A Thermosensitive, Phase-Variable Epigenetic Switch: pap Revisited

SUMMARY

It has been more than a decade since the last comprehensive review of the phase-variable uropathogen-associated pyelonephritis-associated pilus (pap) genetic switch. Since then, important data have come to light, including additional factors that regulate pap expression, better characterization of H-NS regulation, the structure of the Lrp octamer in complex with pap regulatory DNA, the temperature-insensitive phenotype of a mutant lacking the acetyltransferase RimJ, evidence that key components of the regulatory machinery are acetylated, and new insights into the role of DNA binding by key regulators in shaping both the physical structure and regulatory state of the papI and papBA promoters. This review revisits pap, integrating these newer observations with older ones to produce a new model for the concerted behavior of this virulence-regulatory region.

INTRODUCTION

In the United States, extraintestinal pathogenic Escherichia coli (ExPEC) colonization is a leading cause of adult septicemia and the second-leading cause of neonatal meningitis (1, 2). Uropathogenic E. coli (UPEC), a subset of ExPEC, causes the majority of primary and recurrent urinary tract infections (UTIs) (3). Almost 50% of women contract a UTI by their mid-20s (4, 5), leading to a total annual medical expenditure of $1.6 billion in 1995 and a projected $2.8 billion in 2014 (2, 4). A critical factor in the virulence of this uropathogen lies in its ability to adhere to host tissue via a variety of environment-specific fimbria-adhesin complexes, after which it can penetrate the host tissue, colonize tissue surfaces, or further disseminate to other host tissues (6–8). The ascent of UPEC through the human urinary tract involves a selective, dynamic transformation of the bacterial surface through the expression of various fimbriae that interact with host epithelia (9, 10).

Alteration of the surface properties of E. coli requires a significant transition of gene expression over the course of several generations, and completion of the process involves a large energy commitment (9, 10). The regulatory mechanisms that govern bacterial fimbrial expression are sensitive to diverse environmental stimuli, such as temperature, nutrient availability, osmolarity, and pH (11). They are also interlinked to enable differential expression of multiple fimbrial types, an ability that allows for enhanced adhesion to diverse host surfaces (12–16). Within a population of cells, however, the expression of each fimbrial type is rarely homogeneous (17). Instead, the heterogeneous expression of surface properties produces multiple phenotypes, resulting in a varied distribution of invading UPEC subpopulations within the urinary lumen and into intracellular environments (17).

One key clinical subclass of UPEC pathology includes pyelonephritis, an inflammatory infection of the renal pelvis that resides at the junction between the kidney and ureter. Biogenesis of the pyelonephritis-associated pilus (pap) is a virulence mechanism specific to the renal pelvis that exhibits phase-variable switch-like behavior in response to several environmental conditions (9–11, 18–21). For example, temperature has a profound effect on pap gene regulation; in most UPEC strains, pilus gene expression is inhibited at 23°C but active at 37°C (20–23). The switch-like behavior of pap expression is controlled by the 330-bp sequence between the divergent papI and papBA promoters, where papI and papB are transcription factors that help regulate the expression of pap structural genes, and papA is the main pilin structural gene (Fig. 1) (22). Additionally, pap gene expression is controlled by σ⁷⁰, cyclic AMP (cAMP)-CRP, H-NS, Dam, and Lrp, which we will discuss in detail throughout.

FIG 1.

pap promoter region. The 330-bp regulatory region between the divergent papI and papBA promoters contains binding sites for PapB (light blue), CRP (orange), and Lrp (green), shown above the DNA and numbered as in the text. One GATC methylation site (black) is present within the center-most site of each set of Lrp binding sites.

The discovery of pap gene regulation by the leucine-responsive regulatory protein (Lrp), a nucleoid-associated protein that widely associates with DNA, greatly advanced the understanding of the switch-like behavior of pap expression (24, 25). Lrp binds two specific regions that determine papBA expression status (Fig. 1) (24, 26). When bound proximally (relative to the papBA promoter), near the site of papBA transcription initiation, Lrp likely hinders RNA polymerase (RNAP) association with the core promoter elements, inhibiting papBA transcription and leading to an inactive transcriptional state (27). Displacement of Lrp to a more distal set of sites relative to the papBA promoter directly facilitates papBA transcription and leads to an active transcriptional state (26). Understanding the switch-like behavior of Lrp’s positioning on the pap regulatory region was furthered by the identification of distinct methylation patterns and the contributions of several additional transcription factors to the combinatorial logic of pap expression (28). Exploration of the thermosensitive aspects of each transcription factor revealed a high degree of complexity that centers primarily on the interplay of H-NS binding, Lrp localization, and the roles played by methylation of two specific sites within the papBA promoter region (28–31).

Prior literature describing pap regulation typically defines two distinct states of pap gene expression as Phase OFF (no expression from the papBA promoter) and Phase ON (moderate to high expression from the papBA promoter) (32). However, here we propose a new model for pap regulation where these two states are split into three distinct, heritable states of pap expression: fully silenced, inactive, and active, where fully silenced and inactive correspond to Phase OFF and active corresponds to Phase ON (Fig. 2). The fully silenced state (which is formed at low temperatures) consists of H-NS filaments completely covering the pap region and preventing almost all transcription from both the papBA and the papI promoters (Fig. 2A). The inactive state involves basal expression of papBA, with increased levels of papI expression (Fig. 2B). Finally, the active, or Phase ON, state involves the shifting of Lrp to another set of binding sites with the assistance of PapI, allowing the papBA promoter to turn on fully (Fig. 2C). The behavior of the pap promoter is in marked contrast to the regulation of some other pilus-associated operons that only exhibit Phase ON and Phase OFF behaviors, such as the fim operon, where the regulatory region of DNA is excised and inverted to allow for transitions between two distinct states with different DNA sequences (33). We will provide evidence from classical studies and more recent findings to support our new model of pap regulation throughout this review.

FIG 2.

General overview of the silenced, inactive, and active states of pap. (A) The fully silenced state is induced at low (nonhost) temperatures and is characterized by an H-NS filament covering the entire pap DNA region. Here, neither papI nor papBA is expressed. (B) In the inactive state, the Lrp octamer is bound to the three proximal Lrp sites (sites 1 to 3) and the distal GATC site in Lrp site 5 is fully methylated. Basal transcription of papBA can occur, and a molecule of PapB, together with CRP, recruits RNAP and activates transcription of papI. PapI levels are high but papBA transcription is still low, causing the system to remain Phase OFF with regard to pilus expression. (C) In the active state, the Lrp octamer is bound to the three distal sites (sites 4 to 6), while PapI helps to stabilize Lrp in this position. The proximal GATC site within Lrp site 2 is fully methylated. This new positioning of Lrp allows it to bend the DNA to recruit CRP to activate transcription of papBA. Both papI and papBA levels now are high, and the cells are phenotypically Phase ON.

Remarkable dedication by a limited set of laboratories revealed multiple levels of protein-protein interaction, protein-DNA binding, and epigenetic DNA modifications that are integrated to switch between the three pap expression states. These regulatory inputs may change simultaneously, an issue that has served to increase the difficulty in understanding pap regulation. Several broadly encompassing mechanisms have been proposed on the basis of carefully designed studies of the interactions between pap-associated transcription factors and pap DNA. However, the underlying mechanisms responsible for temperature sensitivity and the responses of pap to other environmental conditions have remained difficult to define, particularly with regard to the spatial relationships between pap DNA and most of its associated transcription factors during phase transitions.

In the present review, to synthesize a more complete picture of the regulatory logic governing pap expression, we integrate both fairly well-established aspects of pap regulation with recent findings regarding Lrp, H-NS, and general E. coli physiology. We divide the various factors involved in regulating the papBA promoter by their location of activity, primary roles, and relationship to pap transcription. By integrating recent experimental evidence with that collected prior to the last comprehensive reviews, we intend to produce a new model detailing the concerted behavior of this virulence promoter. We expect that the collected information will provide a strong foundation for future studies of the regulation of the pap operon as well as detailed insight into a regulatory blueprint that is likely shared by a range of other virulence factors.

