Bacteria are in constant conflict with competing bacterial and eukaryotic cells. To cope with the various challenges, bacteria developed distinct strategies, such as toxins that inhibit the growth or kill rivals of the same ecological niche. In recent years, two toxin systems have been discovered — the type VI secretion system and the contact-dependent growth inhibition (CDI) system. These systems have structural and functional similarities and share features with the long-known gram-negative bacteriocins, such as small immunity proteins that bind to and inactivate the toxins, and target sites on DNA, tRNA, rRNA, murein (peptidoglycan), or the cytoplasmic membrane. Colicins, CdiA proteins, and certain type VI toxins have a modular design with the transport functions localized in the N-terminal region and the activity functions localized in the C-terminal region. Despite these common properties, the sequences of toxins and immunity proteins of colicins, CDI systems, and type VI systems show little similarity.
Bacterial toxins are more than prominent virulence factors
Protein toxins are mainly regarded as traits that determine the pathogenicity of bacteria. This applies undoubtly to toxins produced by Vibrio cholerae, Corynebacterium diphtheriae, Bacillus anthracis, and many other toxins that are able to kill their human hosts. Hundreds of bacterial toxins that affect eukaryotic cells have been described (Alouf, 2006). However, toxins also play a great role in intercellular communication between bacterial species. For example, colicin toxins of Escherichia coli only kill E. coli and do not affect eukaryotic cells (Cascales et al., 2007; Jakes & Cramer 2012). Toxins of the contact-dependent growth inhibition (CDI) systems mediate competition between neighboring bacterial cells (Aoki et al., 2010; Hayes et al., 2010). Toxins exported by the type VI secretion system (T6SS) inhibit both bacterial and eukaryotic cells (Cascales & Cambillau, 2012; Silverman et al., 2012). Because these three systems have in common a modular design–export from producing cells and subsequent import into target cells, immunity proteins protecting the producing cells, and the same targets–they will be briefly discussed in this overview with an emphasis on their similarities and differences.
These toxins differ from those belonging to the toxin–antitoxin (TA) systems, which act inside the producer cells. In this latter system, toxins are inactivated by antitoxins, which are either proteins or RNAs. As long as the antitoxins are bound to the toxins, the toxins do not harm cells. Stress-induced degradation of the antitoxins results in activation of the toxins, which kill cells or inhibit cellular growth by various means. The actions of the toxins have important roles in plasmid maintenance, formation of persister cells and biofilms, and multidrug resistance (Yamaguchi & Inouye, 2011).
It is not the intention of the review to provide a comprehensive description of the three systems. For a more detailed description, readers are referred to the literature cited in the references.
Colicin export, import, and mode of action
Colicins are proteins ranging in size from 30 000 to 80 000 kDa. They consist of an N-terminal translocation domain, a central receptor binding domain, and a C-terminal activity domain. The translocation domain is thought to be involved in uptake across the outer membrane, the receptor binding domain serves to attach colicins to the outer membrane receptor proteins, and the activity domain contains the active center for the enzymatic reactions and pore formation.
Colicins are released by the producer cells and enter into sensitive cells. They are encoded on plasmids along with an adjacent immunity gene and a lysis gene (Table 1). No specific colicin export system has been found. Colicins require the small lysis protein for release into the culture supernatant; the lysis protein is found in the membrane fraction. For the lysis protein to release the colicins, it must activate the outer membrane phospholipase A (OmpLA) (Cascales et al., 2007). The lysis proteins of the various colicins have a similar structure and are functionally interchangeable. A few colicin determinants do not encode a lysis protein. These colicins, for example colicins B, M, I, escape from cells by leakage.
