Chapter 8

Periplasm

DONALD B. OLIVER

INTRODUCTION

The periplasmic space lies between the inner and the outer membranes of gram-negative bacteria. A number of processes that are vital to the growth and viability of the cell occur within this compartment. The architecture of the periplasm facilitates cell wall and outer membrane growth and coordinates these processes with cell division. Proteins residing in the periplasmic space fulfill important functions in the detection and processing of essential nutrients and their transport into the cell. Periplasmic proteins promote the biogenesis of proteins entering this compartment along with components destined for incorporation into the peptidoglycan, outer membrane, or capsular layers. The contents of the periplasm provide a microenvironment of small and middle-sized molecules that buffer the cell from changes that occur in its local surroundings. In fulfilling these essential roles, the periplasm is not static but dynamic, capable of regulation to accommodate changes in the external and internal environments that surround it.

In this chapter, I review what is known about the structure, composition, physiology, and biogenesis of the periplasm. I also attempt to point out the limitations in our understanding of this compartment and what questions need to be addressed in the future. These aspects are seen from the perspective of an Escherichia coli cell, although most of what is said is equally true for E. coli‘s close cousin, Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium). I apologize in advance for not citing all of the literature available in this area, but the breadth of the topic made such an approach infeasible. In many cases, the reader is referred to more detailed coverage of specific topics.

PERIPLASMIC STRUCTURE

Electron Microscopic Studies

The basic architecture of the periplasmic spaces of E. coli and S. typhimurium has been inferred from various types of electron microscopy. Studies employing cryofixation methodologies reveal a periplasmic region that is located between the inner and outer membranes and has an estimated width of 13 to 25 nm (48, 67, 79, 115, 116). This variable width may relate to expansion or contraction of the periplasm due to osmotic shifts during sample preparation (see Periplasmic Physiology below), or it may be due to differences in the strains or growth conditions employed by different investigators. Located within this region is the peptidoglycan layer, with an estimated thickness of 5 to 8 nm, which is in close association with the outer membrane (48, 67, 79). Attachment of these two structures is facilitated by the major outer membrane lipoprotein that is murein linked and by strong interaction of the porins with the peptidoglycan (25, 159). Better electron micrographs show the region between the peptidoglycan layer and the inner membrane (the inner periplasmic region) to be regular in thickness and matter content, judging from its density (∼1.05 g cm^–3) and its staining (48, 67, 79). Early studies misrepresent this region in terms of both size and lack of contents, since the techniques used for specimen preparation cause release of the periplasmic contents and/or an altered separation of the inner and outer membranes (thus giving rise to the term "periplasmic space"). This artifact was appreciated in some of the best early studies by DePetris (44).

Hobot et al. (79) proposed a model for the organization of the periplasm in which the peptidoglycan in various states of polymerization and in the form of a loose gel fills the entire space between the inner and outer membranes. It was proposed that peptidoglycan is more highly polymerized near the outer membrane and more loosely polymerized near the inner membrane. This peptidoglycan framework would be filled with an aqueous solution of periplasmic proteins, oligosaccharides, and various small molecules. This model is based on studies which showed that the peptidoglycan has an usually high water content (79) and is consistent with electron microscopic studies that showed that phosphotungstic acid-chromic acid, a peptidoglycan-specific stain, stains the entire periplasmic compartment (115, 116). This model could also explain the low level of lateral diffusion of periplasmic proteins observed by Brass et al. (24) (see Protein Mobility within the Periplasm below). However, both structural and mobility features of the periplasm are consistent also with the very high content of periplasmic protein (see Periplasmic Physiology below). Precise measurements of the amount of soluble peptidoglycan and its hydration state in relation to periplasmic protein are needed before the relative contribution of these macromolecules to conditioning of the periplasmic milieu can be assessed.

Periplasmic Fine Structure and Membrane Adhesion Sites

Some studies have complicated this simple layered structure of the cell envelope by suggesting the existence of zones of adhesion between the inner and outer membranes which may correspond to sites of export of components destined for the outer membrane. Such adhesion sites, or "Bayer patches," were first visualized in the electron microscope by Bayer (13, 14) by plasmolysis of E. coli cells followed by chemical fixation, substitution, embedding, sectioning, and staining (Fig. 1). After shrinkage of the protoplast away from the outer envelope, the inner membrane is attached to the outer membrane at approximately 200 to 400 sites, which cover an estimated 5% of the outer membrane surface. Viewed in close-up, the inner and outer membranes of a typical adhesion site are in apposition for a distance of 20 to 30 nm; the inner leaflet of the outer membrane appears to be thinner than normal, suggesting that some of its constituents are missing along this membrane junction. The extent of fusion between the two membranes and their precise architectures are unclear from such pictures. Adhesion sites were detected only in exponentially growing cells and not in cells grown into stationary phase, in keeping with the proposed biogenic functions of these sites (see Biogenesis of the Periplasm below). It has been suggested recently that the observed junction sites may represent places where new strands of peptidoglycan are covalently attached to the murein sacculus while their growing ends are still attached to bactoprenolpyrophosphate, a component of the inner membrane (proposed in chapter 6 of this volume). A similar proposal could be made for additional classes of adhesion sites created by intermediates of lipopolysaccharide (LPS) or outer membrane protein export.

Fig. 1Visualization of membrane adhesion sites. E. coli B was subjected to plasmolysis in 15% sucrose and subsequent fixation in glutaraldehyde. Magnification is ×62,000.

Source: (Courtesy of M. Bayer.)

The existence of adhesion sites has been challenged by Kellenberger (103; see also chapter 4) on the basis of more recent pictures of periplasmic architecture that utilized cryofixation. The results of Hobot et al. (79), Dubochet et al. (48), and Leduc et al. (115), which showed an equal spacing of 11 to 15 nm between the outer and inner membranes with no contacts, adhesions, or other sorts of membranous connections visible, were used to argue against the existence of adhesion sites. Nor could adhesion sites be visualized by plasmolysis if cryofixation replaced normal chemical fixation (103). It was proposed that the Bayer bridges represent postmortem artifacts of chemical fixation that somehow enhance a stream of proteins trans-migrating across the periplasm or a preexisting but not yet resolved structure involved in macromolecular transport across the two membranes.

This view has been countered by Bayer (15), who showed that adhesion sites can be visualized by cryofixation techniques following plasmolysis provided that plunge-freezing rather than impact-freezing is used to preserve these fragile membrane structures. Such adhesion sites are not simply the result of membrane superposition due to the thickness of the thin section. Two different sorts of adhesion sites were noted: single, isolated sites or multiple sites within regions of inner and outer membrane apposition.

