Screening of K. pneumoniae transposon mutants for alterations in biofilm formation
To identify factors contributing to rapid biofilm formation by K. pneumoniae
AJ218, a 7,000 transposon mutant library was constructed and screened in a polyvinyl-chloride (PVC) microtiter plate assay where biofilm formation was quantified by crystal violet staining 
. Mutants exhibiting reduced or enhanced biofilm ability greater than 15% of K. pneumoniae
AJ218 were then examined individually.
Fifteen biofilm-altered mutants were isolated and the nucleotide sequence immediately flanking each transposon insertion site was identified by Y-linker ligation PCR 
and sequenced to identify the disrupted locus. Seven mutants were subsequently selected for detailed investigation (), based upon differences observed in their ability to express functional type 3 fimbriae (see below). The full-length open reading frame (ORF) of each gene disrupted by the transposon insertion was sequenced and exhibited greater than 99% nucleotide identity to homologs located in the sequenced K. pneumoniae
KTUH-K2022 genome. To confirm the transposon mutant phenotypes, five deletion mutant strains were constructed from wild-type K. pneumoniae
AJ218 whereby the mrkA
, mrkJ and yfiRNB
loci were deleted and replaced with a kanamycin resistance-encoding gene. The ΔyfiRNB
operon deletion was made to compare to the yfiN
transposon mutant, and to analyze mutants that lacked the entire tripartite signalling module. The kanamycin resistance-encoding gene was excised from the ΔmrkH
mutant to avoid polar effects on mrkI
transcription. Each deletion mutant strain exhibited an equivalent defect in biofilm-formation to the initial transposon mutant (). No apparent differences in the planktonic growth rates between wild-type, transposon mutant and deletion mutant strains were observed (data not shown). In complementation analysis, empty vector controls had no impact on biofilm formation for all strains tested (Figure S1
Identification of genetic loci participating in K. pneumoniae AJ218 biofilm formation.
Biofilm formation by K. pneumoniae AJ218.
Type 3 fimbriae of K. pneumoniae are an important mediator of biofilm formation
genes encode the major pilin subunit and periplasmic chaperone of type 3 fimbriae. The mrkC
gene encodes a protein of 828 amino acids, predicted to have the three domains defining the usher translocase, where residues 156-651 conform to the central translocase domain as defined by Pfam00577 
. In contrast to the K. pneumoniae
AJ218 parent strain, any one of the three mutants: ΔmrkA,
, failed to form biofilms () and promote MR/K hemagglutination, indicating loss of type 3 fimbriae expression. Complete sequencing of the mrk
gene cluster of K. pneumoniae
AJ218 showed five ORFs (mrkABCDF
) arranged in the same transcriptional orientation (), consistent with other K. pneumoniae
strains deposited in GenBank.
The mrkABCDF and mrkHIJ loci in K. pneumoniae AJ218.
The mrkH, mrkI, mrkJ and yfiRNB loci contribute to K. pneumoniae biofilm formation
Three other mutants isolated from the screen defined a three-locus cluster (mrkH
) located immediately downstream and transcribed convergently to the mrkABCDF
operon ( and ). The 9 kb region containing the mrkABCDF
clusters is highly conserved amongst the sequenced K. pneumoniae
genomes, with the nucleotide identity between all strains greater than 99%. Analysis of completed bacterial genome sequences showed that the only other species displaying conserved homologs of these two mrk
clusters is Citrobacter koseri
, an opportunistic pathogen 
. Amino acid sequence identities between the Mrk proteins of K. pneumoniae
and C. koseri
BAA-895 range from 82% (MrkD) to 92% (MrkB), while homology between the mrkHIJ
clusters is lower (MrkH
RT-PCR analysis of K. pneumoniae AJ218 cDNA templates spanning intergenic regions between mrkH and mrkI were obtained, but could not be obtained between mrkI and mrkJ (). These results suggest that the mrkH and mrkI genes are co-transcribed in a polycistronic mRNA and mrkJ is transcribed independently of the mrkHI operon.