Table 1. Common properties of colicins, the CDI system, and the T6SS system
In contrast to the nonspecific export of colicins, the import of colicins into sensitive cells is a highly specific process (Fig. 1). Import requires outer membrane receptor proteins to which colicins bind with high affinity. To release the colicins from their receptors and for translocation across the outer membrane, energy provided by the proton motive force of the cytoplasmic membrane is required. Energy coupling between the cytoplasmic membrane and the outer membrane receptor proteins is achieved by proteins that are integrated into the cytoplasmic membrane and extend into the periplasm. The two known energy-coupling devices are the Tol system for group A colicins (proteins TolA, TolQ, TolR, and TolB) and the Ton system for group B colicins (proteins TonB, ExbB, and ExbD). Colicins that are taken up via the Tol system are classified as group A colicins and those that are taken up via the Ton system belong to the group B colicins. Nuclease colicins use two proteins to cross the outer membrane, as exemplified by colicin E3 in Fig. 1a. After initial binding via its central receptor domain to the BtuB receptor, a second binding occurs to the OmpF porin via the unstructured N-terminal translocation domain. Regions of the translocation domain that interact with OmpF have been identified and shown to be located within the OmpF pore by crystal analysis. Then, the colicin binds to periplasmic TolB, which triggers binding of TolB to TolA in the cytoplasmic membrane. Through TolA, the still surface-bound toxin gains access to the proton motive force. Interestingly, dissociation of the immunity protein, which is exported together with the colicin, from the colicin requires the proton motive force (summarized and discussed in Papadakos et al., 2011).
In Fig. 1b uptake of a group B colicin is modeled. Colicin M binds to the FhuA outer membrane receptor. It is unlikely that it is taken up through the lumen of FhuA because in vitro binding of colicin M to FhuA does not expel the plug that tightly closes the β barrel, as evidenced by cysteines introduced into inside regions of the plug that do not become active upon addition of colicin M (V. Braun & F. Endriß, unpublished results). However, the plug must move as a cystine cross-link, which fixes the plug to the barrel abolishes all TonB-dependent FhuA activities (receptor for colicin M, phages T1 and Φ80, transport of ferrichrome, albomycin, rifamycin CGP4832, and microcin J25) but retains the TonB-independent T5 receptor function (Endriß et al., 2003). Movement may be required for exposure of the TonB box at the N-terminal end of FhuA, which is found in all TonB-dependent receptors and colicins. Colicin M and FhuA must interact with FhuA to kill cells (Pilsl et al., 1993). Interaction of the TonB boxes of colicin B and its receptor FepA with TonB (Mende et al., 1990) and colicin Ia and its receptor Cir with TonB (Buchanan et al., 2007) show that the uptake of probably all group B colicins involve a twofold interaction with TonB. Crystal structures of colicin receptor binding domains bound to the receptors show no movement of the plugs out of the receptor β-barrels (Jakes & Cramer, 2012). These findings suggest that the group B colicins are not imported through the lumen of the receptors but use the receptor–lipid interface. However, one has to bear in mind that in the crystal structures, no coupling to the proton motive force via TonB occurs. Movement of the plug out of the β-barrel was derived from cysteine-labeled FepA incubated with colicin B (Devanathan & Postle, 2007), which, however, was challenged by a later study in which FepA plug cysteines were not exposed by binding of colicin B (Smallwood et al., 2009). A recent finding brings the mechanism of group B colicin import closer to the mechanism of group A import. Colicin I uses the Cir receptor twice. An E3–Ia hybrid protein in which the receptor binding domain of Ia was replaced by the E3 binding domain requires the E3 BtuB receptor (Fig. 1a), but killing still depends on Cir and TonB (Jakes & Finkelstein, 2009). Despite great progress in identifying the proteins involved in translocation of colicins across the outer membrane and their interactions, the molecular mechanism of how the activity domains are translocated across the outer membrane to reach their targets in the periplasm, the cytoplasmic membrane, or the cytoplasm remains largely unknown (Kleanthous, 2010; Jakes & Cramer, 2012). Those colicins that form pores in the cytoplasmic membrane spontaneously insert into the cytoplasmic membrane and require an electrochemical potential that is positive outside. Uptake of colicins with targets in the cytoplasm involves proteolytic processing in the cell envelope and FtsH, an AAA+ ATPase/protease in the cytoplasmic membrane (Papadakos et al., 2011; De Zamaroczy & Mora, 2012), which in an ATP-dependent reaction dislocates membrane proteins into the cytoplasm and degrades them (Ito & Akiyama, 2005). Apparently, only the activity domain is required for colicin action in the cytoplasm.
The various colicins function in a variety of ways. Some are nucleases that degrade DNA, tRNA, or rRNA; some form pores in the cytoplasmic membrane and dissipate its electrochemical potential (Cascales et al., 2007); one hydrolyzes murein (pesticin) (Patzer et al., 2012); and colicin M inhibits murein biosynthesis (Braun et al., 2012) (Table 1).
