Genetic Dissection of the Signaling Cascade that Controls Activation of the Shigella Type III Secretion System from the Needle Tip

IpaB must be secreted to exert its regulatory function

To resolve whether IpaB must be exposed on the cell surface to assemble functional TCs, we made an ipaB mutant lacking its first 20 amino acids, predicted to contain the secretion signal⁴⁹. ipaB∆2-20 (Table 1) expressed IpaB at 35% of the level of WT (Fig. 1B), presumably because it binds its chaperone with reduced efficiency. We found that expression of below 20% wild-type IpaB levels in ipaB⁻ leads to maximal fast constitutive secretion while 50% of normal IpaB levels greatly reduces it (Fig. S1). IpaB∆2-20 is not secreted (Fig. 1C, bottom) and ipaB∆2-20 displays fast constitutive secretion and CR unresponsiveness, as in ipaB⁻ (Fig. 1C, top and 1D). Hence, ipaB∆2-20 causes maximal constitutive secretion because it cannot be secreted, leaving the TC immature and hence dysfunctional.

All but one ipaB mutation unresponsive to CR localize to the alacoil

To assess IpaB’s involvement in sensing the activation signal, we searched for ipaB mutants blocked in secretion activation. For this, we screened a library of ipaB mutants based on their color on plates containing CR. Wild-type Shigella are orange on CR plates⁵⁰. This may reflect secretion of early and late effectors in response to CR. Bacteria lacking functioning T3SSs are white⁵¹ and those lacking ipaD or ipaB are red⁴⁶, presumably because they secrete more late effectors.

A library of random ipaB mutants was transformed into ipaB⁻. Around 1.45 × 10⁶ transformants were screened, 241 white clones isolated and 35 white mutants confirmed by sequencing (Materials and Methods; Table 1). Some mutations appeared more than once and occasionally more than one mutation was found in ipaB. Individual mutations were separated to identify which was responsible for loss of CR-sensing capacity (Table 1). Only single point mutations causing the white phenotype, hereafter termed ipaB^*, were further investigated.

No ipaB^* mutant was altered in its expression (Fig. 1E, top), ability to store and leak others Ipas and IpgD (Figs 1E, middle panels and 3F) or to repress expression of the late effector IpaH (Fig. 1E, bottom). However, these mutants showed degrees of reduced sensitivity to CR-induction that, for some, was similar to that seen for previously characterized CR-insensitive ipaD mutants²¹ (Fig. 1G; quantified in Fig. S2). All but one mutation (ipaBN264I) localized to IpaB’s alacoil (Fig. 1H). In total, 48% of the independently identified 43 mutations sequenced localized to the alacoil (amino acids 74–224), which encompasses only 26% of the protein length. This suggests the region is key to secretion initiation. Mutants had strong (ipaBK93N and ipaBN116Y) or mild defects in secretion (ipaBN85I, ipaBQ108L, ipaBK150E, ipaBK188E and ipaBN264I) in spite of similar expression levels (Fig. 1E, top). Therefore, their phenotype is due to a direct effect of the mutations on IpaB function.

The IpaB* mutants are secreted in a constitutive secretor background

To assess if the ipaB* mutations impaired secretion of IpaB, and hence perhaps secretion of the other Ipa/Ipg proteins, we used constitutive secretor ipaD⁻ ipaB^{− 21}. The ipaB* mutants were transformed into this background and their secretion profile analyzed by Western blot. IpaB* mutants were expressed and secreted at the same levels as wild-type IpaB (Fig. S3A), indicating the newly isolated mutations do not affect IpaB’s ability to be secreted.

Most IpaB* mutants form TCs with normal composition

We next used fluorescence-activated cell sorting (FACS) to assess the overall composition of TCs of individual ipaB* mutant cells by immunolabeling the surface of fixed bacteria. The specificity of the antibodies was verified by immunofluorescence (Fig. S4). As negative controls we used mxiH⁻, ipaB⁻ and ipaD⁻, which cannot form needles and/or TCs⁷. As expected ipaB⁻, ipaD⁻, ipaB⁻ ipaD⁻ and mxiH⁻ showed no/very reduced IpaB staining. Six out of seven ipaB* mutants showed normal TC composition. ipaBN264I showed a slightly higher average amount of IpaB at the bacterial surface (statistically significant at p = 0.05 but not at p = 0.02) although it did not show any change in the amount of IpaD (Fig. 2A). Thus, the number of needles and TCs it carries is same as in WT. Therefore, ipaBN264I could affect the accessibility of this mutant IpaB to the antibodies used for FACS, perhaps reflecting its altered conformation in TCs. The data above indicate all IpaB* mutants localize to TCs. Therefore, isolation of CR-insensitive ipaB point mutants suggests IpaB is directly involved in mediating CR responsiveness.