THE pap PROMOTER REGION

As described above, the pap regulatory region has classically been characterized to have switch-like behavior, where expression of the structural pap genes from the papBA promoter is either Phase ON or Phase OFF and each cell within a population displays an all-or-none phenotype (28). However, more recent studies investigating effects of H-NS and temperature on pap expression suggest that there are actually three distinct, heritable states of pap expression, silenced, inactive, and active, where silenced and inactive both appear phenotypically Phase OFF and active is Phase ON, but they differ with regard to the molecular details of regulation in terms of proteins bound to and DNA base modifications of the pap promoter (Fig. 2) (28, 34).

The papI promoter controls transcription of a gene that encodes a small Lrp-binding protein, PapI, while the divergent papBA promoter drives the transcription of the local transcription factor PapB, the main pilin subunit PapA, accessory pilin subunits, and the assembly machinery (Fig. 1) (22). The DNA sequence elements that regulate pap transcription are contained within the 330-bp region between these divergent promoters (Fig. 1). While this entire regulatory region can be bound by an H-NS filament, the pap regulatory region also includes a single cAMP receptor protein (CRP) binding site, two sets of three Lrp binding sites, two 5′-GATC-3′ DNA adenine methyltransferase (Dam) methylation sites in the middle of each set of the Lrp binding sites, and three PapB-binding regions that allow for the autoregulation of the system (Fig. 1) (24, 35).

Several transcription factors have distinct, and, at times, mutually exclusive, roles within the pap promoter region. Aside from RNA polymerase (RNAP), however, no single transcription factor has been found to be essential for papBA transcription (20, 21, 26, 34). As with Lrp, most of these proteins contain at least two separate sites of interaction within the pap regulatory region that help establish the different states of activity. The two histone-like nucleoid-associated proteins, Lrp and H-NS, serve as “structuring” proteins that bend or coat pap DNA and alter pap transcription (36–41). H-NS, in particular, plays an exclusive role in coating the pap regulatory region to give rise to the silenced state (Fig. 2A). Dam and PapI can be considered “phase-modifying” factors, as they function to modulate papBA transcription, primarily by directing Lrp’s localization within the pap promoter (26, 28). Finally, PapB and CRP serve as “potentiating factors” that tune the levels of papI and papBA transcription (42, 43). We will briefly discuss each of these protein binding sites here and in Fig. 1 and then provide more detail on the role of these proteins in pap regulation in the following sections.

The histone-like protein H-NS forms a repressive nucleoprotein complex with the entire pap regulatory region in response to conditions that mimic a nonhost environment (more precisely, a nonrenal pelvis environment), such as low temperature, high osmolarity, glucose as a carbon source, and conditions with high nutrient availability (11). Moreover, H-NS fully blocks methylation of both pap GATC sites by Dam at 23°C and directly prevents the transcription of both papBA and papI by forming large filaments along the pap regulatory DNA (34). This is the basis for the silenced state (Fig. 2A), which will be discussed in more detail below.

When the pap regulatory region is not completely coated with H-NS filaments, a multitude of additional regulatory sites become available for the other key protein players involved in pap regulation. The histone-like, nucleoid-associated protein Lrp can access a total of six known sites upstream of the papBA promoter (Fig. 1) (24). These sites divide neatly into two distinct sets of three sites that, when bound by Lrp, correspond to a particular state of transcription, either inactive (proximal set) or active (distal set), as illustrated in Fig. 3. The most proximal site (Lrp binding site 3) of the repressive proximal set lies between the −35 and −10 sequences of the papBA core promoter, and each sequential Lrp-binding site in the proximal set is 21 bp upstream of the preceding one (Fig. 3) (26, 28). A larger gap of 25 bp exists between the proximal and distal sets (between sites 1 and 6) (26). The distal set of sites is more tightly spaced, with 20 bp between sites 6 and 5 and 19 bp between sites 5 and 4 (Fig. 3) (26). The basis for the specific observed spacing between each Lrp-binding site became clearer when the structure of Lrp was identified, as shown in Fig. 3A (36). The difference in intersite spacing between the distal sites and the proximal sites may contribute to the difference in binding affinities and/or regulatory effects due to strain on either Lrp or the DNA (or both).

FIG 3.

Lrp binding to the pap promoter DNA. (A) General crystal structure of the Lrp octamer generated from PDB entry 2GQQ in the absence of DNA. The green regions correspond to the regulator of amino acid metabolism (RAM) domain, which also allows for Lrp oligomerization, and the blue regions correspond to the helix-turn-helix (HTH) region of Lrp that is responsible for binding to DNA. (B) The Lrp octamer cooperatively binds to sites 1, 2, and 3 in the pap promoter region, causing DNA bending. When in this position, pap is in its inactive, or Phase OFF, state. (C) The Lrp octamer cooperatively binds to sites 4, 5, and 6 in the pap promoter region, causing DNA bending again, but now leaves pap in the active, or Phase ON, state.

Two potential GATC target sites for Dam-mediated methylation are centered 50.5 and 152.5 bp, respectively, upstream of the papBA transcription initiation site (Fig. 1). Each occurs within the central Lrp-binding site of the proximal and distal sites (site 2 and site 5) and can modulate Lrp binding depending on its methylation status (20, 28, 29, 31).

Closer to papI, there is a single CRP binding site between the two promoters, centered 215.5 bp upstream of the papBA transcription start site and 115.5 bp upstream of the papI start site (Fig. 1) (43, 44).

PapB has three clusters of binding sites within the pap regulatory DNA (Fig. 1). The center of the most proximal and highest-affinity PapB binding site is 71 bp upstream of the papI start site, with 7 to 9 more distal and lower-affinity sites immediately adjacent, bordering the CRP binding site; this cluster of sites is referred to as PapB binding region 1 (35). It is thought that oligomeric binding of PapB within PapB binding region 1 activates papI transcription in cooperation with cAMP-CRP (35, 42, 45). Two additional PapB binding regions are present in the pap regulatory DNA: the second (PapB binding region 2), centered −4 bp from the papBA start site, has much lower affinity than region 1 and, thus, requires a much higher concentration of PapB to be occupied (42, 45); the third (PapB binding region 3) lies within the papB gene body and also has a much lower affinity than region 1 (42). Both of these additional sites are thought to inhibit papBA transcription, potentially by interacting with one another to form a loop of DNA that prevents RNAP association (45).

Two proteins have been shown to modulate thermoregulation: the histone-like nucleoid-associated protein H-NS was found to play a chief role in establishing a difference in transcription between 37°C and 23°C (34), and the Nα-acetyltransferase RimJ also was found to have a role, though less clear, in the thermoregulation of pap transcription (46, 47). The use of temperature to establish the roles of these proteins will be discussed throughout.

THE STRUCTURING PROTEINS: H-NS AND Lrp

The overall organization of the pap regulatory region is dictated by the interplay of two histone-like structuring proteins, H-NS and Lrp, both of which can profoundly alter the structure and accessibility of the pap promoter DNA under appropriate conditions. H-NS is a highly abundant (∼20,000 monomers per cell) histone-like global regulatory protein in Gram-negative enterobacteria that binds AT-rich sequences and intrinsically curved DNA (48–53). Upon binding to an AT-rich sequence, H-NS can oligomerize linearly along that stretch of DNA or form bridges with other H-NS/DNA nucleoprotein complexes (52, 54–56). The formation of these stiff nucleoprotein filaments allows H-NS to establish transcriptionally silent regions, including those covering many prophages (57–60). H-NS responds to multiple environmental variables, including temperature, osmolarity, and pH; H-NS expression also autoinhibits as temperatures rise from 23°C to 37°C (38, 61–63). This autoinhibition of H-NS when conditions match that of a human host accounts for changes in expression of over two-thirds of the temperature-regulated genes in E coli K-12 bacteria (64). Consistent with these findings, H-NS inhibits pap transcription in response to low temperature, high osmolarity, glucose as the carbon source (versus glycerol), and enriched (LB) media (11).