Studies by Rothfield and coworkers on periseptal annuli and polar annuli (38, 124, 160; also see chapter 101) argue strongly that at least one class of membrane adhesion sites exists. On the basis of the position of plasmolysis bays present in cells at different stages of division, the existence of a pair of concentric rings, the periseptal annuli, has been proposed. These organelles mark the sites of future cell divisions. Viewed by serial section electron microscopy, an annulus consists of a continuous ring of close apposition of the inner membrane, murein, and outer membrane (Fig. 2). Following septation and cell separation, each daughter cell inherits one of the two periseptal annuli, which remains as a polar annulus at the new pole of the newborn cell. The periseptal and polar annuli divide the periplasm into three types of subcompartments: two polar and two midcell compartments and one compartment at the site of future division. Such subcompartmentalization could play an important role in periplasmic function, particularly at the site of future division. The ability to visualize plasmolysis bays in living bacteria and the apparent inability to detect these structures by the cryofixation procedures outlined by Kellenberger (103) suggest that certain cryotechniques are not compatible with the maintenance of sensitive membrane structures. Further complicating this controversy, however, Woldringh has challenged the existence of periseptal and polar annuli by offering data suggesting that the apparent positioning of plasmolysis bays is due to physical constraints on the membrane imposed by mild plasmolysis and its localized relief during membrane collapse (189a; for an alternative view, see recent work by Cook and Rothfield [38a]).

Fig. 2Diagrammatic representation of surface views (left) and cross sections (right) of periseptal annuli. Cells at different stages of the division cycle are shown. PSA, periseptal annulus; PA, polar annulus; OM, outer membrane; M, murein; IM, inner membrane.

Source: (Courtesy of L. Rothfield.)

PERIPLASMIC PHYSIOLOGY

Electron microscopic and physiological measurements have been utilized to estimate the size of the periplasm. As outlined above, methods of sample preparation for electron microscopy tend to alter the periplasm and thus perturb the measurements of periplasmic width. Measurements of periplasmic width on cells with full periplasmic compartments range from 13 to 25 nm (48, 79). Stock et al. (171) used physiological techniques to perform a classic study of the periplasmic volumes of E. coli and S. typhimurium. The distribution of radioactive compounds capable of penetrating both inner and outer membranes (water), the outer membrane only (sucrose), or neither membrane (inulin) was measured, and the cytoplasmic, periplasmic, and total cellular volumes were determined. These results place the periplasmic volume at 20 to 40% of the cell volume in normal growth media. As pointed out by van Wielink and Duine (184), estimates of periplasmic width and volume are somewhat at odds. A 20% volume would lead to a 32-nm width, which is greater than has ever been observed by electron microscopy. Electron microscopic estimates are more consistent with periplasmic volumes in the range of 8 to 16% (66).

The volume of the periplasmic compartment is an important number, given the abundance of periplasmic proteins and their approximate atomic dimensions. It has been estimated, for example, that at 10% of the cell volume, the periplasm would be essentially full of periplasmic proteins and would approach a highly viscous state with little available free water (184). This gellike state of the periplasm is consistent with the model proposed by Hobot et al. (79), but it indicates that periplasmic protein rather than soluble peptidoglycan polymers may be sufficient to account for the presumed physiological state of this compartment. Such a highly viscous milieu for periplasmic proteins is consistent with measurements of their mobility in this compartment (see Protein Mobility within the Periplasm below). This physiological state may facilitate the creation of microenvironments within the periplasmic compartment as well as being of significance in the enzymology of periplasmic protein function.

Regulation of Osmolality and Periplasmic Volume

Like plant cells, bacterial cells maintain an osmotic pressure greater than that of the surrounding medium. This difference, termed turgor pressure (∼303 kPa in the case of E. coli over a broad range of osmolarities [88]), is carefully regulated by the bacterial cell and is essential for its growth and division. The primary stress-bearing structure is the murein sacculus. As pointed out by Stock et al. (171), the inner membrane has essentially no mechanical strength, and the cytoplasm and periplasm should therefore be iso-osmotic. However, in light of the suggestion that the periplasm has a gellike state, the periplasm may lend mechanical strength (i.e., as an incompressible gel) to the inner membrane under hypo-osmotic conditions. This may explain why the proper periplasmic volume is maintained even when the major osmoprotectants, the membrane-derived oligosaccharides, are absent. Studies that specifically measure periplasmic volume immediately after osmotic downshift would be useful for addressing this point.

The abilities of E. coli and S. typhimurium to grow at a variety of osmolalities without large changes in the periplasmic volume suggest that there are mechanisms that balance the osmolalities of these two compartments (see chapters 77 and 98). At high osmolality, the major dangers to the cell are loss of cytoplasmic water and shrinkage of the protoplast in the absence of turgor pressure. Under these conditions, well-characterized transport systems for uptake of K⁺ and compatible solutes such as glycine betaine, proline, and trehalose are induced to raise the osmolality of the cytoplasm. The systems that respond most rapidly but transiently to changes in osmolality utilize K⁺ and include Kpd, TrkA, TrkF, and Kup as well as at least two efflux systems, KefB and KefC (40). Synthesis of glutamate is also needed in this case to provide a counterion for K⁺. Since these systems respond to hyperosmotic shock and return to normal as cytoplasmic osmolality increases, it has been suggested that they are regulated by a stretch-sensitive membrane protein that senses turgor pressure (40). KdpD and KdpE constitute a typical two component system that may comprise such an osmosensor (174a).

Systems preferred for osmotic adjustment during steady- state conditions transport compatible solutes such as glycine betaine and proline or synthesize the disaccharide trehalose. These systems are under osmotic regulation also, but they make permanent adjustments to osmotic upshift, suggesting a different sort of regulatory mechanism. The sequence of the low-affinity system that transports both glycine betaine and proline, ProP, predicts an integral membrane protein of the transporter superfamily that is likely to catalyze solute-ion cotransport (41). ProP appears to be an osmosensor as well as an osmoregulator, since it is rapidly activated even in membrane vesicles following hypertonic shift (131). The high-affinity system for uptake of glycine betaine, ProU, belongs to the family of membrane traffic ATPases and consists of three proteins (ProV, ProW, and ProX, encoding an ATP-binding protein, an integral membrane component, and glycine betaine-binding protein, respectively) (42, 65, 129). The system that employs trehalose as an osmoprotectant is complex. While the genes that synthesize trehalose from glucose-6-phosphate and UDP-glucose, otsA and otsB, are induced by high osmolality, there appears to be no feedback mechanism for self-regulation (174). The excess trehalose is secreted into the periplasm, where it is hydrolyzed to glucose by an osmotically regulated trehalase, TreA (21, 153). The utilization of trehalose as a carbon source under conditions of low osmolality occurs through a separate pathway (157).

The transcriptional regulatory signals for the ProU and Ots-Tre systems remain controversial, as does the nature of the osmosensors. Because the response of these systems to osmotic upshift is not attenuated by osmotic adaptation, regulation by turgor pressure or by cell wall or membrane stretch seems unlikely in this case. In addition, since these systems respond only to solutes that cannot penetrate the cytoplasmic membrane, their regulation by the absolute periplasmic or cytoplasmic water activity seems unlikely, although a transmembrane protein that measures the water activity differential across the membrane could exist. Proposals for a role for intracellular K⁺-glutamate and DNA supercoiling as direct effectors of osmoregulation have received serious consideration (78, 95, 151, 152, 176).