Some ipaB* mutants form translocons poorly

As IpaB’s membrane-insertion is necessary for epithelial cell invasion, we studied the effect of the ipaB* mutations on pore formation using contact hemolysis. Indeed, Shigella lyses Red Blood Cells (RBCs) upon physical contact with them⁶, due to membrane insertion of IpaB and IpaC, which form a pore within RBC membranes⁶. ipaB^−/+ and WT showed 85–80% of detergent-mediated hemoglobin release, which is set as 100% hemolysis in this assay (Fig. 3A). Some mutants had normal hemolytic capacity (ipaBK93N, ipaBN116Y, ipaBK150E, ipaBK188E), others showed only 60–30% of total hemolysis (ipaBN85I and ipaBQ108L), while ipaBN264I had none. Mutants with the strongest unresponsiveness to CR (ipaBK93N, ipaBN116Y) displayed hemolytic activities similar to WT. Thus, the ability to respond to CR is genetically dissociable from the ability to perform hemolysis.

Was the decrease in hemolytic activity of some ipaB* mutants due to a problem in membrane-insertion of mutant IpaB? For those mutants with reduced hemolytic activity, we examined the composition of the lysed RBC membranes isolated by floatation in a sucrose density gradient. We also studied ipaBN116Y as the mutant with the greatest reduction in CR induction. Since functional IpaB is a prerequisite for membrane insertion of IpaC⁶, no IpaB and little IpaC were detected in RBCs exposed to ipaB⁻ (Fig. 3B). In the membrane fractions of RBCs incubated with ipaBN85I and ipaBN116Y, the amount of IpaB was less (48% ± 16 and 58% ± 25 reduction relative to ipaB^−/+, respectively; Fig. S3B). For ipaBQ108L, the amount of IpaB detected was even less (75% ± 4 reduction). Mutant ipaBN264I showed little IpaB (94% ± 2 reduction) associated with RBC membranes. All mutants showed proportional reductions in IpaC insertion (Figs 3B and S3B). N264 is found in the amphipathic α-helix of IpaB (Fig. 1A), which is important for interaction with lipids vesicles⁵². Its polar side chain seems required for interaction with lipid bilayers.

There is little correlation between the ipaB* mutants’ abilities to sense CR and invade host cells

To evaluate ability of the mutants to invade epithelial cells, we measured protection from Gentamicin upon entry into HeLa cells, since this antibiotic cannot penetrate host cells. ipaBN85I and ipaBQ108L did not complement ipaB⁻ for cell invasion efficiently and ipaBN264I failed to restore invasion (Fig. 3C). Thus, there is fairly good correlation between the capacities of the mutants to perform hemolysis and invasion. In contrast, some showed strong defects in CR responsiveness but were unaffected in hemolysis and invasion (ipaBK93N and ipaBN116Y). This indicates that their capacities to sense CR and host cells are dissociable genetically.

Combinations of ipaB* mutations enhance CR unresponsiveness

To assess whether the IpaB alacoil folds in vivo as it does in the crystal structure, we combined mutations within amino acids nearby in the structure (Fig. 1H) to investigate whether their combination produces stronger phenotypes.

While all combinations of mutants formed had normal TC composition (Fig. 2A), ipaBN85I, K93N and ipaBK150E, K188E (termed ipaBxx from now on) showed slightly enhanced inability to respond to CR relative to each single mutant whereas for ipaBQ108L, N116Y the enhancement was greater (Fig. 3D,E). ipaBQ108L, N116Y, K150E (hereafter termed ipaBxxx) showed a slight, if reproducible, reduction in leakage and complete uninducibility. Since in ipaBxxx the altered amino acids are far apart, the stronger phenotypes are likely not due to the amino acids co-assessed being close in the structure. However, that some IpaB* mutations enhance others suggests they produce incremental, structurally-related effects.

