SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. Note added in proof
  9. References
  10. Supporting Information

Formation of a Bacillus subtilis biofilm community requires an abundant matrix protein, TasA and an exopolysaccharide. The transcriptional regulatory pathways that control synthesis of these structural features are complex and responsive to multiple physiological and population signals. We report herein that an additional layer of co-transcriptional regulation is required for exopolysaccharide (eps) expression. This mechanism is mediated by a novel cis-acting RNA element, coined ‘EAR’, located between the second and third gene of the eps operon. The presence of the EAR element within the eps operon is required for readthrough of distally located termination signals. We also find that the EAR element promotes readthrough of heterologous termination sites. Based upon these observations, we hypothesize that the EAR element associates with RNA polymerase to promote processive antitermination, a process wherein the transcription elongation complex is altered by accessory factors to become resistant to pausing and termination signals. It is likely that this mechanism is required for eps expression to ensure full synthesis of the unusually long transcript (16 kb). We also identify the EAR element in other species within the order Bacillales, suggesting that a similar mechanism is required for synthesis of biofilm and capsular polysaccharide operons in other microorganisms.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. Note added in proof
  9. References
  10. Supporting Information

Many, if not most, bacteria can form tightly associated multicellular communities called biofilms that adhere with surfaces or are formed at liquid-air interfaces. Key to formation of a bacterial biofilm is an extracellular matrix, typically comprised of exopolysaccharides (EPS), specialized proteins and occasionally DNA, which together encase the organisms and permit a surface-associated lifestyle (Branda et al., 2001; Sutherland, 2001; Branda et al., 2005; Aguilar et al., 2007). In a Bacillus subtilis biofilm community, only a subset of cells is able to turn on the expression of genes necessary for production of exopolysaccharides due to a paracrine signalling pathway mediated by a cyclic lipopeptide, surfactin (López et al., 2009c). Other cells are likely to differentiate into distinct cell types that together contribute to the survival of the community (López et al., 2009b). In this organism, a single operon (eps) encodes for synthesis of exopolysaccharides, although the individual chemical constituents of the polymer have not yet been identified. This operon, along with a separate transcript (yqxM-sipW-tasA) that encodes for synthesis and transport of a matrix protein, TasA, are subject to temporal and spatial control within the biofilm community (Branda et al., 2006; Vlamakis et al., 2008; López et al., 2009b). Multiple overlapping signalling pathways are co-ordinated for control of transcription initiation of these operons (Kearns et al., 2005; Chu et al., 2008; Chai et al., 2009; Winkelman et al., 2009). The master regulator for biofilm formation, SinR, directly represses both operons. Under the appropriate conditions for initiating biofilm formation, SinR repression is relieved by expression of multiple antagonist proteins (e.g. SinI and SlrA) that sequester SinR, thereby resulting in the production of exopolysaccharide and TasA. sinI expression is triggered by low levels of phosphorylated Spo0A, a global regulator of gene expression that is responsive to nutrient availability, quorum sensing and the production of secondary metabolites (Bai et al., 1993; Perego and Hoch, 1996;Ratnayake-Lecamwasam et al., 2001; Fujita and Losick, 2005; Kearns et al., 2005; López et al., 2009a). SlrA is part of a minor pathway stimulating matrix production and its expression is negatively controlled by a separate factor, YwcC (Kobayashi, 2008; Chai et al., 2009). A separate repressor, AbrB, also affects eps expression (Hamon et al., 2004; Chu et al., 2008) by binding directly to the promoter region (Murray et al., 2009). Phosphorylated Spo0A represses AbrB expression, thereby relieving AbrB repression of eps under biofilm conditions.

In this manuscript, we show that an additional layer of co-transcriptional genetic control is required for expression of eps genes. We report the discovery of a conserved RNA element located between the second and third genes of the eps operon, which we coin ‘EAR’ (eps-associated RNA). The EAR element is unique from other conserved RNA motifs that have been discovered (e.g. Gardner et al., 2009) and exhibits a conserved overall secondary structure architecture interspersed with conserved primary sequence determinants. It can be identified in other Bacillus species as well as a few additional genera of the order Bacillales, and is always located within biofilm or capsular polysaccharide operons. Moreover, our data demonstrate that the EAR element is required for eps expression in B. subtilis.

Bacteria are known to employ both small noncoding trans-acting regulatory RNAs (Gottesman et al., 2006) as well as cis-acting regulatory RNAs for transcriptional and post-transcriptional gene regulation (Winkler, 2005). The latter are typically located in the 5′ leader region of mRNAs or within intergenic regions of multigene transcripts, where they can fold into two mutually exclusive structural configurations that differentially affect downstream expression. The equilibrium between the conformational states is usually influenced by a cellular signal, such as a specific protein, unrelated RNA, small molecule metabolite, metal ions or a change in temperature (Gollnick and Babitzke, 2002; Stulke, 2002; Henkin, 2008; Dambach and Winkler, 2009; Gutierrez-Preciado et al., 2009; Klinkert and Narberhaus, 2009; Roth and Breaker, 2009; Ramesh and Winkler, 2010). Most often, cis-acting regulatory RNAs control expression of a proximally located downstream gene by affecting accessibility of the ribosome binding site or by influencing formation of a single adjacent transcription termination signal (Winkler, 2005), although one class of metabolite-sensing RNAs (i.e. ‘riboswitch’) regulates mRNA stability (Collins et al., 2007). We report herein that the ‘EAR’ element is similar with these signal-responsive cis-acting regulatory RNAs in its overall structural complexity, but employs an uncommon mechanism for control of transcription termination.