Lrp is an abundant (∼3,000 dimers/cell) DNA-binding protein in E. coli that affects the expression of over one-third of the genome in response to nutrient availability (65–69). The crystal structure of Lrp has been solved, revealing an octameric assembly consisting of a tetramer of dimers (36), as schematized in Fig. 3A. While the structure was solved in the presence of a small double-stranded DNA oligonucleotide corresponding to Lrp binding site 2 in the pap regulatory region, there was no detectable electron density in the structure for the DNA bound to Lrp (36). Interestingly, E. coli Lrp has been shown to exist in equilibrium between an octameric state and a hexadecameric (16mer) state in cells, where excess leucine shifts the equilibrium toward the octameric form (39, 70). This change in oligomeric state may play a role in regulating Lrp activity at target promoters.

Lrp and leucine together regulate target genes via one of six modes: all combinations of Lrp either activating or repressing a gene and leucine acting in a mode that is independent, concerted, or reciprocal (i.e., leucine has no effect, leucine augments Lrp’s effect, or leucine inhibits Lrp’s effect, respectively) (71–73). However, the majority of Lrp-regulated genes identified thus far seem to be regulated via the reciprocal mode (69, 73), in which the presence of leucine or other effectors antagonizes the effects of Lrp. In addition to leucine, alanine and methionine also interact with Lrp to alter gene expression (74, 75). It is not yet clear how Lrp is able to regulate its targets via six different modes, and it is not yet possible to predict how Lrp will act at any given target on the basis of target DNA sequence alone.

It has been widely stated in the literature that Lrp regulates pap expression in a leucine-independent manner, as neither exogenous leucine nor alanine have so far been shown to significantly affect regulation of the pap region (24). However, a mutational analysis of Lrp revealed that mutations to residues E133 and T134 of Lrp block pap phase variation without affecting the ability of Lrp to bind PapI or DNA (76). In the E. coli foo operon, which shares many regulatory similarities with pap but has been shown to be regulated by leucine and alanine, L131 and L136 Lrp mutations prevent repression of phase variation by either leucine or alanine (74). Additionally, the Calvo laboratory discovered that Lrp L136 mutations are insensitive to leucine at the ilvIH promoter but are not affected in their ability to bind ilvIH DNA (77). Given the proximity of L131 and L136 to E133 and T134 in Lrp, the similarities between the regulatory regions of pap and those of many other pap-like fimbrial operons that are responsive to leucine and alanine, and the as-yet-unidentified reason for Lrp E133 and T134 mutations preventing phase variation, it may be worthwhile to revisit the question of whether Lrp is actually insensitive to leucine and alanine at the pap promoter, especially in a uropathogenic E. coli strain.

Activation mutants of Lrp (V76A, F90L, and T119I) prevented the expression of papBA, suggesting that Lrp is required not only to correctly bend the pap regulatory DNA but also to recruit RNAP for transcriptional activation of papBA (78). Indeed, Lrp was thought to be an essential activator of pap transcription. However, following an epistasis analysis with H-NS, it was found to be nonessential (34). Within the context of pap phase variation, Lrp was long studied as a methylation blocking factor due to its ability to protect two separate GATC sites (Fig. 1) from Dam methylation (29, 34). As mentioned above, Lrp specifically binds two sets of three sites that determine pap phase, where Lrp binding to the higher-affinity sites 1 to 3 is repressive and Lrp binding to the lower-affinity sites, 4 to 6, is activating (Fig. 3B and C) (26). Structural analysis reveals that Lrp binding of the proximal and distal sites is almost certainly mutually exclusive; the bending of DNA by the Lrp octamer may render the unbound sites incompatible with further Lrp binding (36, 79). In support of this hypothesis, only very high, nonphysiological concentrations of Lrp result in simultaneous binding of both sets of sites (26, 31).

The highest-affinity Lrp binding sites are 1, 3, and 5, in descending order (Fig. 1), whereas the lower-affinity sites (2, 4, and 6) differ little in their relative affinities (20). The higher-affinity sites have 9 to 12 times greater affinity than the lower ones (20). Lrp’s affinity for the sites in the pap regulatory region (and possibly other sites on the chromosome) can be further modulated by binding of the accessory factor PapI. The repressive proximal set of Lrp binding sites contains the highest-affinity sites, such that Lrp preferentially binds this set without assistance from other factors (20, 24, 26, 28, 80). The presence of stably bound Lrp at high-affinity site 1 likely contributes to the formation of an octamer that binds cooperatively to the regulatory region. Recent examination of the specific relationship of Lrp’s N-terminal DNA binding region to each Lrp binding site demonstrated that Lrp forms specific complexes with pap sites 1 and 3, yielding a 2:1 Lrp-to-DNA molar ratio, independent of PapI (80). In contrast, site 5 forms only nonspecific interactions with Lrp unless PapI is present, whereupon it yields a specific tightly bound complex in a 2:1:1 (Lrp:PapI:DNA) ratio (80). Consistent with this finding is the observation that the addition of PapI substantially increases Lrp affinity for the distal set of sites (which includes site 5) (26, 31, 80). The balance of Lrp affinity between its binding regions within pap DNA suggests that Lrp alone typically will bind the proximal site, but translocation to the distal site can be enhanced and stabilized by PapI.

Since binding at site 3 would occlude the −35 and −10 core promoter elements of papBA, Lrp bound to this region should prevent RNAP binding, providing a basis for the inactive state (Fig. 2B) (27). The activating, more distal set of sites shown in Fig. 2C has 2 to 5 times lower affinity for Lrp, which binds poorly without the assistance of PapI (26, 31). papBA activation (and Lrp’s role therein) is highly complex and will be addressed throughout this review.

An epistasis analysis with Lrp and H-NS demonstrated the role of thermoregulation and pap DNA-protein binding on papBA expression. Some papBA transcription occurred in the hns mutant (10 to 15% of wild-type levels), but no difference was observed between cells grown at 23°C or 37°C (34, 78). In the absence of H-NS, papBA cannot be silenced, but Lrp is likely bound to the proximal sites (the inactive state), greatly inhibiting papBA transcription (Fig. 2B). In an lrp deletion mutant, very little transcription of papBA occurred at either temperature, suggesting that Lrp is needed to displace H-NS and relieve silencing (34, 78). Hence, loss of binding of either nucleoid-associated protein (H-NS or Lrp) resulted in impaired temperature-associated regulation of phase variation. Intriguingly, the lrp hns double mutant exhibited substantial transcription of papBA (60 to 88% of WT) but again lacked phase variability in response to temperature (34, 78). Thus, in the lrp hns double mutant, the papBA promoter is in an active state but exhibits phase-less transcription: papBA can be transcribed because it is not silenced by H-NS, and Lrp does not occlude the core promoter. The level of papBA transcription in a double lrp hns deletion mutant is likely not quite as high as that of the wild type because there is no Lrp to bind the distal region (and, thus, bend the DNA to recruit CRP and RNAP to bind at the papBA promoter), and Lrp cannot assist in recruitment of RNAP directly through its activation surface. Based on these findings, we can infer that in wild-type cells, H-NS coats and silences the entire pap regulatory region (Fig. 2A), and that Lrp can both participate in eviction of the H-NS filament to relieve silencing of pap and (depending on its conformation and accessory factors) act as either a repressor or activator of papBA transcription.

The model described above is also supported by several methylation protection assays (11, 34). Mutants lacking Lrp exhibited enhanced methylation protection of both GATC sites (present at Lrp binding sites 2 and 5) at 23°C; at 37°C, protection was greatly reduced (11, 34). These findings indicate that silencing of pap by H-NS is more effective at lower temperatures and that H-NS likely obscures the entire pap promoter at those temperatures. Thus, H-NS silencing should inhibit not only papBA transcription but also that of papI. Indeed, a deletion of hns (previously known as drdX) was found to increase papI and papBA transcription (44). Conversely, hns mutants demonstrated decreased methylation protection of the distal site at 23°C, indicating that Lrp primarily binds the proximal set at the lower temperature if H-NS is not present, which is consistent with the inactive (but not silenced) state (44).