Given the lack of correlation between CR-sensitivity and host cell sensing ability in ipaB* mutants, is there any correlation between these phenomena for IpaB? To answer this, we tested the invasive capacity of ipaBxx and ipaBxxx (Fig. S5). Unsurprisingly given that ipaBQ108L is non-invasive, ipaBxxx is also. More informatively, ipaBxx is also non-invasive, when both ipaBK150E and ipaBK188E are WT-like for invasion. Thus, several mutations in IpaB’s alacoil region, especially when combined, do affect host cell sensing. This suggests the alacoil is involved in transmission of both the CR and host-cell sensing signals. However, it seems less sensitive to the latter.

There is intramolecular crosstalk between IpaB regions

Could we alter the combined ipaB* mutant’s secretion phenotypes? In ipaBc-terΔ3, IpaB is expressed lacking its last three C-terminal amino acids, making the T3SS constitutively active and weakly inducible by CR²³. Hence, we examined if the secretion patterns of ipaBxxx are altered when expressed in an ipaBc-terΔ3 background. This combined mutant showed a new, intermediate phenotype: reduced constitutive secretion and reduced CR-induction relative to ipaBc-terΔ3 (Fig. 4A,B, left, C). This indicates epistasis between these sets of mutations, suggesting intramolecular crosstalk between these IpaB domains, where the alacoil region acts upstream of the C-terminus.

Dual modification of IpaD’s C-terminus leads to loss of needle tip binding

We previously isolated ipaD mutants with decreased CR responsiveness²¹ and others characterized ipaDΔ330-332, termed here ipaDc-terΔ3, as a constitutive secretor²⁹. To assess the effect of combinations of these mutations on IpaD, the other TC component, we combined ipaDN186I, K291I (ipaDxx²¹), an ipaD* mutant with strongly reduced secretion, in cis with ipaDc-terΔ3. Contrary to what happened in ipaBxxx_c-terΔ3, ipaDxx had no effect on ipaDc-terΔ3 (Fig. 4A,B, right, D): ipaDxx_c-terΔ3 behaved as a constitutive secretor. However, all mutants expressed similar Ipa/Ipg levels (Fig. 4B), ruling out deleterious decreases in protein expression.

To understand these contrasting results, we assessed the TC composition of these mutants by FACS (Fig. 2B). ipaBxxx, ipaBc-terΔ3 and ipaBxxx_c-terΔ3 have the same tip composition as ipaB^−/+. However, ipaDxx_c-terΔ3 displays a strong decrease in IpaD (and hence IpaB) when compared with ipaDxx and ipaDc-terΔ3. Thus, IpaDxx_c-terΔ3 can not bind the needle tip. This may be due to localization of N186 and K291 near or within the C-terminal helix of IpaD. As TC composition is wild type-like for ipaDxx and ipaDc-ter∆3 mutants, this also suggests they are affected in signaling from the needle tip and not a downstream step.

In trans combination of CR-insensitive ipaB or ipaD mutants and C-terminal deletions generates new phenotypes

To assess whether IpaB and IpaD communicate within the TC, we constructed a series of ipaB and ipaD mutants in trans, which we transformed into ipaB⁻ ipaD⁻. To reveal phenotypic changes, we combined mutants showing mild phenotypes, ipaBxx and ipaDK291E (ipaDx) with others displaying strongly impaired secretion, ipaBxxx and ipaDxx. We also combined these mutants with mutants exhibiting constitutive secretion (ipaBc-terΔ3, ipaDc-terΔ3).

To compare the overall phenotype of all mutants, we plated them on CR plates with and without IPTG, which they need for ipaD expression (Fig. S6). Without IPTG, all strains were red due to absence of TCs and constitutive secretion, verifying they all made T3SSs. With IPTG, ipaDwt ipaBwt was orange, as expected for wild-type. All combinations of ipaB* and ipaD* mutants were white, indicating intact, CR-insensitive TCs and suggesting synergy between these mutations. In addition, all combinations of ipaB* or ipaD* mutants with ipaDc-terΔ3 or ipaBc-terΔ3 mutants were orange to red, suggesting at best partial suppression of the constitutive secretion of ipaDc-terΔ3 or ipaBc-terΔ3.