There are two general mechanisms for control of transcription termination. For the first class, typically referred to as ‘attenuation’, formation of an individual terminator signal is controlled via mutually exclusive secondary structure arrangements. The ‘equilibrium’ between these secondary structures is generally influenced by the presence or absence of an effector ligand. For the second class, first discovered for regulation of bacteriophage λ genes (Roberts, 1969; Condon et al., 1995; Roberts et al., 2008), RNAP is modified to become resistant to downstream terminator and pause sites (Weisberg and Gottesman, 1999; Nudler and Gottesman, 2002; Greive and Von Hippel, 2005; Roberts et al., 2008). Because this results in readthrough of multiple terminator sites over long stretches of DNA, it is generally referred to as ‘processive antitermination’. For these mechanisms, short DNA or RNA determinants assist assembly of multiple protein accessory factors with the transcription elongation complex (TEC) that together promote readthrough of termination sites (Fig. S1). The most thoroughly analysed examples affect transcription of the phage λ genome and rRNA genes (rrn) in E. coli. Such mechanisms have only been poorly characterized in non-proteobacterial species. Indeed, only five processive antitermination classes have been described overall and more than a decade has passed since the last new example. Interestingly, a λ-like phage, HK022, employs a two-stem RNA motif of approximately 65 nucleotides in length (‘put’) that alone is capable of inducing antitermination activity (King et al., 1996; Weisberg and Gottesman, 1999; Fig. S1). Also, a DNA-binding protein, RfaH, has been demonstrated to cause processive antitermination of certain operons in proteobacterial species, including genes required for lipopolysaccharide synthesis (Artsimovitch and Landick, 2002). The EAR element is responsible for a new processive antitermination mechanism class, which shares features with the latter two classes (put and RfaH antitermination) in that it is an RNA element that promotes antitermination of polysaccharide genes. Specifically, we find that the EAR element provides readthrough of distally located termination sites within the eps operon and is also capable of promoting readthrough of heterologous terminators. In the absence of this antitermination mechanism eps transcription is truncated at intermediary points of the operon, resulting in an exopolysaccharide-deficient phenotype. Therefore, we hypothesize that a key function of this mechanism is to ensure completed synthesis of the long eps transcript, where each gene is likely to be required for EPS production. Given that the EAR element can be found in other species within the order Bacillales, it is likely that EAR-mediated antitermination is a common requirement for synthesis of biofilm and capsular polysaccharide operons in these microorganisms.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. Note added in proof
  9. References
  10. Supporting Information

Identification of a conserved RNA element located within exopolysaccharide operons

During the course of a bioinformatics-based search for conserved portions of bacterial intergenic regions (data not shown), we identified a particularly intriguing candidate region. It consists of a ∼120-nucleotide region that can be found in a subset of Bacillales, mostly restricted to the Bacillaceae with a few examples in Paenibacillaceae (Fig. 1A). It does not resemble any previously established RNA motifs, such as those catalogued by the Rfam database (Gardner et al., 2009). The motif is nonrandom with regards to its genomic distribution; each individual example is located within or upstream of an operon encoding for biofilm or capsular polysaccharide genes (Fig. 1C). When the motif is located within an intergenic region of a polysaccharide operon, it is always upstream of the biosynthesis genes but downstream of a few uncharacterized putative regulatory genes. Such is the arrangement for B. subtilis, where it is situated within a ∼250 nt intergenic region between the second and third gene (epsB-C) of the 15-gene eps operon (Fig. 1C). For these reasons we designate the motif as the ‘EAR’ element, for eps-associated RNA.

image

Figure 1. Comparative sequence analysis of EAR elements. A. Alignment of EAR elements from Bacillales. Regions exhibiting base-pairing potential are highlighted by different colours (P1-P5). ‘PK’ indicates the presence of a potential pseudoknot. The consensus sequence and secondary structure arrangement are depicted within the last two rows. B. Sequence and secondary structure of the Bacillus subtilis EAR element as predicted by sequence analysis and structural probing (see Fig. 3), located within the epsB-C intergenic region. Red positions are conserved in at least 95% of sequences. Variable loop regions are marked by grey semicircular lines. C. Representative arrangements of EAR elements and their nearest genes from several species, including: B. subtilis, B. cereus, B. halodurans, G. kaustophilus and A. flavithermus. The EAR element is denoted by black circle (●). All eps genes are shown as white rectangles and are labelled by the appropriate COG category (Cluster of Orthologous Group).

Download figure to PowerPoint

The EAR element adopts a complex secondary structure

Comparative sequence analysis revealed many residues that are likely to be conserved at the level of primary sequence, mostly located within helical portions (Fig. 1). Sequence analyses also revealed many co-varying residues that when considered together provide strong evidence for the presence of base-paired segments. From these data we proposed a secondary structure arrangement consisting of five helical segments (P1–P5) and a pseudoknot at the 5′ terminus (Fig. 1B).

To further test this secondary structure model, we performed in-line probing analyses of the B. subtilis EAR element in vitro. This structural probing assay takes advantage of the natural instability of RNA wherein single-stranded or flexible regions are more susceptible to spontaneous cleavage compared with base-paired regions (Soukup and Breaker, 1999; Nahvi et al., 2002; Wakeman and Winkler, 2009). The analysis resulted in probing patterns that closely supported the secondary structural model (Fig. 2). One notable exception is the conserved pseudoknot (PK) region, which appeared to exhibit overall internucleotide ‘flexibility’ in a manner similar to bulged residues and single-stranded regions, suggesting that it is not well-formed for purified molecules in vitro (Fig. 2A). Also, the L1 terminal loop contains a stretch of five nucleotides (nt 24–28), adjacent to the pseudoknot, that appear to be highly conserved at the primary sequence level (Fig. 1). However, it is unlikely that these conserved residues participate in formation of an intrinsic secondary structural element for RNA molecules in vitro given that their spontaneous cleavage frequencies resembled those of bulged or otherwise ‘unstructured’ residues. From this we speculate that the conserved L1 residues may be required in vivo for a functional role other than simple secondary structure formation, such as assisting intermolecular interactions with other cellular components. Similarly, the putatively pseudoknotted region may be structurally enhanced in vivo, perhaps in response to additional factors. We also note from our in vitro structural probing data that the P3 and P4 helices appear to be modestly less structured than the P1 and P5 helices (Fig. 2A and data not shown). Nonetheless, the probing data overall support the secondary structure as proposed by comparative sequence alignments. Finally, these data also demonstrated that mutation of the P3 helix (M3) does not perturb the global secondary structure arrangement in vitro, and instead locally alters the P3 and L3 components. Similarly, mutation of the L3 terminal loop (M4) has essentially no effect on the overall secondary structure arrangement in vitro, including the P3 helix (Fig. 2B and C). We therefore considered these particular mutants to be particularly useful (as described below) in that they reflect mutagenic alteration of conserved residues (M3) and nonconserved residues (M4), while maintaining an overall common secondary structure arrangement.

