Previous work of our group has shown that the peroxisomal processing protease DEG15/GPP is present in two conformations, monomeric and dimeric, with different substrate specificities (Helm et al. 2007). Homodimerization of DEG15 is calcium-dependent and only the dimeric form of DEG15 represents the PST2-specific peroxisomal protease. So far the molecular basis for the calcium-dependency of dimerization is unknown, and we thus analysed the potential influence of the calcium signal mediator CaM on this process.
AtDEG15 interacts with AtCML3
We could show recently that the CaM-like protein AtCML3 is localized in peroxisomes (Chigri et al. 2012), making it a potential mediator of calcium-dependent dimerization of AtDEG15. In order to analyse a potential interaction between AtDEG15 and AtCML3, we used the Y2H system with two different reporter genes. Transcription of the HIS3 reporter leads to growth on His-free selection medium and allows a qualitative or semi-quantitative estimation of protein interaction; transcription of the lacZ reporter leads to expression of β-galactosidase and allows a quantitative estimation of protein interaction dependent on the affinity of the protein partners for each other.
Initially, interaction of full-length AtDEG15 with AtCML3 was tested together with multiple controls (Fig. 1A and supplementary Fig. S1). The full-length coding regions of AtDEG15 and AtCML3 were each cloned both into pBridge and pGAD424, fusing them with the DNA-binding domain or with the DNA-activation domain, respectively. These vectors were then used to double-transform the yeast strain AH109 with either AtDEG15_pBridge and AtCML3_pGAD424 or vice versa. As negative controls, a combination of empty vectors with complementary vectors harbouring one of the constructs were used. While the results of all experiments are shown in supplementary Figure S1, the most relevant data are presented in Fig. 1. While no growth on His-free plates could be observed for the negative controls, co-transformation of AtDEG15 together with AtCML3—in both combinations—lead to growth of the transformants on His-free medium, indicating that AtDEG15 and AtCML3 interact with each other (Fig. 1B and Fig. S1A). When β-galactosidase activity was analysed, the measured activity for AtDEG15/AtCML3-interaction was 20-fold higher than for the negative controls (Fig. 1C and supplementary Fig. S1B), indicating that the interaction between both proteins is very strong.
The interaction between AtDEG15 and CaM was further confirmed in vitro by affinity chromatography on CaM-agarose (Fig. S2). Using IPTG inducible vectors for production of recombinant AtDEG15 in E. coli, expression yielded high amounts of the undegraded AtDEG15 full length protein that, however, was deposited in inclusion bodies. Recombinant 6His-AtDEG15 was isolated from inclusion bodies in the presence of 4 M urea and further purified by His-tag affinity chromatography. Purified AtDEG15 was subsequently incubated with CaM-agarose in the presence of either 0.1 mM CaCl2 or 5 mM EGTA/5 mM EDTA and, after extensive washing, the bound proteins were eluted from the matrix by an excess of commercial bovine calmodulin. Protein analysis from all fractions by SDS-PAGE and Coomassie staining revealed binding of AtDEG15 to the CaM-ligand solely in the presence of Ca2+ (supplementary Fig. S2A). Furthermore, AtDEG15 could be eluted from the column not only with bovine calmodulin but also with recombinant AtCML3 (supplementary Fig. S2B).