We next verified expression IpaB, IpaC and IpaD in these strains was similar to WT (Fig. 5D). We also assessed the levels of IpgD and IpaH. Indeed, the more the T3SS secretes, the higher the expression of IpaH and, to some extent, also of IpgD. The increased levels of IpgD and IpaH expression confirmed that, all combination of ipaD* or ipaB* mutants with ipaBc-terΔ3 or ipaDc-terΔ3, respectively, were constitutive secretors.

No difference was observed in the TC composition of these mutants but one, ipaBxxx ipaDc-terΔ3 (Fig. 2C). Despite normal levels of IpaDc-terΔ3, it had lower levels of surface-localized IpaBxxx. Both ipaBc-terΔ3 ipaDwt and ipaBc-terΔ3 ipaDxx, show similar levels of IpaB but higher, if not significantly different, levels of IpaD.

Those “*” mutants that individually showed normal leakage, when combined with “*” or WT partners were confirmed to display the same phenotype and the same mutants in combination with constitutive secretors (ipaBc-terΔ3 & ipaDc-terΔ3) to display constitutive secretion. Thus, in ipaBc-terΔ3 ipaDxx, contrary to what happened in ipaBxxx_c-terΔ3, ipaDxx did not attenuate constitutive secretion of ipaBc-terΔ3, although it formed compositionally normal tips (Fig. 5A, lane 14 from left). When ipaBxxx was combined with ipaDc-terΔ3 (Fig. 5A, lane 13) a constitutive secretor phenotype was also observed, due to lack of IpaBxxx and presence of ipaDc-terΔ3 in TCs. These data indicate a C-terminal deletion in IpaB bypasses the repressing effects of ipaD* mutations on secretion activation. This suggests IpaB’s C-terminus acts downstream of the upper part of the IpaD C-terminal helix. Furthermore, by comparison with the partial suppression of constitutive secretion seen with IpaBxxx_cterΔ3, lack of suppression of ipaBc-terΔ3 by ipaDxx suggests the globular domain of each protein can signal independently, via its own C-terminus.

Combination of mutants showing mild reductions in inducible secretion, ipaBxx and ipaDx, resulted in decreased CR-inducibility (Fig. 5B,C, lanes 8 and 9), whilst the combination of ipaBxxx with ipaDx resulted in abrogation of the latter’s inducibility (Fig. 5C, lanes 10 and 11). The combination of ipaBxxx and ipaDxx in trans did not enhance their already strong phenotypes (Fig. 5C, lane 12). Thus, mild IpaB* and IpaD* mutations are synergistic. These data, together with our TC reconstruction where IpaD and IpaB are juxtaposed¹², support intermolecular communication between their globular regions during secretion activation. Finally, when the constitutive secretor ipaBc-terΔ3 was co-expressed with ipaDxx, another new phenotype was observed (Fig. 5B,C, lane 15): fast constitutive secretion as in ipaBc-terΔ3 (Fig. 5A) but as little CR-responsiveness as in ipaDxx (Fig. 5B,C). This suggests intermolecular crosstalk between the globular regions and the C-termini of these two molecules during CR sensing.

For ipaBxxx ipaDc-terΔ3, in spite of the constitutive secretion described above, including of IpaD, we observed a reduction in the level of IpaD secreted without or without CR as compared to ipaBwt ipaDc-terΔ3 (Fig. 5B,C, lane 13). This was verified by western blotting (Fig. S7A). This indicates that IpaBxxx uniquely affects fast secretion of IpaDc-terΔ3.

The activation signal travels from IpaB and IpaD to the needle

To examine how IpaB and IpaD interact with the needle component MxiH, we combined ipaBxxx, ipaDxx, ipaBc-terΔ3 and ipaDc-terΔ3 with slow constitutive secretor mutant mxiHQ51A¹⁷. This mutation is located in MxiH’s “head”, i.e. at the very top of the needle, where the C-termini of IpaD and IpaB probably interact MxiH¹². Using FACS, we titrated the IPTG concentration necessary to obtain wild-type levels of IpaB, IpaD and MxiH in the mxiHQ51A background (Fig. S8). Then we confirmed the expression of IpaB, IpaC and IpaD in these strains was similar to WT (Fig. 6). We also assessed the expression of IpgD and IpaH.

