Yeast Snf4 is a prototype of activating γ-subunits of conserved Snf1/AMPK-related protein kinases (SnRKs) controlling glucose and stress signaling in eukaryotes. The catalytic subunits of Arabidopsis SnRKs, AKIN10 and AKIN11, interact with Snf4 and suppress the snf1 and snf4 mutations in yeast. By expression of an Arabidopsis cDNA library in yeast, heterologous multicopy snf4 suppressors were isolated. In addition to AKIN10 and AKIN11, the deficiency of yeast snf4 mutant to grown on non-fermentable carbon source was suppressed by Arabidopsis Myb30, CAAT-binding factor Hap3b, casein kinase I, zinc-finger factors AZF2 and ZAT10, as well as orthologs of hexose/UDP-hexose transporters, calmodulin, SMC1-cohesin and Snf4. Here we describe the characterization of AtSNF4, a functional Arabidopsis Snf4 ortholog, that interacts with yeast Snf1 and specifically binds to the C-terminal regulatory domain of Arabidopsis SnRKs AKIN10 and AKIN11.
Members of the Snf1/AMP-activated protein kinase (AMPK) family play essential roles in the regulation of carbon metabolism, energy housekeeping and stress responses in eukaryotes. In yeast, Snf1 is required for derepression of glucose-regulated genes during starvation and switch from fermentative to non-fermentative carbon source. Snf1 is involved in the regulation of glucose sensing and transport, gluconeogenesis, mitochondrial and peroxisome biogenesis, cell cycle, sporulation, thermotolerance, and stress-controlled accumulation of glycogen ( Carlson, 1998). Animal AMPKs control several key enzymes of ATP-consuming anabolic pathways, but also modulate glucose-regulated gene expression during nutrient starvation and hypoxia ( Hardie & Carling, 1997; Kemp et al. 1999 ). In analogy, plant Snf1-related kinases (SnRKs) regulate the activity of rate limiting metabolic enzymes, including 3-hydroxy-3-methylglutaryl-CoA reductase, nitrate reductase and sucrose phosphate synthase ( Sugden et al. 1999 ), as well as the transcription of glucose and stress-regulated genes ( Bhalerao et al. 1999 ; Purcell et al. 1998 ).
In addition to functional analogies, the Snf1/AMPK/SnRK enzymes show similar subunit organization and share some conserved interacting partners. The catalytic (AMPKα/Snf1) and activator (AMPKγ/Snf4) subunits of yeast and animal enzymes are found in complexes with different β-subunits (SIP1, SIP2, Gal83/AMPKβ1, β2; Hardie et al. 1998 ). A remarkable conservation of plant enzymes is illustrated by the fact that the catalytic subunits of SnRK1 enzymes complement the yeast snf1 mutation ( Halford & Hardie, 1998), recognize the yeast Snf4, Sip1 and Gal83 proteins, and bind to plant orthologs of Snf1/AMPK β and γ subunits ( Bhalerao et al. 1999 ; Bouly et al. 1999 ; Lakatos et al. 1999 ).
Most of our current knowledge on regulatory functions of Snf1-related enzymes is based on genetic and biochemical analysis of glucose signaling in yeast. Glucose depletion, inducing the phosphorylation of Snf1, stimulates stable association of the catalytic subunit with Snf4 and one of the β-subunits ( Carlson, 1999). Dephos- phorylation mediated by glucose-induced interaction with Reg1, a regulatory subunit of protein phosphatase Glc7, leads to autoinhibition of Snf1 by an intramolecular interaction between the N-terminal kinase and C-terminal regulatory domains of the catalytic subunit ( Ludin et al. 1998 ). Mutations affecting Snf4 result in a defect of both Snf1 activation and derepression of glucose repressible genes preventing yeast growth on sucrose and non-fermentable carbon sources, such as glycerol and ethanol. As Snf1 modulates multiple signaling pathways, a similar sucrose non-fermenting (snf) phenotype is caused by mutations in regulatory genes that encode, e.g. the Hap2/Hap3 subunits of CAAT-binding factor, controlling the transcription of respiratory genes, the snf3 and rgt2 glucose sensors, and subunits of RNA polymerase II mediator complex ( Gancedo, 1998). The snf4 phenotype is suppressed by mutations in genes (e.g. Grr1, Reg1, Hxk2, MIG1, Ssn6, TUP1, etc.) required for the induction and maintenance of glucose repression, as well as by overexpression of Snf1, SIP1, Gal83 and Snf1-binding Leu-zipper factor SIP3 ( Carlson, 1998). Several factors controlling the regulation of stress responses, probably in cross-talk with glucose signaling, act as multicopy snf4 suppressors. These include, for example, the stress-response element STRE-binding zinc-finger transcription factors Msn2 and Msn4, the MIG1-binding exportin Msn5, and casein kinase I (Yck1) ( DeVit & Johnston, 1999; Estruch & Carlson, 1993; Robinson et al. 1992 ).