The ΔmrkH mutant was significantly reduced in biofilm formation, both in a static assay that measures initial biofilm formation () and a continuous flow-cell assay that measures mature biofilm development (). For both assays, the mutation could be successfully complemented with wild-type mrkH copies (ΔmrkH [pMrkH]). The product of mrkH contains a putative C-terminal PilZ domain capable of binding the second messenger c-di-GMP. Multiple alignments of the PilZ domain of K. pneumoniae MrkH with other experimentally characterized PilZ domain-containing proteins demonstrated complete conservation of two functionally important sequence motifs in MrkH (). The amino acid residues in these conserved motifs (109RxxxR and 140D/NxSxxG) are known to be crucial for c-di-GMP binding and allosteric regulation of PilZ domain-containing proteins for subsequent downstream functions. We therefore propose that MrkH functions as a c-di-GMP-binding protein.
Flow-cell-cultivated biofilm formation by K. pneumoniae AJ218 after 4 days.
Conservation of PilZ-, EAL- and GGDEF-domain proteins in K. pneumoniae.
Biofilm formation by the ΔmrkI
mutant was moderately deficient compared to wild-type, and could be rescued when supplied with wild-type mrkI
[pMrkI]; ). The mrkI
gene encodes a putative LuxR-like transcriptional regulator that contains a C-terminal helix-turn-helix (HTH) DNA-binding motif. Proteins with LuxR domains can participate in quorum sensing 
, however MrkI lacks the entire N-terminal autoinducer-domain required for binding N-acyl homoserine lactones, suggesting MrkI is a transcriptional activator of gene(s) responding to signals other than those involved in quorum sensing.
The final gene in the three-locus cluster, mrkJ, encodes a putative phosphodiesterase (PDE). This enzyme contains an EAL domain that mediates the hydrolysis of c-di-GMP. Multiple alignments of the EAL domain of MrkJ with other known PDE enzymes demonstrated the conservation of several regions throughout the domain sequence (). Like other enzymes of the same class, the residues ‘ECL’ that form the putative active site of MrkJ varies from the consensus ‘EAL’ sequence. The negative regulatory role of EAL-domain proteins over c-di-GMP levels is consistent with the phenotype of both the ΔmrkJ mutant (enhanced biofilm) and complemented mutant, ΔmrkJ [pMrkJ] (defective biofilm) (). This model for biofilm formation is further supported by the identity of the seventh mutant identified in the transposon mutant library screen of K. pneumoniae AJ218 (). The yfiN gene encodes a putative integral-membrane diguanylate cyclase (DGC) with a GGDEF domain, which functions to synthesize c-di-GMP. Multiple alignment of the GGDEF domain of YfiN with other studied DGC proteins showed conservation of several regions, including the GGDEF residues that form the putative catalytic active site (). Deletion of the entire yfiRNB gene cluster resulted in a significant impairment in biofilm formation and complementation with the yfiRNB operon (ΔyfiRNB [pYfiRNB]) enhanced biofilm formation to levels almost two-fold greater than wild-type ().
We therefore propose that biofilm formation in K. pneumoniae is regulated by the relative availability of intracellular c-di-GMP, coordinated by c-di-GMP turnover enzymes (YfiN and MrkJ), and sensed by a receptor protein (MrkH). When a PDE is inactivated or a DGC up-regulated leading to increased c-di-GMP availability, biofilm formation is stimulated. Conversely, when a DGC is inactivated or a PDE up-regulated resulting in decreased c-di-GMP concentration, biofilm formation is inhibited.
MrkH positively regulates type 3 fimbriae expression in K. pneumoniae
We sought to address whether the activity of MrkH influences type 3 fimbriae expression. A sensitive functional assay for type 3 fimbriae is the MR/K hemagglutination (HA) activity, mediated by the MrkD adhesion located at the tips of type 3 fimbriae. As expected, the ΔmrkA
mutant failed to mediate a visible MR/K HA reaction at the highest bacterial density tested (approximately 1×1010
CFU/mL; ). Likewise, the ΔmrkH
mutant completely lacked MR/K HA activity at the same cell concentration. Upon complementation, strong MR/K HA activity was observed for both strains. The MR/K HA titer for the complemented strains was either reduced to wild-type level (for the ΔmrkA
-complemented mutant) or to approximately six-fold lower than wild-type level (for the ΔmrkH
-complemented mutant). It was shown that empty vector controls had no impact on MR/K HA expression for all strains tested (Figure S2
). A complementary assay for MR/K HA expression utilized a specific anti-MrkA antibody in immunoblot analysis of total cellular extracts. As expected, MrkA could not be detected in the ΔmrkA
mutant, while expression was restored when the mutant was complemented with the complete mrkABCDF
gene cluster (ΔmrkA
[pMrk]; ). Using this assay, MrkA was not detected in wild-type K. pneumoniae
AJ218, indicating that this strain only weakly expresses type 3 fimbriae in planktonic culture. Similarly, MrkA was not detected in the ΔmrkH
mutant, but was expressed above wild-type levels when the mutant was complemented with a plasmid expression construct for MrkH (ΔmrkH
[pMrkH]; ). Therefore, when MrkH is over-expressed in K. pneumoniae
, both the production of the MrkA subunit and the MR/K HA activity is increased. MrkH is therefore a critical, positive regulator of type 3 fimbriae expression.