ipaB⁻ and ipaD⁻ showed more IpaD and MxiH surface staining, respectively, by FACS (Fig. 2D). As neither ipaB⁻ nor ipaD⁻ show longer needles⁶, we investigated whether these mutants upregulate the number of T3SSs and hence TCs they express using electron microscopy analysis of negatively stained, osmotically shocked cells, as previously established¹⁷. We visualized 0.8 ± 1.3 (n = 14), 2.5 ± 2.3 (n = 19) and 3.7 ± 2.7 (n = 27) T3SS basal bodies on the periphery of WT, ipaB⁻ and ipaD⁻ cells, respectively. Therefore, as demonstrated by a Student’s test after ANOVA, ipaB⁻ and ipaD⁻ possess similar number of basal bodies but both possess significantly more than wild-type (p = 0.0162 and 0.0004, respectively). This suggests that the fast constitutive secretion state leads to a 3–4 fold increase the number of T3SS basal bodies. In addition, mxiHQ51A showed a reduction in IpaB surface staining relative to wild-type. This was not previously observed¹²,¹³, but might contribute to increased leakage in this mutant. In a mxiHQ51A background, ipaBc-terΔ3 and ipaDc-terΔ3 showed a significant absence of IpaB, but not of IpaD (Fig. 2D). However, in a mxiHwt background these two mutants show normal TC composition (Fig. 2B). This indicates short C-terminal deletions in IpaDc-terΔ3 or IpaBc-terΔ3 adversely affect IpaB’s interaction with MxiH with a point mutation in its head. Finally, in a mxiHQ51A background, ipaBxxx showed slightly higher (significant at p = 0.05 but not at p = 0.02) and ipaDxx showed higher levels of IpaB surface staining, respectively, suggesting these mutations stabilize IpaB at the TC. These data indicate the IpaB C-terminus interacts with MxiHQ51 and this is affected by mutations in the alacoil and in the C-terminus of IpaD.

No expression of IpaH was detected in mxiHQ51A whole cell extract although this mutant is a slow constitutive secretor (Fig. 6C). This suggests the level of constitutive secretion in mxiHQ51A is not strong enough to activate ipaH transcription. Furthermore, ipaB⁻, ipaBc-terΔ3, ipaD⁻ and ipaDc-terΔ3 when each combined in trans with mxiHQ51A show similarly increased levels of IpgD and IpaH, confirming their fast constitutive secretion.

As expected from their tip composition, ipaBc-terΔ3 and ipaDc-terΔ3 displayed constitutive secretor phenotypes when combined with mxiHQ51A (Fig. 6A left and right panels). However, for ipaDc-terΔ3 this level was higher than in ipaDwt. In ipaDc-terΔ3 mxiHQ51A, IpaB is lost from TCs whereas both proteins are still present intracellularly, again suggesting the former is the cause of fast constitutive secretion. The differing levels of leakage (Fig. 6A,C, top row) displayed by mxiH⁻/mxiHQ51A versus mxiH⁻ ipaB⁻/mxiHQ51A ipaBwt indicate that in the latter, complementation by ipaB is only partial (Fig. 6A, left, lanes 4–6), perhaps because its expression level is lower than in the former (Fig. 6C, lanes 5–7). However, ipaBxxx suppresses the leakage seen in mxiH⁻ ipaB⁻/mxiHQ51A ipaBwt (Fig. 6A, left, lanes 7). Full complementation occurred in mxiH⁻ ipaD⁻/mxiHQ51A ipaDwt (Fig. 6A, right, lanes 4–6) and suppression of leakage by ipaDxx was also seen. That ipaBxxx and ipaDxx both stabilize IpaB in the TC in an mxiHQ51A background (Fig. 2D) may explain the leakage reduction. Both ipaBxxx and ipaDxx suppress mxiHQ51A more strongly under inducible conditions (Fig. 6B, both sides, lanes 9–12). This indicates that IpaD and IpaB signal to MxiH, initially via their globular domains.

Under non-induced conditions ipaBc-terΔ3 mxiHQ51A released IpaC and IpaD in normal amounts, but not IpaB (Fig. S7B), although IpaB is secreted in ipaBc-terΔ3²³. In addition ipaDxx mxiHQ51A could leak IpaD but not secrete it inducibly (compare Fig. 6A, right, lane 7 to 6B, right, lanes 10 & 12). Together with the data showing that ipaBxxx affects secretion of IpaDc-terΔ3, this suggests IpaB and IpaD can sense each other’s status in the TC and regulate aspects of their secretion.

