A lysR-type regulator is involved in the negative regulation of genes encoding selenium-free hydrogenases in the archaeon Methanococcus voltae

Authors


Summary

The archaeon Methanococcus voltae encodes two pairs of NiFe-hydrogenase isoenzymes. One hydrogenase of each pair contains selenium in the active site, whereas the other one is selenium-free. The gene groups for the selenium-free hydrogenases, called vhc and frc, are linked by a common intergenic region. They are only transcribed under selenium limitation. A protein binding to a negative regulatory element involved in the regulation of the two operons was purified by DNA-affinity chromatography. Through the identification of the corresponding gene the protein was found to be a LysR-type regulator. It was named HrsM (hydrogenase gene regulator, selenium dependent in M. voltae). hrsM knockout mutants constitutively transcribed the vhc and frc operons in the presence of selenium. A putative HrsM binding site was also detected in the intergenic region in front of the hrsM gene. Northern blot analysis indicated that the hrsM gene might be autoregulated.

Introduction

The transcriptional apparatus of Archaea resembles that of Eucarya with regard to the RNA polymerase composition, transcription factors and promoter structure (review: Soppa, 1999). In contrast, transcriptional regulation has been found to be more similar to that of Bacteria in many cases (reviews: Bell and Jackson, 2001; Soppa, 2001) and genome searches have yielded a high number of genes encoding regulators with characteristic helix–turn–helix motifs typically seen in bacterial proteins involved in transcriptional regulation (Aravind and Koonin, 1999). As a mode of negative regulation, classical repression has been described, i.e. competition of a negatively regulating protein for the promoter (Cohen-Kupiec et al., 1997; 1999; Bell and Jackson, 2001; Lee et al., 2003; Lie and Leigh, 2003). Alternatively, binding of regulators downstream of the promoter elements was observed, suggesting that recruitment of transcription initiation factors might occur in the presence of the negative regulator but transcription would subsequently be blocked. Some of the negative regulators are autoregulated (Bell and Jackson, 2000; Brinkman et al., 2000; Enoru-Eta et al., 2000;2002). Positive bacterial-like transcriptional regulation has also been described (Brinkman et al., 2002; Ouhammouch et al., 2003). However, not all of the archaeal transcription regulators are similar to bacterial counterparts. In a well-studied case the activator resembles a eukaryotic leucine-zipper protein and is supposed to help recruiting the transcription complex (Krüger et al., 1998; Gregor and Pfeifer, 2001). Small effector molecules were identified which are necessary for the function of regulatory proteins, either as inducer (Lee et al., 2003) or for the binding of a repressor, such as metal ions (Bell et al., 1999) or an amino acid (Enoru-Eta et al., 2000).

We have studied the transcriptional regulation of two hydrogenase operons in the methanogenic archaeon Methanococcus voltae. Methanogenic archaea gain their energy by the stepwise reduction of C1-moieties to methane (reviews Weiss and Thauer, 1993; Thauer, 1998; Deppenmeier, 2002).

Most of these organisms can use CO2, which is reduced with the help of electrons gained from molecular hydrogen by the action of hydrogenases. Two types of NiFe-hydrogenases are involved in this process. One of the enzymes transfers the electrons to the deazaflavine coenzyme F420, whereas the electron-acceptor(s) of the other type, called F420 non-reducing, are not completely known. One reaction in which an enzyme of this type appears to be involved is a terminal step of methanogenesis, the heterodisulphide reductase reaction in Methanothermobacter marburgensis (Stojanowic et al., 2003). In M. voltae a selenium-containing F420 reducing hydrogenase plays a role in the same reaction (Brodersen et al., 1999).

In this organism two isoenzymes of both the F420 reducing NiFe-hydrogenase, called Fru and Frc, and of the F420 non-reducing hydrogenase, called Vhu and Vhc, are encoded in the genome (Halboth and Klein, 1992). One enzyme of each type, Fru and Vhu, contains selenocysteine in the hydrogen activating reactive site. The corresponding isoenzymes, Frc and Vhc, have a cysteinyl residue in the homologous positions. The two NiFeSe-hydrogenases are constitutively expressed. The operons vhc and frc encoding the selenium-free enzymes are only transcribed under selenium limitation (Berghöfer et al., 1994). They are connected by a common intergenic region comprising both promoters and positive and negative regulatory sequence elements which were defined by mutational analyses employing a reporter gene system (Noll et al., 1999). A putative activator protein has been identified (Müller and Klein, 2001). The relative positions of the genes and their regulatory elements are shown in Fig. 1.