image

Figure 2. In-line probing of wild-type and mutant EAR molecules. In this assay, bands qualitatively correlate with unpaired residues. These data suggest that the overall global configurations do not differ between wild-type, M3 and M4 molecules. A. Quantification of the in-line probing analysis. Nucleotide position is shown on the x-axis as depicted in the secondary structure, while y-axis represents the degree of cleavage at a particular position as calculated using SAFA (Das et al., 2005). B–C. The band patterns overall are nearly identical for wild-type, M3 and M4, with the exception of the highlighted region between G49 and G73 (B). This region of the molecule was mapped to the P3 helix, the location of the M3 and M4 mutations. Specifically, the M3 construct exhibited a significant disruption of the P3 helix compared with wild-type (notice the appearance of bands between G49 and G73), while the M4 mutation, which is located within the P3 terminal loop, showed no significant changes of the P3 helix (C).

Download figure to PowerPoint

The EAR element is important for EPS production

To test whether the EAR element is functionally required for expression of downstream genes, we dissected the 126 nt region encompassing the EAR element from the B. subtilis epsB-C intercistronic region. For this experiment we utilized a genetic approach that would not incorporate any additional genomic changes, such as introduction of antibiotic resistance cassettes. Also, because common laboratory strains of B. subtilis are unable to form biofilms (Aguilar et al., 2007; Earl et al., 2008), all genetic changes were introduced into the most frequently utilized ‘undomesticated’ strain, NCIB3610. This revealed that deletion of the EAR element resulted in decreased biofilm formation, as indicated by loss of pellicles (floating biofilm communities) and colony rugosity (Fig. 3A–C). Similarly, site-directed mutagenesis (e.g. M3, M5, M8) of conserved nucleotides and helical portions also resulted in a biofilm-deficient phenotype. In contrast, mutagenesis (M4) of a nonconserved terminal loop (L3) had no effect. Compensatory mutations (M6, M7) were then introduced into the RNA element in an attempt to restore base-pairing integrity to helices that had been disrupted by mutagenesis while still maintaining an altered primary sequence. Restoration of helical integrity appeared to be sufficient to restore biofilm formation at one location (P1) but not another (P3), suggesting that a combination of secondary structure and primary sequence is likely to be important for EAR function. Scanning electron microscopy of biofilm communities (e.g. Branda et al., 2006) was also used to qualitatively assess whether mutation of the EAR element appeared to affect EPS production. Electron micrographs of wild-type and the neutral mutant, M4, revealed the presence of a coating consistent with biofilm EPS, whereas the deleterious M3 mutant exhibited an apparent lack of this material (Fig. 3D). Thus, the EAR element is functionally required for EPS production, presumably through control of downstream gene expression.

image

Figure 3. The EAR element is required for B. subtilis biofilm formation and exopolysaccharide production. A. Secondary structure arrangement of the B. subtilis EAR element. Coloured in red are positions that are conserved in at least 95% of sequences. Regions that can form base pairing interactions are labelled ‘P1’ through ‘P5’. ‘PK’ denotes a potential pseudoknot. Positions of site-directed mutagenesis (labelled as ‘M3’ through ‘M8’) are denoted in blue. B. Complex colony morphology of bacterial strains containing mutations of the EAR element. Complex colony architectures are characteristic of biofilm communities (Branda et al., 2001; Aguilar et al., 2007). C. Top-down view of pellicle formation for wild-type B. subtilis NCIB 3610 and ΔEAR. D. Scanning electron micrographs of the surface of biofilms for the wild-type strain and cells containing mutations in the EAR element.

Download figure to PowerPoint

The EAR element exerts regulatory control over distally located genes

It is generally expected that cis-acting regulatory RNAs alter expression levels of an immediately downstream gene in response to specific stimuli (Irnov et al., 2006). To assess the importance of the EAR element on expression of the most proximal downstream gene, epsC, we measured the relative abundances of eps genes for wild-type, ΔEAR, M3 and M4 strains (Fig. 4A and B). Surprisingly, mutation or deletion of the EAR element resulted in virtually unchanged levels of the most proximal downstream gene, epsC. Furthermore, levels of epsD and epsE remain unchanged for all strains. In contrast, more distally located genes (epsF, epsH and epsM) exhibited substantially decreased abundance for ΔEAR and M3 strains but not for the neutral mutant, M4. As an independent test of these data, total RNA was extracted from ΔEAR and wild-type strains and subjected to microarray analysis (Fig. 4C). The latter analysis revealed a marked and specific decrease in expression of epsF-O genes for the ΔEAR strain, further supporting a regulatory relationship between the EAR element and expression of the second half of the eps operon. Therefore, the EAR element affects expression levels of distally located genes within the eps operon, without affecting proximal genes. From these data we hypothesized that the EAR-mediated regulatory mechanism specifically affects the efficiency of transcription elongation at distal locations within the eps locus, most likely by promoting processive antitermination (Fig. S1).

image

Figure 4. The EAR element promotes expression of distally located genes. A. Schematic of the eps operon with coloured lines representing the general locations that were chosen for qPCR analyses. B. Relative expression level of eps genes in EAR mutants as compared with NCIB 3610. The relative expression of epsA was used to normalize the data from each sample (see Supplementary Methods for more details). C. Microarray analysis of ΔEAR as compared with NCIB3610. Total RNA was extracted from a total of three 48 h colonies cultured on MSgg solid medium (see representative colony pictures in Fig. 4) and used for microarray analysis. Differences in expression of eps genes between ΔEAR and NCIB3610 are shown above. MSgg is a medium condition that has been previously observed to promote biofilm formation in B. subtilis.

