The flagellum and the injectisome are two of the most complex and fascinating bacterial nanomachines. At their core, they share a type III secretion system (T3SS), a transmembrane export complex that forms the extracellular appendages, the flagellar filament and the injectisome needle. Recent advances, combining structural biology, cryo-electron tomography, molecular genetics, in vivo imaging, bioinformatics and biophysics, have greatly increased our understanding of the T3SS, especially the structure of its transmembrane and cytosolic components, the transcriptional, post-transcriptional and functional regulation and the remarkable adaptivity of the system. This review aims to integrate these new findings into our current knowledge of the evolution, function, regulation and dynamics of the T3SS, and to highlight commonalities and differences between the two systems, as well as their potential applications.
1. The flagellum and the injectisome share a type III secretion system
Far from being ‘bags of enzymes’, bacteria have evolved to employ remarkably complex structures and pathways to ensure survival and spreading. Two of the most impressive examples of large macromolecular complexes span both membranes of Gram-negative bacteria: the bacterial flagellum, a rotating filament used in chemotaxis, and the injectisome, an injection device used for interkingdom transfer of proteins.
Although these two machineries clearly differ both in overall structure and function, at their core, they both consist of a conserved machinery for protein export, the type III secretion system (T3SS). In the flagellum, the T3SS is used to export the distal flagellar components and build the extracellular filament. Within the injectisome, the T3SS is at the centre of the export machinery and enables both the formation of the extracellular needle and the direct transfer of substrates from the bacterial cytosol into the host cells. While the term ‘type III secretion system’ is often applied to the whole injectisome, we will use it for the export machinery within both systems (and, accordingly, for statements valid in both cases) and use ‘flagellum’ or ‘injectisome’ to specify the respective complete system.
The flagellum is a rotating semi-rigid helical filament that is anchored within the bacterial membranes and driven by the influx of protons or Na+ ions. It allows bacteria to move within fluid environments, including through liquid films on surfaces. Usually, the rotational direction of the flagellum is coupled to chemosensory receptors, allowing taxis, the movement of bacteria within gradients of chemicals, oxygen and even temperature [1,2]. As most bacteria are too small to sense spatial gradients, they use temporal sensing, comparing conditions at a given time with those a few seconds ago and bias their overall swimming in a favourable direction. The flagellum is a target of the chemotaxis pathway. In the well-studied case of Escherichia coli, the flagella usually rotate in a counterclockwise direction, forming a bundle of flagella which leads to forward movement of the bacterium. Movement away from favourable conditions leads to fewer effector molecules being bound to the cluster of membrane spanning chemoreceptors, and this increases the activity of a kinase, CheA, which phosphorylates a small 14 kDa response regulator, CheY. CheY ∼ P binds to the switch complex of the flagellar motor causing the rotor to switch its rotation direction. This leads to dispersal of the flagellar bundle and tumbling, which randomly reorients the bacterium, causing it to swim in a new direction when the bundle reforms. If this is an improved direction the cell will swim for longer before tumbling, leading to a biased random walk with an overall movement towards favourable conditions. Although the detailed mechanism and integration into the chemotaxis network differs between species, similar mechanisms are found in most flagellated bacteria . While biasing the overall swimming pattern towards an improving environment appears to be the main function of the flagellum, there is evidence that in some species the flagellum can have additional roles. It is known to remodel in response to changes in external load or ion concentration [4,5] and there is increasing evidence of motors being slowed or stabilized by proteins activated in response to the growth state [6–9]. The flagellum may also directly act as a sensor, for example to detect surface proximity by higher load or slowed rotation, which can then lead to a switch in the bacterial lifestyle towards a sessile, biofilm-like mode [10,11].
Injectisomes are transmembrane complexes used by Gram-negative bacteria to translocate so-called effector proteins into eukaryotic host cells, where the effectors influence host behaviour in favour of the bacterium [12,13]. In many species, this process is essential for pathogenesis; however, it is also implicated in symbiosis [14–16]. The pool of effector proteins greatly varies between different bacterial species, reflecting the needs of the respective bacterium. In most studied cases, protein translocation1 is a tightly regulated process that is only initiated upon contact with the host cell.
In some cases, bacteria harbour more than one type of injectisome (reviewed in ). The most prominent example is Salmonella, which uses its different injectisomes, encoded in two different Salmonella pathogenicity islands (SPI), for entry into host cells (in the case of SPI-1) and the subsequent survival, replication and dissemination therein (in the case of SPI-2). While a number of bacterial species have the genetic information for more than one set of flagella, there are only very few cases, most prominently some Vibrio species, in which more than one type is expressed at any given time [18–20].2 However, quite often, flagella and injectisomes are present at the same time and both systems have been shown to heavily modulate each other, suggesting crosstalk between systems.
The aim of this review is to compare the two types of T3SS and to show commonalities and differences between them. There are excellent earlier reviews on the T3SS [12,13,17,22,23] and we will particularly focus on new aspects of T3SS that have emerged in the last few years and key open questions.
2. Structural and functional homologies among the T3SS and other complexes
The common core of all T3SS resides in the transmembrane and cytosolic parts (table 1 and figure 1) and includes five proteins forming the export apparatus3 and the surrounding MS ring in the inner membrane (IM), as well as the cytosolic ATPase, its negative regulator, the proposed stalk at the cytosolic interface and the cytosolic C-ring.