We repeated the affinity chromatography on CaM-agarose using recombinant AtDEG15 expressed in E. coli under the control of an arabinose-inducible vector (Schuhmann et al. 2008). Production of 6His-AtDEG15 was extremely low, but the protein was produced in a soluble form. However, under these conditions recombinant AtDEG15 protein exhibits the characteristics of the monomer as a general protease with an intrinsic self-cleavage activity. Purification of the N-terminally His-tagged protein consistently resulted in various amounts of degradation products further reducing the amount of full length AtDEG15 (Schuhmann et al. 2008). When His-tag purified AtDEG15 protein was immediately analysed by western blot analysis using antibodies directed against the Xpress-epitope located between the His-tag and the N-terminus of AtDEG15 protein (Fig. S3), the full length AtDEG15 (78 kD including tags) could be detected as well as an ~58 kD protein that still bears the intact N-terminus, since it reacts with the antibodies directed against the Xpress-epitope. This supposingly represents an AtDEG15 variant lacking the 93 amino acids C-terminal to the protease domain, which could be relatively protease-resistant due to its folding (supplementary Fig. S3). When the same purified protein fraction was analysed by affinity chromatography on CaM-agarose (Fig. 2A), the full length AtDEG15 as well as the N-terminally intact AtDEG15 variant bound to the CaM ligand in the presence of Ca2+ (Fig. 2A) but not in the presence of EGTA/EDTA (Fig. S2C). Storage of the very same His-tag purified AtDEG15 protein batch over night at 4 °C produced even more N-terminal as well as C-terminal degraded AtDEG15 variants (supplementary Fig. S4A, protein loaded). While all N-terminal intact proteins bound to CaM-agarose, the N-terminal truncated variants did not bind thereby serveing as an internal negative control (Fig. S4A).
Together, these data confirm the specific and Ca2+-dependent interaction of AtDEG15 with CaM and suggest an involvement of the N-terminal part of AtDEG15 in this interaction.
Identification of the AtDEG15 CaM-binding domain
To clearly identify regions important for the interaction with CaM, several N- and C-terminally truncated variants of AtDEG15 were prepared (Fig. 1A; supplementary Fig. S1, inlay) and interaction with AtCML3 was analyzed by Y2H. As already shown above, full-length AtDEG15 interacted with AtCML3, resulting in growth on His-free medium and strong expression of β-galactosidase. Interaction could also be observed with a truncated AtDEG15 variant comprising only the 373 amino acids long N-terminal domain, thus lacking both the protease and the C-terminal domain (AtDEG15N) (Fig. 1B; supplementary Fig. S1A). Quantitative analysis showed that the interaction was strong and at a level similar to the full length AtDEG15 (Fig. 1C; supplementary Fig. S1B). The interaction of AtDEG15_N with CaM in a Ca2+-dependent manner was confirmed in vitro by affinity chromatography on CaM-agarose using the AtDEG15_N (aa 1–327) recombinantly expressed in E. coli (Fig. 2B; supplementary Fig. S2D).
AtDEG15 variants missing the C-terminal domain or just missing the plant specific loop within the protease domain also interacted with AtCML3 (see Y3H analysis below). In contrast, various N-terminally truncated AtDEG15 constructs did not interact with AtCML3 (Fig. 1B, C; supplementary Fig. S1), confirming that the CaM-binding site is located in the N-terminal part of AtDEG15. Typical CaM binding sites comprise about 25 amino acids (Calmodulin Target Database CTDB: http://calcium.uhnres.utoronto.ca/). We thus narrowed down the CaM-binding domain by deleting the first 25, 50, 75 or 100 amino acids of AtDEG15, respectively (Fig. 1 and Fig. S1). Indeed, already the removal of the first 25 amino acids was sufficient to completely abolish interaction of AtDEG15 with AtCML3 suggesting that the CaM-binding site is located in the outmost N-terminus of AtDEG15 (Fig. 1B, C; Fig. S1).
To ensure that the results observed with N-terminally truncated variants are not caused by lack of expression or protein degradation, we analysed the presence of these variants by western blot. Western blot analysis of the most relevant constructs using antibodies directed against the BD in pBridge confirmed that the AtDEG15 recombinant proteins are indeed being expressed in yeast and are of the expected size (supplementary Fig. S5).
