D gels (arrowhead in Figures 6A,B). This confirms that most of labeled β1 remains unaltered upon salt stress. Taken together, these data confirm a pI/MW shift for labeled β5 upon salt treatment.

Because of the shifts of denatured labeled β2 and β5 subunits when separated on 1D and 2D gels, we hypothesize that the MW is caused by a PTM of the labeled β2 and β5 subunits. Phosphorylation and glycosylation are two PTMs described for the proteasome (Bose et al., 2004; Zong et al., 2008; Scruggs et al., 2012). To test if the MW shifts are caused by phosphorylation, we treated the labeled samples with alkaline phosphatase. However, treatment of MVB072-labeled sample with alkaline phosphatase did not affect the shifted signal (Figure 7A). Probing the same extract with an antibody against phosphorylated MAP kinases showed a strong reduction, indicating that protein dephosphorylation worked in this assay (Figure 7B).

To test if the MW shifts are caused by differential glycosylation, we incubated the MVB072-labeled sample with PNGase F to remove N-glycans or with deglycosylation mix to remove both N- and O-glycans using a mixture of PNGase F, O-Glycosidase, Neuraminidase, β1–4 Galactosidase, en β-N-Acetylglucosaminidase. Neither of these treatments affected the labeling profile, suggesting that the MW shift is not caused by N- or O-glycosylation (Figure 7C). By contrast, Bovine Fetuin, a glycoprotein standard used as a positive control, did shift upon both treatments, indicating that deglycosylation worked in this assay (Figure 7D).

Post-translational modification is a likely molecular mechanism underpinning the shift in MW and pI of catalytic subunits. In animals, multiple PTMs have been described for proteasome subunits such as phosphorylation, glycosylation, ubiquitination, oxidation, glutathionylation, and nitrosylation (Bose et al., 2004; Ventadour et al., 2007; Zong et al., 2008; Scruggs et al., 2012; Cui et al., 2014). PTMs can regulate the stability of the proteasome, alter the assembly of different subunits and change the degradation pattern of proteasome or activity profile of catalytic subunits (Bose et al., 2004).

First, protein phosphorylation can cause both MW and pI changes (Zong et al., 2008). Phosphorylation of a serine residue in the α6 subunit has been studied in root tips of rice (Umeda et al., 1997). Phosphorylation also likely causes a mass increment of β1din during the induction of defense in tobacco (Suty et al., 2003). There are several predicted phosphorylation sites in catalytic subunits. The β2 subunit contains eight Ser, five Thr, four Tyr residues, whereas the β5 subunit contains 10 Ser, one Thr, and three Tyr residues (Supplementary Figure S3). However, searches or our MS data allowing phosphate modifications, did not reveal any phosphorylated peptides from the β2 or β5 subunits. In addition, phosphatase treatment did not affect the MVB072 activity profile in tomato roots upon 250 mM NaCl treatment (Figure 7A) suggesting that phosphorylation is not the underlying mechanism of the altered proteasome upon salt stress. However, phosphorylation cannot be ruled out because some phosphorylations can endure alkaline phosphatase treatment.

Second, proteasome subunits can be regulated by glycosylation. For instance O-Glycosylation reversibly inhibits proteasome function via modification of Rpt2 ATPase in the 19S regulatory particle of the proteasome in animals (Zhang et al., 2003). Likewise, N-Glycosylation is a key PTM of specific proteins during osmotic stress adaptation in plants (Koiwa et al., 2003). However, deglycosylation by PNGase or deglycosidase mix had no effect on the MW shift of β5 (Figure 7C) indicating that also glycosylation is not the underlying mechanism.

Third, ubiquitination is common on proteasome subunits. Both β2 and β5 catalytic subunits are ubiquitinated at Lys residues (Ventadour et al., 2007; Kim et al., 2013). However, the predicted MW shift of mono- or polyubiquitinated β2 and β5 is too large (>8.5 kDa) to explain the observed ∼2 kDa MW shift in the altered proteasome. The same argument excludes sumoylation as a PTM underlying the observed shifts in MW.

In addition, ROS production leads to accumulation of oxidatively modified proteins (e.g., carbonyl compounds) that alter the function of enzymes (Basset et al., 2002). There is evidence of carbonylation of the 20S proteasome in response to carbon starvation in maize root tips (Basset et al., 2002). Furthermore, glutathionylation has been detected on each of the β5 subunits of the plant proteasome (Dixon et al., 2005). β2 has four and β5 has two Cys residues which may explain a 2 kDa upon S-glutathionylation. Finally, Cys residues can also be modified by reactive nitrogen species, which are released during stress (Hess et al., 2005). However, the use of reducing agents during our sample preparation would probably remove oxidation, glutathionylation, and nitrosylation from Cys residues, so these PTMs are not likely to explain the MW/pI shifts that we detected. However, many additional PTMs are known and they can also be combined in several ways. Also a combination of PTMs might result in the observed MW/pI shifts in the activity profile of β2 and β5 catalytic subunits.