Hypo-Osmolality, MDOs, and Donnan Potential

During osmotic downshift, the volume of the periplasmic compartment will decrease unless the periplasm is largely incompressible or unless compensatory solutes are manufactured and transported to this space. Such solutes, however, must be sufficiently large and hydrophilic that the outer membrane is impermeable to them. Only one system, that responsible for synthesis and transport of membrane-derived oligosaccharides (MDOs), has been described (see chapter 70). MDOs are a family of closely related periplasmic oligosaccharides, each containing approximately 6 to 12 glucose units joined by β-1–2 and β-1–6 linkages that are variously substituted with sn-1-phosphoglycerol, phosphoethanolamine, and succinic O esters. Their synthesis is osmoregulated, and under conditions of low osmolality they can account for up to 5% of dry cell weight. Elucidation of the details of their biosynthesis is ongoing, and a tentative scheme is presented in chapter 70. A number of steps that occur in the periplasmic compartment have been defined. Passage of carrier-bound MDOs from the cytoplasmic side to the periplasmic side of the inner membrane involves an unknown transport system analogous to similar steps in LPS or peptidoglycan synthesis. Two periplasmically accessible phosphoglycerol transferase activities have been detected. Phosphoglycerol transferase I (MdoB) is an integral inner membrane protein that transfers phosphoglycerol residues from phosphatidylglycerol to carrier-bound MDOs only (91). Phosphoglycerol transferase II is a periplasmic enzyme specific for soluble MDOs and apparently utilizes carrier-bound MDOs as a source of phosphoglycerol residues (63). mdoB mutants produce MDOs that lack phosphoglycerol substitutions. The mdoA locus of E. coli consists of two genes, mdoG and mdoH, both of which are necessary for MDO biosynthesis (123). MdoG is a 97-kDa membrane-spanning protein that is necessary for normal glycosyl transferase activity, while MdoH is a 56-kDa periplasmic protein whose function remains to be determined.

MDO osmoregulation is somewhat similar to the ProU and Ots-Tre systems in that MDO biosynthetic rates need to remain high even after osmotic adaptation. Analysis of mdoH-lacZ fusions and mdoGH-specific RNA leads to the conclusions that the MDO osmoregulatory system is transcriptionally controlled and that regulation does not depend on the presence of MDOs (114). The nature of the osmosensor is currently unknown. However, osmosensors that respond to turgor pressure or to cell wall or membrane stretch do not appear to be likely, since they would return to normal after osmotic adaptation.

Despite the great abundance of MDOs in cells grown at low osmolality, the MDO system is dispensable under laboratory conditions. mdoA mutants completely deficient in MDO synthesis grow normally in media of low osmolality (104). However, an mdoA mutant showed a decreased number of membrane adhesion sites, implying some perturbation of envelope structure (83). These results suggest either that another system for regulating the osmolality of the periplasm exists (though previous lack of detection of it seems unlikely, given the size and abundance constraint of the putative solute) or that the structure of the periplasm is relatively incompressible and additional osmotic buffering is not needed.

The strongly anionic nature of MDO and its abundance gives rise to a Donnan potential across the outer membrane. Stock et al. (171) measured this Donnan potential by the unequal distribution of radioactive Na⁺ and Cl^– ions across the outer membrane and found it to be approximately 30 mV in cells growing in M63 minimal medium. The presence of a Donnan potential creates a particular ionic composition within the periplasm. Since MDO is the major fixed anion in the periplasm and is dispensable, the Donnan potential itself appears to be largely dispensable. However, no data on measurement of the Donnan potential in mdoA mutants are available.

Protein Mobility within the Periplasm

Only two studies have been performed to measure the mobility of proteins within the periplasmic compartment. Brass et al. (24) used a photobleaching recovery method to measure the diffusion rates of fluorescently labeled proteins that were introduced into the periplasmic space by Ca²⁺ permeabilization. The lateral diffusion rate of proteins within the periplasm is 1,000-fold lower than comparable measurements in aqueous medium and nearly 100-fold lower than expected cytoplasmic diffusion rates. Foley et al. (57) utilized this same methodology to demonstrate compartmentalization of the periplasmic space by periseptal and polar annuli. They showed that fluorescence recovery is incomplete when periseptal or polar regions are thoroughly photobleached. Plausible factors reducing diffusion rates within the periplasm include (i) extreme viscosity due to a high content of protein and unpolymerized peptidoglycan and a high state of hydration of these components, (ii) poor mobility within the murein sacculus due to a sieving effect as well as to its high state of hydration, and (iii) severely reduced aqueous space within the periplasm (e.g., 30,000 molecules of maltose-binding protein with dimensions of 4 by 4 by 7 nm would occupy 40% of the periplasmic compartment at an estimated thickness of 12 to 15 nm [cited in reference 24]).

PERIPLASMIC PROTEIN CONTENT

Periplasmic Protein Isolation

Soluble periplasmic proteins occupy different locations within this compartment: they may associate peripherally with the peptidoglycan layer or with the inner or outer membrane; alternatively, they may diffuse freely within the aqueous space within the periplasm. More complex topologies are possible also in the case of proteins that are technically inner or outer membrane proteins but that possess a domain that resides within the periplasmic space (e.g., see TonB and TolA below). Most often, however, the term "periplasmic protein" is defined operationally by methods that selectively release the contents of the periplasmic compartment. Obviously, such methods represent a compromise between conditions that are stringent enough to disrupt peripheral associations of periplasmic proteins with the cell envelope and those that are mild enough to prevent release of integral membrane or cytoplasmic proteins.

Three methods are widely used to release periplasmic proteins: spheroplast formation (20, 125), osmotic shock (137), and chloroform treatment (4). In addition, some periplasmic proteins are released simply by washing cells with Tris-KCl, Tris-EDTA, or sucrose-Tris-EDTA (for references, see reference 16). Occasionally, highly selective release of particular periplasmic proteins can be achieved by employing one of these conditions (for example, see reference 121). Spheroplasts are usually prepared by treatment of cells with lysozyme-EDTA in a concentrated sucrose solution to act as an osmotic buffer. Presumably, degradation of the peptidoglycan layer allows more efficient leaching of periplasmic proteins through the outer membrane whose structure has been perturbed by EDTA (LPS present in the outer leaflet of the outer membrane requires Mg²⁺ counterions to preserve its packing). Osmotic shock involves pretreatment of cells with a concentrated sucrose solution and EDTA followed by rapid dilution into a medium of low osmolality. It is presumed that periplasmic proteins are expelled by the sudden expansion of the inner membrane against the cell wall. Chloroform treatment involves treatment of cells directly with chloroform followed by subsequent dilution into buffer. It is hypothesized that a chloroform-sensitive "plug" located in the outer membrane is perturbed by this treatment, allowing release of the periplasmic contents. In these procedures, the cells are removed after appropriate treatment by sedimentation. Very similar but nonidentical periplasmic protein profiles are obtained by these different procedures. When these methods are used, it is imperative to perform control experiments with protein markers whose locations have been established in order to monitor the effectiveness of the procedure with a particular strain.