Notably, several of these core T3SS components are structurally similar to components of F/V-type ATPases; most significantly, the ATPase FliI/SctN, the central stalk FliJ/SctO and the negative regulator FliH/SctL resemble the α/β subunits, the γ subunit and the b/δ subunits of the F0F1-ATP synthase, respectively [33,35–37] (figure 1). Assuming that the T3SS ATPase itself does not rotate, this suggests that the FliJ/SctO stalk might rotate and be linked to substrate export [29,33].
The outer membrane (OM) ring structures are not related between the two systems. While flagella include a P-ring spanning the peptidoglycan layer and an L-ring transversing the OM, injectisomes contain a secretin-type ring, probably acquired from various sources later in evolution , that spans both of these layers. The reason for this divergence is unclear; however, the flagellar P- and L-rings need to function as bushings for the rotating flagellum, whereas secretins are very stable structures, shown to seal the OM until their conformation is changed to an open state [39,40], possibly by the growing needle.
The extracellular filaments of T3SS clearly vary in structure. The needles in injectisomes of animal pathogens and the flagellar hook differ in size and structure, but have a very similar helical arrangement . In plant pathogens, the needle is replaced by a pilus, which can be up to several micrometres long, and is thought to be able to penetrate the thick plant cell wall. Beyond the injectisome needle the structures vary. In most animal pathogens there is a small tip structure built by a hydrophilic translocator protein  thought to allow pore formation in the host cell by two interacting hydrophobic translocators, but the injectisomes in enteropathogenic E. coli (EPEC) have a flexible sheath of up to 0.6 µm length and are able to span the mucous layer in the intestine . The flagellar hook is extended by a long (up to 10 µm) semi-rigid helical filament made of flagellin; hook and filament are joined by two junction proteins. The C- and N-terminal domains of flagellins are conserved and allow the folding and polymerization into an 11 start helical tube, while the middle domain, exposed on the outer surface of the filament, varies between species .
With the possibility of imaging T3SS in situ by cryo-electrotomography, the interest in structural studies of the T3SS has been revived. A study of flagellar motors from different bacteria revealed that while a structural core of the conserved proteins was visible in all cases, the overall appearance of the motors as well as peripheral structural elements showed striking differences, highlighting the adaptability of the flagellum . Recent studies have visualized the injectisome in situ under similar conditions [27,46,47]. They showed a certain flexibility within the structure and differences to structures based on purified injectisomes , and allowed the comparison of the structures of the T3SS in the flagellum and the injectisome [27,47]. A particularly interesting structural detail is the C-ring, which is clearly visible as part of the switch complex in the flagellum. While there are homologous proteins in the injectisome, no C-ring structure was detected in the first injectisome cryo-ET studies [27,46]. However, six cytosolic ‘pods’, that were absent in strains lacking the C-ring component, were visible in a recent cryo-ET structure , suggesting that the architecture of the corresponding structure is significantly different in the injectisome and the flagellum.
Flagellar motors spin against a ring of stators, which are not present in hook-basal body isolations or visible in cryo-ET of species with extracellular filaments. However, they are visible in spirochaetes with periplasmic filaments , which rotate at high load. No equivalent proteins have been detected in injectisomes, and although there is a report of rotation of the injectisome needle tip , this observation has so far not been repeated or functionally linked to any T3SS component.
3. Injectisome-T3SS derived from flagellar ancestors
The evolutionary relationship between the flagellum and the injectisome has been the subject of considerable debate. Flagella exist in both Gram-positive and Gram-negative bacteria, while injectisomes have so far only been found in Gram-negative species, suggesting a flagellar T3SS as ancestor [24,54]. However, early phylogenetic studies suggested that both systems share a common ancestor, and have since evolved differently from each other . With the ever-growing number of genome sequences [56,57], it became clear that flagellar components are almost always encoded on the bacterial chromosome and co-evolved with the rest of the genome to a certain degree [58,59], while the genes coding for the injectisome are often encoded on virulence plasmids or pathogenicity islands, and are more closely related to each other than their flagellar counterparts. They are distributed independently of the phylogeny of the respective species, and have probably been frequently transferred between species [38,55,60]. Latest analyses suggest that the ancestral T3SS was used for locomotion and was similar to current-day flagella, while modern injectisomes were derived through a series of gene losses and subsequent acquisitions via a non-flagellar ancestor that did not translocate proteins, leading to a protein translocation machinery  (figure 2).
Within the injectisomes, eight subfamilies can be distinguished (table 2). While, as mentioned above, there is no clear correlation between the phylogenetic tree of bacteria and their injectisome, injectisome subfamilies do confer host specificity. While the Ysc, SPI-1, SPI-2 and Chlamydia subfamilies target animal cells, Hrp1, Hrp2 and Rhizobia have plant host cells, and the more distantly related Myxococcus subfamily is present in free-living, non-pathogenic or symbiotic bacteria [12,38]. The Myxococcus subgroup lacks the secretin ring in the OM and any detectable needle subunit [38,56,61]. While it might form a fully functional T3SS by associating with a secretin of one of the type II secretion systems present in Myxococcus, it was also proposed to belong to an intermediate ancestral form of T3SS (figure 2) specialized in transport across the IM [38,61]. Interestingly, many pathogens harbour two different injectisomes, which almost always belong to two different subfamilies [17,38]. As an example, many Yersinia strains have the genes for the Ysc T3SS expressed in animal cells at 37°C and the Ysa T3SS which is not expressed under these conditions. Recently, the Ysa system was shown to be involved in intracellular survival and to be required for the infection of Drosophila S2 cells at lower temperatures, suggesting a role in an alternative insect host [62,63].