Download figure to PowerPoint

EAR-mediated readthrough of intrinsic terminators within the eps operon

Except for rRNA antitermination, the known processive antitermination systems all promote readthrough at intrinsic terminators (Roberts et al., 2008). Furthermore, it has been previously suggested that intrinsic termination is the predominant mechanism of termination in Firmicutes (De Hoon et al., 2005; Sierro et al., 2008). Assuming the EAR element is also proficient in antitermination activity, our data suggested that the target termination sites were likely to be located between epsE and epsF, based upon the drop in expression that occurs at this juncture for EAR mutants (Fig. 4). Consistent with this hypothesis, a search for intrinsic terminator hairpins (G-C rich stem loop followed by a run of U residues) revealed several reasonable candidates, although we note that one of the putative terminators (‘1’) is a particularly poor candidate (Fig. 5A; Fig. S2). To test for EAR antitermination at this locus, we fused the EAR element to the upstream half of epsF, which was in turn fused to a lacZ reporter gene. We then measured LacZ activity for constructs containing wild-type or M3 EAR elements. Expression levels of lacZ were substantially higher (∼10-fold) for constructs containing a wild-type EAR (Fig. 5B). The decreased lacZ expression that occurs upon introduction of the M3 mutation suggests that, indeed, EAR is likely to promote readthrough of intrinsic terminators in epsF. This experiment was repeated with truncated versions of the epsF putative terminator cluster. Only the region that included the strongest terminator candidate (‘2’ in Fig. 5A) was required for antitermination activity in this assay. The weak terminator candidate (‘1’ in Fig. 5A) was present in all constructs; therefore, we cannot rule out the possibility that this terminator is also a target for EAR antitermination. However, it is also possible that this weak terminator candidate is not relevant in vivo and that an individual intrinsic terminator in epsF (‘2’) is a primary target for EAR antitermination.

image

Figure 5. EAR-assisted readthrough of intrinsic terminators in epsF. A. Putative intrinsic termination sites located within the epsF gene (shown within the coding sequence in Fig. S2). B. Constructs containing different truncations of the epsF region, located downstream of the EAR element and upstream of a lacZ gene, are shown schematically in the left panel. Mutation of the EAR element results in lowered lacZ expression for each of the constructs that contain terminator sites, suggesting that the EAR element promotes readthrough of intrinsic terminators within epsF. C. Expression for a subset of the lacZ reporters was monitored during exponential and stationary phase conditions for cells cultured in both rich and defined media. The fold-difference between expression of wild-type and M3 EAR constructs is shown within white ovals. Expression of the EAR-epsF-lacZ reporter fusion is identical between NCIB3610 and 168 (data not shown), correspondingly, these particular experiments were conducted in strain 168.

Download figure to PowerPoint

As a preliminary assessment for whether cellular conditions may affect EAR antitermination, expression of wild-type and M3 EAR-epsF-lacZ reporter fusions were monitored under varying growth conditions (Fig. 5C). As a control for this experiment, we also measured expression of lacZ reporter constructs lacking the epsF terminator sites but containing either a wild-type or M3 EAR element. Modest antitermination activity (2.6-fold) was observed during exponential phase growth in rich medium. The apparent antitermination activity increased in stationary phase (17-fold). In contrast, no appreciable change in expression was observed between exponential and stationary growth phases for the wild-type and M3 constructs lacking epsF terminator sites. When cultured in defined minimal medium typically used to promote biofilm formation (MSgg), the apparent antitermination activity was increased during exponential phase growth (10-fold) and modestly further increased during stationary phase growth (18-fold). However, it is important to note that the increase in apparent antitermination activity was due primarily to decreased expression of the M3 construct, rather than increased expression for the wild-type construct. Therefore, these data do not necessarily prove that EAR-mediated antitermination activity is increased under these conditions but may instead suggest that intrinsic termination activity is generally improved during stationary phase or in growth in minimal medium. Because B. subtilis populations are known to be heterogenous, especially in stationary phase and in minimal medium, these data also do not address whether antitermination activity might vary more significantly for certain cellular subpopulations during growth and development. It is possible that some cells (e.g. EPS producing bacteria) exhibit a greater overall change in antitermination activity than detected by our reporter assays, which is masked by the averaging of lacZ activity during bulk culture conditions. Future experimentation at the single-cell level will be required to address these points.

EAR-mediated readthrough of heterologous intrinsic terminators

A hallmark of processive antitermination mechanisms is the readthrough of heterologous termination sites (Weisberg and Gottesman, 1999; Nudler and Gottesman, 2002; Roberts et al., 2008). To test this possibility for the EAR element, it was placed upstream of a well-characterized, strong intrinsic terminator (‘T7t’; Reynolds and Chamberlin, 1992) followed by lacZ (Fig. 6). Introduction of the M3 mutation into this construct lowered lacZ expression approximately fivefold as compared with the wild-type version. This experiment was then repeated for constructs containing instead a tandem array of three different strong, heterologous, transcription terminators (‘T7t-λtR2-rrnB T1’; Artsimovitch et al., 2000) (Fig. 6). The M3 mutant again resulted in significantly decreased lacZ expression relative to wild-type (∼30-fold), suggesting that the EAR element is capable of promoting readthrough of multiple, heterologous terminators, akin to other processive antitermination systems.

image

Figure 6. The EAR element promotes readthrough of heterologous intrinsic terminators. A. The heterologous intrinsic terminators used in this study are shown schematically. B. A schematic representation of the antitermination assay construct is shown. The EAR element was placed upstream of a well-characterized, strong intrinsic transcription terminator from the T7 bacteriophage (T7t) (Reynolds and Chamberlin, 1992), which itself was upstream of a lacZ reporter. Expression was measured for constructs with a wild-type or M3 EAR element. C. Similarly, a wild-type or M3 EAR element was placed upstream of a tandem arrangement of three different intrinsic transcription terminators (T7t-λtR2-rrnB T1; Artsimovitch et al., 2000), also upstream of lacZ.