Type 3 fimbriae expression by K. pneumoniae AJ218.
Regulation of type 3 fimbriae expression in K. pneumoniae by YfiRNB and MrkJ
We also examined the levels of MrkA subunit production and MR/K HA activity of ΔyfiRNB (DGC) and ΔmrkJ (PDE) mutant strains (). The MrkA subunit was not detected in the ΔyfiRNB mutant, but steady-state levels of MrkA were higher than wild-type in mutants complemented with the wild-type yfiRNB operon. Conversely, MrkA levels were increased in the ΔmrkJ mutant, but absent when complemented with wild-type mrkJ gene copies. The MR/K HA activity of these strains was consistent with the immunoblot results (). These observations suggest that YfiRNB and MrkJ function, respectively, as positive and negative regulators of type 3 fimbriae expression. The phenotypes of the mutants are consistent with their roles in modulating the intracellular levels of c-di-GMP.
MrkI, the LuxR-like regulator, mediates type 3 fimbriae functionality in K. pneumoniae
The hemagglutination tests demonstrated that the ΔmrkI mutant, shown previously to be deficient in biofilm formation, also had decreased MR/K HA activity. This is consistent with a reduced amount of type 3 fimbriae on the cell surface. These results suggest that MrkI functions as a positive regulator of type 3 fimbriae. Counter-intuitively, the ΔmrkI mutant appeared to express more MrkA subunit than the wild-type strain. Moreover, when complemented, the over-expressed MrkI strain appeared to produce even greater amounts of MrkA subunit. Why would the increased levels of MrkA produced in the ΔmrkI mutant not be assembled into functional fimbriae to facilitate biofilm formation and MR/K HA?
We hypothesize that this discrepancy is due to MrkI exhibiting multiple roles within the cell. It could function as a minor transcriptional activator of type 3 fimbriae synthesis, in conjunction with a role to regulate another component of the fimbriae assembly pathway (e.g., the usher translocase, MrkB). Alternatively, MrkI could regulate other cell surface factors, such as the polysaccharide capsule, which may in turn affect the function of type 3 fimbriae, as previously described for type 1 fimbriae 
In this scenario, a decrease in type 3 fimbriae assembly when MrkI is absent could lead to intracellular accumulation of MrkA, seen as an apparent increase in the steady-state level of MrkA by immunoblot (), resulting in the observed deficiency in MR/K HA and biofilm formation. When MrkI is over-expressed, the functionality of MrkA is restored and type 3 fimbriae expression and biofilm formation is enhanced greater than wild-type levels. The increased levels of MrkA subunit produced in the ΔmrkI complemented strain is suggestive of MrkI functioning as a strong activator of mrkA transcription when present in high numbers.
Using real-time PCR, only a slight decrease in mrkA gene transcript expression is observed in the ΔmrkI mutant, consistent with MrkI having a partial role in type 3 fimbriae activation (). However, we failed to see a dramatic increase in mrkA RNA levels in the ΔmrkI complemented strain that could account for the large elevation in MrkA protein production seen in the cell extracts. That MrkB, the periplasmic chaperone, and MrkC, the usher translocase, are required to assemble and export the MrkA protein might account for this. However, we cannot rule out that MrkI also positively regulates other gene(s) that encode factors which participate in the correct folding of the MrkA fimbrial subunit.
