The discovery of SycO highlights a new function for type III secretion effector chaperones (original) (raw)
Introduction
Type III secretion (T3S) allows pathogenic and symbiotic Gram‐negative bacteria to deliver bacterial effector proteins straight into the cytosol of an eukaryotic host cell (Cornelis and Wolf‐Watz, 1997). Translocation of the effectors is a very rapid process (Schlumberger et al, 2005) triggered by host cell contact (Pettersson et al, 1996) and achieved in one step by a sophisticated nanomachine called the injectisome or needle complex (Kubori et al, 1998; Mueller et al, 2005). The assembly and operation of the injectisome involves small (12–18 kDa) chaperones that remain in the bacterial cytosol (Wattiau and Cornelis, 1993; Menard et al, 1994; Feldman and Cornelis, 2003; Parsot et al, 2003; Ghosh, 2004). These chaperones represent a special family of chaperones found only in T3S systems. Some of these chaperones are involved in the assembly of the injectisome (class III) or the translocation pore (class II), whereas others are ancillary to effectors (class I). Class‐I chaperones are acidic (p_I_: 4–5), usually dimeric, proteins, which bind their cognate effector within their first 100 amino acids, just downstream of the short N‐terminal secretion signal. They are often, but not always, encoded next to the gene encoding their partner protein. They present a low sequence similarity, but their structure is quite well conserved. Several three‐dimensional structures of T3S chaperones have been solved during the last years, showing that they all adopt a similar fold consisting of five β‐strands and three α‐helices (Birtalan and Ghosh, 2001; Luo et al, 2001; Stebbins and Galan, 2001; Evdokimov et al, 2002; Trame and McKay, 2003; Phan et al, 2004; van Eerde et al, 2004; Buttner et al, 2005; Locher et al, 2005). Structures of chaperones in complex with their effector proteins have shown that the chaperone‐binding domain (CBD) of effectors wraps around the homodimers in an extended, horseshoe‐like conformation with some secondary structure organization in α‐helical structures (Stebbins and Galan, 2001; Birtalan et al, 2002; Phan et al, 2004). Chaperones are removed from effectors, before their translocation into host cells, by the ATPase that is part of the injectisome (Woestyn et al, 1994; Akeda and Galan, 2005). Although some chaperones of T3S effectors have been intensively studied and characterized, their essential function remains a matter of controversy. Some studies suggest that chaperones could be three‐dimensional targeting factors or even that they play a role in the setting of a secretion hierarchy (Boyd et al, 2000; Birtalan et al, 2002). Other observations suggest that they are required for the storage of effectors in the bacterial cytosol before secretion (Page et al, 2002) and even that they prevent folding during storage (Stebbins and Galan, 2001; Page et al, 2002).
The Ysc (for Yop Secretion) T3S system allows pathogenic Yersinia (Y. pestis, Y. pseudotuberculosis and Y. enterocolitica) to escape the innate immune defenses. It is encoded on a 70‐kb virulence plasmid, called pYV in Y. enterocolitica, and it injects six effector proteins, called Yops, into the cytosol of macrophages. YopH, YopE, YopT and YopO disturb the cytoskeleton dynamics and inhibit phagocytosis (Cornelis, 2002). Three of these Yops, namely YopE, YopH and YopT, have a dedicated chaperone called SycE (Wattiau and Cornelis, 1993; Birtalan and Ghosh, 2001), SycH (Wattiau et al, 1994; Phan et al, 2004) and SycT (Iriarte and Cornelis, 1998; Buttner et al, 2005; Locher et al, 2005), respectively. No chaperone has been described so far for YopO (called YpkA in Y. pestis and Y. pseudotuberculosis), a broad‐spectrum serine threonine kinase (Galyov et al, 1993) that becomes activated upon contact with actin (Juris et al, 2000). It is localized to the plasma membrane of the target cell (Hakansson et al, 1996; Dukuzumuremyi et al, 2000) and it interacts with Rho and Rac (Barz et al, 2000) but its physiological target is still unknown. The gene encoding YopO (YpkA) is part of an operon encoding also the effector YopP (Cornelis et al, 1987; Galyov et al, 1994; Iriarte and Cornelis, 1999) (YopJ in Y. pestis and Y. pseudotuberculosis), which interrupts pro‐inflammatory signalling cascades and leads macrophages to apoptosis (Mills et al, 1997; Zhou et al, 2005). The first open reading frame of the operon, orf155, encodes a protein that presents all the characteristics of a T3S chaperone (Iriarte and Cornelis, 1999), but its role could not be demonstrated so far (Trulzsch et al, 2003).