Download figure to PowerPoint

Investigation of EAR antitermination in vitro and within a heterologous host

In general, processive antitermination systems have not been studied in Firmicutes and other Gram-positive bacteria. Transcription elongation factors, which are required for rrn and λN antitermination activity, have also been only well-studied in E. coli. It is possible, if not likely, that some of their features and requirements will significantly vary in Gram-positive microorganisms. In contrast, the antitermination mechanism mediated by the put RNA element does not require these elongation factors. Therefore, one might assume that cofactor-independent processive antitermination mechanisms, such as put, might function in a heterologous host. To test this hypothesis for EAR antitermination, we examined expression levels of lacZ reporter constructs containing wild-type and M3 EAR elements within an E. coli host (Fig. 7A). Expression was nearly identical between the constructs in E. coli, suggesting that EAR antitermination does not function within this organism. This result may suggest that a Bacillus-specific signalling factor is required for EAR antitermination. However, it could also result from a missing or incompatible transcription elongation cofactor. Finally, it is also possible that the failure of EAR antitermination within the E. coli host results from key structural differences between bacterial RNA polymerase enzymes. Assuming that the EAR element directly associates with RNA polymerase, akin to the put antitermination element, it is possible that minor structural differences at this RNA-protein interface could disrupt EAR function.

image

Figure 7. EAR antitermination may require additional cellular cofactors. A. A schematic representation of the antitermination assay construct is shown. A wild-type or M3 EAR element was placed upstream of a tandem arrangement of three different intrinsic transcription terminators (T7t-λtR2-rrnB T1; Reynolds and Chamberlin, 1992; Artsimovitch et al., 2000), also upstream of lacZ. Expression of these constructs was then assessed within an E. coli heterologous host. B. We also assessed the ability of the EAR element to promote antitermination in vitro, in the absence of additional cellular components. A schematic representation of the PCR-generated DNA template for the in vitro antitermination assay is shown. Representative data are shown (in replicates) for single-round, in vitro transcription reactions using B. subtilis RNA polymerase. Termination at all three termination sites is marked accordingly, as are transcripts resulting from run-off transcription. Quantification of wild-type and M3 EAR constructs revealed a similar overall level of run-off transcription for both, suggesting that antitermination activity is impaired in vitro.

Download figure to PowerPoint

To further test whether the EAR element alone is sufficient for antitermination activity, as is the case for the put antitermination system, we assessed antitermination activity in vitro. Specifically, we performed in vitro transcription reactions that included purified B. subtilis RNAP, nucleotides, and PCR-generated DNA templates containing the EAR element followed by intrinsic termination sites (Fig. 7B). We examined the degree of run-off transcription (i.e. corresponding to readthrough of the intrinsic termination sites) for wild-type and M3 constructs. These data demonstrated essentially no differences in the amount of run-off transcription for wild-type and M3 EAR constructs, suggesting that EAR antitermination is likely to require assistance from cellular cofactors. Therefore, although put and EAR antitermination both rely on RNA elements (e.g. Fig. S1), they are likely to be mechanistically distinct.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. Note added in proof
  9. References
  10. Supporting Information

Characterization of systems in E. coli and λ have revealed some of the events that occur during processive antitermination (Weisberg and Gottesman, 1999; Roberts et al., 2008). First, RNA polymerase successfully transitions from an initiation complex into a TEC. For some processive antitermination systems, the TEC then encounters a promoter-proximal pause site, which allows time for accessory factors to associate with discrete elements in the nascent transcript and the TEC. In the event that binding of these accessory antitermination factors is unsuccessful within the lifetime of the paused complex, elongation is likely to cease upon encountering downstream termination sites. However, when successfully bound by accessory factors, modified TEC complexes transcribe faster than a typical TEC and are more resistant to pause and termination sites. Presumably, these properties are maintained for as long as the TEC-binding antitermination factor(s) remain associated. However, the antitermination-proficient TEC eventually encounters a terminator site strong enough to cause dissociation, presumably located at the operon terminus. An interesting exception to this overall pathway is the put antitermination element, which does not require additional factors. The put RNA element (e.g. Fig. S1) interacts directly with a portion of RNA polymerase near to the RNA exit channel and is alone sufficient for inducing readthrough of pause and terminator sites in vitro (King et al., 1996; 2004; Weisberg and Gottesman, 1999; Sen et al., 2002; Komissarova et al., 2008).

In this study, we present the discovery of a new processive antitermination mechanism class, which is required for biofilm exopolysaccharide expression in B. subtilis. We find that the EAR element can promote transcriptional readthrough of endogenous and heterologous intrinsic termination sites. This antitermination activity is even maintained for a tandem arrangement of three ‘strong’ heterologous intrinsic terminator sites. From these data, we predict that the EAR antitermination mechanism will share certain features with the previously described processive antitermination mechanism classes. However, it is likely that the EAR antitermination mechanism exhibits features that are unique from these other systems as well. For example, the EAR antitermination mechanism is one of only a few classes that are bacterially encoded. Also, EAR antitermination requires an RNA element that is both unusual in its phylogenetic distribution and structural sophistication. Antitermination by the bacteriophage put element also requires a structured RNA element (a ∼65 nucleotide dual-hairpin motif.) However, the EAR mechanism appears to demand more complex secondary structure and primary sequence features overall. Also, unlike put antitermination, the EAR element alone is not sufficient to promote antitermination activity in vitro.

What could be the functional role(s) of EAR antitermination? For the λ bacteriophage, processive antitermination by λN and λQ proteins is required for high-level expression of early and late phage transcripts respectively. These mechanisms ensure both robustness of λ gene expression and completed synthesis of these unusually long operons (Weisberg and Gottesman, 1999). For example, estimates based upon prior in vitro quantitative analyses suggest that a ‘normal’ length operon (< 5000 nt) is likely to result in 80–90% of transcripts being fully transcribed. This contrasts with only 30–60% of the λ late operon (∼23 000 nt) being fully synthesized without aid of the λQ processive antitermination system, which restores full transcript synthesis to ∼90% (Von Hippel and Yager, 1991; 1992). Similarly, rrn antitermination increases transcriptional efficiency and overall abundance of rRNAs to accommodate fast-growing bacteria (Condon et al., 1995). Indeed, recent quantitative analyses of rrn antitermination demonstrate that the high rate of rrn transcription required for rapidly dividing cells results in dense RNAP ‘traffic’ along the transcripts (Klumpp and Hwa, 2008). Specific incorporation of multiple Rho-dependent termination sites within these transcripts helps to rapidly remove RNAP complexes that have not been altered into the terminator-resistant form, resulting in increased transcription elongation efficiency overall.