Evidence of increased methylation (decreased protection) of both GATC sites in the pap regulatory region was present in both lrp and hns mutants (34). This resulted in the DNA being resistant to cleavage by GATC-specific, N⁶-methyladenine (N⁶mA)-sensitive restriction endonucleases, as revealed by the presence of a large band of pap DNA corresponding to methylation at both GATC sites, while the corresponding band was minimally present in wild-type strains (34). In the lrp hns double mutant, the fully methylated band was the only one detected, supporting the notion that no GATC protection from Dam methylation within pap occurs in the absence of these two nucleoid-associated proteins. The presence of this band in the hns mutant at 37°C, however, indicates that Lrp is unable to outcompete Dam in all cells. lrp mutants also yield methylation patterns consistent with Dam methylation at both GATC sites; given the low level of papBA transcription by lrp mutants, it is likely that H-NS still binds methylated pap DNA (34).

The antagonistic relationship between the nucleoid-associated proteins H-NS and Lrp helps define the phase-variable behavior of pap. H-NS oligomerization within pap promoter DNA, a prerequisite for pap silencing, is clearly more favored at 23°C than at 37°C (34). This view is supported by the observation that transcriptionally active cells grown initially at 37°C and then transitioned to 23°C become completely silenced after one generation; even cells retaining an active methylation pattern become transcriptionally silent at the papBA promoter (34). A similar example of antagonism between H-NS and Lrp is observed at the ilvlH operon, where H-NS was demonstrated to directly inhibit Lrp-mediated activation (81). This antagonistic relationship between Lrp and H-NS may help to explain the mechanism for the transition between the silenced and inactive states, which we will discuss later in the review.

PHASE-MODIFYING ENZYMES: Dam AND PapI

DNA adenine methyltransferase (Dam) is responsible for recognizing and methylating over 15,000 genomic 5′-GATC-3′ sites within a single generation and, in doing so, plays critical roles in chromosomal replication, gene expression, mismatch repair, and nucleoid structure (82). It is important to recognize that, due to its palindromic recognition motif, each Dam “site” necessarily includes two potential points of methylation (the adenine present on each strand) and, thus, at any given time each GATC site can be unmethylated, hemimethylated, or fully methylated. The level of Dam is tightly regulated such that each cell contains about 130 molecules (83). During mismatch repair, MutS, MutH, and MutL recognize a lesion in the DNA and chew it back to the closest hemimethylated GATC site, allowing the repair machinery to determine which strand was the parental strand (84). Additionally, there is an abundance of GATC sites near the origin of replication, and the precise timing of Dam methylation here coordinates the timing of chromosomal replication initiation by DnaA (85, 86). The presence of GATC sites at promoters and regulatory regions also allows Dam methylation to regulate gene expression by altering the affinity of transcription factors (e.g., CRP and Lrp) to that region of DNA relative to the unmethylated or hemimethylated GATC sequence (87, 88). Indeed, the pap regulatory region is a canonical example of the regulatory role played by Dam methylation in establishing and maintaining transcriptional states.

As mentioned above, the pap regulatory region contains two GATC sites within the central site of each set of Lrp binding sites (i.e., sites 2 and 5 in Fig. 1). Binding of Lrp to either the proximal or distal set (which are likely mutually exclusive) inhibits the ability of Dam to methylate that respective GATC site (89). Conversely, Dam methylation at either GATC site inhibits the ability of Lrp to associate with that site (89, 90). Recent evidence suggests that although methylation affects Lrp binding to sites 2 and 5, it does not completely preclude binding. Instead, it alters binding affinity, which provides a potential explanation for how the binding state of Lrp on pap DNA may change (20).

To test the effects of Dam and Lrp on each other and on papBA expression, genomic papB-lacZ fusion strains were grown on M9 glycerol agar plates containing 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal), where a blue colony indicates the original cell was active and a white colony indicates an inactive initial cell, and pap DNA methylation from these colonies was assessed via digestion with methylation-sensitive restriction enzymes (28). Replating of inactive single colonies grown at 37°C, a temperature known to limit the methylation-blocking effects of H-NS, exhibited minimal papBA transcription, as indicated by 99% of resulting colonies being white (28). Consistent with the homogeneity of low papBA expression, these inactive colonies had a homogenous methylation pattern consisting of a fully methylated distal GATC site and fully unmethylated proximal GATC site (28, 31). Similarly, mutations that prevent methylation of the proximal GATC site (mutation to GCTC) resulted in a homogeneous, inactive state (90).

In contrast, replated active single colonies grown at 37°C exhibit a heterogeneous mixture of methylation patterns (28, 31). Up to 40% of the methylation patterns derived from active colonies resembled those of inactive cells, while an equal amount of the methylation patterns derived from active colonies were of the opposite, active pattern (i.e., fully methylated proximal site, unmethylated distal site) (28). The remaining patterns contained some combination of hemimethylated or unmethylated GATC sites and may be due to recent DNA replication (hemimethylation) or residual H-NS silencing (fully unmethylated) (28). Despite the drastic heterogeneity of methylation states derived from active colonies, replating single active colonies and growing at 37°C resulted in 75% transcriptionally active (blue) colonies (28). Mutations preventing methylation of the distal GATC site (by mutating to GCTC), generating only the active methylation patterns, resulted in a homogenous, transcriptionally active state (all resulting colonies were blue) (90). This discrepancy in methylation pattern and LacZ activity of replated single colonies may be mostly due to a combination of DNA replication and cell division affecting methylation state over time and the high sensitivity of the blue-white screen for low levels of lacZ transcription. However, a few other factors may also contribute to this discrepancy of methylation homogeneity between inactive and active cells.

The effects of methylation on Lrp binding to the proximal and distal sets are not equivalent: Dam methylation at site 5 decreases the affinity of Lrp binding to the distal set by 5 to 7 times; this compares with only a minimal decrease (10%) in Lrp affinity for the proximal set following methylation of site 2 (20, 80). Consistent with this finding, footprints of pap DNA fully methylated at both GATC sites demonstrate Lrp binding at only the proximal, inactive set (90). However, the homogeneity of inactive methylation patterns suggests that Lrp is also able to bind and protect the proximal GATC site, blocking Dam methylation. PapI, a positive regulator of papBA transcription, may allow Lrp to bind more efficiently to a hemimethylated distal site shortly after DNA replication, activate papBA transcription, and yield the heterogeneous patterns associated with activation (26, 28, 31, 34). Presumably, if Lrp is stabilized at the distal site, subsequent cell divisions would allow for Lrp to continue protecting the distal site from methylation and render the proximal site open to full methylation by Dam.

PapI is an 8.8-kDa local transcription factor with a helix-turn-helix motif that affects Lrp affinity for each central binding site of the proximal and distal set of binding sites of the pap regulatory region (i.e., sites 2 and 5, respectively) (26). PapI binds poorly to Lrp or pap promoter DNA individually (30, 91). However, the affinity of PapI for Lrp-DNA complexes is significantly greater (20, 30, 80, 91). Once bound, PapI forms a specific complex with both Lrp and pap DNA that is distinct from the Lrp-DNA complex (80, 91). PapI forms these complexes with greatest affinity at Lrp binding sites 2 and 5 (Fig. 1), with site 5 (distal) having approximately twice the affinity of site 2 (proximal) (26, 30, 80, 91).

Despite the higher affinity for the distal site, it is probable that the PapI-Lrp-pap DNA complex initially forms over site 2. Whether the pap DNA is unmethylated (as would be the case immediately after eviction of H-NS) or methylated at both GATC sites, Lrp has a higher affinity for the proximal set of binding sites (90). Thus, if starting with inactive cells, PapI likely encounters Lrp first at site 2. In the meantime, site 5 would have been unprotected from Dam methylation, yielding the methylation and binding pattern associated with inactive cells (Fig. 2B). As discussed previously, Dam methylation alters the affinity of Lrp binding to the proximal and distal sets of sites (31). The fully methylated distal site has very low affinity for Lrp alone (20, 31). In the presence of PapI, the affinity of Lrp for a methylated distal site increases 5 to 7 times, while the affinity of Lrp for a fully methylated proximal site decreases in the presence of PapI (20). Furthermore, Lrp bound the distal site of wild-type pap DNA only in the presence of PapI and methylation of both GATC sites (90). In mutants that prevent proximal GATC methylation and, thus, force a locked inactive state, PapI overexpression shifts methylation protection over to the distal GATC site (90). Therefore, PapI brings the difference in Lrp affinity between the two sets of binding sites to a minimum, increasing the probability of cells transitioning from a proximal-bound to distal-bound Lrp.