Colicin-producing cells are protected by cosynthesis of small immunity proteins. Nuclease colicins in the cytoplasm are inactivated by the extremely tight and highly specific binding of their respective immunity proteins (Papadakos et al., 2011). Pore-forming colicins are inactivated in the cytoplasmic membrane by binding to their immunity proteins. Colicins affecting the murein structure are inactivated by their cognate immunity proteins in the periplasm (Braun et al., 2012).
Contact-dependent growth inhibition (CDI) systems
In contrast to growth inhibition by colicins, which are released into the culture medium and diffuse from the producer cells to the target cells, growth inhibition by CDI requires contact between the interacting cells (Aoki et al., 2005; Ruhe et al., 2013). Through CDI, a single E. coli EC93 cell can inhibit the growth of several hundred target cells within a few hours of co-culture. The species specificity of colicins is very high but that of the CDI system has not been sufficiently studied. Preliminary results indicate that the E. coli EC93 CDI system shows no activity against Salmonella enterica Typhimurium, and Enterobacter aerogenes (Aoki et al., 2011).
CDI toxins exhibit a modular design consisting of a very large conserved N-terminal domain, in some cases well over 3000 residues long, and a highly variable C-terminal activity domain comprising 230–360 residues. The N-terminal domain is exposed at the cell surface, and one may speculate that it serves not only to contact via the C-terminal domain BamA of target cells but acts more generally as adhesin and mediates self-agglutination of the producer cells and hemagglutination. This prediction is derived from the similarity between the predicted structure of CdiA and the crystal structure of FHA, the Bordetella pertussis hemagglutinin. The secreted FHA domain consists of stacks of repeated β-strands (Clantin et al., 2007), which is also predicted for CdiA proteins.
The CdiA toxins are secreted by the two-partner secretion system (Type Vb secretion), which involves the Sec machinery for secretion across the cytoplasmic membrane and a specific protein, designated CdiB, for secretion across the outer membrane (Fig. 2). Sensitivity of target cells to CdiA of E. coli EC93 requires BamA and AcrB in the target cells (Aoki et al., 2008). BamA is an outer membrane protein that is essential for outer membrane protein assembly. AcrB is a cytoplasmic membrane component of a multidrug efflux pump that forms a complex with AcrA and TolC and uses the proton motive force to export small toxic molecules. BamA seems to serve as receptor that binds CdiA at the cell surface. The function of AcrB is unclear. CdiA via BamA might open a pore in AcrB that dissipates the proton motive force which is observed in cells exposed to CdiA. Whether AcrB is also essential for growth inhibition by nuclease CdiA toxins (Table 1) with targets in the cytoplasm remains to be studied.
Translocation of the activity domain of the CdiA toxins into sensitive cells requires contact between producer cells and target cells, which is mediated by the N-terminal domain of the toxin and BamA. The next step, which follows the initial contact and involves uptake of the activity domain (toxin), is not known. It is assumed that the activity domain is released from the N-terminal domain by proteolysis and taken up into the cytoplasmic membrane or the cytoplasm of the recipient cells, where the targets are located. CdiA of E. coli EC93 was found to be proteolytically processed from 319 kD to fragments of 284 and 195 kD (Aoki et al., 2005). Unlike initially hypothesized, killing involves no signaling from the binding site at the cell surface into the interior of the cells, but rather requires import of the activity domains into cells to form pores or to degrade tRNA and DNA (Table 1).
Producer cells of CDI toxins are protected by small immunity proteins, designated CdiI, that specifically bind to the cognate toxin. There is no cross-immunity among CdiA toxins. Each CdiI protein binds to the C-terminal activity domain of the respective CdiA, which results in CdiA inactivation (Morse et al., 2012).
Although Cdi systems are widely distributed among gram-negative Alpha-, Beta- and Gammaproteobacteria, only certain strains of any given species contain cdiABI homologs (Aoki et al., 2010; Zhang et al., 2012). The size of CdiA proteins varies from 1400 to 2000 residues in Neisseria and Moraxella strains to over 5600 residues in Pseudomonas and Dickeya strains. The well-studied CdiA of E. coli EC93 consists of 3132 residues, of which the C-proximal 250 residues exhibit the toxic activity. A conserved peptide motif among CdiA proteins, VENN, is located 200–250 residues from the C-terminus and demarcates the transition from the relatively conserved N-terminal segment to the highly variable C-terminal segment, designated CdiA-CT. Functional CdiA chimeras may be produced by experimentally fusing heterologous domains. CdiA-CT of Yersinia pestis CO92 fused to the CdiA N-terminal domain of E. coli 536 kills E. coli 536 but not Y. pestis. Immunity to the chimera is conferred by CdiI of Y. pestis and not by E. coli 536 (Aoki et al., 2011). The host specificity is determined by the N-terminal domain, and immunity is specified by the C-terminal domain.