All ipaB and ipaD mutations studied are epistatic over mxiHQ51A in that they change TC composition and/or affect aspects of secretion regulation. Given their location atop needles¹², this demonstrates both proteins regulate secretion upstream of the needle, i.e. from the TC. Our overall results, summarized in Fig. 7, now better mechanistic understanding of the TC’s intriguing functionalities.

IpaB is involved in sensing the secretion activation signal at the needle tip

All ipaB* mutants are partially defective in CR-sensing although they show normal TC composition. Furthermore, they are stably expressed, able to prevent premature secretion and do not have intrinsic secretion defects. IpaB∆2-20’s cytoplasmic sequestration provokes maximal constitutive secretion and uninducibility. There is presently no evidence that IpaB exerts any direct regulatory function inside the bacterium, although its secretion is signaled by its “empty” chaperone. This is supported by the phenotype of ipaBc-terΔ3, which has a normal N-terminus but is still a constitutive secreter because of its needle interaction defect. Short C-terminal truncations in IpaB may alter its conformation at needle tips, leading to constitutive secretion with some inducibility²¹. Thus, IpaB’s presence at the tip is essential for T3SS regulation and inducibility. Therefore, 1) our IpaB* mutants are acting from the TC and 2) TCs containing only pentameric IpaD are a secretion-deregulated assembly intermediate that exists prior to IpaB recruitment or due to the temporary loss of the hydrophobic IpaB from the needle tip¹².

Some combinations of ipaB* mutants display additive phenotypes. The triple mutant ipaBxxx abrogates inducibility although it is stably at the tip. The fast constitutive secretor and partially inducible phenotype of ipaBc-terΔ3 was reduced by in cis addition of ipaBxxx. These two sets of mutations are in different regions but, when localized in the same molecule, they alter each other’s effect. Observations of synergism and epistasis support the notion that the activation signal travels within IpaB.

IpaB’s involvement in events beyond host cell sensing

All previously published ipaB mutants carry substantial deletions, leading to pleiotrophic effects²³,²⁴,⁴¹. Here, we used an unbiased method to obtain single mutations, focusing on IpaB’s role in secretion regulation. Selected mutants were also used to analyze IpaB’s interaction with eukaryotic cells. ipaBN264I is strongly impaired in hemolysis and invasion because it is unable to insert IpaB and IpaC in RBC membranes. ipaBQ108L also displays substantial reductions in membrane-inserted IpaB and IpaC. In contrast, ipaBN116Y shows no reduction in hemolysis or invasion although it displays less membrane-inserted IpaB and IpaC. Why the latter two mutants are affected in these steps is not clear. ipaBN85I is mildly affected in CR-sensitivity, hemolysis and translocon insertion but to a greater extent in invasion. This mutant might thus be defective in events after translocon insertion, such as effector release¹⁷,⁵³.

Role of IpaB’s alacoil in signal transduction

All but one ipaB* mutation are located in its alacoil²². In this region, IpaB and its Salmonella SPI1 T3SS homolog SipB show structural similarity with the receptor-binding domain of E2, E3/E9 and Ia colicins²². Colicins are secreted by some strains of Escherichia coli and lethal for others⁵⁴. The colicin sequence similarity does not extend into the amphipathic helix and two transmembrane regions of the IpaB family, although the topologically equivalent portion of pore-forming colicins (such as Ia) has similar features. These hydrophobic features are shared too with the Bcl-2 family of apoptosis regulators⁵⁵ and with the membrane insertion domains of diphtheria toxin⁵⁶, which both also form membrane pores. Recently, the structure of central region of an IpaB homolog from Aeromonas, was solved in combination with its chaperone. AopB’s fold is very reminiscent of the hydrophobic region of pore-forming colicins and Bcl-2 proteins, with the hydrophilic and amphipathic helices wrapping around and shielding the hydrophobic ones⁵⁷. This suggests IpaB/SipB adapted mechanisms used by pore-forming toxins to insert into membranes.