An interesting mechanism was suggested for translocation of Lrp between generations; in the presence of PapI, Lrp was found to have increased affinity for hemimethylated pap DNA (20). PapI specifically increased Lrp affinity for hemimethylated DNA over fully methylated DNA (∼8-fold) and also had greater affinity for hemimethylated bottom-strand pap DNA (∼4-fold over top strand) (20). Thus, the period of time between replication of the pap regulatory region and full methylation of the distal site provides a period during which binding of Lrp to the distal sites is particularly likely, providing a path to alteration of pap phase if the proximal site is transiently unoccupied or if the proximal Lrp subsequently dissociates. Given that the two hemimethylated strands have different affinities for PapI-Lrp, a heterogeneously active population of daughter cells is possible and potentially establishes bet-hedging with regard to pap expression.

An alternative hypothesis that does not require replication involves the precise balance of Lrp affinity between proximal and distal sets of binding sites to allow a single Lrp octamer to enter a pathway for translocation without complete dissociation from the pap regulatory DNA. This balance is specifically established by the innate affinity of pap DNA for Lrp and is significantly modified by the actions of PapI and Dam. This hypothesized translocation mechanism is supported by low-affinity Lrp binding between the two sets of Lrp binding sites (i.e., between sites 1 and 6) (26). This low-affinity binding indicates that a single Lrp octamer binds to the intermediate region of pap promoter DNA and translocates between the proximal and distal sets. If such a mechanism is possible, a single Lrp octamer would presumably sample all possible binding states (proximal versus distal, with versus without PapI), with the equilibrium between states dictated by methylation status and PapI concentration.

Evidence for intermediate region binding specifically arises from DNase I footprints, comparing protection of wild-type and mutant pap DNA by various levels of Lrp in the presence or absence of PapI (26). A small region of wild-type pap DNA (the intermediate region), located between Lrp binding sites 1 and 6, was protected at 1.5 μg/ml Lrp without PapI and 2.0 μg/ml Lrp with PapI. A mutation within Lrp site 2, which locks cells in an active state, resulted in the reverse phenotype with protection of this intermediate region at 1.75 μg/ml Lrp without PapI and 1.5 μg/ml with PapI. Conversely, a mutation in site 3, which also results in a “locked on” phenotype, showed better protection of the intermediate region without PapI (at 1.0 μg/ml Lrp) than with PapI (protection at 1.5 μg/ml Lrp). Lastly, a mutant of site 5 with a “locked off” phenotype exhibited poor protection of the intermediate region (at 2.0 μg/ml Lrp) that did not change with the addition of PapI. Thus, it appears that protection of this intermediate region is inhibited either by preventing Lrp octamer formation over the proximal set of sites or by limiting transitions between the two sets, although the changes in intermediate region protection are generally small in magnitude and occur at far higher Lrp concentrations than binding of the main sites. The fact that high Lrp concentrations can produce intermediate binding is still relevant, because these concentrations are mimicked by the persistence of the Lrp octamer within the pap promoter, where it can engage a multitude of binding configurations as a consequence of its proximity established through high affinity to pap-DNA.

Activation of papBA transcription likely requires that the Lrp octamer first forms over the proximal set and then transitions to the distal set. The proximal set (with or without methylation) has higher affinity for Lrp, as stated previously (80). Consistent with this, a mutation in the intermediate region, between sites 1 and 6, locked cells in the inactive state (26). If the octamer could initially form over either set with reasonable efficiency and transitioning from proximal to distal binding sets was not required, the intermediate region mutation should have produced a heterogeneous mixture of active and inactive cells. Although a footprint analysis of this specific mutant was not shown, the presence of a homogenous phenotype consisting of transcriptionally inactive cells supports the hypothesis that the intermediate region is required for papBA activation. In all likelihood, the actual mechanism for Lrp translocation between the proximal and distal sites consists of both DNA replication to generate hemimethylated GATC sites and the transition of Lrp directly from the proximal to the distal sites by transiently binding to the intermediate region, allowing for phase switching of pap expression.

Despite PapI’s tendency to balance Lrp affinity between the proximal and distal sets of binding sites, Lrp affinity for the proximal set appears greater in all cases, regardless of methylation status (20). Thus, some other interaction seems necessary to stabilize Lrp at the distal sites to prevent transitions back to the proximal site prior to the association of RNAP. As will be introduced in the next section, the actions of PapB and CRP may combine to establish a cooperative mechanism with Lrp to favor the formation of fully activated papBA transcription (Fig. 2C).

POTENTIATING ELEMENTS: CRP AND PapB

PapB is an 11-kDa transcription factor that regulates adhesin expression, including the repression of type 1 fimbriae and activation of the P-pilus (45, 92). Most studies of PapB were performed with multicopy vectors and, thus, should be judged with some care; however, their results lead to the following conclusions. PapB concentration determines papBA transcription: at low to intermediate levels, PapB functions as an activator; once its concentration reaches a certain threshold, PapB represses papBA transcription (42, 45). Within pap, PapB binds three distinct stretches of DNA (Fig. 1); the first covers 40 to 90 bp upstream of the papI promoter; the second is near the papBA start site, spanning only about 26 bp; and the third, also about 26 bp, is downstream of the papBA start site within the papB open reading frame (35, 42, 45).

The first PapB binding region is thought to activate papI transcription and has 8 to 10 sequential binding sites. The first PapB binding site within this region, which lies most proximal to the papI start site, was found to activate papI transcription (35, 42, 45). PapB was also found to oligomerize between the papI start site and the CRP binding site, binding 7 to 9 more monomers, depending on the PapB concentration (35, 42, 45). Deletions of the first PapB binding region resulted in reduced transcription of both papI and papBA (44). The second and third binding regions likely inhibit papBA expression via steric hindrance of RNAP at the promoter by looping DNA (35, 42, 45). PapB has a much lower affinity for these two inhibitory binding regions than the first PapB binding region, helping establish a sequence of activation followed by repression within pap that tracks with PapB concentration (Fig. 2) (42).

CRP regulates expression of over 100 genes in response to cAMP levels by either activating or repressing transcription (93, 94). The cAMP-CRP complex plays a significant role in the activation of both papBA and papI transcription in response to cAMP levels. Given the locations of the high-affinity PapB site (centered at −71) and the CRP site (−115.5), it is likely that these two proteins make direct contact with the C-terminal domain of the α subunit (αCTD) of RNAP, stabilizing the RNAP-promoter complex (95). The interaction between CRP and PapB may also enhance binding to their respective DNA sites. The degree to which PapB oligomerization in its first binding region facilitates the formation of this papI promoter complex remains unknown. However, deletion of the first PapB binding region abolishes papI transcription and reduces papBA transcription by 75% (44). Deletions of the CRP binding site also abolish papI transcription and dramatically decrease papBA transcription to only 1% of wild-type levels (44). In cells that lack H-NS, however, papBA transcription is restored even in the presence of these CRP binding site mutations (44).

Given that the interaction of CRP with the papBA promoter requires PapI to allow Lrp translocation to its distal binding site, it is conceivable that CRP interaction with the papI promoter occurs first as cells transition from an inactive to an active state. Therefore, the best signal for assessing H-NS displacement would be papI transcription. From studies of hns deletions, it is evident that CRP plays a major role in activating papI transcription (44). However, Lrp deletions were also found to result in a 6-fold decrease in papI transcription (31). It is possible that Lrp binding to relieve H-NS silencing has the secondary effect of potentiating papI via trace production of PapB even in the papBA inactive state (Fig. 2B). Increased levels of PapI would, in turn, increase papBA transcription by modifying Lrp’s affinity to favor the active distal position. Rising levels of PapB result in an extended oligomer within the first PapB binding region that could stiffen the DNA such that CRP no longer makes direct contact with RNAP to activate papI transcription (Fig. 2C); the result would be diminished papI transcription.