4. Transcriptional and post-transcriptional regulation of the T3SS
It appears obvious that both the expression and the function of T3SS must be regulated. Environmental inputs control not only whether a bacterium needs a flagellum, but also how many, while injectisomes are generally only useful within the host organism.
Expression of flagellar genes is tightly regulated, ensuring sequential production of proteins for ordered assembly. In E. coli, a class 1 master operon, flhDC, regulated by metabolic state and cell cycle, controls the expression of class 2 operons, which encode for the proximal structural components of the flagellum including its T3SS core. After completion of the T3SS, the anti-σ28 factor FlgM, a T3SS substrate, is exported, which allows the expression of class 3 genes, including the filament subunit, by the alternative sigma factor σ28 . In addition, the expression levels of flagellar components are often regulated as part of a switch between a free swimming lifestyle and a sessile or biofilm-associated lifestyle. A key player in this switch is the small cytoplasmic signal molecule cyclic-di-GMP (c-di-GMP) [65–67]), which itself may be influenced by the cell cycle, linking flagellar biosynthesis to cell division. In general, high c-di-GMP levels lead to downregulation of flagellar transcription [68,69].
For the injectisome, a related transcription hierarchy exists. However, in this case, all structural components belong to class 2 operons, with the effectors corresponding to class 3. Effector transcription is strongly upregulated upon contact to host cells. While the underlying mechanisms seem to differ between organisms , this generally involves the export of a regulatory protein, leading to release of its cytosolic binding partner, which then directly or indirectly induces transcription.
Interestingly, transcription of the two types of T3SS often influences each other, and in the majority of the cases flagella and injectisomes are counter-regulated. In species with both systems, flagellar genes are often activated only at temperatures below 30°C, while injectisome genes are transcribed within the host organism, which for animal pathogens is sensed by a temperature of 37°C [70–74]. In addition, there is direct counter-regulation involving the main transcription factors FlhDC and VirF [75–79], in agreement with a general switch between motility and virulence. As the export signals for flagellar and injectisome substrates are non-exclusive (see below), such a strategy might help to prevent the export of flagellum substrates through the injectisome (which in the case of flagellin might induce host cell defence) and vice versa.
It has become obvious that the carbon storage regulator (Csr) system  is a key factor in the post-transcriptional regulation of T3SS components. Csr is a regulatory complex in which non-coding Csr RNAs titrate the binding of CsrA to mRNA targets, leading to degradation of these targets. The Csr system has pleiotropic effects on the expression levels of injectisome components in Salmonella [81,82], EPEC  and Yersinia (reviewed in ), and was recently shown to downregulate the injectisome in persistent Yersinia infections . CsrA also governs the co-regulation of the injectisome with other virulence mechanisms. In Pseudomonas aeruginosa, the RetS sensor protein, another system involved in regulating Csr, switches between expression of the T3SS and the type VI secretion system by regulating c-di-GMP levels .
CsrA also controls motility, with CsrA upregulating the expression of flagella by stabilizing the transcript of the flagellar master operon flhDC [87,88]. Interestingly, CsrA not only regulates motility in this post-transcriptional manner, but is also directly involved in flagellum biosynthesis by suppressing flagellin biosynthesis prior to completion of the flagellar T3SS . Based on the fact that the CsrA gene is mainly present in genomes that encode flagellin, this has been postulated to be the primordial function of the Csr system . In addition, CsrA directly regulates the expression of several main regulators of c-di-GMP levels, connecting the two main pathways influencing flagellar biosynthesis [90,91].
Both T3SS also can sense and react to the proximity of host cells or other surfaces. Flagella can directly sense the increased load on the motor, once a flagellum is stuck because of contact to a surface, which can lead to a ‘swim-or-stick switch’, resulting in the downregulation of expression of proteins needed for motility and induction of biofilm formation (reviewed in ). This effect has been clearly demonstrated in Caulobacter crescentus, where production of holdfasts, adhesive polysaccharides at the attached cell pole, could be visualized within minutes after attachment and flagellar arrest [10,93], and in Bacillus subtilis, where inhibition of flagellar rotation leads to biofilm formation by triggering the DegS–DegU two-component signal transduction system and other pathways [11,94,95].
Injectisomes also react to approaching cells, often upregulating expression in preparation for effector translocation upon cell contact. To detect this proximity, both changes in local oxygenation [96,97] and local pH  can be sensed. Similarly, intracellular pathogens can determine the correct time for expression of the injectisome, as shown for Salmonella SPI-2, which is upregulated and activated by the decreased pH in the vacuole , sensed by acidification of the bacterial cytosol . Interestingly, some species have also been shown to display bimodal expression of injectisome subunits during infection, leading to injectisome-positive and -negative subpopulations of the same strain, as was shown for Salmonella SPI-1 , EPEC  and Shigella , which might allow a subset of bacteria to benefit from the effects of the T3SS while circumventing the associated energy expense. The regulation of the T3SS is summarized in figure 4.