Figure 1.

Relative position of the vhc and frc hydrogenase operons and the regulatory cis-elements. Top: the operons and their intergenic region are shown schematically. The promoters (triangles) are positively (+) or negatively (–) regulated via activator (grey boxes) or repressor (open box) binding sites. Bottom: the sequence of the 453 bp intergenic region is shown. The TATA box and initiator elements of the divergent promoter are given in underlined italics. The negative regulatory region as defined by deletion analysis is underlined. The activator binding sites defined by site-specific mutagenesis are highlighted by grey shading.

Up to now the negative regulator has been unknown. Here we report that it is a protein belonging to the LysR family of prokaryotic regulators (Schell, 1993). It is likely that the protein also regulates its own expression.

Results

Affinity purification of a LysR-type regulator specifically binding to DNA containing a regulatory cis-element involved in the regulation of hydrogenase promoters

As a binding site for the putative negative regulator of vhc and frc transcription had been identified we attempted to purify the protein by DNA-affinity chromatography employing matrix-bound oligonucleotides which either contained or lacked the binding sequence.

Figure 2 shows the protein patterns on SDS–polyacrylamide gels of eluates from affinity columns carrying DNA either comprising or lacking the sequence required as cis-element for the negative regulation. A prominent 33 kDa protein is noticeable only in the eluate from the binding site-containing column. The protein was excised from the gel and N-terminally sequenced. Data base searches yielded a conserved sequence in polypeptides from Methanocaldococcus jannaschii and Methanococcus thermolithotrophicus annotated as LysR-type regulators (Fig. 3)

Figure 2.

Eluted proteins from streptavidin agarose carrying DNA without (lanes 1) or with the negative regulatory sequence (lane 2) of the vhc/frc intergenic region. Portions of the two fractions were mixed in 1 + 2. A silver-stained SDS polyacrylamide gel is shown. Masses of marker proteins and their positions are shown on the left. The ∼33 kDa protein (marked by arrows) seen only in the fraction eluting from the column containing the putative repressor binding site was further analysed.

Figure 3.

Identification of homologous sequences from methanogenic archaea by comparison with the N-terminal sequence obtained from the excised 33 kDa protein band shown in Fig. 2. Closer inspection of the sequence from M. thermolithotrophicus (M. thermolith) indicated that the codon translated into valine (v) most probably serves as a GTG initiation codon, because a Shine–Dalgarno sequence is clearly discernible in the DNA sequence in front of the putative 5′-terminal part of the open reading frame.

Using primers derived from conserved amino acid sequences part of the M. voltae gene was amplified and a partial sequence with high sequence identity of the derived polypeptide sequence to the homologous proteins was evident. The total gene sequence and that of the intergenic region in front of the gene was then established using a random fragment genomic library of M. voltae as described in detail in the Experimental procedures. Figure 4 shows a comparison of the derived polypeptide sequence of the HrsM protein the homologous product of M. jannaschii, and the bacterial LysR-type regulators CysB from Klebsiella aerogenes and OxyR from Escherichia coli. The archaeal sequences showed very strong similarity to the bacterial consensus, from which the two bacterial sequences deviate much more.

Figure 4.

Alignment of the HrsM amino acid sequence and that of the M. jannaschii homologue with the consensus sequence of archaeal LysR-type regulators. This alignment is also compared with the consensus sequence of bacterial LysR-type regulators and two of its representatives, the CysB regulator protein from Klebsiella aerogenes and the OxyR protein from Escherichia coli. In the two consensus sequences identical, functionally similar and charged amino acids are given bold face capital letters. The helix–turn–helix domains (Schell, 1993) are shaded.

The HrsM protein is involved in the negative regulation of the frc and vhc hydrogenase promoters

In order to assess whether the protein was indeed involved in the transcriptional regulation of the vhc and frc operons insertion mutants were obtained in strain M. voltae V1 which contains a chromosomal uidAβ-glucuronidase reporter gene driven by the vhc-promoter preceded by the frc/vhc intergenic region (Pfeiffer et al., 1998) comprising all regulator sequences involved in the regulation of the vhc and frc promoters (Beneke et al., 1995). The obtained transformants were tested for the sites of integration of the insertion vector. As shown in Figure 5, site-specific integration was observed in different clones.