By expression of an Arabidopsis cDNA library in yeast, we have identified 10 classes of heterologous multicopy snf4 suppressors, some of which show either functional or structural similarity to yeast genes uncovered by genetic screens using sucrose non-fermenting mutants. Here we report on the characterization of AtSNF4 that can partially replace the function of yeast Snf4 and interact with Snf1 and plant SnRKs.
Identification of heterologous snf4 suppressors encoded by Arabidopsis cDNAs
A yeast strain carrying the snf4Δ2 mutation ( Celenza & Carlson, 1989) was transformed with cDNA libraries made from Arabidopsis cell suspension and light-grown seedlings, respectively, in expression vectors pYX112 ( Umeda et al. 1998 ) and pFL61 ( Minet et al. 1992 ). Ura3+ transformants enriched on 2% glucose were subjected to standard selection on minimal medium (SD-GEG) containing 3% glycerol, 2% ethanol and 0.05% glucose to isolate snf4 suppressors ( Jiang & Carlson, 1996). From each colony, plasmid was isolated and recurrently transformed into the snf4Δ2 host to assay the suppression of snf4 growth defect on SD-GEG and minimal medium containing 2% glycerol ( Rose et al. 1990 ). Characterization of 68 snf4Δ2 suppressors by sequencing of the cDNA inserts identified three AKIN10 and 43 AKIN11 cDNAs that promoted growth of the snf4Δ2 mutant on both SD-GEG and 2% glycerol media, as well as complemented the snf1Δ10 mutation ( Table 1). This result was consistent with earlier observations showing that yeast Snf1 and its Arabidopsis orthologs AKIN10 and AKIN11 suppress the snf4 mutation ( Bhalerao et al. 1999 ; Celenza & Carlson, 1989). The remaining 22 clones failed to suppress the snf1Δ10 deficiency and differed in their capability to suppress the snf4Δ2 growth defect on 2% glycerol ( Table 1).
Table 1. Functional similarities between Arabidopsis suppressors of snf4Δ2 and yeast proteins
Number of clones
Similarity to yeast function
Suppressors of snf4Δ2 on 2% glycerol
Hexose transporter, Mss1
Casein kinase I
Suppressors of snf4Δ2 only on 3% glycerol, 2% ethanol and 0.05% glucose
Arabidopsis cDNAs suppressing the snf4Δ2 growth defect in both SD-GEG and 2% glycerol media encoded proteins with some similarity to known multicopy snf4 suppressors and elements of yeast stress signaling pathways. Thus, ZAT10 carried two Cys2His2 zinc-fingers similar to those present in the yeast snf4 suppressors Msn2 and Msn4 (data not shown). ZAT10 had been demonstrated to also suppress the Li+/Na+ sensitivity of yeast mutants deficient in the Ca2+/calmodulin-dependent protein phosphatase calcineurin ( Lippuner et al. 1996 ). Other Arabidopsis snf4Δ2 suppressors identified an ortholog of Hap3 (a subunit of CAAT-binding complex required in yeast for activation of glucose-repressed respiratory genes ( Edwards et al. 1998 ; Xing et al. 1993 ), casein kinase I (a yeast snf4 suppressor; Ali et al. 1994 ; Estrada et al. 1996 ; Robinson et al. 1992 ), and a novel hexose transporter (Mss1) with structural similarity to Snf3. Finally, Myb30 showed a relationship to Cdc5 that is a binding partner of a fission yeast ortholog of PRL1, a WD-protein inhibitor of Arabidopsis Snf1-related kinases AKIN10 and AKIN11 ( Bhalerao et al. 1999 ; McDonald et al. 1999 ). A second class of Arabidopsis cDNAs, suppressing the snf4Δ2 growth defect only in SD-GEG, but not in 2% glycerol, encoded a ZAT10-like zinc-finger factor, AZF2, and novel orthologs of SMC1-cohesin (Mss2), calmodulin (Mss3), UDP-galactose transporter (Mss4) and yeast Snf4 (AtSNF4) ( Table 1). To demonstrate that these heterologous snf4 suppressors did not represent artifacts, one of the latter ‘weak’snf4Δ2 suppressors, AtSNF4, was subjected to detailed characterization.