Based upon these prior observations, it can be presumed that EAR antitermination may have evolved for similar reasons: (i) full synthesis of long operons and (ii) improved efficiency of transcription where robust expression and strict stoichiometry is required. At approximately 16 kb, the eps operon in B. subtilis is indeed substantially longer than the average bacterial transcripts (Dam et al., 2007). Furthermore, ensuring that individual cells exhibit particularly high levels of eps expression may also be important during B. subtilis biofilm development. Recent data have suggested that EPS-producing bacteria may comprise the cellular precursors for endospore formation within the B. subtilis biofilm community (Vlamakis et al., 2008). For example, mutations in key eps genes resulted in lowered sporulation rates within the biofilm community, despite the fact that none of the eps genes directly participate in the sporulation developmental pathway. EPS production, it has been hypothesized, is wasteful enough that nutrient levels within the microniche surrounding EPS producers plunge to concentrations that encourage endospore formation, a developmental programme restricted to conditions of energy depletion. Thus it is possible that the robust eps expression afforded by EAR antitermination may have evolved in part to ensure metabolic ‘specialization’ of EPS-producing bacteria within the biofilm community, although further experimentation will be required to test this hypothesis.

One possible model predicts that the EAR element functions in ‘constitutive antitermination’ (Fig. 8B). However, that this activity cannot be recapitulated in vitro using purified components suggests that additional cellular factors are required, such as the elongation factors (e.g. ‘Nus’ factors) that are important for other antitermination systems. This overall mechanistic model would conceptually suggest that EAR antitermination essentially acts as an extension of regulatory mechanisms controlling transcription initiation. In other words, once transcription of eps has been initiated the EAR element would function as a constitutive ‘gas pedal’ to ensure full transcript synthesis.

image

Figure 8. Transcriptional and co-transcriptional regulation of the eps operon. A. The eps operon is directly repressed by the master regulator for biofilm formation, SinR. Under biofilm-promoting conditions, inactivation of SinR (by SinI and SlrA) and reduced expression of AbrB leads to increased transcription of the eps operon. Activation of EAR antitermination allows for complete synthesis of the long eps transcript. In the absence of EAR antitermination, most of the transcripts will be terminated at the epsF locus. B. From the data presented herein, we hypothesize that the EAR element promotes eps synthesis via two possible models, which are not mutually exclusive. Constitutive antitermination could result from cellular cofactors that assemble with EAR and the transcription elongation complex to promote processive antitermination. Alternatively, processive antitermination may require an unknown signal in addition to the EAR element. t = intrinsic terminator sites within epsF.

Download figure to PowerPoint

Alternatively, it is also possible that EAR antitermination requires participation of a physiological or population-based signal produced during biofilm formation (Fig. 8B). Intriguingly, we are only able to identify the EAR element within functionally related operons, perhaps suggesting that it shares a specific regulatory relationship with biofilm and capsular polysaccharide gene clusters. Our data also revealed that, in the absence of EAR antitermination, a major block for eps transcription is located inside the epsF coding region. It is possible that, for B. subtilis, EAR antitermination could be co-ordinated with the functional roles of genes located upstream and downstream of epsF. For example, the immediately upstream gene, epsE, is a glycosyltransferase that acts as a ‘molecular clutch’ by halting flagellar movement upon its expression (Blair et al., 2008). Downstream of epsF are mostly genes required for polymerization and extracellular assembly of the exopolysaccharides. Therefore, although derepression of SinR and AbrB under biofilm conditions would permit transcription initiation of eps, transcript elongation would still prematurely terminate within the epsF locus (Fig. 8A). This could result in a population of nonmotile cells that await accumulation of an unknown physiological or quorum-based signal before allowing full EPS production.

Regulation of biofilm formation in B. subtilis has been well studied at the transcriptional level (Aguilar et al., 2007). Previous data revealed a cascade of transcription factors (e.g. Spo0A, DegU, SinR, SlrR, RemA, RemB) that together contribute to the activation of eps and yqxM-sipW-tasA operons (Branda et al., 2001; Kearns et al., 2005; Verhamme et al., 2007; Chu et al., 2008; Chai et al., 2009; Winkelman et al., 2009). Our data together introduce a new layer of co-transcriptional regulation that is mediated by the EAR element and is essential for biofilm formation in certain Bacillales. These data also suggest that structurally complex, ‘riboswitch-like’ RNA elements may be employed for processive antitermination mechanisms (Fig. 8; Fig. S1). Given the rapidly expanding collection of new regulatory RNA candidates (e.g. Barrick et al., 2004; Corbino et al., 2005; Weinberg et al., 2007; Gardner et al., 2009; Xu et al., 2009), this discovery may indicate that other RNA-mediated processive antitermination mechanisms also await discovery. It has already been established that certain regulatory RNAs are widespread for control of transcription initiation (e.g. 6S RNA) (Wassarman, 2007); it remains to be determined whether RNA elements controlling elongation efficiency are similarly common.

Experimental procedures

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. Note added in proof
  9. References
  10. Supporting Information

Chemicals and oligonucleotides

All chemicals and enzymes, unless otherwise noted, were obtained from Sigma-Aldrich and New England Biolabs respectively. DNA oligonucleotides were purchased from Integrated DNA Technologies. DNA oligonucleotides used in these studies are shown in Table S1.

Strains and growth conditions

All B. subtilis NCIB3610 and 168 strains (Bacillus Genetic Stock Center, Ohio) were cultured in 2xYT (16 g tryptone, 10 g yeast extract, 5 g NaCl per litre) broth or on Tryptone Blood Agar Base (TBAB, Difco) plates at 37°C. For β-galactosidase activity assays, cells were cultured in modified MSgg medium [100 mM MOPS (pH 7), 5 mM potassium phosphate (pH 7), 0.5% glycerol, 0.5% glutamate, 2 mM MgCl2, 700 µM CaCl2, 50 µM MnCl2, 100 µM FeCl3, 1 µM ZnCl2, 2 µM thiamine; Branda et al. 2001] at 37°C to an OD600 = 0.6–0.8 in the presence of 80 µM IPTG. MSgg medium is traditionally used when investigating biofilm conditions for B. subtilis. To analyse colony architecture, 5 µl of an overnight culture in 2xYT were spotted onto MSgg plates supplemented with 1.5% Bacto agar (Difco) and incubated at 30°C for 72 h. Images of B. subtilis colonies were captured at 6–10× magnification using a Zeiss AxioCam Mrc 5 camera equipped with a 0.63× objective lens. When appropriate, antibiotics were included at the following concentrations: 100 µg ml−1 spectinomycin, 5 µg ml−1 chloramphenicol and 1 µg ml−1 erythromycin plus 25 µg ml−1 lincomycin. 50 µg ml−1 tryptophan was added to MSgg medium above for all experiments with strain 168 and its derivatives.