5. What determines the number and cellular distribution of T3SS?
The production and assembly of injectisomes and especially flagella, with around 20 000 flagellin subunits in the filament , come at a high energetic cost. Perhaps even more importantly, the number and localization of T3SS is crucial for their physiological function: although a recent study has shown remarkable robustness of E. coli swimming to variations in the number of flagella , both the number and the relative positioning of the flagella will directly influence the movement of the bacterium. Likewise, the chance of a bacterium establishing contact to a host cell with an injectisome directly depends on their number and distribution.
Depending on the species, flagella can be distributed peritrichous (distributed across the bacterium), medial or polar, either as a single flagellum or in tufts. While peritrichous flagella are usually evenly spread out across the bacterium (with the possible exception of the tips) and their number largely correlates with the cell size within a species, the insertion of new polar flagella is controlled by the cell cycle and often requires a polar marker . Examples for such markers and their relation to the cell cycle are the GTPase FlhF and its negative regulator, the MinD-like ATPase FlhG (reviewed by ), found in many species, and the c-di-GMP receptor TipF and its interactor TipN in C. crescentus [107,108]. Interestingly, in both cases, the localization of flagella involves an interaction with components of the flagellar switch complex [108,109], suggesting that the final localization of the flagellum is at least influenced by the cytosolic components. Notably, c-di-GMP is involved not only in flagellar positioning but also in direct regulation of motility (see below, section ‘Functional regulation and dynamics’).
Injectisomes are more difficult to visualize than flagella. However, based on electron micrographs, the number of injectisomes strongly varies between different systems. The Salmonella SPI-2 system was found to be present in very low numbers (one to a few ), whereas 10–100 SPI-1 needle complexes  and 50–100 Shigella needle complexes  were observed per bacterium. The distribution of injectisomes was determined by fluorescence microscopy of functional fluorescently labelled components in Yersinia enterocolitica [113,114]. All analysed components were distributed in relatively evenly spaced foci around the bacterium, which is also the case in Shigella . Recently, it was shown that these foci in Y. enterocolitica may correspond to clusters of injectisomes, which themselves tend to maximize the distance to their next neighbours, and that the number of injectisomes per cluster, rather than the number of clusters, increased upon activation of the T3SS . The average number of injectisomes per Y. enterocolitica cell ranged from 10 to 30 in this study.
Although these numbers may generally represent the lower limits of possible ranges, due to the possible loss, non-visualization, or clustering of injectisomes, these data indicate that number and distribution of injectisomes strongly vary between different systems, based on their requirement. As an example, the difference between the number of SPI-1 and SPI-2 T3SS in Salmonella may reflect the different roles of these systems in the infection process . SPI-1 injectisomes are formed by extracellular Salmonella and a large number conceivably increases the chance of establishing host cell contact, while SPI-2 is activated by bacteria tightly surrounded by the vacuolar membrane, where a single injectisome seems to be sufficient to establish contact and translocate the corresponding effectors.
The distribution of T3SS is probably linked to where they are integrated into the membranes and, importantly, the peptidoglycan (PG) layer; therefore, PG-modifying enzymes, such as PG lyases, play an important role in the number and positioning of T3SS. Interestingly, many such PG lyases or muramidases are encoded within T3SS gene clusters [117–121] and a recent report even indicated an interaction between a lyase and the injectisome inner rod . Similarly, in the flagellum, the PG lyase FlgJ has been shown to assist in the formation of the inner rod with its N-terminus, while the C-terminus exhibits a muramidase function [122,123]. However, PG lyases appear to be functionally redundant as not all T3SS have a dedicated PG lyase, and while the absence of single PG lyases may influence the positioning of T3SS, it does not lead to strong overall T3SS phenotypes.
6. Transcription hierarchy, assembly and activation of the T3SS
In the flagellum, the assembly order roughly reflects the three-tiered expression hierarchy mentioned earlier. Comparing the structure of flagellar precursors in strains lacking single components, Kubori et al.  concluded that the actual assembly of the flagellum starts with the MS ring in the IM. Next, the C-ring/switch complex can form. The integration of the export apparatus could not be visualized in these experiments. The presence of the large export apparatus component FlhA has been shown to induce and stabilize MS ring formation, arguing for an early role of FlhA in the assembly of the flagellum , although it is not absolutely required for MS ring formation . For more detailed reviews on the assembly of the flagellum refer to [127–129].
In the injectisome, all structural components belong to class 2 operons and the proteins that make up the injectisome are therefore likely to be expressed and present at the same time. The assembly of the machinery (reviewed in ) is therefore thought to be mainly governed by protein interactions. While studies in Salmonella SPI-1 suggest an inside-out assembly of the membrane rings [130–132], similar to the model for the flagellum, the secretin was shown to nucleate membrane ring assembly in Y. enterocolitica , similar to type IV pili . Independently, the export apparatus can assemble within the IM [114,134]. After the concurrent assembly of the cytosolic complex, requiring all of its components SctN, SctL, SctQ and SctL (but not the stalk SctO), the T3SS is functional and the needle can be formed [113,135].