Bioinformatic analyses supported the presence of a putative CaM-binding site at the very N-terminus of AtDEG15 (Fig. 3). A characteristic of CaM-binding domains is the spacing between two bulky hydrophobic residues and the formation of a basic amphiphilic helical structure and the latter is formed between Val-3 and Glu-20 of AtDEG15 (Fig. 3A). A high probability for a CaM-binding motif is indicated within the same first 20 N-terminal amino acids of AtDEG15 (Fig. 3B; Calmodulin Target Database, CTDB), as well as both a so-called 1–8–14 and a 1–14 motif (Fig. 3C). To confirm these bioinformatic data, we performed crosslinking experiments with CaM and a synthetic peptide comprising the first 21 amino acids of AtDEG15 (H2N-MDVSKVVSFSRNFAVLVKVEG-CONH2). A cross-linking product with CaM can be observed solely in the presence of both the synthetic peptide and calcium (supplemental Fig. S6A; +calcium). No cross-linking product occurs in the absence of calcium (Fig. S6A; −calcium) or when CaM is replaced by egg albumin (Fig. S6B). This supports that the first 21 amino acids of AtDEG15 are able to specifically interact with CaM.
A sequence alignment of various plant DEG15 orthologous revealed highly conserved 1–8–14 or 1–14 motifs within the N-terminal 20 amino acids of all proteins from dicots and monocots (Fig. 3C; supplemental Fig. S7). By contrast, the moss sequence (PhpaDEG15) has no clear CaM binding motif at the beginning of its long N-terminus (Fig. S7). The alignment shows that PhpaDEG15 contains the conserved methionine that is positioned at the start of the CaM binding motif in higher plants, whereas several of the hydrophobic and basic amino acids that are necessary and conserved in higher plants for formation of the amphiphilic helix are missing within this stretch of amino acids (Fig. S7). As a consequence, the prediction programs do not indicate a CaM binding motif within the moss sequence.
Together these results suggest that the first 20–25 amino acids of AtDEG15 are necessary and responsible for CaM-binding. No CaM-binding motif is found within the N-terminus of peroxisomal processing proteases from animals and cellular slime molds or from the moss Physcomitrella patens (Fig. S7), indicating that CaM-binding together with Ca2+-dependent regulation of DEG15 dimerization is conserved among and specific for higher plants (monocots and dicots).
Homodimerization of AtDEG15 requires calmodulin
Y2H analysis (Fig. 1 and Fig. S1), affinity chromatography on CaM-agarose (Fig. 2 and Fig. S2), bioinformatic analyses (Fig. 3) and CaM-AtDEG15 peptide interaction (Fig. S6) all support an interaction between AtDEG15 and AtCML3 due to CaM-binding at the most N-proximal 20 amino acids of AtDEG15. Since it was previously shown that homodimerization of DEG15/GPP is calcium-dependent (Helm et al. 2007), a potential role of AtCML3 in this process was elucidated by Y3H analysis using several deletion variants of AtDEG15 (Fig. 4A) in the presence or absence of AtCML3.
Homodimerization of full length AtDEG15 was strong in the presence of AtCML3, reaching 8.92 Miller units (Fig. 4B; DEG15/DEG15, 0 mM Met). By contrast, homodimerization of full length AtDEG15 in the absence of AtCML3 was at the level of the empty vector control. Interestingly, both AtDEG15 molecules need to contain the CaM-binding domain, since no effect of AtCML3 on homodimerization could be observed, if a full length AtDEG15 was tested against the construct AtDEG15ΔCML3-bd lacking the N-terminal 25 amino acids long CaM binding domain (Fig. 4B; DEG15Δ25aa/DEG15). Consequently, no homodimerization in the presence of AtCML3 was observed either, when both AtDEG15 molecules lacked this domain (DEG15Δ25aa/DEG15Δ25aa) or in all other constructs lacking this part of the protein (Fig. 4B). By contrast, the N-terminus of AtDEG15 always promoted homodimerization in the presence of AtCML3 (Fig. 4B; DEG15Δloop/DEG15; DEG15ΔC/DEG15), even if present alone (Fig. 4B; DEG15N/DEG15).