The colicin E9 alacoil region opens up for it to insert into target membranes⁵⁸. Perhaps IpaB also “opens up” upon insertion of its hydrophobic regions into host membranes, thus initiating the signaling cascade (Fig. S9)? The clustered location of our “*” mutants within the alacoil and their synergistic effects suggest conformational changes in this region trigger secretion activation. However, if preventing such an alteration were how our new CR-insensitive mutations act, the IpaB* mutants should to be unable to sense host cells as well. More likely therefore, and in view of their synergism with our IpaD* mutants, the mutations render the IpaB-IpaD interface more resistant to disruption by CR, whilst the host cell is sensed differently. Given IpaB’s physically distal position in the TC, this may occur through IpaB alone, initially via an area other than its alacoil. However, the signal would be transduced through the alacoil and via IpaD since ipaBxx and ipaD* mutants are largely insensitive to host cells. ipaBN264I, in the amphipathic helix, impairs insertion of IpaB in host membranes and is also partially CR-insensitive. It therefore highlights a connection between changes in the amphipathic helix, which lies parallel to the membrane in liposome-inserted IpaB⁵²,⁵⁹, and signal transmission via the alacoil, where the other CR-insensitive mutations map.

Directionalities for signal transduction

Single amino acid substitutions in IpaB deregulate secretion activation. Combining deregulated mutants demonstrated intramolecular communication within IpaB (Figs 3D,F and 4) and intermolecular communication between TC components (Fig. 5). Moreover, ipaBxxx and ipaDxx both suppress mxiHQ51A constitutive secretion (Fig. 6). This indicates IpaB and IpaD act in the same pathway and via MxiH to regulate secretion activation. Considering TC morphology¹², we conclude IpaB is the major host cell sensor while IpaD works it with to transmit the activation signal down the needle.

How is the signal transmitted down the TC to the needle? Localization of our IpaB mutations to the amphipathic helix and alacoil suggests these regions act coordinately. We uncovered further crosstalk between different regions of IpaB by combining ipaB* alacoil mutations in cis with some in the IpaB C-terminus²³. IpaB-IpaD, IpaB-MxiH, IpaD-MxiH (this study and²⁸,²⁹) and MxiH-MxiH¹⁷ interactions are all important. IpaD-IpaD interactions are difficult to probe. Without an atomic resolution model for the TC and needle, the mechanism of this signal transduction event is unapproachable in further detail.

Bacterial strains and culture

Bacterial strains and plasmids used are listed in Tables 1 and 2. Shigella flexneri strains were maintained on Congo red (CR, 100 μg/ml; Serva) agar plates and grown at 37 °C in Trypticase Soy Broth (Becton Dickinson) supplemented with antibiotics when necessary (100 μg/ml ampicillin, 50 μg/ml kanamycin, 10–20 μg/ml chloramphenicol, 5 μg/ml tetracycline). IPTG and arabinose were used at the concentrations indicated in the figure legends.

Construction of ipaBΔ2-20

PCR site-directed mutagenesis using pDR1²³ as a template was carried out with Pfx Platinum polymerase (Invitrogen). ipaB was amplified using forward primers 1 and 2 (Table 3) to create ipaBwt and ipaBΔ2-20, respectively. Primer 5 was used as reverse primer. After digestion with HindIII and PstI, fragments were cloned into pUC19, thereby removing eight additional N-terminal amino acids introduced by earlier subcloning⁶⁰ that could have an undesirable effect on the signal peptide and giving rise to plasmids pIMC34 and pIMC35. This vector had the same levels of IpaB expression as pDR1, which also yields functional IpaB protein⁶⁰ and which was previously used to isolate the point mutants below.

To assess the level of IpaB necessary for complementation a new plasmid was constructed where ipaB was placed under the control of an arabinose inducible promoter, giving rise to pIMC30. We used pBAD_myc_His cut by NcoI and PstI to clone ipaB amplified by PCR with primers 4 and 18.

Construction of the ipaB mutant libraries

PCR mutagenesis was carried out with Taq DNA polymerase (New England Biolabs) using error prone reaction conditions (3 mM and 4 mM Mg²⁺), primers 3 and 4 and pDR1 as template. PCR fragments were purified, digested with HindIII and PstI and ligated into plasmid pDR1²³. Ligation mixtures were electroporated into E. coli DH5α carrying an empty pACT3 to produce LacI and repress otherwise toxic ipaB expression or into XL2 for the same reason. Transformations were incubated for 16h at 37 °C, plasmid DNA was then extracted and kept at −20 °C.