Figure 5.

Insertional knock-out of the hrsM gene. The insertion vector pNPAC- h which carries an internal fragment of the hrsM gene was used for the disruption of the gene on the chromosome by homologous integration as shown in (A). The expected change in the Hind III restriction fragment lengths compared to the wild-type pattern is indicated. The chromosomal DNA of puromycin-resistant clones was analysed and compared to the parent strain V1 by Southern hybridization after Hind III digestion, as shown in (B). The insert of the integration vector was used as the probe. Clones with the fragment lengths expected after gene disruption (NMV2, 3, 7) and clones with the wild-type fragment lengths (NMV1, 4, 5, and 6) were detected. They represent spontaneous resistant mutants. M, size marker. The properties of the insertion vector are described in detail in the Experimental procedures.

The reporter gene activity was then assayed in extracts from wild-type and mutant cultures grown with or without selenium. In the transformants with interrupted regulator genes activity was found under both conditions, whereas extracts from the V1 control strain showed only very little constitutive glucuronidase activity when the cells had been grown in the presence of selenium (Fig. 6). The spontaneous resistant mutants were also tested for glucuronidase activity in the presence of selenium. They behaved like wild-type cells (data not shown).

Figure 6.

Beta-glucuronidase activities in extracts from wild-type or transformed Methanococcus voltae V1 cells grown with or without selenium. Average results of four measurements each of specific glucuronidase activities are shown with standard deviations. Strain V1 contains the wild type hrsM gene which was disrupted in strains NMV 2, 3 and 7. Results obtained in extracts of cells grown without selenium are shown in dark grey. Light grey bars represent results obtained in cell extracts after cultivation in the presence of 10 µM sodium selenite. 100% equal a specific activity of 24.5 nmol nitrophenol µg protein−1 min−1.

It is noticeable that full activity was not reached in the extracts from transformant cells grown in the presence of selenium compared to wild-type or mutant extracts after cultivation of the cells under selenium limitation. This is in line with the previous assumption that the two promoters are both positively and negatively regulated (Noll et al., 1999), as it is not expected that the positive regulation would be affected by the knockout of the gene for the negative regulator.

In order to directly show the influence on the transcriptional activity of the hydrogenase promoters RT-PCR was performed with RNA isolated from wild-type or mutant cells with primers encompassing sequences of the frcA or vhcG genes, the promoter-proximal genes in the frc and vhc operons, respectively. The results (Fig. 7) confirmed that the knockout mutant led to derepression of the hydrogenase gene transcription in the presence of selenium.

Figure 7.

RT-PCR analysis of hydrogenase transcripts in RNA from wild-type and mutant cells grown with or without selenium. Total RNA from V1 wild type (lanes 1, 2, 4, 5) or the V1 derivative NMV2 carrying an insertion in the hrsM gene (lanes 3, 6, 7, 8) was subjected to RT-PCR after DNAse I treatment and heat inactivation of the DNAse. Primer pairs were used which bracket parts of the indicated genes encoding subunits (vhcG and frcA) of the two selenium-free hydrogenases or, as a control, of one of the selenium-containing hydrogenases (vhuG). The cells had been grown in the presence or absence of selenium as indicated. The expected product sizes were: vhcG, 625 bp; frcA, 946 bp; vhuG, 810 bp.

As expected, the mutation had no influence on the known constitutive transcription of the vhuG gene encoding a subunit of one of the selenium-containing hydrogenases. A control experiment with the wild-type strain V1 was also carried out separately. No change in the constitutive expression was seen in the presence or absence of selenium. Control reactions performed without reverse transcriptase yielded no products.

We attempted to directly analyse the binding properties of the HrsM protein. However, the protein eluting from the affinity column did not rebind to the vhc/frc intergenic region. The same was true for HrsM protein heterologously expressed as a histidine-tagged protein in E. coli and purified to homogeneity regardless of whether the his-tag was cleaved off (data not shown). We also attempted to obtain binding activity by adding a small molecular fraction of crude cell extracts from cells grown in the presence of selenium, which might have contained the binding cofactor(s). This approach also failed.