CRP also appears to directly stabilize RNAP at the papBA promoter. Mutants lacking CRP show only basal papBA transcription even with the addition of a plasmid that constitutively expresses papI, supporting the hypothesis that CRP plays a PapI-independent role in activating papBA transcription (43). Mutations within Activation Region 1 (AR1) of CRP nearly eliminate papBA transcription, whereas mutations within the AR2 region have no effect (43). These results indicate that papBA contains a CRP-dependent class I promoter (43).

However, the CRP binding site is centered 215.5 bp upstream of the papBA transcriptional initiation site, too far away for AR1 to interact directly with the αCTD. This problem may be solved by the dramatic bending of pap DNA caused by the Lrp octamer bound to the distal active set of sites; such bending could bring CRP into close proximity to the αCTD (36, 43). The hypothesis that Lrp facilitates CRP’s interaction with RNAP at the papBA promoter is supported by the finding that insertion mutations between the most distal Lrp binding site and the CRP binding site result in variable inhibition of papBA transcription based on the length of the insertion; both 1- and 2-helical turn insertions had little effect on transcription, whereas 0.5- and 1.5-helical turn insertions resulted in complete inhibition. The importance of specific spatial relationships between CRP and Lrp suggests that a specific promoter complex is formed to regulate papBA transcription (43). The papBA promoter architecture may exert further stabilizing effects on Lrp in the distal position, particularly if Lrp is also making contact with RNAP, which could help prevent Lrp translocation to the inactivating proximal sites.

Taken together, the actions of CRP and PapB within the pap promoter appear to regulate a balance between the activation of either papI or papBA in a mutually exclusive manner. CRP appears to function with Lrp, either independently or cooperatively, to displace H-NS filaments and relieve silencing of both papI and papBA. Although divergent, the two promoters are interdependent, and each produces a factor that potentiates the activity of the other. Simultaneous removal of silencing from both promoters, via the actions of CRP and Lrp, would allow efficient activation of papI transcription in the presence of low PapB concentrations. The degree of oligomerization within the first PapB binding region (which responds proportionally to PapB concentration) could mediate a switch between the divergent promoters, with CRP serving as a mutually exclusive, activating element. Since Lrp-mediated bending of DNA (itself modulated by PapI and methylation status) appears required for CRP to interact with RNAP at the papBA start site, the activating role of Lrp may in part be to overcome the structural limitation of the promoter architecture and enable CRP-mediated activation. Because Lrp also potentially binds RNAP, this interaction with RNAP and indirectly with CRP may further stabilize Lrp in the distal position and inhibit translocation of Lrp back to the proximal sites.

RimJ AND pap PROTEIN ACETYLATION

Perhaps the most surprising element involved in pap regulation is the protein RimJ, which was originally characterized as a ribosome assembly factor that binds the 30S ribosomal subunit protein S5 and acetylates the N-terminal alanine to aid in 30S ribosome maturation (96). Interestingly, RimJ was shown to have an important role in the regulation of papBA transcription in response to multiple environmental variables (47, 97). In wild-type cells, papBA transcription is active at 37°C and silenced by H-NS at 23°C. In contrast, rimJ mutants lose thermoregulation, showing significantly impaired repression of papBA transcription at 23°C (47). At 37°C, wild-type and rimJ mutants have near-equal papBA activity. In wild-type cells, papBA transcription is inhibited by rich (LB) media, glucose (11 mM), or high osmolarity (310 mM NaCl versus 8.5 mM NaCl); the inhibitory effects of these stimuli are also reduced in rimJ mutants (47). Because similar stimuli also operate through H-NS, RimJ appears to function within the same global environmental response system, favoring pap expression in low-nutrient, low-salt environments. However, H-NS and RimJ likely function independently at low temperature, as a rimJ hns double mutant exhibits a greater loss of low-temperature-induced inhibition of pap transcription than either single mutant alone, and this double rimJ hns mutant is synthetically lethal in minimal media (47). An experiment examining RimJ overexpression demonstrated that RimJ-induced repression of papBA was only enhanced at 23°C (47), which coincides with the presence of H-NS. At 37°C, RimJ had no effect on papBA transcription, even when expressed at a level 3 orders of magnitude greater than that which produced 99% of RimJ-induced inhibition at 23°C (47).

How does RimJ exert control over papBA transcription? One plausible explanation is that RimJ acts through the ribosome to provide indirect control over translation of a pap-associated protein or proteins. Indeed, rimJ mutants produce much less total protein per cell than wild-type cells and, thus, have a significantly longer doubling time than wild-type cells, even in LB media (98). Furthermore, rimJ mutants have an impaired ability to initiate chromosomal replication, likely through reduced levels of DnaA due to impaired translation ability (98). It seems possible that the stress from impaired ribosome function in rimJ mutants allows for altered expression of Lrp, CRP, and/or H-NS, which could explain the ability of rimJ mutants to no longer repress papBA transcription at 23°C. It is also possible that the slower chromosomal replication rate in rimJ mutants at lower temperatures affects Dam levels in the cell, thereby affecting the rate at which GATC methylation can occur and, thus, preventing phase variation. More experiments are needed to test these hypotheses.

An alternative hypothesis for the role of RimJ on pap regulation involves its Nα-acetyltransferase (NAT) activity, which catalyzes the transfer of an acetyl group from acetyl-coenzyme A (acetyl-CoA) to the amino terminus (Nα) of a protein. There is some evidence that RimJ has acetylation substrates other than S5, although the exploration of additional RimJ substrates has been extremely limited since the mid-1980s (99, 100). Considering that RimJ and H-NS were shown to act independently, and that both CRP and Lrp appear to work in opposition to H-NS with regard to pap regulation, these proteins could be RimJ acetylation substrates. Indeed, rimJ mutants exhibited elevated papBA transcription at low temperatures but only if CRP was present (47). Recalling that CRP pap DNA binding site deletions in an hns deletion background showed greater papBA transcription than did the wild type, and that deletion of CRP abolishes papI transcription, one hypothesis is that RimJ inhibits CRP’s ability to disrupt H-NS oligomerization. Alternatively, considering the 6-fold decrease in papI transcription associated with Lrp deletions, RimJ may affect Lrp’s ability to compete with H-NS for binding of pap DNA. Further study is required to determine whether Lrp, CRP, or some other protein involved in pap regulation is a direct RimJ acetylation target, and, if so, what regulatory role the resulting acetylation plays.

Another form of protein N-acetylation involves the transfer of an acetyl group to the ε-amino (Nε) group of a lysine residue within a protein. Nε-lysine acetylation is a major regulatory modality that governs histone behavior in eukaryotes. Given the involvement of the histone-like nucleoid-associated proteins H-NS and Lrp, this form of protein acetylation could have analogous effects on pap transcription. In bacteria, two mechanisms for Nε-lysine acetylation have been identified: enzymatic and nonenzymatic. The former is catalyzed by an Nε-lysine acetyltransferase (KAT), such as YfiQ/PatZ, using acetyl-CoA as the acetyl donor (101–105). In contrast, nonenzymatic Nε-lysine acetylation occurs without the help of a distinct enzyme and uses acetyl phosphate (acP) as the acetyl donor (101, 102, 105–107).

Nα-acetylation is thought to be relatively uncommon in bacteria, as few substrates have been identified (104, 108). In contrast, Nε-lysine acetylation appears to be abundant. Recent global surveys reveal Nε-lysine acetylation across diverse bacterial phyla (105), including almost 20% of the E. coli proteome (106, 107, 109). Acetylated lysines have been reproducibly detected on three pap-associated proteins: CRP, H-NS, and Lrp (107, 109–111). Seven CRP lysines have been detected as acetylated; of these, four are regulated by acP. Six H-NS lysines have been detected as acetylated; of these, one is regulated by acP and three are sensitive to deacetylation by the deacetylase CobB (108).