Downstream of full-length cdiB cdiA cdiI genes, many strains encode truncated cdiA genes and complete cdiI genes (Poole et al., 2011; Zhang et al., 2012). The CdiA fragments comprise the C-terminal domain up to the VENN motif. The truncated cdiA genes usually lack a translation initiation site but lead to toxicity when cloned in an expression vector. The truncated CdiA proteins are specifically inactivated by the immunity proteins, which are encoded downstream of the cdiA fragment genes. The cdiA fragment genes show sequence similarity to the full-length cdiA of the same strain. In addition, sequence similarity to cdiA genes of other bacteria is quite frequent. For example, one of four CdiA-CT fragments of Yersinia pseudotuberculosis is 88% identical over 111 residues to a CdiA-CT of Citrobacter rodentium ICC168, and 87% identical over 107 residues to CdiA-CT of E. coli AO 34/86 (Poole et al., 2011). The silent truncated cdiA genes might serve as repertoire for novel cdiA genes created by recombination with complete cdiA genes in the expression loci.
CDI systems have also been found in the gram-positive genera Bacillus (Table 1), Listeria, Clostridium, and Streptococcus (Holberger et al., 2011). Single strains may encode several CdiA homologs. For example, Bacillus subtilis 168 encodes six predicted CdiA proteins, which share conserved N-terminal regions but are variable in the last 120–150 C-terminal residues. The N-terminal regions contain sequences that have sequence similarity also to CdiA proteins of other species, for example, between E. coli O175:H7 strain EC869, Listeria innocua, and Bacillus pumilus.
Several of the CdiA proteins from the gram-positive species have RNase activity. When cloned into E. coli, these proteins inhibit growth provided they are not specifically inactivated by their cognate CdiI immunity proteins. The cdiI genes are encoded adjacent to the cdiA genes. The basic design of the CDI systems of gram-positive and gram-negative bacteria is genetically and functionally similar. However, the gram-positive CdiA proteins are shorter than those of the proteobacteria, for example, YobL of B. subtilis consists of 600 residues.
It is not known whether gram-positive CdiA proteins are secreted. If secretion occurs, it differs from secretion of the gram-negative CdiA proteins. The gram-positive Cdi systems lack CdiB, which functions in two-partner secretion (TPS). The N-terminal CdiA regions are related to the ESAT-6/WXG100 superfamily. The ESAT-6/WXG100 motif is a secretion signal for the type VII secretion systems of mycobacteria and bacilli (Zhang et al., 2011). It will be interesting to learn whether the CdiA proteins are anchored to the cell wall of gram-positive bacteria, whether the N-terminal domain is used for anchoring, whether the CdiA proteins are secreted, and whether host proteins function in CdiA import into target cells.
Many of the gram-positive and gram-negative CdiA toxins are nucleases (Table 1). The activity resides in the C-terminal segment that is highly polymorphic, in contrast to the N-terminal regions, which display similarities among different species. Recent comprehensive genomics predict many additional enzymatic functions for CDI toxins, such as deaminases, ADP-ribosyl transferases, and variety of peptidases (Zhang et al., 2012).
Toxins secreted by the type VI secretion system (T6SS)
The type VI secretion system T6SS was defined in a Vibrio cholerae infection model study investigating the cytotoxicity of V. cholerae strains for amoebae of Dictyostelium discoideum (Pukatzki et al., 2006). Virulence required the extracellular translocation of proteins lacking the N-terminal hydrophobic leader sequences via T6SS. This system also mediated cytotoxicity to a mammalian macrophage cell line. T6SS is also widely distributed in bacterial strains not known to be pathogenic. In these bacteria and in pathogenic bacteria, T6SS mediates cytotoxicity against related and unrelated bacteria. Besides its self-killing, V. cholerae kills certain strains of E. coli, Salmonella Typhimurium, and Citrobacter rodentium but not Pseudomonas aeruginosa PAO1 (MacIntyre et al., 2010). Pseudomonas aeruginosa kills E. coli but not B. subtilis. The toxins delivered by the T6SS of Serratia marcescens Db10 kill E. coli, Pseudomonas fluorescens, and Enterobacter cloacae (Murdoch et al., 2011).