Since periplasmic proteins are defined operationally by these fractionation methods, additional criteria should be used to substantiate a periplasmic localization. Such methods can include (i) histochemical or immunological localization by electron microscopy, (ii) activity measurements (for enzymes) with substrates or inhibitors that penetrate the outer but not the inner membrane, (iii) selective labeling with cross-linkers that can penetrate the outer but not the inner membrane, (iv) use of gene fusion methodologies to enzymatically tag a protein with a reporter that is enzymatically active only when translocated across the inner membrane (e.g., alkaline phosphatase or β-lactamase [27, 126]), and (v) selective release of periplasmic proteins by particular mutants possessing a leaky outer membrane (e.g., envA [193], lpp [172], lky [189], and tolA and tolB [188]). Only the first and last methods distinguish a periplasmic location from a cell envelope one. No localization method is free of potential artifacts, and convincing data can be acquired only by using more than one method.

Classification of Periplasmic Proteins

Periplasmic proteins can be divided into several categories that are based on their functions. These categories are listed in Tables 1 and 2. They include (i) solute or ion-binding proteins that function in conjunction with ABC transporters or chemotaxis receptors for the sensing and uptake of sugars, amino acids, peptides, vitamins, and ions; (ii) catabolic enzymes that degrade complex molecules into simpler ones for transport across the inner membrane; (iii) detoxifying enzymes that serve a protective role for the cell; and (iv) enzymes or proteins that promote the biogenesis of major envelope proteins or proteinaceous appendages, peptidoglycan, LPS, capsules, or MDOs. For discussion of periplasmic biogenesis, the reader is referred to the next section (Biogenesis of the Periplasm) as well as chapters 6, 9, 61, 69, 70, and 150.

Table 1Periplasmic binding proteins

Table 2Other categories of periplasmic proteins

Periplasmic binding proteins are a large and extensively studied group of proteins. Binding proteins are effective in concentrating the solute to be transported, since they have a high affinity for their ligand (generally in the range of 0.1 to 1.0 μM) and are present at a very high concentration in the periplasm (on the order of 1 μM in certain cases). Binding proteins are single-chain proteins generally consisting of two globular domains with a composite molecular mass of approximately 20 to 40 kDa. The peptide-binding proteins DppA and OppA and the nickel-binding protein NikA are an exception to this rule, since they each contain a third domain that accounts for their larger sizes (1, 136, 179). Crystal structures for 10 different binding proteins have been determined (reviewed in reference 149; also see references 56, 139, 140, 179, 187, 192, and 194), and additional structures are forthcoming (for example, see reference 49). These structures reveal a common paradigm: the proteins are ellipsoidal and consist of two lobes connected by a flexible hinge (except for DppA, OppA, and NikA, which contain a third lobe that is not involved in ligand binding). The lobes are apart in the unliganded state, and they come together to trap and bind the solute. Hydrogen bonds play a dominant role in solute recognition, with both lobes contributing to the observed pattern of stereospecificity.

Genetic and biochemical analyses have revealed subtleties in the way that binding proteins interact with their homologous membrane permease components to promote release of the solute and transport of it across the inner membrane. In one of the best-studied cases, that of maltose transport, it appears that solvent-accessible amino acid residues on both domains of maltose-binding protein (MalE) interact with the two integral membrane permease subunits, MalF and MalG, to promote solute release from MalE and uptake by the transporter (84). Furthermore, it appears that interaction of binding proteins with the membrane transporter can mediate a type of signal transduction not unlike that affected by the binding proteins during chemotaxis. For example, in reconstituted systems for maltose and histidine transport, ATP hydrolysis and solute uptake were dependent on the interaction of the binding protein with the membrane transporter. In mutants that acquired the ability to transport solute in the absence of the binding protein, ATP hydrolysis was uncoupled from solute uptake, suggesting that the transporter no longer requires the normal signaling mechanism to promote the ATP-driven conformational cycles needed for solute uptake (43, 143). In the phosphate regulon, the phosphate-binding protein, PstS, has also been proposed to play a role in the signal transduction pathway that promotes proper gene regulation in response to the external phosphate concentration. It is presumed that interaction of solute-bound PstS with the membrane transporter is required to generate the repression signal that is sent to the sensor kinase-response regulator pair, PhoR and PhoB proteins (39). For additional coverage of the membrane components of binding protein dependent systems and their mechanism of action, the reader is referred to chapter 7.

Four binding proteins play roles in chemotaxis: MalE interacts with the Tar transducer, MglB and RbsB interact with the Trg transducer, and DppA interacts with the Tap transducer in E. coli (2). S. typhimurium lacks the Tap transducer and therefore does not carry out peptide chemotaxis. For detailed coverage of chemotaxis, the reader is referred to chapter 10.

Three different systems are utilized for peptide transport. The oligopeptide permease system displays the broadest specificity, since it can transport any peptide containing two to five amino acid residues. This system even transports peptides containing d-amino acids, and in this capacity, it plays an important role in the recycling of peptides generated during cell wall turnover (64). The crystal structure of OppA reveals the basis for sequence-independent peptide binding: like other binding proteins, two lobes enclose the oligopeptide ligand within the interior of the protein. The oligopeptide is found in an extended conformation in which parallel and antiparallel β-strands of OppA interact with it to satisfy completely the hydrogen-bonding potential of the oligopeptide backbone while forming a salt bridge with the a-amino group of the oligopeptide. Different amino acid side chains of the oligopeptide are accommodated within voluminous hydrated cavities in the OppA structure (179). The tripeptide permease system has the highest affinity for hydrophobic tripeptides, but it can transport dipeptides to some extent. The tripeptide permease is anaerobically induced and depends on the function of OmpR and EnvZ for positive regulation, although it is not under osmotic control (62). The dipeptide permease system has the highest affinity for dipeptides containing only l-amino acids or glycine, but it can transport some tripeptides also (127). Recently, it has been shown that the dipeptide permease can transport the important heme precursor 5-aminolevulinic acid (52, 185). Strains lacking these three different peptide permeases are deficient in peptide uptake, suggesting that no additional high-affinity peptide transport systems exist.

TonB and TolA

Although TonB and TolA are tethered to the inner membrane by a short membrane anchor and as such are formally inner membrane proteins, their importance to periplasmic function makes a discussion of them appropriate. As a discontinuity between the inner and outer membranes, the periplasm creates a special problem for gram-negative bacteria. How can the uptake of larger molecules across the outer membrane occur in an energy-dependent fashion in the absence of direct access to energy sources? This problem has been solved by the evolution of receptor-mediated uptake systems for iron-bearing siderophores and vitamin B₁₂ that utilize the proton motive force generated in the cytoplasmic membrane and channel it across the periplasmic space to the outer membrane protein receptors via the TonB protein (for a review, see reference 147; also see chapters 5 and 7). Group B colicins and various bacteriophages also utilize these TonB-dependent pathways to gain entry into the cell. TonB appears to consist of three domains: (i) a short amino-terminal segment that spans the cytoplasmic membrane serves as an uncleaved signal peptide and interacts with two cytoplasmic membrane proteins, ExbB and ExbD, to stabilize TonB and facilitate its energization (101, 102, 183); (ii) a proline-rich region proposed to be up to 10 nm in length presumably spans the periplasmic space (53); and (iii) a carboxy-terminal region that, based on genetic and biochemical criteria, interacts with outer membrane receptors (17, 167). TonB action appears to be independent of membrane adhesion sites, since uptake across the outer membrane is readily uncoupled from inner membrane transport (50, 96, 154). In addition, these transport pathways utilize periplasmic binding proteins and inner membrane permeases, suggesting a two-step mechanism of entry into the cell.