Here, we demonstrate that orf155 indeed encodes a chaperone for YopO and we called it SycO. We show that SycO prevents the intrabacterial aggregation of YopO by covering a membrane localization domain (MLD). As many T3S effectors, which have a chaperone, have also been shown to be membrane associated, these observations suggest that covering a MLD could be the primary function of the T3S effector chaperones.
Results
orf155 encodes a protein that has the characteristics of chaperones of effector proteins
In the three pathogenic Yersinia species, the operon encoding YopO (YpkA) and YopP (YopJ) contains a small open reading frame (orf) encoding a 142‐ (Y. pestis KIM) to 155‐ (Y. enterocolitica E40) residue acidic protein (Galyov et al, 1994; Iriarte and Cornelis, 1999) (Supplementary Figure S1). Upstream of residue 133, the three proteins are 98% identical. Programs searching sequence similarities and fold recognition revealed low but significant similarities between Orf155 and various T3S class‐I chaperones (Supplementary Figure S1). Based on this, a three‐dimensional structure, very similar to that of class‐I chaperones, could be derived (Supplementary Figure S1 and Supplementary data 1).
orf155 encodes SycO, the chaperone of YopO
The entire orf155 was deleted from the pYV plasmid of Y. enterocolitica E40(pYV40) giving E40(pML4001) (Δ_orf155_ mutant). Yop secretion was triggered by chelating Ca2+ ions (Cornelis et al, 1987). Under standard induction conditions (4 h), all the Yops were detected in the culture supernatant (not shown). However, when the time allowed for secretion was reduced, there was less YopO (YpkA) in the culture supernatant of mutant bacteria than in the culture supernatant of wild‐type (wt) bacteria (Figure 1A). No other Yop, including YopP, encoded downstream of orf155 and yopO was affected, indicating that the mutation was non‐polar (data not shown). As known for the other T3S chaperones, SycO itself was not secreted (data not shown).
Figure 1

The alternative text for this image may have been generated using AI.
The Δ_orf155_ mutant is impaired in secretion and translocation of YopO. (A) Plasmid pISO56 encoding YopOflag was introduced into wt E40(pYV40) bacteria and into Δ_orf155_ mutant bacteria E40(pML4001). Yop secretion was induced by incubating at 37°C in a medium deprived of Ca2+ ions. Samples of the supernatant were taken at different time points after shift of the culture from 28 to 37°C. YopOflag was detected by immunoblot. Top, wt; bottom, Δ_orf155_ mutant. The lower background band shows that an equivalent amount of supernatant has been loaded in each lane. (B) Translocation of the YopO‐Cya reporter into infected J774 macrophages. Plasmid pCD10 encoding YopO143‐Cya was introduced into E40(pYV40) wt bacteria, into Δ_orf155_ knockout bacteria E40(pML4001) and into the translocator‐deficient E40(pCNK4008) (called ΔHOPEMNB). The low‐copy plasmid pML4 encoding sycO complemented the Δ_orf155_ mutation, whereas the vector alone did not. Cellular levels of cAMP were assayed after infection.