The membrane rings and IM export apparatus are exported via the Sec pathway. However, once the core T3SS is assembled, it starts to independently export early substrates, the flagellar hook FlgE and the hook-capping protein FlgD in the case of the flagellum and the needle subunit in the case of the injectisome. In both cases, the correct length of the filament or needle is measured, in a so-far incompletely understood way, by a ruler protein (FliK/SctP) [136–138]. In interaction with an export apparatus component, the type III secretion substrate specificity switch (T3S4) protein FlhB/SctU, it induces a change in substrate specificity, allowing secretion and assembly of the flagellin subunits and the needle tip proteins, respectively (figure 3).
While the injectisome is functional at this stage, the secretion of effectors is still suppressed, until contact to a host cell has been established (figure 3). At the moment, it is unclear which molecular events lead to effector export, but the large variety in experimentally determined activation signals in different species (apart from host membrane contact, effector secretion is controlled by extracellular Ca2+ levels in Yersinia, O2 levels in Shigella  and the local pH in Salmonella SPI-2 [98,139]) suggest that this process might be triggered by species-specific signals. The existence of various signalling-deficient mutants in the tip and the needle [140–145] suggests that at least some of these signals are sensed at the tip, and transferred via the needle. Deletions of the tip protein in P. aeruginosa and Shigella lead to uncontrolled secretion [142,146], implicating a plug at the tip of the needle. However, this is not the case for all injectisomes and a cytosolic protein complex present in all injectisomes, termed the cytosolic plug or gatekeeper complex, is involved also in substrate control. Absence of either component of this complex of cytosolic proteins leads to uncontrolled secretion of effectors [147–151]. A similar role has been shown for a regulatory protein, LcrG, in Yersinia, which binds to the tip protein in the cytosol [152–154]. Constitutive secretion in the absence of any of these cytosolic components suggests an active prevention of secretion of late substrates prior to the second switch. Interestingly, it has been recently shown that the secretion control by the tip plug and cytosolic gatekeeper complex may be a part of the same mechanism, which possibly involves the inner rod protein [155,156].
Formation of the flagellar filament, which is the equivalent substrate class in the flagellum, does not seem to require an additional signal. Also, in contrast to the hook, which has a controlled length, the length control of the filament may vary from species to species, with some species appearing to have a defined length and others continuously growing filaments [157–161]. In all studied cases, the filament rapidly regrows if it is sheared despite the loss of the cap proteins catalysing filament growth.
7. Export signals
Export signals ensure that cargo is delivered to and exported by the correct secretion system. In contrast to other export pathways like Sec or Tat, the T3SS export signals are remarkably vaguely defined. While an uncleaved and generally unstructured N-terminal sequence seems to be common for all T3SS substrates, the length of the minimal signal greatly varies, additional signals seem to be required for some classes of substrates and, despite increasingly sophisticated algorithms [162–166], it is still difficult to precisely predict T3SS substrates.
While T3SS export signals appear to exclude export via unrelated export systems, there are numerous reports that the signals are interchangeable between different T3SS, and that flagellar substrates can be exported by injectisomes and vice versa [167–173]. Although flagella and injectisomes are rarely coexpressed in vivo, these findings highlight that the respective signal sequences are not exclusive.
The injectisome export signal has been studied in more detail. Most studied T3SS effectors have one N-terminal signal sequence located within the N-terminal 15–40 amino acids. While this sequence is sufficient to target even unrelated proteins for export through the injectisome , in some cases, the signal required for translocation into the host cells was found to be longer than the one sufficient for export via the T3SS [175,176]. Studies from various organisms show numerous silent mutations within the signal sequence have no or little effect on export, implying that the export signal is predominantly encoded in the amino acid sequence [177–179]. However, it has also been shown that frameshifted mRNA export sequences can still support the export of substrates and that the mRNA sequence around the start codon plays a role in export ([180,181], reviewed in ), indicating that the effects of mRNA and protein sequence on targeting may be additive.
While most effector proteins have one N-terminal secretion signal, other substrate classes appear to have additional export signals: the Yersinia ruler protein features a central stretch that is also required for export , and translocators as well as some effectors were shown to require both N- and C-terminal signals [184–186].
In the flagellum, the N-terminal parts of the flagellin or hook protein have been identified as secretion signals, albeit the very N-termini were not strictly required for secretion [187–189]. Both the identified minimal export signal of the flagellin and full-length anti-σ28 factor FlgM have been successfully used as export signals for substrates which could subsequently cleaved by TEV protease [190,191]. As in the injectisome, both the peptide sequence and an mRNA signal seem to play a role in substrate recognition .
Export signals are intrinsically unstructured and unconserved, and it has not been possible to identify a common interaction partner at the T3SS. Rather, the export signal has been proposed to increase the local concentration of the target protein at the flagellum. While this process appears to be sufficient to allow the export of (usually overexpressed) non-native secretion substrates, it has been suggested that the targeting of native substrates is further enhanced or rendered more specific by the binding of their chaperones (box 1).