Quantitative RT-PCR analysis of mrkA RNA levels.
MrkH positively controls the transcription of the mrkA promoter
We sought to test whether MrkH activates transcription of the mrkABCDF
operon by stimulating a promoter(s) located in the upstream region of this gene cluster. The sequence of this upstream region is shown in . To test this hypothesis, we established an assay system in Escherichia coli
using a lacZ
reporter positioned downstream of specific regulatory regions of the mrkA
gene (Materials and Methods
). Initially, two mrkA-lacZ
fusions were constructed: mrkA
-1 and mrkA
-2, spanning positions −759 to +109 (mrkA-lacZ-
1) and from −295 to +109 (mrkA-lacZ
-2) relative to the mrkA
translational start codon. These two plasmids, along with the “promoterless” control plasmid pMU2385, were each transformed into E. coli
K12 strain MC4100 (ΔlacZ
) containing either vector pACYC184 (MrkH–
) or its derivative pMrkH (MrkH+
) and assayed for β-galactosidase expression.
Transcriptional analysis of the mrkA regulatory region.
Compared to the negative control (pMU2385), both mrkA-lacZ-1 and mrkA-lacZ-2 produced low but significant levels of β-galactosidase activity (12 U) in the MrkH– background (). In both cases, MrkH stimulated expression of mrkA-lacZ-1 and mrkA-lacZ-2 more than 300-fold. These results demonstrate the presence of a promoter(s) in the upstream region of mrkA, which is highly activated by MrkH. Given the similar regulatory patterns of the two constructs, both the promoter and operator elements are present within the 404 bp fragment carried on mrkA-lacZ-2.
To verify MrkH-mediated activation of mrkA transcription directly in K. pneumoniae, the 404 bp mrkA fragment (as carried by mrkA-lacZ-2) was inserted upstream of the cat reporter gene in plasmid pKK232-8. The resulting plasmid, mrkA-cat, was then introduced into three isogenic K. pneumoniae strains: the wild-type AJ218 strain (MrkH+ haploid), the ΔmrkH mutant (MrkH–) and wild-type AJ218 carrying pMrkH (multi-copy MrkH+). As expected, ΔmrkH mutants express barely-detectable CAT activity (9.8 U; ). In contrast, when K. pneumoniae carried the wild-type mrkH gene, CAT expression increased 49-fold to 448 U. Further evidence of the stimulatory role of MrkH on transcriptional activation comes from the CAT expression measured (2000 U) when multiple copies of the mrkH gene are present.
Mapping the mrkA transcriptional start site and the region required for MrkH activation
To map the transcriptional start site(s) of mrkA, we performed a primer extension experiment. Total cellular RNA was isolated from E. coli MC4100 strains containing pMrkH with either pMU2385 (control) or mrkA-lacZ-2. Following hybridization of the RNA with 32P-labelled primer (Px1mrkARev) and extension with CMV reverse transcriptase in the presence of dNTPs, the samples were analyzed on a sequencing gel. A single extension product was evident from the mrkA-lacZ-2 sample (). The data mapped the start site of transcription to 204 bp upstream of the putative start site of translation of mrkA (). Inspection of the sequence revealed the presence of the hexanucleotides TATATT centered at -10.5 relative to the start site of transcription, which is a good match to the consensus sequence of the -10 region of a bacterial σ70 promoter. The only possible -35 sequence (TTAATG), which matches poorly to the consensus sequence, was found 15 bp upstream of the putative -10 region. The combination of a poor -35 region and imperfect spacing may contribute to the very weak promoter activity observed in the MrkH− background.
Mapping the start site of transcription of the mrkA promoter by primer extension.
Mapping the mrkA promoter allowed us to make more precise deletion constructs: mrkA-lacZ-3 and mrkA-lacZ-4, in order to localize the region responsible for MrkH-mediated activation of mrkA transcription. As shown in , construct mrkA-lacZ-3, in which most nucleotides downstream of the start site of transcription were deleted, exhibited the same degree of activation by MrkH as mrkA-lacZ-2, indicating that the deleted downstream sequence (between +40 and +313) was not required for MrkH activation. In the case of mrkA-lacZ-4, however, removal of the upstream region between -91 and -56 caused a 5-fold increase in basal level promoter activity in the MrkH− background and completely abolished the MrkH-mediated transcriptional activation of the mrkA promoter in the MrkH+ background (). From these results, we determined that a cis-acting element responsible for MrkH-mediated activation was located between positions -91 and -56.