Strain construction

Unless otherwise noted, all mutant strains were generated by directly transforming competent B. subtilis NCIB 3610 cells with plasmid DNA (∼5–10 µg) prepared from methylation-deficient E. coli strains (INV110, Invitrogen). DNA was transformed into B. subtilis using a modified version of a previously published protocol (Anagnostopoulos and Spizizen, 1961; see Supplementary Materials for more details).

In-frame deletions: ΔEAR.  To generate the marker-less ΔEAR deletion construct, the region upstream and downstream of EAR was PCR amplified from NCIB 3610 chromosomal DNA using primer pairs irv126/308 and irv293/129 respectively. The two fragments, containing ∼20 nucleotides overlap, were then subjected to a second round PCR reaction using primer pair irv126/129. The DNA sequence that was deleted using these oligonucleotide primers is indicated at the end of this section. The resulting PCR was digested with EcoRI and BamHI and sub-cloned into pMAD (Arnaud et al., 2004), which carries a temperature sensitive origin of replication, erythromycin resistance cassette and constitutively active lacZ gene. Correct clones were confirmed by DNA sequencing. The plasmid was used to transform NCIB 3610 strains at the permissive temperature for plasmid replication (30°C) with selection for resistance to erythromycin (1 µg ml−1 and lincomycin 25 µg ml−1). Isolates resulting from Campbell integration were obtained after culturing the cells overnight in 2xYT broth at the restrictive temperature (37°C) and plating serial dilutions at 37°C with selection for resistance to erythromycin and lincomycin. These isolates were blue when grown on solid medium containing bromo-chloro-indolyl-galactopyranoside (Xgal). To cure the plasmid, the strain was incubated overnight in 2xYT broth without shaking and without antibiotic at 30°C, followed by shaking incubation for 5 h at 30°C and 3 h at 37°C. The cells were then serially diluted and plated on TBAB at 37°C in the absence of antibiotic. Individual colonies were patched onto TBAB plates and TBAB plates containing erythromycin and lincomycin. Isolates that were sensitive to antibiotics, and were white on Xgal-containing medium, were presumed to result from recombination-loss of the integrated plasmid. Chromosomal DNA from these isolates was isolated and used as templates for diagnostic PCR reactions and subsequent DNA sequencing reactions to confirm mutagenesis of the targeted genomic locus.

EAR mutant strains: M3, M4, M5, M6, M7, M8.  All pMAD plasmids (Arnaud et al., 2004) containing EAR mutations were transformed into ΔEAR as a recipient strain, instead of NCIB 3610, using the same protocol that was employed for making in-frame gene deletion strains (described above). The EAR mutations were created via QuikChange mutagenesis (Stratagene) using the following primer pairs for the initial amplification by Pfu polymerase: irv321/322 (M3), irv323/324 (M4), irv335/336 (M5), irv339/340 (M6), irv341/342 (M7), irv343/344 (M8).

EAR-T7t-lacZ transcriptional fusion.  The B. subtilis EAR region and lacZ gene were amplified using the oligonucleotide pairs irv364/365 and irv366/367 respectively. Oligonucleotides irv365 and irv366 each contain a portion of the previously characterized T7 terminator sequence, ‘T7t’, which could be reassembled after a subsequent PCR-sewing step (Reynolds and Chamberlin, 1992). The two fragments were subjected to a second round PCR-sewing reaction using primer pair irv364/365 in order to generate an EAR sequence that is followed by T7t and the lacZ gene. The resulting PCR was digested with HindIII and SphI and subcloned into pHyperSpank (gift from Gurol Suel, Dallas, TX). These oligonucleotide primers also incorporated SalI and NheI restriction sites into the region between the EAR-T7 terminator sequence and the T7 terminator-lacZ region, respectively, which were then used for subsequent plasmid constructions. NCIB 3610 cells were then transformed with the resulting plasmid with selection for resistance to spectinomycin. Individual colonies were screened by diagnostic PCR reactions for integration of the plasmid into the amyE locus via double homologous recombination. The remaining EAR-terminator-lacZ fusion strains were created using similar to that described above. Further descriptions and sequences for each construct and oligonucleotides are included in the Supplementary Materials.

β-galactosidase activity assays

Cells grown in MSgg media (B. subtilis) or LB (E. coli) were collected at mid-exponential growth phase by centrifugation and β-galactosidase activity assays were performed in triplicate using a protocol similar to that described previously (Miller, 1972). Briefly, cell pellets were resuspended in 1.0 ml of Z buffer (Miller, 1972). 10 µl of toluene was added and the cell suspension was vortexed for exactly 30 s followed by evaporation of toluene in a fume hood. 0.15 ml of orthonitrophenyl-β-D-galactoside was added to 0.75 ml of permeabilized cell suspension to initiate the enzymatic assay (Ti). Reactions were terminated (Tf) by addition of 0.375 ml of 1M Na2CO3, centrifuged briefly and analysed for absorbance at A420. Data were analysed relative to control reactions containing Z buffer alone. Miller Units were calculated using the equation [(1000 × A420)/((Tf − Ti) × 0.75 ml × A600)].

Total RNA extraction

To isolate total RNA from bacterial colonies, 5 µl of an overnight culture in 2xYT was spotted onto MSgg plates and incubated at 30°C for 48 h. The cells were lifted off of the solid medium, transferred into LETS buffer (0.1 M LiCl, 10 mM EDTA, 10 mM TrisHCl, 1% SDS) and passed through a 22G syringe four times to break up the extracellular matrix. The resulting cellular suspension was then used for extraction of total RNA as described elsewhere (Collins et al., 2007).