It was recently reported that the binding of Lrp to DNA is affected by lysine acetylation. Three Lrp lysines have been detected as acetylated; of these, K10 is in the unstructured N-terminal tail (108). The N terminus of Lrp was not crystallographically resolved, consistent with the presence of a disordered N-terminal tail similar to the N termini of eukaryotic histones (36). Deletions or mutations within or near this unstructured N terminus significantly impair oligomerization and reduce DNA binding, both of which lead to decreased Lrp-dependent transcription (112). Another lysine located just C-terminal to this tail, K25, is nonenzymatically acetylated by acP (106); this acetylation is induced by exposure to glucose (109). K25 was not targeted by YfiQ/PatZ or CobB, and mutations to K25 did not significantly affect DNA binding (113). In contrast, K36 is reported to be reversibly acetylated by YfiQ/PatZ and CobB, and mutations to K36 significantly impaired DNA binding (113). Given that enzymatic and nonenzymatic acetylation of other proteins have been shown to influence protein stability and function, one can speculate that the acetylation of one or more lysines could affect the behavior of Lrp and alter pap activity. It is particularly intriguing that there are two acetylated lysines in close proximity; one depends on reversible acetylation through enzymes, and the other depends only on the flux of carbon through central metabolism.

Taken together, the effects of RimJ and lysine acetylation portray acetylation as a key mode of pap regulation that could enable high sensitivity to environmental and metabolic changes. In the following section, however, the explicit role of acetylation (both ribosomal and on lysine residues) will not be included within our model for pap regulation; in contrast to other pap-associated transcription factors, acetylation-linked transcription factors are much less studied in the context of pap, and very little is understood regarding the role of RimJ. RimJ will be included to demonstrate its known effects within pap, without mention of a direct or indirect substrate. Further study of RimJ and protein acetylation, both enzymatic and nonenzymatic, as well as the role of deacetylation, may help to explain the most intriguing properties of pap, such as thermoregulation and Lrp and CRP binding.

FORMING A MODEL FOR pap REGULATION

Revisiting the pap promoter after several important discoveries helps to further demonstrate its deep intricacy while at the same time raising new questions regarding how pap regulation is integrated with the cell’s surroundings and internal state. Figure 4 shows several well-supported aspects of pap regulation. A quick glance is enough to conclude that many interactions occur within this modest 330-bp stretch of DNA. In this context, a few general principles should be considered when forming a basis to model pap transcription. Chiefly, the topology of pap DNA and the variable binding of pap-associated transcription factors drive the transcription and activation of both papI and papBA. Each unique state of activity consists of a distinct bending of pap promoter DNA. Our model attempts to simplify the relationships between each particular regulatory state by organizing the activity of pap into a set of transitions that respond to environmental stimuli and the concentration of several pap-associated transcription factors.

FIG 4.

Molecular model for pap regulation. (A) Beginning from the fully silenced state, the H-NS nucleoprotein complex with the pap DNA is destabilized as the temperature increases upon host entry and is replaced by binding of CRP and/or Lrp, depending on nutrient conditions. Both papBA and papI transcription are off. (B) An Lrp octamer binds to the three proximal Lrp sites, while Dam methylates the distal GATC site. Basal levels of transcription of papBA allow for the production of a few molecules of PapB. The pap promoter is in the inactive state and still phenotypically Phase OFF. (C) PapB binds to its highest affinity site (PapB binding region 1) near the papI promoter and bends the DNA such that CRP can recruit RNAP for transcription of papI. PapI levels increase. The cells are still Phase OFF. Upon cell division, the fully methylated distal GATC site becomes hemimethylated, and Lrp may translocate along the DNA and transiently bind the three distal binding sites. PapI binds to the Lrp-distal site complexes and stabilizes Lrp in this position. (D) The movement of Lrp to the distal sites and stabilization therein permits Dam to access and methylate the proximal GATC site. Because Lrp is no longer obstructing the papBA promoter, the papBA levels increase and additional PapB begins to form a filament in PapB binding region 1. This state marks a transition point between the inactive and active states from the standpoint of regulatory logic. (E) Increased levels of PapB allow it to form a filament near the papI promoter, effectively shutting off the expression of papI and locking the cells in the active state. Lrp bound at the distal sites allows for bending of the DNA such that CRP can recruit RNAP for activation of papBA. The papBA expression is now at its highest, and the cells are phenotypically Phase ON. (F) PapB levels continue to rise until PapB binds to PapB binding regions 2 and 3, effectively shutting off papBA expression.

The proximity of the papI and papBA promoters makes it impossible to affect one promoter without impacting the other; a shared CRP binding site required to activate either promoter exemplifies this concept. Furthermore, both divergent promoters yield mRNAs coding for a protein that binds within pap, leading to differential expression of papI and papBA. This relationship can be more easily explored by defining conditions for the fully silenced state regarding the concentration of PapB and PapI and predicting the consequences on pap promoter activities as the levels of these proteins change during the process of pap activation. The steps outlined in this model likely occur in reverse as cells transition from the active to the inactive state. However, given that cells grown at 23°C for just a single generation resulted in the silencing of papBA, we suspect H-NS filaments can form at any time over the entire pap promoter, given appropriate environmental stimuli, to generate the fully silenced state (34). Presumably the formation of such an H-NS filament erases all previous aspects of the regulatory state related to binding of other proteins, and any state information contained in the methylation pattern will be rapidly diluted out over subsequent cell divisions.

We begin our proposed model for pap regulation with the fully silenced state (Fig. 4A). H-NS binding and subsequent oligomerization silences the entire pap promoter region. When H-NS is bound, both PapI and PapB levels should be at their lowest. To enable transcription from either papI or papBA, H-NS must be displaced. Mutants containing partial or complete deletions of the papB gene showed a significant reduction of H-NS-mediated silencing (44, 45); therefore, the papBA side of the H-NS oligomer likely extends to include part or all of the papB open reading frame. Regarding the papI side of the H-NS filament, no evidence has been presented to suggest that DNA downstream of the papI transcription start site affects H-NS binding within pap. Thus, within our model, the papI side of the H-NS filament was placed slightly upstream of the papI start site to reflect the known inhibitory effect that H-NS oligomerization exerts on papI transcription (44). Based on the independent epistasis analyses of Lrp and CRP with H-NS, we conclude that any protein that binds within pap is an H-NS competitor by default. However, given that neither PapB nor PapI is present under conditions of H-NS silencing, the displacement of H-NS filaments must require binding of the primary activating factors, Lrp and/or CRP, depending on nutrient availability. RimJ, conversely, appears to function similarly to H-NS, as its deletion results in decreased silencing of the pap promoter under conditions that normally enhance the activity of H-NS. Whether RimJ acts directly on CRP, Lrp, and/or an unknown acetylation substrate, or indirectly through impaired DNA replication and reduction in translation ability, has not been determined and remains a promising question for future research.

Once relieved of H-NS silencing through a combination of increasing temperature and Lrp/CRP binding, the activity of both promoters depends on the concentration of PapB and PapI. In the absence of H-NS binding, papBA must have a basal rate of transcription for PapB to bind PapB binding region 1 and activate papI in conjunction with CRP (inactive state) (Fig. 4B and C). Shortly after the relief of H-NS silencing, the affinity of Lrp to the proximal sites should be much higher than that of the distal binding sites, because PapI is present at low levels. As a consequence, Dam has a greater opportunity to methylate the unprotected distal GATC site; this establishes the inactive papBA state (Fig. 4B). The exact timing of Dam methylation relative to papI activation remains unclear, and in practice there may not be a single unique order of events leading from the states shown in Fig. 4A to C. Once transcription of papI becomes activated, PapI levels rise and increase the likelihood that Lrp will translocate from its proximal position to the distal position by stabilizing Lrp at the distal sites (Fig. 4C and D). There is strong evidence to suggest that DNA replication must occur to generate hemimethylated pap DNA to allow Lrp to translocate from the proximal to the distal sites (whether by rolling along the DNA or by unbinding the proximal sites and binding to the distal sites). Upon replication and cell division, as PapI levels continue to increase, PapI helps to stabilize Lrp at the distal site (Fig. 4D). While CRP is not activating papBA yet, Lrp is no longer occluding binding of RNAP to the papBA promoter, allowing for moderate papBA expression. At this point, however, papBA is not fully active because the close proximity of the CRP binding site to the papI promoter combined with DNA bending afforded by high-affinity PapB binding favors activation of the papI promoter by CRP, and CRP activation of papI and papBA is mutually exclusive.