These findings support the conclusion that CaM binding to AtDEG15 at the N-terminal CaM-binding domain is a decisive feature for homodimerization of AtDEG15 and that AtCML3 is a potential mediator of this process.
Full length AtDEG15 is required for restoration of PTS2 processing in atdeg15 mutant in planta
As shown above, CaM-binding at the N-terminus of AtDEG15 is a pre-requisite for homodimerization. This is important since only the DEG15 dimer is a specific PTS2 processing protease (Helm et al. 2007). We thus transformed AtDEG15 variants corresponding to those used in the Y3H experiments (Fig. 4A) into atdeg15 knockout mutant plants of A. thaliana (SALK_007184) under control of the 35S promoter. All AtDEG15 variants had the C-terminal PTS1–SKL attached in order to ensure their correct targeting into peroxisomes, and transcription of all constructs was confirmed by RT-PCR (Fig. 4C a). Successfully transformed atdeg15 knockout plants homozygous for the insertion of the different AtDEG15 variants were then used to analyze processing of the PTS2-containing enzymes pre-gMDH and pre-thiolase in planta (Figs. 4C, 5).
Processing of pre-gMDH to the smaller mature gMDH is easily visible by Western Blot analysis probed with a gMDH-specific antiserum when wild type protein extracts (Fig. 4C b; Col0 WT) are compared to knockout mutants of atdeg15 (Fig. 4C b; compare WT, SALK_007184 and GABI-Kat line GK_237G09). Transformation of the SALK_007184 mutant plant with a full-length AtDEG15 construct restored the processing of pre-gMDH to mature gMDH (Fig. 4C b; deg15_DEG15). Obviously the full length AtDEG15 is present and functional in the transformed plants as can be recognized by its processing activity, thus confirming the feasibility of this experimental approach. Transformation of the SALK_007184 mutant plant with diverse AtDEG15 variants lacking the calmodulin-binding domain or with the DEG15 protease domain alone could not restore the pre-gMDH to gMDH processing (Fig. 4C; deg15_DEG15ΔCML3-bd, deg15_DEG15Δ1/3, deg15_DEG15Δ2/3, deg15_DEG15ΔN, deg15_DEG15Δpd).
AtDEG15 proteins missing the domain C-terminal to the protease domain or missing the plant-specific loop within the protease domain also did not process pre-gMDH (Fig. 4C; deg15_DEG15ΔC, deg15_DEG15Δloop) although they contain the calmodulin-binding domain and were able to homodimerize as shown by Y3H analysis (Fig. 4B). Furthermore, AtDEG15 constructs lacking different amounts of the N-terminal domain but containing the calmodulin-binding domain artificially attached at the N-terminus were also not able to process pre-gMDH in planta (Fig. 4C; deg15_DEG15Δ1/3Ν + CML3-bd, deg15_DEG15Δ2/3Ν + CML3-bd, deg15_DEG15ΔΝ + CML3-bd). This shows that multiple domains of AtDEG15 are required for proper processing enzyme activity. The AtDEG15-N-terminus alone was not transformed into the atdeg15 knockout mutant (SALK_007184), since processing activity is not to be expected without the protease domain.
Processing of pre-thiolase to the smaller mature thiolase is also easily visible by Western Blot analysis probed with a thiolase-specific antiserum. The most relevant protein extracts used above were thus also analyzed for pre-thiolase processing ability (Fig. 5). Again, the deg15 knockout mutant plant could not process pre-thiolase (Fig. 5B). Transformation of the deg15 knockout mutant plant with the full length AtDEG15 construct restored processing, whereas the various deletion constructs did not restore processing enzyme activity (Fig. 5C).