A second protein spanning the periplasmic discontinuity is TolA, which is required for the uptake of group A colicins and filamentous phage DNA after their initial binding to cell surface receptors (reviewed in reference 188). The Tol system is required also for some aspect of outer membrane biogenesis or integrity, since most tol mutants are hypersensitive to detergents, dyes, and certain antibiotics, and tolA and tolB mutants leak periplasmic proteins into the media (188). The core elements of the Tol system, TolA, TolQ, and TolR, appear to be homologous to those of the TonB system. TolQ and TolR are homologous to ExbB and ExbD, respectively, with respect to their sequence, membrane topology, and ability to promote cross-complementation (26, 98, 99, 100, 135). Although TolA is substantially larger than TonB and does not display any obvious sequence similarity, the two proteins are predicted to possess a similar three-domain structure (147, 188). The amino-terminal domain of TonB can be functionally substituted for that of TolA (101), while the carboxy-terminal domain of TolA appears to interact with the amino-terminal regions of substrate proteins to promote their uptake (similar to the manner in which TonB interacts with its outer membrane receptors) (18, 117, 188). Unlike the TonB-dependent systems, it is unclear whether Tol-dependent import occurs by (i) a two-step mechanism with a periplasmic intermediate, (ii) a one-step mechanism at membrane adhesion sites, or (iii) both mechanisms, depending on the substrate being imported. Resolution of this issue is likely to take a combination of genetic, biochemical, and structural studies. In addition, more attention needs to be focused on defining the component(s) that is acted upon by the Tol system in order to promote outer membrane stabilization (presumably its real function).

BIOGENESIS OF THE PERIPLASM

As proteins destined for the periplasm emerge on the outer surface of the inner membrane, they must fold into their native state, acquire posttranslational modification where necessary, and assemble into quaternary structures in the case of oligomeric proteins. Proteins destined for the outer membrane may be released as periplasmic intermediates before being inserted into the outer membrane. Alternatively, such proteins may be translocated directly from the inner membrane to the outer membrane via the proposed membrane adhesion sites. The periplasm also plays an important role in the terminal steps of LPS and capsule biogenesis, although the existence of periplasmic intermediates in the assembly of these polymers remains controversial. First, I review the periplasmic proteins that play a role in protein biogenesis in this compartment. Then, I review what is known about the assembly of proteins in the periplasm and the regulation of this assembly. Finally, I address the question of whether periplasmic intermediates of outer membrane proteins, LPS, and capsular polysaccharide exist. For more comprehensive reviews of these topics, the reader is referred to chapters 5, 9, 64, and 69.

Disulfide Bond Formation

It is now clear that the bacterial cytoplasmic compartment is a reducing environment, while the periplasmic compartment is oxidizing, allowing newly secreted proteins to form disulfide bonds after passage across the inner membrane. This fact explains why secretory proteins such as alkaline phosphatase that require disulfide bonds for activity remain inactive when retained within the cytoplasm (45). This state of affairs presumably influenced the evolution of tertiary structures of a number of envelope proteins whose presence in the cytoplasm would be harmful to the cell if present in an active state.

Two separate pathways that catalyze disulfide bond formation within the periplasm are known. One system consists of DsbA and DsbB (3, 11, 12, 97), while a second system consisting of DsbC has been described recently (133, 164). dsb mutants are defective in disulfide bond formation and accumulate reduced forms of periplasmic and outer membrane proteins. Such folding intermediates are often protease sensitive and degraded or chase slowly into stable, oxidized forms of the mature protein. DsbA and DsbC have been purified and shown to be periplasmic disulfide oxidoreductases (3, 12, 97, 133, 164). Both proteins possess a Cys-X-X-Cys motif that is the site of the catalytically active disulfide bond. DsbB is an inner membrane protein that appears to promote DsbA reoxidation (11). The component that catalyzes the reoxidation of DsbC is presently unknown, as is the ultimate source of oxidizing potential for the two systems. It has been proposed that both systems are normally required for the oxidizing state of the periplasm and that they operate in parallel, acting on similar substrates. This conclusion is supported by studies that showed that (i) dsbA and dsbC mutants are defective in the oxidation of similar proteins; (ii) dsbA dsbC or dsbB dsbC double mutants have more severe phenotypes than dsbA, dsbB, or dsbC single mutants; and (iii) multicopy suppression studies show that the dsbA and dsbC genes can functionally substitute for each other (133).

Proline Isomerase

E. coli and S. typhimurium contain two distinct peptidyl-prolyl cis-trans isomerases, one cytoplasmic and the other periplasmic (37, 74, 122). This localization presumably reflects the need to catalyze protein folding in both of these cellular compartments. The two proteins from E. coli are approximately 50% identical in primary amino acid sequence, and they show approximately 35% identity with the human peptidyl-prolyl cis-trans isomerase, known as cyclophilin, that is the target of the immunosuppressive drug cyclosporin A (74). The periplasmic enzyme from E. coli has a high catalytic efficiency k _catK_m identical to that of human cyclophilin (122). The solution structure of the E. coli periplasmic enzyme has been determined recently and is similar in its global fold to human cyclophilin, but conformation at the active site differs substantially for the two enzymes (36). This observation explains why the periplasmic proline isomerase from E. coli is not significantly inhibited by cyclosporin A (37, 122). The periplasmic proline isomerase is not essential for growth at either 35 or 42°C, but analysis of envelope protein folding in this mutant has not been reported (138).

Periplasmic Chaperones

Since the periplasm lacks energy in the form of nucleoside triphosphates, chaperones that reside in the periplasm cannot be similar to the ATP-dependent chaperones of the Hsp60 and Hsp70 families (see chapters 61 and 88). Nor would periplasmic chaperones for monomeric proteins be similar to the cytoplasmic chaperone, SecB, that serves an antifolding function. In contrast, like SecB, they should recognize nonnative structures, but they need to promote folding into native structures. Chaperones for the oligomeric proteins composing pili have been proposed to bind to hydrophobic, interactive surfaces and to sequester them until association with the appropriate acceptor protein can be achieved.

SecD Protein.

It has been proposed that SecD serves a late function in protein export, perhaps as a chaperone to promote the proper folding and release of secretory proteins from the inner membrane into the periplasmic compartment. Both SecD and SecF are integral inner membrane proteins that possess substantial periplasmic domains and share a region of sequence homology (61, 145). The phenotypes of secD, secF, or secD secF null mutants are identical (cold sensitive with poor viability and severe protein export defects even at the permissive temperature [144]), suggesting that both components act at a similar step in protein secretion. Matsuyama et al. (128) showed that treatment of spheroplasts with anti-SecD antibody inhibits the release of MBP and other secretory proteins from the inner membrane, where the proteins remain in a trypsin-sensitive (nonnative) state. Inhibition is specific for anti-SecD antibody, as a similar treatment with anti-SecE or anti-SecY antibody directed against periplasmic regions of these proteins does not interfere with protein secretion or release. These results imply that SecD fulfills a late function in protein export, but more direct studies are needed to define the precise biochemical nature of this function, which remains controversial. Arkowitz and Wickner (6) suggested that SecD and SecF are required for the maintenance of the proton motive force across the inner membrane, while Kim et al. (105) suggested that these proteins may modulate a cycle of membrane insertion and deinsertion of SecA protein, the translocation ATPase that catalyzes protein export.