AtSNF4 is an Arabidopsis ortholog of yeast Snf4
All four AtSNF4 cDNAs identified as ‘weak’snf4Δ2 suppressors encoded a protein of 41.8 kDa that carried four cystathionine β-synthase (CβS) repeats characteristic for the yeast Snf4 and animal AMPKγ proteins ( Fig. 1a; Kemp et al. 1999 ). The corresponding gene, AtSNF4, consisted of 12 exons and proved to be identical with the database entry BAC F7G19.11 (AC000106) in chromosome 1-8.7, although this accession included a wrong prediction and annotation of ORF suggesting a similarity to the β-subunit of Rattus AMP-activated kinase (X95577). Blast-alignments with the Blossum 62 algorithm, Clustal-W alignment, and tree-analysis showed that AtSNF4 shared a low, but significant, sequence identity with yeast Snf4 and rat AMPKγ. In contrast, no significant sequence identity was detected between AtSNF4 and AKINγ, a predicted Arabidopsis ortholog of Snf1/AMPK γ-subunit ( Bouly et al. 1999 ), although both proteins carried CβS-repeats ( Fig. 1a,b). Low-stringency Southern DNA hybridization with the AtSNF4 cDNA probe indicated that AtSNF4 is a single-copy gene in Arabidopsis ( Fig. 1c). Northern hybridization analysis of steady-state RNA levels suggested that AtSNF4 is preferentially transcribed in dividing cells of Arabidopsis cell suspension, as well as in shoot apex and flower buds, but expressed at a lower level in differentiated tissues of roots, and leaves of rosette and inflorescence ( Fig. 1d).
To demonstrate that AtSNF4 can indeed interact with Arabidopsis Snf1-related kinases, a pull-down assay was performed with AKIN10 and AKIN11 that were purified to homogeneity in form of N-terminal glutathione-S-transferase (GST) fusion proteins ( Bhalerao et al. 1999 ). From the AtSNF4 cDNA template coupled 5′-upstream to a T7-promoter, 35S-methionine labeled AtSNF4 protein was synthesized by in vitro transcription translation and incubated with a control GST protein, as well as with GST-AKIN10 and GST-AKIN11, bound to the glutathione–Sepharose matrix. Following equal 35S-AtSNF4-loading, only GST-AKIN10 and GST-AKIN11 retained the probe after stringent washes indicating that AtSNF4 tightly bound to the Arabidopsis Snf1-related kinases in vitro ( Fig. 1e).
Glucose-regulation of AtSNF4-binding to Snf1 and plant SnRKs in yeast
Qualitative ( Fig. 2a) and quantitative (data not shown) growth assays showed that AtSNF4 conferred a slower growth in the presence of 3% glycerol, 2% ethanol and 0.05% glucose than AKIN10 and AKIN11 in the snf4Δ2 mutant. This observation suggested that the interaction of AtSNF4 with the yeast Snf1 kinase catalytic subunit was somehow compromised. Therefore, we tested whether AtSNF4 shows a similar glucose-regulated interaction with yeast Snf1 and plant SnRKs as seen earlier for Snf4 in the yeast two-hybrid system ( Bhalerao et al. 1999 ; Jiang & Carlson, 1996). AtSNF4 was fused in frame with the Gal4 activation domain (GAD) in the yeast two-hybrid prey vector pACT2 and combined by co-transformation with pAS2 bait vectors encoding fusions of the yeast Snf1, tobacco NPK5 ( Muranaka et al. 1994 ), and Arabidopsis AKIN10 and AKIN11 kinases with the Gal4 DNA-binding domain (GBD). Activation of the Gal4-regulated LacZ reporter gene in at least 12 independent transformants, carrying each bait and prey combination, was monitored by measuring β-galactosidase levels after growing the cells in SD-medium with either 2% glucose or 3% glycerol, 2% ethanol and 0.05% glucose. Interaction of AtSNF4 with the catalytic subunits of yeast and plant kinases was only stimulated two- to fourfold under glucose derepression ( Fig. 2b) contrasting an over 100-fold activation of interaction between yeast Snf1 and Snf4 ( Jiang & Carlson, 1996), and a 32–42-fold enhancement of Snf4-binding to AKIN10 and AKIN11 observed earlier under similar conditions ( Bhalerao et al. 1999 ). The data also indicated that binding of AtSNF4 was stronger to yeast Snf1 than to its plant orthologs, and was also well detectable under glucose repression. Yeast Snf4 fused to GBD did not recognize GAD-AtSNF4 in control experiments confirming the assumption that the yeast Snf4 protein present in the two-hybrid system could not be a mediator of interactions between yeast Snf1, plant SnRKs and AtSNF4. Other control experiments showed that the PRL1 inhibitor of AKIN10 and AKIN11, as well as its homolog PRL2, did not bind to AtSNF4.