T6SS translocates toxic effector proteins directly from donor cells into recipient cells by physical contact between the two cell types. Bacterial host specificity may be determined by the proper delivery of the effector proteins to their recipient targets. It is presently unknown whether specific receptors are involved, but this might be the case if the T6SS translocation mechanism is similar to the phage infection mechanism (see below). Phages require highly specific receptors, particular proteins, lipopolysaccharides, and teichoic acids, which occur only at the cell surface of bacteria. Resistance is most likely not caused by the lack of substrates in target cells because these substrates occur in all cells (Table 1).
The structure and function of only a few toxins have been identified. The VgrG1 protein confers toxicity of V. cholerae to eukaryotic cells. It carries a C-terminal extension (evolved VgrG1) that irreversibly cross-links actin. The extension is not required to kill E. coli by cell lysis (MacIntyre et al., 2010). VgrG1 of Aeromonas hydrophila ADP-ribosylates actin. VgrG1 forms the tip of the secretion apparatus and is released either into the medium or into recipient cells. It serves a dual function, it is part of the T6SS apparatus, and it is toxic to eukaryotic cells. Antibacterial effector proteins of T6SS have been identified in P. aeruginosa. Of the three effector proteins Tse1–Tse3 in this organism, Tse1 acts as a murein (peptidoglycan) amidase that cleaves the γ-d-glutamyl-l-meso-diaminopimelic acid bond, and Tse3 hydrolyzes the N-acetylmuramic acid-N-acetylglucosamine bond (Russell et al., 2011). Tae3 of Salmonella Typhi and an amidase of Burkholderia thailandensis cleave between meso-diaminopimelic acid and d-alanine. Tae4 of Salmonella Typhimurium hydrolyzes the γ-d-glutamyl-meso-diaminopimelic acid bond (Russell et al., 2012). The Ssp2 protein of Serratia marecescens expressed in E. coli inhibits growth of E. coli. The cognate Rap2 immunity protein is localized in the periplasm of S. marcescens (English et al., 2012), which suggests that the toxin affects murein in the periplasm. In general, the toxin producer cells are protected from being harmed by the toxins by immunity proteins that very specifically bind and inactivate the toxins.
The T6SS gene locus encodes 13 proteins that form the T6SS apparatus, which spans the cell envelope of the toxic cells. Exposed to the cell surface are two proteins, Hcp and VgrG, which are secreted by T6SS (Fig. 3b). They share structural homologies with the tail and puncturing device of bacteriophage T4 (Fig. 3a). Hcp is a homolog of the phage tail tube protein, and VgrG is a homolog of the phage spike (needle) complex. When T4 infects cells, the tail sheath is contracted and propels the tail tube toward the cell interior. The VipA/VipB proteins of V. cholerae T6SS form tubules that morphologically resemble contracted phage tail sheath (Leimann et al., 2009). VipA forms long straight structures in the cytosol; formation of the fibers depends on the presence of VipB (Basler et al., 2012). VipA sheaths assemble in c. 20–30 s μm−1. They contract in < 5 ms to about 50% of the extended length. Then, they disassemble in 30–60 s, and disassembly requires ClpV and ATP. Released VipA quickly reassembles to form new extended sheaths. In the general T6SS nomenclature, VipA is named TssB and VipB is named TssC. Contraction of the TssBC proteins provides the force for pushing the Hcp tube outside the cell. The contraction is powered by the ClpV AAA+ ATPase. Type VI protein secretion involves a highly dynamic apparatus that functions analogous to DNA translocation by contractile phage tails (Fig. 3b). However, the orientation of the two processes differs topologically. Phages attack cells from the outside and transfer DNA into the cytosol, whereas the T6SS apparatus is located in the cytosol and in the cell envelope and translocates proteins into the culture medium and into recipient cells.