Lipoprotein-Specific Chaperone.

A recently identified 20-kDa periplasmic protein, p20, is required for release of lipoproteins from the inner membrane and association with the outer membrane (S.-I. Matsuyama, T. Tajima, and H. Tokuda, personal communication). p20 promotes the release of lipoprotein translocation intermediates from the inner membrane of spheroplasts by forming a water-soluble complex with a one-to-one stoichiometry. This carrier is specific for outer membrane lipoproteins, as no complex between p20 and OmpA was detected. It has been suggested that an outer membrane receptor for this carrier exists, since proteolysis of the outer membrane inhibited the in vitro incorporation of lipoproteins into this fraction. The carrier must facilitate an extremely rapid rate of flux between the inner and outer membranes, since the estimated 200 to 400 p20 molecules present in the periplasm would need to transport over 10⁶ molecules of lipoproteins to the outer membrane per cell generation.

Pilus-Specific Chaperones.

A large family of periplasmic chaperones is known to be required for the assembly of pili or fimbriae, appendages that mediate attachment of enteric bacteria to appropriate target cells of eucaryotic hosts. Most genes encoding pilus-specific chaperones are located in pilus operons that are contained on plasmids or chromosomal segments unique to pathogenic isolates of E. coli and S. typhimurium. A listing of these is given in chapters 11 and 150. PapD and related chaperones make up a large family whose members are similar in amino acid sequence (50% homology) and overall structure and display cross-complementation (19, 82, 94). Unlike cytoplasmic chaperones such as GroEL, DnaK-DnaJ, and SecB, which recognize their substrates in unfolded states, pilus-specific chaperones maintain their substrates in a conformation similar to that of their native state (as assessed by the presence of disulfide bonds, surface reactivity to monoclonal antibodies, and native binding specificity [86, 112]). The pilus-specific chaperones appear to operate by capping interactive surfaces present on their protein substrates, thereby preventing the aggregation and degradation of the proteins within the periplasmic compartment (86). Such chaperones also possess an effector function, since they facilitate targeting of these subunits to an outer membrane protein that controls the assembly reaction.

The best-characterized member of a pilus-specific chaperone is PapD (see chapters 11, 61, and 150). Structurally, PapD consists of two globular domains, each composed of a β-barrel structure with a topology identical to that of an immunoglobulin fold (81). The intersection of the two domains forms a cleft, which is the location of invariant surface-exposed amino acid residues that constitute, at least in part, the binding site for substrates with conserved carboxy-terminal regions (85, 168). Cocrystalization of PapD with a peptide composed of the carboxy-terminal 19 amino acid residues of PapG, the adhesin subunit, demonstrated that the peptide was aligned in an extended conformation along a cleft β-strand of PapD and that the carboxy terminus and the conserved amino acid residues of the peptide formed hydrogen bonds and participated in hydrophobic interactions with invariant amino acid residues of PapD to promote the observed binding specificity (113). Release of a given pilus subunit from PapD occurs when the complex interacts with an outer membrane protein, PapC, that has been likened to a molecular usher, since it determines the order of assembly of the various pilus subunits. The basis for this specificity appears to reside in the relative affinities of the usher for the different pilus subunit-PapD complexes (47). Complexes having the highest affinity for the usher are assembled first; those with the lowest affinities for the usher are assembled last. Subtle conformational changes produced when a given subunit interacts with the usher may allow release of PapD and its reutilization.

HlpA Protein.

HlpA has been proposed to act as a chaperone to promote protein folding within the periplasmic space or sorting to the outer membrane, although the evidence in favor of this hypothesis is scanty. This protein was discovered independently in several laboratories under the name of HlpA, Skp, or OmpH (80, 109, 181, 186). It is a highly conserved, strongly cationic protein that is often associated with anionic components such as DNA, ribosomes, or LPS. However, HlpA was localized to the periplasm by careful cell fractionation methods (182). Previous studies suggesting that HlpA plays a role in protein translocation are in error, since its site of action in the in vitro protein translocation system employed was cytoplasmic rather than periplasmic (181). However, the biochemical activity detected in this case is suggestive of some type of chaperone activity, and HlpA has homology to GRP94, a chaperone of the endoplasmic reticulum (J. Coleman, personal communication). HlpA appears to be nonessential for cell growth, since a strain containing a gene disruption grows normally in rich media (161). Further work will be required to determine whether HlpA has a role in periplasmic or outer membrane biogenesis.

Biogenesis of Periplasmic Proteins

It is now generally accepted that targeting of proteins to the periplasm represents a default pathway whereby a preprotein can gain entry into the Sec pathway simply by possessing sufficient information. Normally, a functional signal peptide is required for this process, but the ability to export periplasmic proteins completely lacking signal peptides (albeit at reduced efficiencies) in certain secY (prlA) mutants suggests that features other than the signal peptide provide important targeting information (46). Such information may reside in the absence of positive charges within the early mature region of the preprotein and in their folding pathway, which delays acquisition of a highly ordered, export-incompetent structure, thus promoting their interactions with chaperones like SecB (5, 32, 73).

Proteins undergoing translocation through the protein-conducting channels located in the inner membrane (93, 166) must be in a largely unfolded state, since only 20 to 30 amino acyl residues are contained within such channels during each catalytic translocation cycle (51, 162). Upon reaching the external side of the inner membrane, periplasmic proteins must begin to fold into their native structures, acquire necessary posttranslational modifications (e.g., cleavage of the signal peptide and acquisition of disulfide bonds), and be released from this membrane in a soluble state. Little information about the order of these events is available, and it is not unreasonable to assume that such ordering will be determined by the particularities of the folding and modification pathways for a given protein. In the case of β-lactamase, release from the membrane appears to require extensive protein folding and modification. Thus, when folding was inhibited, either transiently at low temperature (132) or permanently by truncation of the protein (108), a secretion intermediate of mature β-lactamase bound to the external surface of the inner membrane was detected. Blocking disulfide bond formation by replacement of either cysteine with tyrosine also led to a buildup of this intermediate (55). A similar paradigm has been found for maltose-binding protein: truncated forms that acquire a protease-resistant (i.e., folded) state were efficiently released from the inner membrane, whereas those that remained protease sensitive were not (90). It is likely that chaperones that facilitate such folding or release steps exist, but their identities remain unknown (see SecD Protein above).

Role of the Periplasm in Outer Membrane Protein Biogenesis

Three types of models have been proposed to explain protein routing to the outer membrane: (i) the proteins are released directly into the periplasm, where they bind to and insert into the outer membrane; (ii) the proteins remain associated with the inner membrane until they interact with membrane adhesion sites which facilitate their passage to the outer membrane; and (iii) the proteins are secreted directly to the outer membrane at adhesion sites without the existence of an inner membrane intermediate. Considerable controversy has surrounded this problem. Insight into the mechanism of outer membrane biogenesis has been slowed by several factors: (i) analysis of membrane adhesion sites has been problematic (for example, see reference 89), and no biochemical insight into these structures has been reported; (ii) intermediates in outer membrane protein biogenesis are transient and unstable, and present procedures of cell fractionation are likely to perturb such associations or promote artifactual events; (iii) most workers in the field have been reluctant to confirm localization results by both cell fractionation and immunoelectron microscopic methodologies; and (iv) details of outer membrane protein structure necessary for the interpretation of genetic and biochemical studies have been slow in coming.