The AtSNF4 binding site was mapped by assaying the interaction of GAD-AtSNF4 with a series of GBD-baits carrying N- and C-terminal deletions of AKIN10 and AKIN11 ( Fig. 2c). The binding domain of AtSNF4 was confined to C-terminal sequences located between amino acid positions 349 and 512 of AKIN10 and AKIN11 that were previously shown to be insufficient for binding of yeast Snf4 ( Bhalerao et al. 1999 ). AtSNF4 showed weaker binding to AKIN11 sequences located between positions 399 and 512, and failed to interact with a more distant C-terminal domain defined by positions 469 and 512. The minimal AtSNF4 binding site was thus localized between AKIN11 positions 399 and 469.
A starting point of described experiments was the observation that the Arabidopsis Snf1-related protein kinases AKIN10 and AKIN11 can replace the Snf1 kinase function and interact in a glucose-regulated fashion with the activating Snf4 subunit in yeast ( Bhalerao et al. 1999 ). Based on the assumption that conserved domains of yeast Snf1 may recognize similarly conserved subunits and regulators derived from plants, Arabidopsis cDNAs were expressed in yeast to search for suppressors of the snf4Δ2 mutation. In addition to AKIN10 and AKIN11, 10 different classes of Arabidopsis snf4Δ2 suppressors were identified. Some of these showed a similarity either to known yeast snf4 suppressors or to elements of stress signaling pathways, including AtSNF4, a potential Arabidopsis Snf4 ortholog.
To illustrate the usefulness of the heterologous suppressor screen as a genetic approach, it was demonstrated that AtSNF4 is a novel functional ortholog of yeast Snf4 which interacts with Snf1 and plant SnRKs in the two-hybrid system and in vitro. Intriguingly, AtSNF4 showed very little similarity to AKINγ, another Snf4-like Arabidopsis protein that is known to bind to AKIN11, although it cannot suppress the snf4 mutation and interact with Snf1 in yeast ( Bouly et al. 1999 ). Despite their considerable sequence divergence, both AtSNF4 and AKINγ carry well-recognizable CβS domains as the yeast Snf4 and animal AMPKγ proteins. It is therefore likely that the interaction of AKINγ and AtSNF4 with plant SnRKs is mediated by structurally conserved CβS domains.
AtSNF4 binds to a region in the C-terminal regulatory domain of both AKIN10 and AKIN11 that is located downstream of the yeast Snf4-binding site and overlaps with part of the binding domain for the β-subunits ( Jiang & Carlson, 1997). The fact that this domain also shows an interaction with other factors, such as the PRL1 WD-protein, suggests that AKIN10 and AKIN11 may occur in different regulatory complexes. Unlike AKINγ, which is similarly expressed in non-reproductive organs ( Bouly et al. 1999 ), the highest levels of steady-state AtSNF4 transcript were detected in flower buds, shoot apex and actively dividing cell cultures. Thus, if both AKINγ and AtSNF4 perform Snf4-like functions, the redundancy of these factors could provide a means for differential regulation of SnRKs in Arabidopsis.