C-di-GMP facilitates binding of MrkH to the mrkA regulatory region
To test whether MrkH was able to bind directly to the mrkA
regulatory region, we expressed and purified recombinant MrkH (MrkH-8×His) and used it in an electrophoretic mobility gel shift assay (EMSA). The purity of the MrkH-8×His preparation is shown in Figure S3
. The mrkA
fragment which spanned between −91 and +160, relative to the start site of transcription, was end-labeled with 32
P and incubated with varying amounts of MrkH-8×His in the absence or presence of 200 µM c-di-GMP at 37°C for 20 min. The samples were then analyzed on native polyacrylamide gels. In the absence of c-di-GMP, no shift of DNA was seen at the MrkH-8×His concentration of 125 nM (). Increasing the protein concentration to 250 or 500 nM resulted in the partial shift of DNA; however, no discrete protein-DNA band was obvious. In the presence of c-di-GMP, the majority of the DNA was shifted to form a major protein-DNA complex (C1) and a minor complex (C2) at the MrkH-8×His concentration of 125 nM. At the higher protein concentration of 250 nM, the intensity of the larger C2 complex was enhanced. Increasing the protein concentration to 500 nM led to a complete shift of DNA and the formation of a single, even larger protein-DNA complex (C3; ). The right panel of showed that the addition of specific cold competitor DNA can out-compete the binding of MrkH to the labeled probe, demonstrating that MrkH binds specifically to the mrkA
promoter region. Two additional control experiments were carried out in which the mrkA
promoter fragment was incubated with either the wild-type MrkH-8×His in the presence of GTP or the mutant MrkH-8×His (113R-A, for details see below) in the presence of c-di-GMP. EMSA analysis showed that under these conditions neither the wild-type nor the mutant MrkH can form a protein-DNA complex with the mrkA
promoter fragment (Figure S4
Analysis of the binding of MrkH-8×His to the mrkA regulatory region by EMSA.
The results of EMSA demonstrated that (i) MrkH is a DNA binding protein, (ii) c-di-GMP facilitates the binding of MrkH to the mrkA regulatory region and (iii) MrkH can oligomerise on DNA to form a very large MrkH:DNA complex.
C-di-GMP positively controls the activity of the MrkH protein
To provide further evidence that c-di-GMP positively regulates MrkH function, we performed mutational analysis of the mrkH
gene. Construct mrkH
:113R-A contained a point mutation in which the conserved arginine residue at position 113 within the putative c-di-GMP-binding PilZ domain was replaced with an alanine residue ( and ). An immunoblot of a C-terminal His8
-tag fusion with MrkH:113R-A demonstrated that this mutant construct was stably expressed in E. coli
MC4100 (Figure S5
Analysis of a PilZ-domain mutation on MrkH-mediated transcriptional activation, biofilm formation and type 3 fimbriae expression.
To examine the ability of MrkH:113R-A to activate the transcription of mrkA, pACYC184 carrying the mrkH:113R-A mutation, along with pACYC184 carrying wild-type mrkH (MrkH+) and pACYC184 alone (MrkH−), were transformed into E. coli MC4100 containing the single-copy reporter plasmid mrkA-lacZ-2. These strains were then assayed for β-galactosidase activity following mid-log growth in LB media. As expected, the wild-type MrkH+ exerted more than 300-fold activation of the mrkA promoter (). Consistent with our prediction that MrkH is essential for c-di-GMP-mediated positive control of mrkA transcription, the substitution mutation within the PilZ domain completely destroyed the ability of MrkH to activate transcription from the mrkA promoter.
To examine the effect of the PilZ-domain mutation on biofilm formation and type 3 fimbriae expression of K. pneumoniae AJ218, we introduced the mrkH:113R-A construct described above into the ΔmrkH mutant and assayed the resulting strain in static biofilm formation and MR/K HA activity. Although strong biofilm formation () and MR/K HA activity () was observed by the ΔmrkH mutant upon complementation with wild-type mrkH, the mrkH:113R-A construct was completely ineffective and failed to rescue the ΔmrkH mutant phenotypes.