We then monitored translocation of YopO into macrophages during the infection of a cell culture, using the adenylate cyclase (Cya) reporter method (Sory and Cornelis, 1994). We introduced plasmid pCD10, encoding the first 143 amino acids of YopO fused to Cya, into wt Y. enterocolitica E40 as well as into the Δ_orf155_ mutant and various control strains and we infected cultured J774 macrophages. Cells that were infected by wt bacteria accumulated cAMP. In contrast, cells infected with the Δ_orf155_ mutant did not produce more cAMP than cells infected with translocation‐deficient bacteria ΔHOPEMNB (Figure 1B). Complementation of Δ_orf155_ mutant bacteria with plasmid pML4 restored translocation of YopO up to the wt level (Figure 1B). These experiments showed that the product of orf155 is required for translocation of YopO and hence, given its similarity to Syc chaperones, it was tentatively called SycO.
To ascertain that translocation of YopP was not dependent on SycO, we monitored apoptosis of macrophages infected by wt Y. enterocolitica E40(pYV40) and by Δ_orf155 Y. enterocolitica_ E40(pML4001). No difference could be seen between the two infected cultures (data not shown), confirming that SycO is not the chaperone of YopP, as shown earlier (Trulzsch et al, 2003).
SycO binds to residues 20–77 of YopO
To confirm that SycO is the chaperone of YopO, we tested whether the two proteins interact and how. The sycO‐yopO genes were coexpressed in Escherichia coli BL21 and SycO was co‐purified with YopOhis (data not shown) suggesting that the two proteins form a complex when synthesized together in E. coli. To ascertain this observation, a sample of co‐purified proteins was loaded on an analytical gel filtration column. YopO and SycO eluted together around fraction 9 (Figure 2A), whereas purified SycO loaded alone eluted around fraction 14 (Figure 2B). To localize the CBD of YopO, three different YopO‐Cya hybrid proteins were produced in the multi‐effector knockout Y. enterocolitica ΔHOPEMT together with GST‐SycO. YopO143‐Cya and YopO77‐Cya co‐purified with SycO on glutathione Sepharose, whereas YopE20‐Cya did not (not shown). These experiments led to the conclusion that the CBD of SycO is localized within the 77 N‐terminal residues of YopO. As SycE binds to residues 15–75 of YopE (Woestyn et al, 1996; Birtalan et al, 2002) and SycH binds to residues 20–70 of YopH (Woestyn et al, 1996), we made the assumption that SycO could also bind immediately after the secretion signal of YopO. To test this hypothesis, we constructed a GST‐YopO hybrid, a variant where residues 20–77 of YopO were deleted (GST‐YopOΔ22–77) and a hybrid containing only residues 20–77 of YopO (GST‐YopO20–77). All three hybrid proteins were expressed in parallel in Y. enterocolitica E40. SycO was pulled down together with GST‐YopO and with GST‐YopO20–77 but not with GST‐YopOΔ20–77 (Figure 2C). This indicated that residues 20–77 of YopO include the core of the CBD.
Figure 2

The alternative text for this image may have been generated using AI.
Characterization of the YopO‐SycO interaction. (A) YopOhis and SycO encoded by pML9 were synthesized in E. coli BL21 and co‐purified on Ni2+ Sepharose. The purified complex (L) was loaded onto a Superdex™ 200 gel filtration column and the eluted fractions (4–15) were analyzed by Coomassie‐stained SDS‐PAGE. YopOhis and SycO eluted together around fraction 9. (B) Purified SycO protein (see Supplementary data 4), loaded on the same column, eluted around fraction 14. (C) Plasmids pML10, pML11 and pML17 encoding GST‐YopO, GST‐YopOΔ20–77 and GST‐YopO20–77, respectively, were introduced into Y. enterocolitica E40 wt producing SycO from the pYV40 plasmid. GST hybrid proteins were pulled down from crude extracts, analyzed by SDS‐PAGE and the presence of SycO was monitored by immunoblotting. L, lysate; U, unbound; W, wash; E, eluted.