T3SS chaperones are proteins assisting assembly and/or export of their cargo, effectors (class I chaperones), hydrophobic translocators (class II chaperones) or the components of the needle or flagellum (class III) . Class I chaperones are separated into two subclasses. While class IB chaperones (also called general chaperones) bind a variety of effectors (similar to the class II chaperones that often can bind both hydrophobic translocators), the class IA chaperones, representing the majority of T3SS chaperones, have one specific target protein and are often encoded next to this protein. They are relatively small acidic proteins that form dimers. Different, non-exclusive roles have been assigned to T3SS chaperones: (i) they might keep effectors in a (partially) unfolded state, facilitating the unfolding prior to export; (ii) they have a role in signalling, indicating the export of their binding partner [192–199]; (iii) they can mask a localization domain for targeting in the host cell, preventing mislocalization within the bacterium ; and (iv) they may play a role in targeting of their cognate effector and introduction of a secretion hierarchy. Concerning the last role, chaperones have been found to specifically target effector proteins to the injectisome rather than the flagellum , to impose a secretion hierarchy for the injectisome through decreasing affinities to a cytosolic ‘sorting platform’  and, in a similar way, but with another interaction partner, the export apparatus component FlhA, to coordinate the assembly of the flagellum .
8. Type III secretion export is a multi-step process with many unknown features
Despite intense research, the export process through the T3SS is not well-understood. A myriad of interactions among the proteins involved at the cytosolic interface of the T3SS, and between these proteins and export cargo have been reported (reviewed in ). In addition, the interaction network of the cytosolic components might well change between different functional states of the T3SS. A candidate for such a switch is FliH/SctL, the negative regulator of the ATPase, which can additionally interact with both the major export apparatus component and the C-ring [27,28,203].
Even if not all of these interactions are necessarily relevant in vivo, their multitude suggests a multi-step binding and export process rather than one or two decisive interactions. Beside the low specificity of export signals, this notion is supported by the fact that loss of some conserved cytosolic components of the T3SS can be compensated for. As an example, the flagellar ATPase FliI is not absolutely required for the formation and rotation of flagella, especially in the absence of its negative regulator FliH and at increased proton-motive force (PMF) [204,205], an effect that was to some extent also observed for a catalytically inactive injectisome ATPase . Similarly, absence of the flagellar C-ring component FliMN could be compensated by overexpression of the flagellar main transcription factor FlhDC  or even the ATPase FliI alone , although similar experiments did not lead to the same outcome for the injectisome (A Diepold 2009, unpublished data).
Although the path of cargo is thus unclear, the narrow diameter of both the injectisome needle4 and the flagellar filament (20–25 Å, accommodating a maximum of one to two α-helices) [211,212] means that proteins travelling through these appendages will have to be at least partially unfolded prior to export . In agreement with this assumption, very stably folded artificial T3SS substrates such as ubiquitin, GST or GFP were shown to be rejected by the export machinery [214,215] or even block the machinery in the cases of fusions to the ruler protein . A recent study exploited this fact to visualize a trapped export substrate in the injectisome needle  and showed that presence of this substrate leads to distinct conformational changes within the basal body.
The main player in unfolding the substrate (and the concurrent removal of the chaperone) is the ATPase , which has been shown in many studies to engage with effectors and/or their chaperones. Recent structural data showed that some effector/chaperone complexes can form hexamers, which would fit the hexameric ATPase structure . Interestingly, it was shown that targeting to the ATPase and secretion are largely independent events: while chaperone binding, rather than presence of the secretion signal, was required for association of a substrate with the ATPase, this was insufficient for secretion of the substrate .
Besides its role in the flagellar switch complex, the C-ring is also involved in the export process. Many reported interactions with both cargo proteins and machinery components indicate a multifaceted role in cargo export. More specifically, in the Salmonella injectisome, the C-ring protein participates in a cytosolic ‘sorting platform’ (together with SctL and SctK), proposed to ensure the secretion hierarchy through different binding affinities to T3SS chaperones, which bind cargo proteins . In the Yersinia injectisome, the C-ring protein exchanges with a cytosolic pool  and since the exchange rate correlated with effector export, the C-ring might be involved in the functional regulation of the T3SS .
9. Energy transduction for export
Most transport processes in bacteria use energy provided by either the PMF across membranes or by ATPases. For the T3SS, both factors have been shown to be important—while mutations in the T3SS-associated ATPase FliI/SctN abolished type III secretion [220–222], the same was true for loss of the PMF [213,223,224] (reviewed in [23,225]). So far, the specific roles of both factors are not completely understood, although the finding that absence of the ATPase can be overcome to a certain degree by overexpression of other flagellar components, mutations in the export apparatus components FlhA and FlhB, or increased PMF [204,223] suggests that the PMF is more central to energizing export. The main role of the ATPase could therefore be to detach chaperones and unfold the cargo protein prior to export . On the other hand, such unfolding may directly contribute to the energy balance of translocation, as the entropic force stored in the unfolded subunits could be used to direct transport through stepwise refolding after export. Two variants of this latter model have been proposed for flagellin export: cargo might enter the flagellum through single-file diffusion  or, more efficiently, by coupling subunits: the folding of the N-terminus of one filament subunit would generate a pulling force allowing the C-terminus of the same subunit to pull along the N-terminus of the next subunit, and so on . Such a model is compatible with a recent publication showing a constant growth rate of the flagellum at the distal tip , in contrast to earlier work observing an exponential decrease in growth rate with length [158,228,229], which fits a model where flagellin subunits were pumped into the flagellum from the base and slowed down by protein compression and friction . Which of the models best describes export, and whether the export process is energized similarly for all cargo classes, remains to be shown.