Indication for negative autoregulation of the HrsM protein

Because LysR-type bacterial regulator genes are frequently autoregulated, we analysed the region upstream of the hrsM gene. It is preceded by a 327 bp intergenic region which also contains a sequence with very high similarity to the putative HrsM binding site in the vhc/frc intergenic region (Figs 8). Deletion of the central oligo-dT-tract of this sequence in the vhc/frc intergenic region leads to partial derepression of the hydrogenase promoters (our unpublished results). This was an indication that the hrsM transcription might be regulated in a similar way as vhc and frc operons, which would suggest its selenium-dependent autoregulation. In order to check whether the expression of the gene was indeed dependent on the selenium supply a Northern blot analysis was performed using RNA from cells grown with or without selenite and employing a labelled hrsM probe. Figure 9 shows that the transcript indeed occurs in higher amounts in selenium-limited cells compared to those grown in medium containing selenite.

Figure 8.

Conserved sequence in the promoter regions of the genes encoding the selenium-free hydrogenases and the lysR-type regulator HrsM.

Figure 9.

Northern analysis of steady state levels of the hrsM transcript in cells grown in the presence of selenium or under selenium limitation as indicated. Total RNA from cells grown with or without selenium was separated on a denaturing agarose gel and transferred to a nylon membrane which was probed with a randomly 32P-labelled hrsM gene probe. The filter stained with methylene blue (left) shows the rRNA bands for calibration of the amounts of RNA in the two lanes. The autoradiogram of the filter after Northern hybridization is presented on the right side. Size marker positions are indicated.

Discussion

It has been known that M. voltae can adapt to selenium deprivation by derepression and activation of the transcription of the gene groups frc and vhc, encoding selenium-free isoenzymes for two types of NiFeSe- hydrogenases necessary in the reductive methanogenic energy generating pathway. Both positive and negative regulatory sequence elements were localized in the intergenic region connecting the two operons (Noll et al., 1999) and a putative activator has been described (Müller and Klein, 2001). However, the negative regulator remained elusive. Our present data show that it is a LysR-type regulator. It was purified on the basis of its binding specificity and, as predicted, a knockout mutant in the hrsM gene led to a derepression of the vhc and frc promoters. The binding site of the HrsM protein includes the dyadic sequence TNA-N7-TNA which is a classical motif of LysR-type regulator binding sites (Schell, 1993). In the present case a 6 bp inverted repeat separated by 7 bp is seen. Preliminary experiments in vitro have indicated that the HrsM protein can dimerize (data not shown). This could create a symmetric complex at the binding site, which would make the observed apparent symmetric negative regulatory action on both promoters from the central binding site in the intergenic region plausible.

The inspection of the M. jannaschii genome yielded a sequence similar to the conserved sequence in M. voltae, which the HrsM protein is supposed to recognize as its binding site. It is located in front of the hrsM homologue. However, the gene at the other end of the intergenic region in front of the M. jannaschii hrsM homologue is different. This would still allow for the assumed autoregulation of the LysR-type regulator in M. jannaschii. The hydrogenase operons encoding selenium-free enzymes of the Frc or Vhc type are not linked in M. jannaschii. Inspection of the upstream regions of these operons did not yield the conserved putative HrsM binding motif. Therefore, the function of the HrsM homologue in M. jannaschii can only be partially identical to that in M. voltae.

It is remarkable that the derepression in the hrsM mutant did not lead to the same level of promoter activity observed in the absence of selenium. This agrees well with the previous observation that the removal of the negative regulatory element by deletion also did not result in full activity of the concerned hydrogenase promoters. This observation led us to postulate additional activation of the transcription (Noll et al., 1999).

LysR regulators are supposed to have originated in the kingdom of Bacteria and to have entered Archaea by lateral gene transfer (Pérez-Rueda and Collado-Vides, 2001). The HrsM protein shows striking conservation of amino acid residues when compared to the consensus sequence of the bacterial proteins, both in the N-terminal helix–turn–helix domain as well as in the C-terminal domain frequently involved in binding of effector molecules. In comparison, the bacterial LysR-type regulators show much more variability, i.e. individual proteins deviate to a higher degree from each other and the consensus. This may reflect lesser diversification of the gene type in archaea, where LysR regulators are not abundant.