As PapB levels rise, secondary to the increased affinity of Lrp to the distal position due to the stabilization by PapI, extended PapB oligomerization between the CRP binding site and the papI transcription start site will reach a point at which CRP cannot contact RNAP at the papI promoter. Consequently, CRP becomes available to interact with or recruit RNAP at the papBA promoter; distally bound Lrp helps bend the DNA to allow for this interaction (Fig. 4E). It is also possible that Lrp helps recruit RNAP to the papBA promoter given the inactive papBA state of Lrp activation mutants (78). Given the mutual exclusivity of CRP activation within the pap promoter region, papBA activation implies inactivation of papI; therefore, PapI levels would be at their highest immediately following the inactivation of papI, maximizing the affinity of Lrp to the distal sites and, thus, the likelihood that Lrp will be bound there. Dam then has a greater opportunity to methylate the proximal GATC site and stabilize the fully active state of papBA transcription (Fig. 4D). Again, the timing for Dam methylation of the proximal GATC site is unknown relative to other steps in this model but presumably can occur at any point following activation of papI transcription, when PapI stabilizes Lrp at the distal sites, protecting the distal GATC site from Dam methylation. PapB levels eventually rise to a maximal concentration that enables binding of PapB to PapB binding regions 2 and 3 (Fig. 4F). PapB bound within this region is thought to produce a loop of DNA that prevents RNAP association, effectively halting further papBA transcription (42, 92). This may function as a switch to halt transcription while preserving the CRP-Lrp-papBA promoter architecture, which would effectively provide a feedback loop fixing the maximum level of papBA expression.

The utility of this step may be most relevant during periods of replication and growth. Assuming that a daughter cell inherits the final concentrations of both PapB and PapI from a fully active parent, prior to replication, the volumetric expansion of the daughter cell may eventually cause the concentration of PapB to fall below the threshold required for binding and looping DNA between PapB binding regions 2 and 3. Therefore, pilin expression can be matched to growth, utilizing the maximal PapB concentration as a threshold to guide papBA expression in active progeny. In addition, the inherited active methylation pattern would further serve to direct progeny of active cells to become active postreplication. In the event that pilin biogenesis becomes dysregulated and leads to cell stress, the response regulator, CpxR, has been shown to bind and repress the entire pap promoter region regardless of methylation status, displacing all pap-associated activation machinery (114, 115). Phosphorylated CpxR has also been shown to bind and repress pap transcription under alkaline conditions, providing another means of incorporating environmental stimuli into pap regulation (115). However, once recovered from stress, the regulation of pap may resume as a function of PapB and PapI concentration and any residual methylation patterns present.

Another possible mode to halt inherited behavior (not shown in the model) was recently identified as a small RNA known as papR, activated by Lrp, that inhibits PapI production by binding papI transcripts (116). papR is not conserved and is only present in a subset of uropathogenic E. coli strains (116); little else is known about the regulation of papR and its role in pathogenesis. Additional modes of regulating pap activity, independent of PapI and PapB, are exciting clues about the sensitivity of pilin expression to environmental and cellular conditions.

CONCLUDING REMARKS

The pap promoter region presents an interesting study for how the regulation of virulence factors can be matched to environmental conditions. Gaining insight into the finely regulated activity of pap has required an extraordinary effort from researchers working in diverse areas of molecular biology, genetics, and structural biology. The present work has organized past discoveries in pap and generated a model that aims to stimulate and support further work. The temporal relationships between the interactions of pap-associated transcription factors with each other and with pap DNA are critical to understanding how the activation and inheritance of an important virulence phenotype occurs with impressive consistency.

Exploring the initial conditions of pap-associated transcription factors during and after silencing will help further differentiate the H-NS-induced silenced state from the papBA inactive state; we predict that PapI levels and overall accessibility of the pap DNA will serve as the primary distinguishing factors between these two states. Correlating PapB and PapI levels will further characterize the initial conditions of the postsilenced state and may expose conditions under which CRP switches from favoring activation of the papI promoter to instead interacting with RNAP at the papBA promoter. Direct, in vivo experimentation on specific concentrations of PapB and PapI may be difficult. However, expanding on current stochastic models will be helpful in guiding the exploration of specific PapI and PapB levels by identifying unique combinations that can alter pap activity, as well as the likely paths of transition between states when the order is uncertain (e.g., the relative ordering of changes in methylation and Lrp binding at the proximal versus distal sites).

Establishing conditions to study the inheritance of variable pap conditions in successive sets of progeny may help explain how epigenetics affect pap transcription. For example, heterogeneous active methylation states may represent a transition toward a more homogeneously active population of cells, and exposing successive progeny to activating conditions may yield a more homogeneous active methylation state. The role of enzymatic and nonenzymatic acetylation on one or more pap-associated transcription factors may also have an important role in affecting the inherited state to respond to changes in cellular metabolism and the environment.

Moving forward, the study of thermoregulation in pap expression could be expanded to include more temperature points, improving the understanding of both the machinery that underlies pap behavior and E. coli’s general relationship with its environment. Given the amount of effort that has been invested in identifying and characterizing the full set of regulators binding to pap, the development of an integrated, quantitative model to assess the completeness of our current understanding of the pap system would be highly informative. Other key areas for potential future work also include assessment of coverage by H-NS and other structuring proteins across different conditions (perhaps assessed by H-NS chromatin immunoprecipitation sequencing and/or nonspecific methods, such as in vivo protein occupancy display [117]) and potentially higher resolution of investigation of the biophysical changes occurring in the pap DNA in the presence of different proteins (perhaps via single-molecule methods, such as DNA curtains [118]). The relationship between RimJ and H-NS activity would be greatly aided by identifying alternative RimJ substrates or further elucidating RimJ’s role in DNA replication and protein translation. The characterization of each acetylated Lrp and CRP lysine, and how each acetylation impacts protein and DNA binding (at differing temperatures), could greatly extend our current model. Lastly, the identification of any additional epigenetic factors, either in pap or elsewhere, that are affected by environmental variables or factors involved in pap transcription would be important for deeper understanding of how a specific, highly pathogenic population can be established and maintained.

As we have shown here, the pap regulatory region presents an intricate mixture of the regulation of a virulence process by structuring proteins, local regulators, epigenetic marks, and the structure of the DNA itself. We expect that it will serve as a paradigm for understanding other complex, environmentally sensitive regulatory regions in bacteria, and that further investigation of pap itself will permit the development of appropriate frameworks for both conceptualizing and quantitatively modeling the behavior of such regulatory regions.

Biographies

Mario Zamora, B.S.E., M.D., is a fourth-year neurology resident at the University of Miami/Jackson Memorial Hospital Health System. He received his medical degree from Loyola University Chicago, Stritch School of Medicine, where he worked with Alan Wolfe, Ph.D., to characterize the effects of posttranslational modifications in E. coli and improve molecular models affecting transcription. In 2021, he will begin vascular neurology/endovascular neurosurgery fellowship training at the University of Tennessee Memphis.

Christine A. Ziegler is a fourth-year Ph.D. student in the Freddolino laboratory in the Department of Biological Chemistry at the University of Michigan. She received both a bachelor’s of science in chemistry with highest distinction and a bachelor’s of science in molecular genetics from the University of Rochester. Her thesis work aims to utilize techniques in microbiology, biochemistry, genetics, and bioinformatics to characterize the molecular mechanisms by which Lrp senses nutrient availability and subsequently regulates its target genes accordingly.

Lydia Freddolino, B.S. (Hons), Ph.D., is an Assistant Professor in Biological Chemistry at the University of Michigan. She trained at the California Institute of Technology, University of Illinois at Urbana-Champaign, Princeton University, and Columbia University. She combines approaches from biophysics, microbiology, and bioinformatics to study how microbial regulatory networks function and the fitness implications of the regulatory decisions that they implement.

Alan J. Wolfe, Ph.D., Professor of Microbiology and Immunology at Loyola University Chicago, is a fellow of the American Academy of Microbiology. He trained at Muhlenberg College, the University of Arizona, the California Institute of Technology, and Harvard University. He has spent much of his career trying to understand how high-energy acetate metabolites allow central metabolic status to regulate protein function. More recently, he built the team that showed that the human urinary bladder is not sterile.