We then investigated the stoichiometry of the YopO‐SycO complex and observed that SycO dimerizes and binds YopO as a dimer (Supplementary data 2), as known for the other class‐I chaperones.
We finally compared the CBD of YopO to that of YopE in silico and found that they share a similar secondary structure, meaning that the YopO CBD is likely to wrap around the dimer of SycO as the YopE CBD wraps around the dimer of SycE (Supplementary data 3).
Binding of SycO to YopO does not prevent folding of the whole protein
It has been proposed that class‐I chaperones could keep the effectors in an unfolded secretion competent state (Stebbins and Galan, 2001). To test this hypothesis for YopO, we monitored the autophosphorylating activity of the purified kinase (Galyov et al, 1993), with and without the CBD. As YopO was poorly soluble without its chaperone but was more soluble as a GST fusion or without the CBD, we compared the kinase activity of GST‐YopO to that of YopOΔ20–77 and YopO‐(SycO)2. As a negative control, we engineered an inactive GST‐YopO hybrid with the D267A substitution in the catalytic site (Juris et al, 2000). As expected, the kinase activity was dependent on the presence of actin (Juris et al, 2000) (Figure 3). Importantly, YopOΔ20–77 was at least as active as GST‐YopO (Figure 3), showing that the CBD is not needed for the catalytic activity of YopO. In addition, the YopO‐(SycO)2 complex turned out to be as active as GST‐YopO, showing that SycO binding does not prevent folding of the whole protein. Interestingly, SycO itself was phosphorylated. This has no physiological relevance, as chaperones are not translocated into host cells (Wattiau et al, 1994), but it confirms the broad spectrum activity of YopO (Juris et al, 2000).
Figure 3

The alternative text for this image may have been generated using AI.
Binding of SycO to YopO does not prevent the catalytic activity of YopO. GST‐YopO, GST‐YopOD267A, YopO(SycO)2 and YopOΔ20–77, expressed in E. coli BL21 from plasmids pML10, pML12, pML10+pML15 and pML11 respectively, were semipurified (Supplementary data 4), incubated with [γ‐32P]ATP in the presence and absence of actin and analyzed by Coomassie‐stained SDS‐PAGE (A) and autoradiography (B).
YopO_Δ_CBD can be exported by the Ysc T3S system and translocated into host cells
To further investigate the role of SycO, we removed the CBD within YopO and first tested the consequences in terms of secretion triggered by Ca2+ chelation. We compared secretion of YopOflag and YopOΔ20–77flag after 3 h of incubation at 37°C in the presence and absence of SycO. As expected, the absence of SycO had a negative effect on secretion of YopO but not on secretion of YopOΔ20–77. Unexpectedly, there was more YopOΔ20–77flag than YopOflag in the supernatant and in the lysate of wt bacteria (Figure 4A). However, it has been shown in Salmonella that the chaperone prevents the effectors SptP and SopE to be secreted by the flagellum rather than by the injectisome (Lee and Galan, 2004). To ensure that YopOΔ20–77flag was indeed secreted by the injectisome, we monitored secretion by Y. enterocolitica W227(pSW2276) mutated in yscN, the gene encoding the ATPase energizing the T3S pathway. As shown in Figure 4A, the protein was not secreted, ruling out secretion by the flagellum. These results show that, in the absence of SycO, YopO is exported better without its CBD. This suggests that the CBD itself creates the need for SycO.
Figure 4

The alternative text for this image may have been generated using AI.