At present, the evidence in favor of a periplasmic route to the outer membrane is conflicting and depends on the particular protein studied. Periplasmic intermediates for a number of outer membrane proteins have been detected by (i) production of genetically altered proteins defective in outer membrane assembly (OmpA and PhoE) (22, 58, 106, 107; but for a different result with LamB, see reference 170), (ii) overproduction of outer membrane proteins (OmpA, TonA) (59, 92), (iii) perturbation of the membrane with phenylethyl alcohol (Lpp and TolG) (70), and (iv) release of putative intermediates following spheroplasting with lysozyme-EDTA (OmpF) (163). In all of these cases, however, it could be argued that alteration of the protein or cell perturbed the normal route of secretion. These results contradict the results of earlier studies that suggested that porins were transported via adhesion sites, since newly synthesized outer membrane porins appeared over these structures (169). However, it is likely that adhesion sites perturb outer membrane structure (14) and in so doing could serve as acceptor sites for periplasmic intermediates of outer membrane proteins. This line of reasoning is supported by in vitro studies which showed that destabilization of membrane structure (by strong curvature of liposomes or use of detergents) is necessary for penetration of OmpA and OmpF into the bilayer (163, 175). It has been suggested that LamB assembles into metastable trimers in the inner membrane before passage to the outer membrane and bypasses the periplasm entirely (170). However, this conclusion rests on the characterization of an unusual "inner" membrane subfraction that may in fact be derived from the outer membrane or membrane adhesion sites.

One should be cautious about trying to invoke the same model of biogenesis of all outer membrane proteins, given their differences in structure, oligomeric state, and association with LPS (155). The data as a whole, however, tend to support a model that routes proteins such as OmpA through the periplasm prior to stable association with the outer membrane. The recent identification of a periplasmic carrier for lipoproteins (see Lipoprotein-Specific Chaperone above) suggests that more general chaperones of this type may soon be identified. Furthermore, although discussion of specialized proteins that are secreted across both bacterial membranes (e.g., hemolysin, heat-stable enterotoxin, and ones made by other gram-negative bacteria) is beyond the subject of this chapter (for a recent review, see reference 148), it is worth mentioning that a two-step secretion pathway involving a periplasmic intermediate appears to be the norm for many of these systems (for recent examples, see references 146, 180, and 191).

Role of the Periplasm in LPS and Capsule Biogenesis

Three possibilities have generally been considered for the transport of LPS from the outer face of the inner membrane to the outer face of the outer membrane. These include (i) transport via membrane adhesion sites, (ii) transport via a periplasmic intermediate, and (iii) transport via membrane vesicles (150). Studies employing immunoelectron microscopy showed that newly synthesized LPS appearing on the outer membrane is located over membrane adhesion sites (134), suggesting that these structures may act as carriers of LPS between the inner and outer membranes. However, these results do not exclude alternative models in which adhesion sites serve to promote localized outer membrane destabilization, thereby allowing insertion of periplasmically contained LPS at junction regions. If LPS is routed through the periplasm, then the existence of a carrier protein that would promote release of the lipid A portion of LPS from the inner membrane and ensure its solubility in the periplasm seems almost certain. The recent discovery of the periplasmic carrier of lipoproteins (see Lipoprotein-Specific Chaperone above) points to the feasibility of such a scheme. The existence of transport vesicles within the periplasm seems remote, given the narrowness of this compartment and the lack of any direct evidence for such vesicles, which would have to be quite numerous, given the rate of outer membrane growth.

Transport of K1-type capsular polysaccharide across the inner membrane requires the KpsM and KpsT proteins, which appear to encode an ABC transport system for this polymer (142, 143a). KpsD is a periplasmic protein that is required for transport of capsular polysaccharide to the outer membrane (see chapter 9). In kpsD mutants, the capsular polysaccharide accumulates within the periplasm (190). Similar mutants have been detected for the K5-type capsular polysaccharide (111). This suggests that transport of capsular polysaccharide normally occurs via a periplasmic intermediate, with KpsD either serving as a carrier protein or helping promote entry of capsular polysaccharide at particular regions of the outer membrane. However, it could be argued that KpsD is needed for stabilization of capsular polysaccharide with the proposed membrane adhesion sites. Since periplasmic intermediates of capsular polysaccharide have not been detected in wild-type strains and the function of KpsD remains unknown, these data do not allow a clear picture of capsule biogenesis to be drawn.

Global Regulation of Periplasmic Protein Synthesis

A global mechanism controls the rate of synthesis of a number of major periplasmic proteins. This mechanism appears to operate by coupling the rate of synthesis of periplasmic proteins to a step in their export. A comparable system appears to operate for global regulation of outer membrane protein synthesis (28, 34, 35, 156). The system regulating periplasmic protein synthesis was first discovered by noting that overproduction of a carboxy-terminally truncated form of the periplasmic enzyme GlpQ' dramatically reduced the synthesis of a subset of periplasmic proteins (e.g., MBP, RBP) but had no effect on the synthesis of outer membrane proteins (e.g., LamB) (75, 76). Although GlpQ' was translocated and processed and did not jam the secretion machinery, it was not released from the external face of the inner membrane. The synthetic block was found to be at the translational level. It was speculated that GlpQ' titrates a component of the secretion machinery that normally senses the rate of export of periplasmic proteins and somehow relays an appropriate regulatory signal to the translational machinery. Presumably, all genes under such control must contain a conserved feature present on their mRNA or nascent protein sequence that serves as a recognition element, although a functional signal peptide is not required for regulation. Epistasis experiments showing that the sensor was located before or at the step catalyzed by SecA protein may be wrong because of the leakiness of most secA mutants. It would be more logical for the sensor to be located after divergence of the export pathways of periplasmic and outer membrane proteins (i.e., on the outer surface of the inner membrane). This logic is consistent with the finding that protein substrates that downregulate either system are blocked at a late step in protein export (28, 75). It has been argued recently that the regulatory system for outer membrane proteins is controlled more at the level of mRNA stability than at that of protein synthesis (77). However, it must be remembered that failure to translate bacterial mRNAs often destabilizes them, thus complicating interpretation of the data. The location and identity of the sensor for each system need to be identified along with the mechanisms of signal transduction and effector control. Recent attempts to identify components that catalyze late steps in protein export may help elucidate these points.

PERIPLASMIC PROTEOLYSIS AND ITS REGULATION

Given that the periplasm is a thoroughfare for protein biogenic pathways, there may be a special need to regulate proteolysis in this compartment. Thus, proteins that have been translocated across the inner membrane must rapidly assume native structures or associate with appropriate chaperones lest they become substrates for periplasmic proteases. Given the absence of high- energy phosphate in the periplasm, the rate-limiting periplasmic proteases must be different from the ATP-dependent proteases of the cytoplasm, Lon and Clp. Relatively little information is available about the control of periplasmic proteolysis, and identification of discrete proteases that carry out endoproteolysis of proteins in this compartment is quite recent. The reader is referred to chapter 62 for a comprehensive treatment of proteolysis.