These results strongly indicated that the c-di-GMP binding region within the PilZ domain of MrkH is critical for transcriptional activation of the mrk operon and subsequent biofilm formation via type 3 fimbriae expression.
MrkH activity is influenced by MrkJ- and YfiN-mediated control of c-di-GMP expression levels
If the above conclusion is correct, we would expect that a decrease in intracellular c-di-GMP levels by enhanced expression of a phosphodiesterase causes a reduction in MrkH activity. Conversely, increasing the endogenous c-di-GMP concentration by enhancing the expression of a diguanylate cyclase would lead to greater MrkH activity.
To test this hypothesis, we introduced pBR322 derivatives carrying either the wild-type mrkJ gene, the yfiRNB operon, or pBR322 alone into E. coli MC4100, which contained pACYC184 carrying wild-type mrkH (MrkH+) and the reporter plasmid mrkA-lacZ-2. To confirm the roles of MrkJ and YfiN as c-di-GMP-specific phosphodiesterase and diguanylate cyclase, respectively, we also generated site-directed mutant constructs of these enzymes in their conserved c-di-GMP hydrolysis/synthesis catalytic sites and tested their activity alongside the wild-type constructs. Construct mrkJ:36ECL-AAA contained substitution mutations in which the EAL domain of MrkJ (starting at residue 36) was replaced with alanine residues ( and ). In addition, construct yfiRNB:328DEF-AAA carried substitution mutations in which the DEF residues of the GGDEF domain of YfiN (starting at residue 328) were replaced with alanine residues ( and ).
Analysis of EAL- and GGDEF-domain mutations on MrkH-mediated transcriptional activation, biofilm formation and type 3 fimbriae expression.
These strains were then assayed for β-galactosidase activity following mid-log growth in LB media. Enhanced expression of wild-type MrkJ led to a complete inability of MrkH to activate transcription from the mrkA promoter (). Conversely, mutation of the EAL domain of MrkJ led to enhanced MrkH-mediated transcriptional activation. This result suggested that an inactive EAL domain renders MrkJ unable to effectively hydrolyze c-di-GMP, permitting c-di-GMP accumulation and increased MrkH activity. The increased expression of wild-type YfiRNB resulted in a small but significant enhancement in MrkH-mediated activation of mrkA transcription. However, this increase was not observed from the YfiN construct containing the mutated GGDEF domain, suggesting that the lowered MrkH-mediated activation of the mrkA promoter was caused by YfiN impairment in c-di-GMP synthesis. The site-directed mutations of the EAL and GGDEF domains of MrkJ and YfiN, respectively, also had pronounced effects on biofilm formation and type 3 fimbriae synthesis by K. pneumoniae AJ218. The mrkJ:36ECL-AAA and yfiRNB:328DEF-AAA gene constructs also failed to effectively complement the respective ΔmrkJ and ΔyfiRNB mutant strains in static biofilm formation () and MR/K HA activity ().
Taken together, these results indicate that MrkJ and YfiN are c-di-GMP-specific phosphodiesterase and diguanylate cyclase, respectively. Moreover, we have clearly demonstrated that c-di-GMP is a positive cofactor essential for MrkH-mediated transcriptional activation of the mrk operon for type 3 fimbriae synthesis and biofilm formation by K. pneumoniae.
Detection of phosphodiesterase activity for the MrkJ protein
From the data presented, MrkJ represents a c-di-GMP-specific phosphodiesterase that has pronounced negative effects on K. pneumoniae
biofilm formation. To further characterize the kinetics of c-di-GMP hydrolysis by MrkJ, we expressed and purified MrkJ (MrkJ-8×His) and analyzed its enzyme activity by High-Performance Liquid Chromatography (HPLC). The purity of the MrkJ-8×His preparation is shown in Figure S6
. We demonstrated that a significant amount of c-di-GMP was hydrolyzed by MrkJ-8×His to form the degradation product 5′-pGpG after 10 sec (). The c-di-GMP was completely converted to 5′-pGpG after 30 min. These results confirmed that MrkJ possesses very strong phosphodiesterase activity, a significant feature for an inhibitor of biofilm formation.
MrkJ displays strong phosphodiesterase activity.