Further similarities across the systems
Besides the discussed common modular nature of the toxins, some of their structures have a few additional similarities. R-type pyocins of P. aeruginosa belong, like colicins, to the bacteriocins. They resemble, like the T6SS protein complex, contractile phage tails (Nakayama et al., 2000; Michel-Briand & Baysse, 2002). The genes encoding the R-type pyocins are sufficiently similar to those of the phages P2 and ΦCTX that they can be assigned to the phage tail proteins, from which their morphological role can be derived. The genomic island SPI-21 of a Salmonella strain encodes a so-called evolved VgrG1 T6SS protein that contains a C-terminal extension similar to the S-type pyocins of P. aeruginosa (Blondel et al., 2009). S-type pyocins are single proteins that degrade DNA (S1–S3), tRNA (S4), or form pores (S5). BTH_i2691 of Burkholderia thailandensis and colicin I have sequence similarity (Russell et al., 2012). Some CDI toxins contain sequences with similarities to the DNase colicins E7, E8, and E9 (Zhang et al., 2011). The C-terminal 96 residues of CdiA E478 of Burkholderia pseudomallei shares 56% sequence identity with the tRNase domain of colicin E5 (Nikolakakis et al., 2012). These examples suggest an evolutionary relationship between these toxin systems.
The recently discovered antibacterial protein toxins of the CDI and T6SS systems add to the huge number of highly diverse bacterial toxins that affect prokaryotic and eukaryotic cells. In contrast to the toxins that act inside the producer cells (Yamaguchi & Inouye, 2011), the secreted toxins are frequently composed of several domains that are each specifically involved in uptake into target cells or their killing. The toxins discussed in this review — colicins, the CDI system, and the T6SS system — have many characteristics in common and some differences. All inhibit growth and eventually kill bacteria. Colicins, CdiA proteins, and specialized (evolved) VgrG proteins have a modular structure. The uptake domains are localized to the N-terminal segment, and the activity (killing) domains are localized to the C-terminal segment. The C-terminal fragments of the nuclease colicins and the nucleases of the CDI systems are proteolytically released from the complete proteins and are taken up into the cytoplasm. Immunity proteins protect the producer cells from being killed by the toxins. The immunity proteins are small and specifically bind to the activity domains and inhibit the toxic functions. The genes encoding the immunity genes are adjacent to the activity genes. One strain may encode more than one toxin of the same kind. The toxins kill cells by hydrolyzing DNA, RNA, or murein, or by forming pores in the cytoplasmic membrane. The toxins differ in that colicins are encoded on plasmids, whereas the toxins of the CDI and T6SS systems are encoded on genomes. They further differ in their release from producer cells. Colicins are not exported by specific secretion systems. Colicins escape from cells by partial cell lysis, in most cases facilitated by lysis proteins encoded adjacent to the colicin activity and immunity genes. From those colicins, which are not synthesized together with lysis proteins, only a fraction is released from cells. Nevertheless, they are found in the culture supernatant of cells and act from there. In contrast, toxins of the CDI and T6SS systems are transferred from cell to cell upon contact. During cell contact, activated T6SS within one cell triggers T6SS activation in its neighbors (LeRoux et al., 2012). Some toxins of the CDI and T6SS systems can be found in the cell culture supernatant. CDI uses the two-partner secretion mechanism (type Vb secretion), which involves the Sec translocon and a specific outer membrane protein, CdiB. T6SS uses a complex protein apparatus similar to the phage T4 infection apparatus. The T6SS apparatus contracts like the phage infection apparatus, and the puncturing device is used for secretion of the toxins from the producer cells and for entering into target cells.
New toxins may be created by duplication of activity and immunity genes and subsequent recombination within one toxin system but also in rare cases between different toxin systems. Recombination between toxin genes leads to a great toxin variety. This is particularly obvious for colicins, where the exchange of DNA fragments that encode receptor binding domains, translocation domains, or activity domains yield new colicins (Braun et al., 2002; Cascales et al., 2007). Increase of toxin diversity by creation of different toxins and various means to transfer the toxins into competing cells increases the ability of the toxic bacteria to compete with other bacteria. Competition among cells creates a strong selective pressure on toxins and their immunity proteins for diversification resulting in a huge variety of toxins (Zhang et al., 2012).
We thank Andrei Lupas for the generous hospitality in his department. The authors' work was supported by the Max Planck Society, the Deutsche Forschungsgemeinschaft (BR330/25-1), and the Fonds der Chemischen Industrie.