Several proteins have been shown to influence the energy transduction, most interestingly FliJ/SctO, which in P. aeruginosa has been recently proposed to harness the PMF together with another cytosolic regulator, PcrD . However, as SctO also influences other processes in T3SS and has been shown to interact with and positively regulate the ATPase activity in EPEC [33,231] and to be an important component in the determination of substrate hierarchy through interactions with FlhA and other components [232,233], it is very difficult to distinguish between its direct and secondary effects in energy transduction. Energization of T3SS export is described in more detail in a recent review .
Type III secretion export is very efficient. Injectisomes can export hundreds to thousands of effector molecules per second [234–236], while flagella grow with a rate of about 10 flagellin subunits per second, corresponding to about 5000 amino acids . Interestingly, the high rate of export in injectisomes correlates with strong inhibition of growth and division in different species [100,237]—in fact, the correlation of Yersinia growth and pathogenicity at 37°C (with the absence of Ca2+ inducing effector secretion and preventing growth under these conditions) was one of the earliest indications of type III secretion [238–240]. Despite this, it is still not known what exactly leads to this growth inhibition. It is unlikely to be the cost of producing injectisomes, as under non-secreting conditions, a considerable number of injectisomes (about half the number compared to secreting conditions ) are produced with little effect on growth. Similarly, effector production and export, or an adverse effect of effector secretion on the bacterium are unlikely to be the sole cause, as growth restriction under secreting conditions does not correlate well with presence or absence of effectors. Also, growth inhibition and secretion could be experimentally uncoupled under certain conditions . This suggests either crosstalk between T3SS and bacterial metabolism  or a specific mechanism resulting in loss of division upon type III secretion activation . In either case, the involved pathways remain to be identified.
10. Functional regulation and dynamics
The activity of T3SS must be under tight control—chemotaxis is only possible if flagellar rotation can be adjusted, and tightly regulating the translocation activity of injectisomes is crucial to ensure an optimal amount of protein is translocated into the host cell without eliciting host responses that are detrimental to the survival of the bacterium, especially in the case of persistent infections. However, in contrast to the transcriptional regulation of T3SS, the functional regulation is less well-understood. The best-studied case is the control of flagellar rotation by the chemotactic pathway, which is mediated by the response regulator CheY, which in its phosphorylated form can bind the C-ring protein FliM , leading to reversal or braking of the flagellum. The same pathway is used for bacterial thermotaxis . Additionally, c-di-GMP plays an important role in the functional control of flagella, in addition to its effects on the transcriptional and post-transcriptional regulation of motility: the c-di-GMP-binding protein YcgR represses motility by binding to the MotAB stator [6,9], as well as to the C-ring [7,8]. In addition, high c-di-GMP levels lead to accumulation of extracellular cellulose in Salmonella, which sterically reduces motility . Heterogeneity in the cellular levels of c-di-GMP, caused by the activity of a phosphodiesterase, which in turn is controlled and asymmetrically partitioned by binding to the chemotaxis histidine kinase CheA in P. aeruginosa, accordingly leads to different motility and chemotactic response between the two daughter cells after division [245,246].
In the case of the injectisome, activation of secretion can be triggered by an increase in local oxygen levels (occurring close to the gastrointestinal tract mucosa, the site of Shigella invasion) , a drop in pH (in the vacuole activating Salmonella SPI-2)  or extrabacterial Ca2+ depletion (which might mimic the low levels of free Ca2+ in the host cytosol for Yersinia). In addition, it is conceivably beneficial to regulate the amount of effectors that are translocated into a host cell as well as the long-term activity of the system, especially in persistent infections. Indeed, it was shown that previous secretion of certain effectors into a host cell suppresses the T3SS activity of bacteria that subsequently attach to the same cell [146,247], suggesting an active suppression of excess translocation.
The rapid development of fluorescence microscopy techniques has allowed the composition of T3SS to be monitored in real time, and revealed that T3SS are by no means static complexes, but can exchange subunits while functioning (reviewed in ). In the flagellum, various components of the cytosolic interface were shown to be dynamic: MotAB, the flagellar stator complex, exchanges with an average dwell time of about 30 s , and indeed is only maintained in place by a conformational change caused by the ion flow [4,250]. In the E.coli flagellum, there can be up to 11 MotAB stator complexes and their number increases with the external force on the filament [4,251–254].5 While the PMF is required for the association of stator complexes with the rotor, individual stators only stay associated for about 30 s before exchanging with stators in the membrane, even at full PMF. In the case of Shewanella, H+-driven MotAB stator complexes exchange with their Na+-driven PomAB counterparts if the extracellular concentrations of H+ and Na+ ions change . Interestingly, the rotating flagellar C-ring components FliMN also exchange within the working flagellum [255–259]; in this case, turnover depends on the rotation direction of the motor, which in vivo is regulated by the presence of the activated response regulator CheY [255,258]. The number of FliM and FliN proteins in the C-ring of the counterclockwise rotating flagellum is 26–29% higher than in the clockwise rotating flagellum [258,259]. These results suggest that remodelling of the flagellar motor is involved in its adaptation, which has been shown to account for the extreme sensitivity of the motor to environmental changes . FliI, the flagellar ATPase, was also proposed to dynamically bind to the C-ring and other components of the flagellar cytosolic interface in a complex with its regulator FliH, based on binding studies ; recently, this exchange has been shown directly .