Screening of ipaB mutant libraries

To identify non-inducible mutants, DNA from each mutant library was electroporated into ipaB⁻ (red colony) and screened for white colonies on TCS agar plates containing 100 μg/ml CR. Putative ‘white’ colonies were selected, their plasmids isolated and retransformed into the ipaB⁻ mutant to ensure the white color was not due to loss-of-function mutations elsewhere within the T3SS-encoding operons. Candidate plasmids were sequenced to identify the mutation(s) responsible for the mutant phenotype (Table 2).

Combination of double and triple ipaB mutants

We constructed the new mutants using two-step PCR followed by fragment purification, digestion with HindIII and SpeI and subsequent cloning into digested pDR1. As mutations in pIMC40 (ipaBN85I, K93N) and pIMC41 (ipaBQ108L, N116Y) are close they were introduced together in the PCR primers. The first step of PCR consisted in two reactions, one reaction used primers 3 and 6 and second reaction primers 5 and 7 for pIMC40, where primers 6 and 7 overlap and contain the mutation, and pairs 3 and 9 and 5 and 8 for pIMC41, where primers 8 and 9 overlap and contain the mutation; pDR1 was used as template. The two PCR fragments were used as a template for the second step PCR, which was performed using primers 3 and 5 for both pIMC40 and pIMC41.

For pIMC42 (ipaBK150E, K188E), as both mutations are farther apart, mutation K188E was introduced by two-step PCR into pIMC4 already encoding K150E. The first step PCR was carried out with primer pair 3 and 11 and primer pair 5 and 10. The second step was as for pIMC40 and pIMC41, with primers 3 and 5. The triple mutant pIMC43 (ipaBQ108L, N116Y, K150E) was constructed as for pIMC41 but using pIMC4 as template.

ipaBxx, ipaBxxx, and ipaBwt were later introduced into pUC18_oc plasmid¹², which is insensitive to LacI, to combine them in trans with the IPTG inducible plasmid pACT3. The ipaB mutants were amplified by PCR using primers 12 and 13, using pDR1, pIMC42 and pIMC43 as templates. PCR fragments were then digested with NdeI and BamHI and cloned into pUC18_oc given rise to plasmids pIMC46, pIMC51 and pIMC57.

Construction of truncated ipaBΔ3 and ipaDΔ3 and derivatives

We cloned ipaBΔ3 into pUC18-oc by digesting pDR2 with SpeI and PstI and ligated into pIMC51 given rise to pIMC56. The SpeI-PstI digested fragment was also cloned into pIMC43 generating the new plasmid pIMC47 containing ipaBxxx in cis with ipaBΔ3.

We also constructed a series of ipaD mutants in pUC18-oc¹². We cloned the new ipaD constructs into pDR6-oc plasmid. We first reintroduced the native C322 into ipaD by digesting pipaD with EcoRV/PstI and ligating into pDR6-oc, creating pIMC61. We then constructed pIMC62 by ligating ipaDN186Y, K291I digested from pIMA237 with EcoRV and PstI into pDR6-oc. Then we made pIMC63 by deleting the last 3 aa of IpaD by PCR using pipaD as template and primers 14 and 15. PCR fragment was digested with EcoRV and BamHI and ligated into pDR6-oc. These primers were also used to obtained ipaDxxD3 by PCR, although in this case, pIMA237 was used as a template. The PCR fragment was digested and ligated into pDR6-oc given rise to pIMC64.

Combination of ipaB and ipaD mutants

To combine ipaB and ipaD in trans we constructed a series of ipaD mutants in pACT3. By PCR, using primers 16 and 17, we amplified ipaDx and ipaDxx from pIMA233 and pIMA237, respectively. PCR fragments were digested and ligated into pACT3 giving rise to pIMC58 and pIMC60. To amplify ipaDΔ3 from pIMAC49, we used primers 15 and 16. After digestion with SacI/BamHI and ligation into pACT3, pIMC59 was obtained. ipaDwt was amplified from pipaD by PCR using primers 16 and 17. The PCR fragment was digested and cloned into pACT3 giving rise to pIMA246.

All plasmids created were verified by sequencing (Eurofins).

Expression of mxiQ51A in trans with ipaB and ipaD mutants

A series of ipaB and ipaD mutants cloned into pUC18-oc were co-expressed in trans with mxiHQ51A, expressed from pACT3.

Analysis of protein synthesis and secretion

Characterization of the cellular interactions of ipaB mutants