Quantitative real-time RT-PCR

Quantitative real-time RT-PCR (qPCR) data were determined from at least three independent experiments, each individually performed in triplicate. Approximately 4 µg total RNA was mixed with two units DNase (Promega) and 0.5 mM MgCl2 in 20 µl of reactions and incubated at 37°C for 30 min, followed by 75°C for 10 min. cDNA libraries were then synthesized from this material using the iScript cDNA synthesis kit (Bio-Rad) as per manufacturer instructions. Approximately 0.5–1.0 µg DNase-treated total RNA and random hexamer primers were included in these cDNA synthesis reactions. Control reactions lacking reverse transcriptase were prepared for each RNA sample. All reactions were prepared in triplicate in MicroAmp optical 384-well reaction plates (ABI) and contained 150 nM primers, 5 µl of 2× SYBR Green (Bio-Rad) master mix and 25 ng of cDNA templates in a total volume of 10 µl. The cycling parameters used for these experiments were: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min, using an ABI 7900HT Fast Real-Time PCR instrument. DNA oligonucleotides used in these analyses [epsA (irv356/357), epsC (irv358/359 and irv430/431), epsD (irv432/433), epsE (irv434/435), epsF (irv368/369 and irv436/437), epsH (irv360/361), epsM (irv370/371)] were designed using Primer3Plus software such that they amplified a ∼100 base pair amplicon (Untergasser et al., 2007). Data were analysed using Sequence Detection Systems (SDS) version 2.2.2 software (Applied Biosystems) with automatic baseline and threshold determination. The expression of different eps genes was normalized to the amount of transcription of the eps operon (as measured by epsA expression level) and the fold-change relative to wild-type (ΔΔ<ι>Ct) was calculated using by the following formula:

  • image

Standard deviation values were calculated as described elsewhere (Bookout et al., 2006).

Microarray analysis

Total RNA was harvested from wild-type B. subtilis NCIB 3610 and ΔEAR colonies as described above. Samples were processed and hybridized to GeneChip B. subtilis Genome Arrays (Affymetrix) by the UT Southwestern Microarray Core Facility following the manufacturer's recommendations. Expression analysis was performed using the GeneChip Operating System (GCOS) software.

Scanning Electron Microscopy (SEM)

Bacillus subtilis colonies were prepared on solid MSgg medium as described above and fixed and prepared for SEM by the UT Southwestern Molecular and Cellular Imaging Facility using standard protocols. Representative images were captured using a FEI XL30 ESEM microscope at 10 000× magnification.

RNA structural probing

DNA templates for in vitro studies of different RNAs were created by PCR using appropriate oligonucleotide primers, irv378/irv379. PCR products were then prepared using the QIAGEN PCR clean-up kit. RNAs were synthesized at 37°C in vitro from 25 to 50 µl of reaction mixtures that included ∼10–30 pmol templates, 30 mM Tris-HCl (pH 8.0), 10 mM DTT, 0.1% Triton X-100, 0.1 mM spermidine-HCl, 2.5–5.0 mM each NTP (Roche), 40 mM MgCl2 and ∼50 µg ml−1 T7 RNA polymerase. Reactions were terminated after 2.5 h with 2× volume 8 M urea. Products were resolved by denaturing 6% PAGE and excised, cut into ∼ 1 mm cubes, and equilibrated in 200 mM NaCl, 10 mM Tris-HCl (pH 7.5) and 10 mM EDTA (pH 8.0) for < 2 h at 23°C. Passively eluted RNAs were then ethanol precipitated and quantified by A260. An extra G was added at the 5′-terminus to facilitate T7 transcription.

RNA substrates for in-line probing were dephosphorylated using calf intestinal alkaline phosphatase (New England Biolabs) and 5′-radiolabelled using T4 polynucleotide kinase (New England Biolabs) and γ-32P ATP (Amersham). Reactions contained ∼1 nM RNA, 50 mM Tris-HCl (pH 8.3) and 100 mM KCl. RNAs were incubated at 25°C for ∼40 h and products were resolved by 10% PAGE adjacent to control lanes containing: partial digestion by RNase T1 (cleavage after G), a hydroxyl cleavage ladder (cleavage at every position), and an aliquot of unreacted RNA.

All polyacrylamide (National Diagnostics) gels were visualized using a PhosphorImager (Amersham) and quantified with SAFA (Das et al., 2005) and ImageQuant software (Molecular Dynamics).

Single-round transcription assay

DNA templates for in vitro transcription assays were created by amplifying wild-type and M3 EAR-T7t, λtR2,rrnB T1 constructs using oligonucleotide pairs irv55/irv373 from the lacZ fusion plasmid (see Supplementary Materials for details). A promoter sequence from B. subtilis glyQ gene was added through a forward primer (irv55). Elongation complexes were formed with a 40 nM concentration of DNA template, 44 nM RNAP,100 nM SigA and transcription buffer (200 mM Tris-HCl pH 8.0, 20 mM NaCl, 14 mM MgCl2, 0.1 mM EDTA, 14 mM β-mercaptoethanol) in a total volume of 100 µl. Elongation complexes were halted at nucleotide position 17 when transcription was initiated in the absence of CTP but the presence of 100 µM ApU dinucleotide, 2.5 µM ATP, 2.5 µM GTP, 1 µM UTP and [α-32P] UTP. The halted complexes were formed upon incubation at 37°C for 10 min. Transcription was restarted by the addition of all four nucleotides at 200 µM concentration and heparin to 100 µg ml−1. All reactions were quenched after 15 min incubation at 37°C by an equal volume of 2× urea loading buffer (45 mM Tris-borate, 4 M urea, 10% sucrose [w/v], 5 mM EDTA, 0.05% SDS, 0.025% xylene cyanol FF, 0.025% bromophenol blue), and resolved by 6% denaturing polyacrylamide gel electrophoresis.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. Note added in proof
  9. References
  10. Supporting Information

This work was funded by The University of Texas Southwestern Medical Center Endowed Scholars Program and the Welch Foundation (I-1643). Research by I. I. is supported by a Sara and Frank McKnight predoctoral fellowship. We are grateful to Jeffrey Barrick for helpful discussions. We are grateful to the UTSW electron microscopy core facility for their help in analysing B. subtilis colonies via SEM and the UTSW microarray core facility for assistance in analysing microarray data. We are grateful to Irina Artsimovitch for the gift of purified B. subtilis RNA polymerase.

Note added in proof

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. Note added in proof
  9. References
  10. Supporting Information

Discovery of the EAR element was also independently reported in another manuscript where it is referred to as the epsC motif (Weinberg et al., 2010).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. Note added in proof
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. Note added in proof
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
MMI_7131_sm_supp_info.pdf1413KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.