In the injectisome, less is known about the behaviour of the cytosolic components. Recently, however, the C-ring component SctQ has been shown to exchange with a cytosolic pool in the functioning T3SS , and the SctQ exchange rate correlated with the export process.
In all of these cases, the T3SS appears to react differently to a variety of signals, suggesting that these dynamic components allow modulation of the function of the T3SS in response to internal and external cues, ensuring both efficiency and adaptivity of these important molecular machines.
Figure 4 summarizes the influence of external signals on the expression and function of the T3SS.
11. Applications of the T3SS in medicine and biotechnology
Both the injectisome and the bacterial flagellum are targets for the development of novel antimicrobial agents, preventing essential steps in colonization. On the other hand, they can also be exploited. The most obvious application of the T3SS is to export specific proteins. Proteins of choice can easily be targeted to the machinery by a short N-terminal signal (which may be cleaved off at a later stage), and are specifically exported into the supernatant, which leads to significant enrichment of soluble protein without additional purification steps. Both types of T3SS have been used for this purpose: the injectisome was used to export various proteins, including spider silk proteins , and has recently been shown to be a useful tool for the production of biopolymer-forming proteins, with increased export levels upon overexpression of the master transcriptional regulator . Similarly, the flagellar T3SS was shown to be able to export a variety of different protein types  and subsequently optimized for protein secretion .
The injectisome, a system evolutionarily optimized for protein translocation, has also been used to directly transport proteins into host cells, especially to deliver vaccines [264–267], but also for immunomodulatory cytokines . Traditionally, attenuated strains were used for this purpose, but recently, achromosomal, but T3SS-active minicells were also shown to be functional for translocation . Besides such medical approaches, injectisomes might be useful carriers for the delivery of proteins into cultured eukaryotic cells [270,271].
Beyond the exploitation of the T3SS for protein transport, its application as a biosensor is conceivable. Possible future applications include the use of the flagellum as a sensor for biofilm formation, or, similarly, the injectisome to sense cell contact.
12. Open questions
Despite intense research, the molecular events that govern the export of substrates via the T3SS remain largely unknown. Many interactions between T3SS components and substrates and their chaperones have been recorded, but so far no clear picture emerges. Given the recent findings that several core parts of the machinery are not absolutely essential for export, it is possible that T3SS export relies on a series of local enrichment steps. Cross-linking studies and single-molecule fluorescence microscopy experiments for different classes of cargo under various conditions might help to uncover the path of cargo through the T3SS.
A closely linked question is the energy transfer for T3SS export. While it appears clear that the PMF plays an important role in export, no conclusive theory for how the PMF is linked to export has yet been provided.
An important player in the T3SS is the export apparatus in the IM. Owing to the high proportion of transmembrane helices in the export apparatus, both structural and functional knowledge of its components is limited, but with the application of cryotomography and selective cross-linkers, important insights into the structure of the export apparatus and possible homologies to other export systems are to be expected.
While the influence of many global regulatory networks on the expression and, in some cases, function of both types of T3SS has been revealed in the last years, it is largely unknown how these signals are processed and implemented at the T3SS. This is especially true for the injectisome, where very little is known about the functional regulation of the system.
Understanding the regulation and molecular mechanism of the T3SS will not only help us to appreciate the function of these complex nanomachines, but may also help to control and interfere with the T3SS, both in preventing the function of the system to specifically prevent pathogenicity, and to tailor the T3SS for biochemical applications from targeted translocation to locomotion.
We declare we have no competing interests.
We thank EMBO for a long-term fellowship (ALTF 170-2011) to A.D. BBSRC is thanked for continuous funding of the work in the Armitage laboratory. This review does not claim to be comprehensive. The authors tried to preferentially highlight recent studies that have not been mentioned in previous reviews and apologize for any omissions.
One contribution of 12 to a theme issue ‘The bacterial cell envelope’.
↵1 In this review, we define ‘translocation’ as the process of transporting a protein from the bacterial cytosol into a host cell, while we use ‘secretion’ to describe the mere export of a protein from the bacterium into the extracellular milieu.
↵4 Effector export is generally believed to occur through the needle, as visualized for the trapped ruler protein in . However, based on the presence of translocators and effectors on the surface of the bacterium prior to host cell contact and the observation that such extracellular effectors could be translocated into the host cell in a separate (but still T3SS-dependent) step, Wolf-Watz and co-workers proposed a binary AB toxin-like export mechanism translocation model [209,210]; the significance of this finding in vivo is so far unclear.
↵5 In agreement with this observation, species swimming under high load like spirochaetes, as well as Helicobacter, have more than 11 stators, supporting the idea that a higher number allows increased torque generation [45,48].
- Accepted July 23, 2015.
- © 2015 The Author(s)
Published by the Royal Society. All rights reserved.