The function of LysR regulators is frequently dependent on effector molecules. Such an effector for HsrM is expected to be a signal substance possibly containing selenium, which has not yet been identified. This probably explains our finding that HsrM protein expressed in E. coli with a histidine tag and purified to homogeneity did not bind to DNA containing the defined negative regulatory sequence even after cleaving off the tag (our unpublished results). It has recently been shown that in the closely related organism M. maripaludis sulphur-containing isoenzymes of several selenoproteins are upregulated after inactivation of the selB gene (Rother et al., 2003). This gene encodes an essential translation factor for the incorporation of selenocysteine. The authors concluded that free selenium is not involved in regulation but rather a successional compound.

LysR-type regulator genes are frequently linked to genes which are under their control. These features have been thought to have facilitated their lateral gene transfer. We have indeed detected a gene that is linked to the hrsM gene by a common intergenic region and divergently transcribed. However, this linkage is not seen in the hrsM homologue in Methanocaldococcus jannaschii, a closely related methanogen. Many LysR-type bacterial regulators are autoregulated. Our finding that selenium deprivation of M. voltae resulted to an increased hrsM transcript level is in accordance with this notion. Rigorous proof will, however, need a future more detailed functional analysis of the hsrM promoter.

Experimental procedures

Strains and plasmids

The strains used in this study are listed in Table 1.

Table 1. . Strains used in this study.
Strain/plasmidRelevant genotypeConstruction, source or reference
M. voltae DSM 1537wild typeGerman Collection of Microorganisms
and Cell Cultures (Braunschweig)
M. voltae V1carries uidA gene linked to the vhc/frc
intergenic region and driven by the vhc promoter
Pfeiffer et al. (1998), laboratory stock
E. coli recA1 endA1 gyrA96 thi-1 hsdR17 supE44 Stratagene (Amsterdam)
XL-1 Blue relA1 lac[F′proAB lacIqZ M15 Tn10 (Tetr)] 

Accession number

The hrsM gene sequence is accessible (GenBank accession number: AY 352049)

Plasmids used

Figure 10 shows the vectors used for retrieval of the hsrM gene or insertion mutagenesis.

Figure 10.

Vectors used for retrieval of the hsrM gene (A) or insertion mutagenesis (B).The insertion vectors were based on a previous construct (Thomas et al., 2002). The E. coli vector pSL1180 carrying a ColE1 replication origin (Brosius, 1989) was used a backbone. The previously used puromycin resistance selection marker pac for M. voltae was replaced by a modified new N-pac gene with optimized codon usage for the use in organisms with high A + T contents (GenBank accession number AY438700). It is under the control of the S-layer gene promoter (Kansy et al., 1994) and the terminator of the methylreductase (mcr) operon (Müller et al., 1985) of M. voltae. The additional promoter of the M. voltae hmvA gene (Agha-Amiri and Klein, 1993) was introduced to avoid polarity after the insertion of the vector into the M. voltae chromosome via homologous recombination.

Primers used in this study

The primers used in this study are listed in Table 2.

Table 2. . Primers used in this study.
NameSequenceUsage
hvm5
hvm3
CGCAAGGTACCGTTAGTAATCA
GTATTTCACGAGTTCCTGAACC
Generation of internal hrsM fragments and amplification of
sequences out of genomic library vectors
pSLEcoRI-3′21GAGCGGATAACAATTTCACACAGGAAmplification of hrsM sequences out of genomic library vectors

General methods

Media and procedures for culturing M. voltae and selenium limitation were described before (Whitman et al., 1982; Berghöfer et al., 1994). Cloning and hybridization methods as well as polymerase chain reactions were performed as described earlier (Chomczynski, 1992; Sambrook and Russel, 2001). DNA sequencing was done according to the method of (Sanger et al., 1977) using a LICOR model 4000 sequencer with fluorescently labelled primers purchased from MWG Biotech AG (Ebersberg, Germany).

Reverse transcriptase PCR

Four µg total cellular RNA in H2O was treated with RNAse-free DNase I (20 units µg−1 RNA) in a total volume of 20 µl for 2 h at 37°C. The DNase was inactivated by heating at 75°C for 10 min. The RNA was then used as template in a reaction employing the one-tube AccessQuickTM RT-PCR System (Promega GmbH, Mannheim) according to the manufacturer's instructions.