Recombinant AtDEG15 exhibits the characteristics of the monomer as a general protease with an intrinsic self-cleavage activity
We tried to confirm the necessity of CaM for AtDEG15 PTS2 processing activity in vitro using the recombinant, enzymatically active AtDEG15. Unfortunately, the enzyme exhibited only the general protease activity described for the monomeric form. Digestion of bovine β-casein with recombinant AtDEG15 revealed several peptides as identified by MALDI-TOF (supplemental Fig. S8), deriving especially from cleavage within the C-terminal part of β-casein. When the recombinant AtDEG15 was incubated with 35S-methionine labeled pre-gMDH, pre-gCS or pre-thiolase, it processed pre-gMDH to mature gMDH, albeit independent of the presence or absence of calcium, while neither pre-gCS nor pre-thiolase were processed under either condition. It is important to notice, that pre-gMDH is not an indicative substrate in vitro since processing at the correct cleavage site can be achieved also by other unrelated proteases such as the ricinosome-localized KDEL-tailed cysteine endopeptidase involved in programmed cell death (Gietl et al. 1997; Schmid et al. 1998). This is probably due to an unusual distribution of amino acids at the cleavage site (Helm et al. 2007).
As mentioned above, the recombinant AtDEG15 also seems to digest itself leading to multiple N-and C-terminal truncated variants of the protein (Figs. 2A, S3 and S4A). By contrast, the AtDEG15 N-terminus lacking the protease domain and C-terminus was stable for a much longer time and also interacted with the CaM-ligand (Figs. 2B and S4B). Unfortunately, it was not possible to isolate the AtDEG15 full length protein and remove all cleavage products by size exclusion chromatography since storage or further analysis of the isolated AtDEG15 full length protein resulted again in degradation variants.
Since the monomeric form of AtDEG15 is a general protease activated by denatured proteins (Helm et al. 2007), we would thus assume, that correct folding of the entire recombinant AtDEG15, especially of the long N-terminus as a pre-requisite for dimeric PTS2-specific enzymatic activity cannot be ensured in vitro, thereby resulting in protein auto-degradation.
Role of CaM/AtCML3 in homodimerization of AtDEG15 and in peroxisome metabolism
The data presented so far indicate that CaM binding promotes homodimerization of AtDEG15, which in turn is a necessary pre-requisite for its PTS2 processing activity. While AtCML3 can mediate AtDEG15 homodimerization, it is unclear whether AtCML3 or a different CML full-fills this role in planta. The atcml3 knockout mutant line (pst16586, RIKEN) has a transposon insertion directly in front of the PTS1 signal—SNL of AtCML3. This would prevent peroxisomal targeting presumably leading to mislocalization of AtCML3 to the cytosol as shown in our previous publication for a YFP-AtCML3 fusion construct lacking SNL (Chigri et al. 2012). Nevertheless, PTS2 processing of pre-gMDH to the smaller mature gMDH and of pre-thiolase to the smaller mature thiolase, respectively, is unimpaired in the atcml3 knockout mutant line as shown by Western Blot analysis of protein extracts probed with the gMDH- and the thiolase-specific antiserum, respectively (Fig. 5A, B).
However, atcml3 knockout plants do show defects in normal peroxisomal metabolism. Treatment of wild type plants with the herbicide precursor 4-(2,4-dichlorphenoxy) butyric acid (2,4-DB), which is converted in peroxisomes to the toxic herbicide 2,4-dichlorphenoxyacetic acid (2,4-D) by beta-oxidation, results in a strong reduction of root growth. A resistance to 2,4-DB can also be observed for homozygous atcml3 mutants when grown on solid medium containing 2,4-DB (Fig. 6). Interestingly, it was shown recently that the lack of PTS2 processing by AtDEG15 results in an increased resistance to the 2,4-DB, the only obvious phenotype observed so far for homozygous atdeg15 mutant plants (Schuhmann et al. 2008).
Thus, not only seems AtCML3 to be involved somehow in peroxisomal β-oxidation; the phenotypical similarity between for atcml3 and atdeg15 mutants corroborates a functional correlation between these two proteins in vivo.