AtSNF4 binds to Snf1, tobacco NPK5 and Arabidopsis AKIN10 and AKIN11 also under glucose repression and its interaction with these kinase catalytic subunits is only increased to about two- to fourfold during glucose depletion in yeast. These data suggest that considerable differences may exist between the regulation of yeast Snf1 and plant SnRKs. Nonetheless, the snf4 suppressor screen indicated a potential conservation of some regulatory functions between yeast and Arabidopsis. Suppression of the snf4 mutation by the zinc finger factors ZAT10 and AZF2 is thus intriguing because ZAT10 has been shown to activate salt-stress tolerance controlled in yeast by the STRE-binding Msn2 and Msn4 factors that are efficient snf4 suppressors ( Estruch & Carlson, 1993; Lippuner et al. 1996 ). Although the regulatory function of CK1 is still obscure in plants, the observation that the bifunctional serine/threonine and tyrosine specific casein kinase I (CK1, Ali et al. 1994 ) suppresses the snf4 mutation, as its yeast orthologs Yck1 and Yck2, also suggests a role for CK1 in SnRK-linked signaling. On the other hand, the finding that overexpression of Arabidopsis Hap3b confers an snf4 suppressor phenotype is less surprising, since Arabidopsis Hap2 cDNAs were isolated by complementation of the corresponding yeast mutation, and sequence analysis confirmed a remarkable conservation of their interacting Hap3 pairs ( Edwards et al. 1998 ). However, it remains a challenging task to clarify how overexpression of Arabidopsis orthologs of calmodulin, hexose/UDP-hexose transporters, SMC1-cohesin and Myb30 suppress the block of Snf1 activation in the snf4Δ2 mutant. Further research, including the isolation of second site mutations restoring the snf4 phenotype in the presence of heterologous suppressors and characterization of corresponding gene mutations in Arabidopsis, is in progress to answer this question.
Isolation of snf4Δ2 suppressors
Arabidopsis cDNA libraries made from cell suspension and seedlings in Ura3+-selectable expression vectors pYX112 ( Umeda et al. 1998 ) and pFL61 ( Minet et al. 1992 ) were transformed into the yeast strain MCY 1853 (Mata,snf4Δ2, his4-539, lys2-801, ura3-52, SUC2, Celenza & Carlson, 1989) using a standard LiCl transformation protocol ( Ausubel et al. 1999 ). The transformants were grown for 5 days on selective Ura– synthetic minimal SD-medium ( Rose et al. 1990 ), then harvested and plated at a density 105 cfu/plate on 30 plates containing SD-GEG-medium (SD-medium with 3% glycerol, 2%ethanol and 0.05% glucose). From colonies appearing on the plates during 10–14 days at 30°C, DNA was prepared as described previously ( Robzyk & Kassir, 1992) to rescue the cDNA expression vectors by transformation of E. coli strain MC1061 (hsdR, mcrB, araD139Δ(araABC-leu), 7679ΔlacX74, galU, galK, rpsL (thi) using ampicillin selection. To confirm and classify the snf4Δ2 suppressors, plasmid DNAs prepared from E. coli were transformed into yeast carrying either the snf1Δ10 (strain MCY 1846; Celenza & Carlson, 1989) or snf4Δ2 mutation. The transformants were selected on Ura– SD-medium, then tested for growth on SD-GEG-medium and SD-medium containing 2% glycerol, as sole carbon source. MCY 1846 and MCY 1853 strains, carrying the empty expression vectors pYX112 and pFL61, were used as negative controls. Strains harboring the same vectors with cDNAs encoding the Arabidopsis SnRK1-type AKIN10 and AKIN11 protein kinases served as positive controls. cDNAs of confirmed snf1Δ10 and snf4Δ2 suppressors were subjected to DNA sequencing using an automatic ABI377 sequencer (Applied Biosystem).