Transfection of M. voltae

The method first described by Metcalf et al. (1997) was applied to M. voltae with modifications: cells were grown overnight to an OD600 of 0.5–0.8. All further steps were performed in an anaerobic chamber under H2/CO2/N2 (5/20/75% v/v) atmosphere. All media and buffers used were anaerobic and contained 1 µg ml−1 resazurin as a redox indicator. One ml of the exponentially growing culture was spun down and the supernatant was carefully removed. The cells were resuspended in 300 µl 0.68 M sucrose solution for protoplasting. The lyophilized DNA used for transformation was dissolved in 25 µl 20 mM Hepes pH 7.0. 1 µl DOTAP ESCORT liposome reagent (Sigma, Munich) per µg DNA employed was added to the DNA solution and the mixture kept for 15 min at room temperature. It was then gently mixed with the protoplast suspension and was kept in the anaerobic chamber for at least 1.5 h. The mixture was then inoculated into 10 ml of growth medium without puromycin and incubated at 37°C for at least 5 h before adding puromycin or plating on 10 µg ml−1 puromycin-containing plates. Colonies formed after 5–10 days.

Beta-glucuronidase assay

Beta-glucuronidase was assayed in crude cell extracts as previously described (Müller and Klein, 2001).

Retrieval of the hrsM gene

A genomic library in pNPAC (Fig. 9A) carrying random ∼ 600 bp Tsp 509I fragments in the Eco RI site was used in PCR reactions employing primers hvm3 or hvm5 in combination with primer pSLEcoRI-3′21 to obtain overlapping partial sequences covering the hrsM gene and adjacent sequences. The amplification products were sequenced and the total sequence obtained from the overlapping partial sequences.

Preparation of cell extracts

Extracts for the enrichment of the HrsM protein were prepared from cells grown under selenium limitation. 5 ml of an exponential culture of M. voltae were used to inoculate a 1 L culture which was grown to an OD600 of 1–2. The cells were collected by centrifugation and lysed by the addition of 10 ml H2O. The lysate was centrifuged for 30 min at 12 000 g and 4°C. The supernatant was frozen.

DNA affinity chromatography

Biotin-labelled double-stranded oligonucleotides were generated by annealing the biotin-labelled single-stranded oligonucleotide with the complementary non-labelled single strand.

In order to activate the carrier material, 2 ml of strepavidin agarose (Novagen, Madison, Wisconsin) were centrifuged at 6000 g for 1 min and washed three times in 1 ml of coupling buffer (10 mM Tris-HCl, 1 mM EDTA, 1 M NaCl pH 7.5). The biotin-labelled double-stranded oligonucleotide was bound to the pretreated material by mixing 2 ml of 50% streptavidin agarose in binding buffer, 100 µl of 50 pmol µl−1 oligonucleotide and 100 µl of 2 × binding buffer. The mixture was gently mixed for 2 h at room temperature on a rotary shaker and subsequently washed three times with 1 ml of binding buffer. The DNA agarose was suspended in 1 ml of protein binding (PB) buffer (20 mM Tris-HCl, 2 mM EDTA, 1 mM DTT, 200 mM NaCl, 10% glycerol, 0.1% Triton X100, pH 8.0) and incubated at room temperature with gentle shaking for 30 min.

Two millilitres of the DNA agarose were mixed with 10 ml of cell extract and the mixture was shaken for 20 min at room temperature. The agarose was spun out, resuspended in 10 ml of PB buffer and poured into a 1.5 × 14 cm column. The column was washed three times each with 10 ml of PB buffer containing 200 mM, 300 mM or 350 mM NaCl. The remaining protein was eluted from the column by applying four times 120 µl each of PB containing 1 M NaCl. The eluate was kept frozen at −20°C. Before use, it was desalted by spinning it through a Mobispin S-200 column (MoBiTec GmbH, Göttingen, Germany).

Conserved domain search

The analysis of the HrsM domain structure was performed using the NCBI conserved domain search program (http://www.ncbi.nlm.gov/BLAST).

Polypeptide sequencing

N-terminal polypeptide sequencing was performed at the service unit of the Institute of Physiological Chemistry and Pathobiochemistry of the University of Münster.

Acknowledgements

We thank Diana Kruhl for technical assistance and Dr S. Curtenaz for critical comments. This work was supported by the Deutsche Forschungsgemeinschaft (SFB 395) and Fonds der Chemischen Industrie.

Ancillary