DegP (or HtrA) appears to be an important endoprotease that recognizes periplasmic proteins with abnormal structures often generated by heat or in vitro genetic methods. It is essential for cell survival only at temperatures above 42°C (118, 173). The gene is under heat shock control but is independent of rpoH (σ ³²) regulation. Instead, its promoter appears to be recognized by another transcription factor, σ ^E, that controls rpoH transcription (119). DegP degrades certain periplasmic fusion proteins and mutant forms of maltose-binding protein in vivo (172). In vitro, DegP is a highly specific endoprotease that degrades only casein among a number of protein and peptide substrates tested (120). Its inhibition by diisopropylfluorophosphate suggests that it is a serine protease. The cellular location of DegP remains somewhat uncertain. It contains a cleaved signal peptide, and phoA fusion analysis suggests that most of the protein is exported across the inner membrane (173). Fractionation data for a strain that overproduces DegP revealed a cytoplasmic and inner membrane location for this protein, with essentially none present in the periplasm (as detected by chloroform release) (118). A careful localization study that uses a wild-type strain and attempts to devise conditions that will release DegP from the inner membrane is needed to define the relationship of DegP to the inner membrane and the periplasm more precisely. It has been suggested that DegP is required to degrade improperly folded or denatured proteins that accumulate during heat shock, since it is essential only at temperatures above 42°C and since strains which leak periplasmic proteins because of a lipoprotein deficiency are less dependent on degP function for survival at high temperature (173). Bypass suppressors of degP have been isolated, but their mechanism of action has not been determined (8).

Protease III or protease Pi is a large metalloprotease found primarily in the periplasm that was originally detected by its activity against small peptides of less than 7 kDa such as insulin, glucagon, or a fragment of E. coli β-galactosidase (29, 177). The enzyme is an endoprotease with high sequence specificity. Null mutations have been constructed within the structural gene for protease III, ptr, with no obvious growth defect (10, 54). It has been suggested that protease III also recognizes larger proteins with abnormal structures since it has been implicated in the in vivo breakdown of a protein A–β-lactamase chimera (10). DNA sequence analysis of the region upstream of ptr suggests that it contains a nitrogen-regulated promoter (33).

Protease I has a somewhat checkered history, and a recent report suggests that it is in fact a periplasmic thioesterase and not a protease. The enzyme was originally purified by its ability to cleave a model substrate, N-acetyl-dl-phenylalanine-2-naphthyl ester (141). However, whatever minor proteolytic activity was originally detected appears to have been due to an impurity, since the purified enzyme shows essentially no activity against casein or small peptides (110). The enzyme is nonessential, since E. coli or S. typhimurium mutants lacking protease I activity and a more recently constructed null mutant grow normally and show no alteration in protein turnover rates (87, 110, 130). Protease I has recently been shown, on the basis of identical gene sequences, amino-terminal protein sequences, and subcellular locations, to be identical to a periplasmic thioesterase I (30, 87). Careful enzymological studies revealed a single active site on the protein that was responsible for both activities and showed that protease I is unable to cleave nonactivated oxygen esters such as those present in normal proteins or peptides (31). Therefore, protease I should now be designated thioesterase I. The only thioesters that are known substrates for this enzyme, however, are cytosolic, so the function of this periplasmic enzyme remains unknown.

Ptr, or Tsp, appears to be an endoprotease that selectively degrades proteins with nonpolar carboxy-terminal regions and to be responsible for cleavage of the carboxy-terminal 11 amino acid residues of penicillin-binding protein 3 in vivo (72, 165). Both minicells and Ptr-overproducing strains indicate that this protein is located both in the periplasm and in the inner membrane, although further localization studies with wild-type strains are needed to clarify this picture (72, 165). A ptr null mutant that is defective in carboxy-terminal processing of penicillin-binding protein 3 has been isolated. This mutant does not form colonies on salt-free rich media at 42°C, does form filaments under these conditions, and is defective in the heat shock response (72). Recently, Tsp has also been proposed to play a role in potentiating long-chain fatty acid transport, although this effect is not pronounced and could be indirect (7). Clearly, further studies are needed to clarify the role of Ptr in protein processing within the periplasmic space, the heat shock response, and periplasmic transport processes.

Although localized to the outer membrane, the OmpT protease (E protein or PgtE in S. typhimurium [69]) has been suggested to play a role in the breakdown of abnormal proteins located within the periplasm. The presence of an ompT mutation stabilizes certain heterologous, periplasmic fusion proteins that are normally subject to endoproteolysis (9, 71). In one case, OmpT-mediated breakdown appeared to occur in vivo, suggesting a role for OmpT in turnover of periplasmic proteins with nonnative structures. However, this finding is inconsistent with the suggestion that the active site of OmpT is on the external side of the outer membrane (68). The topology of OmpT protease and its role in periplasmic proteolysis need further clarification.

CONCLUDING REMARKS

While an overall understanding of the structure and contents of the periplasmic space has been obtained, the periplasm still remains the most controversial and poorly defined compartment in the gram-negative bacterial cell. The exact size and internal architecture of the periplasm remain areas of dispute, and the fine structure that leads to low mobility of periplasmic proteins needs further clarification. While progress on the periseptal annulus, a new organelle that subdivides the periplasm for purposes of cell division, has recently been made, the complex biochemistry leading to the topologically correct polymerization of the peptidoglycan layer within this space and its coordination with cell growth and division have yet to be elucidated. The more general role of membrane adhesion sites and their involvement in macromolecular transport to the outer membrane remain unresolved issues. Isolation of mutations in distal steps of outer membrane protein, LPS, and capsular polysaccharide transport and biochemical and structural analyses of appropriate mutants should help identify the components of these secretion machineries and clarify the mechanism of biogenesis of these polymers. If a genetic handle on the problem of membrane adhesion sites could be obtained, and if reliable protein markers for these structures become available, then important progress could be made on this long-standing problem of gram-negative cell biology.

Recent progress has started to reveal the major principles behind periplasmic protein biogenesis and to identify the cast of characters that mediate specific steps in the folding, modification, and release of proteins into the periplasmic space. Continued studies of this problem should refine our knowledge of this process as well as reveal how additional steps such as protein oligomerization are accomplished. Studies of protein folding and assembly reactions may help shed light also on the definition of native structure and what determines whether a protein will be subject to proteolysis. Such studies could generate well-defined substrates for characterization of periplasmic proteases, allowing better definition of proteases that are responsible for physiological protein turnover within the periplasm. Finally, studies have often neglected the fact that the periplasm is a dynamic structure whose size and content reflect to some extent the environment and metabolic conditions of the cell. More attention needs to be focused on the different forms of communication between the periplasm and the cytoplasm which help coordinate the activities of these two compartments. Elucidating the mechanisms that regulate MDO biosynthesis by osmolality or that coordinate periplasmic protein synthesis with export should serve to promote this latter cause.

ACKNOWLEDGMENTS

I thank Deborah Alexson for assistance with the references. This work was supported by Public Health Service grant GM 42033.