To create a fusion between the Gal4 activation domain (GAD) and AtSNF4 in the yeast two-hybrid prey vector pACT2 ( Durfee et al. 1993 ), the AtSNF4 cDNA sequence (in bold) was modified by PCR-amplification (25 cycles of 30 sec 90°C, 30 sec 58°C, 1 min 72°C, followed by 5 min at 75°C) using a primer pair 5′-ACCCGGGAATTCATATGGTTCCTGCTGGTTTTA-3′ and 5′-CGAGAATTCCCGGGTCAAAGACCGAGCAGGAATTG-3′. The PCR product was cloned in pGEM-T (Promega), sequenced and moved into pACT2 by EcoRI. To construct AKIN11 fusions with the Gal4 DNA-binding domain (GBD) in the two-hybrid bait vector pAS2–1 (Clontech), AKIN11 cDNA (sequences in bold; GenBank accession number X99279) deletions were similarly generated by PCR-amplification using the following 5′-primers with EcoRI and 3′-primers with SalI cleavage sites (codon position corresponding to Fig. 2c are underlined):
(5′, position 1-) 5′-ACCCGGGAATTCATGGATCATTCATCAAATAG-ATT-3′
(5′, position 98-) 5′-ACCCGGGAATTCGAGTATGTCAAGTCTGGA-GAG-3′
(5′, position 399-) 5′-ACCCGGGAATTCTCTCATGCGCATCCTCG-TGA-3′
(5′, position 469-) 5′-ACCCGGGAATTCTCACCCACTGTCATCAAA-TTTGA-3′
(3′, position − 66) 5′ ACCCGGGTCGACCACTTTCTCTCCATCTCC-ATGT-3′
(3′, position − 271) 5′-ACCCGGGTCGACGAACCAACGGTGTTGAC-GGAT-3′
(3′, position − 521) 5′-ACCCGGGTCGACTCAGATCACACGAAGC-TCTGT-3′
Detection of protein interaction in the yeast two-hybrid system
Protein interactions in the two-hybrid system were tested by LacZ filter-lift assays and monitoring growth on SD-medium containing 50 m m 3-aminotriazole after cotransformation of pAS2-1 (GBD, Trp+) bait and pACT2 (GAD, Leu+) prey plasmids into the yeast strain Y190 ( Durfee et al. 1993 ). The interactions were repeatedly assayed by mating of the yeast Matα strain Y187, carrying the pAS2-1 GBD-constructs, with strain Y190 (Mata), harboring the pACT2 GAD-plasmids ( Harper et al. 1993 ). Glucose regulation of protein interactions was assayed by measuring β-galactosidase enzyme activity in at least 12 independent transformants grown either in SD-medium with 2% glucose or in SD-GEG-medium containing 3% glycerol, 2% ethanol and 0.05% glucose as described previously ( Jiang & Carlson, 1996, 1997).
In vitro protein binding assays
Expression and purification of glutathione-S-transferase (GST) fusion proteins GST-AKIN10 and GST-AKIN11 were according to Bhalerao et al. (1999) . PCR-amplified AtSNF4 cDNA linked 5′-upstream to a T7 polymerase promoter was used as a template to synthesize 35S-methionine-labeled AtSNF4 protein in a coupled transcription-translation system (Promega). Equal amounts of 35S-AtSNF4 protein were incubated with glutathione–Sepharose coupled to GST-AKIN10, GST-AKIN11 and control GST proteins (5 μg of each) in 200 μl binding buffer (20 m m Tris–HCl, pH 7.5) for 1 h at 4°C. After removal of the supernatant, the beads were washed four times with 10 volumes of binding buffer, then the matrix-bound proteins were eluted with 4× sample buffer and, along with the supernatant fractions, separated on a 10% SDS-polyacrylamide gel to detect the 35S-AtSNF4 protein by autoradiography ( Németh et al. 1998 ).
DNA and RNA hybridizations
Arabidopsis genomic DNA was prepared according to Dellaporta et al. (1983) , whereas total RNA from cell suspension and different plant organs was isolated by guanidium thiocyanate extraction ( Sambrook et al. 1989 ). DNA and RNA samples (30 μg of each) were fractionated on native and formaldehyde agarose gels, blotted to Hybond-N (Amersham) filters and hybridized at high stringency with an AtSNF4 cDNA probe excised from the expression vectors as EcoRI fragment and labeled with an oligolabelling kit (Pharmacia) as described ( Sambrook et al. 1989 ). As loading control, the RNA blots were hybridized with α-actin Act2 cDNA probe. Southern hybridization at low stringency ( Fig. 1c) was carried out in the presence of 37% formamide at 37°C followed by washing the filters with 2 × SSC, 0.1% SDS at 37°C, 2 × SSC, 0.1%SDS at 42°C, and twice with 1 × SSC, 0.1%SDS at 42°C.
This article is dedicated to Prof. Jeff Schell with our appreciation for the unique chance to enjoy his continuous support, interest and scientific input in our work as members of his research group at the MPIZ Cologne. We thank Sabine Schäfer for her excellent technical assistance. This work was supported by the Deutsche Forschungsgemeinschaft (DFG Ko 1438/3–1) and a NFR grant provided to R.B.