YopOΔ20–77 can be secreted and delivered into host cells. (A) Plasmid pISO56 encoding YopOflag or pML16 encoding YopOΔ20–77flag was introduced into wt Y. enterocolitica E40(pYV40), Δ_sycO Y. enterocolitica_ E40(pML4001) and yscN Y. enterocolitica E40(pSW2276). Yop secretion was triggered by Ca2+ chelation for 3 h and the supernatants were analyzed by Western blotting. (B) Translocation of the YopO143Δ20–77‐Cya reporter into infected J774 macrophages. Plasmid pML31 encoding YopO143Δ20–77‐Cya (ΔCBD) was introduced into the multi‐effector knockout mutant E40(pIML421) (called ΔHOPEMT) and into the translocator‐deficient E40(pCNK4008) (called ΔHOPEMNB). Plasmid pCD10 encodes YopO143‐Cya (wt). Cellular cAMP levels were assayed after infection.
As a protein that is secreted is not necessarily delivered into cells (Figure 1), we removed the CBD from the YopO143‐Cya reporter protein and monitored translocation. Again, the observations on in vitro secretion and translocation were somewhat different. YopO143Δ20–77‐Cya was efficiently delivered into cells by ΔHOPEMT bacteria (Figure 4B) but not by wt bacteria (not shown). Thus, for translocation, we can conclude that the presence of the CBD and the chaperone facilitates the delivery of the effector but targeting can occur in their absence.
SycO prevents the intrabacterial aggregation of YopO mediated by the CBD
In order to understand why the presence of the CBD reduces the amount of YopO secreted in vitro, we compared the solubility of YopO and YopOΔ20–77 in E. coli. In the absence of SycO, YopOΔ20–77his was abundant and mostly soluble, whereas YopOhis was present in very small amounts and exclusively in the insoluble fraction (Figure 5A). In the presence of SycO, there was more YopOhis than in the absence of SycO and most of it was soluble. This suggested that YopO is less soluble than YopOΔ20–77 and that SycO helps solubilizing YopO.
Figure 5

The alternative text for this image may have been generated using AI.
The CBD causes aggregation of YopO. (A) YopOhis, YopOΔ20–77his and YopOhis+SycO were overexpressed in E. coli BL21 from plasmids pML7, pML8 and pML9, respectively. The cleared lysates (see Materials and methods) were centrifuged at 100 000 g to separate soluble (Sn) from insoluble (P) proteins and analyzed by Western blotting. (B) GST‐YopO, GST‐YopOΔ20–77 and GST‐YopO(SycO)2 were synthesized in E. coli and purified as described in Supplementary data 4. After removal of the GST tag, the proteins were analyzed on a Superdex™ 200 gel exclusion chromatography. Purified SycO (see Supplementary data 4) was analyzed in parallel.
To confirm this observation, we purified YopO and YopOΔ20–77 and the YopO‐(SycO)2 complex using GST tags that were cleaved off at the last stage. Again, the yield of YopO was much lower than the yield of the two others (not shown). The three preparations as well as SycO alone were analyzed by size‐exclusion chromatography (Figure 5B). Whereas SycO, YopO‐(SycO)2 and YopOΔ20–77 were eluted as expected from their size, YopO appeared in the void volume, indicating that it was aggregated. Thus, the CBD creates the need for SycO because it drastically reduces the intrabacterial solubility of YopO.
The SycO‐binding domain is a cell membrane targeting domain
The previous experiments demonstrated that SycO is required because of the presence of residues 20–77 in YopO and that these residues are not involved in the catalytic activity. As YopO is targeted to the membrane of the host cell (Hakansson et al, 1996; Dukuzumuremyi et al, 2000), we tested whether the residues 20–77 would not act as an MLD. To investigate this possibility, the 3′ end of different yopO constructs was fused to the EGFP gene and HEK293T cells were transfected with the recombinant plasmids. Whereas EGFP was distributed in the cytosol and the nucleus, YopO‐EGFP was preferentially localized at the plasma membrane, as expected. Furthermore, cells transfected with plasmid pML1, encoding YopO‐EGFP, became rounded (Figure 6), indicating that the YopO‐EGFP hybrid protein had kept an activity in the cell. A YopOD267A‐EGFP catalytic mutant also localized to the membrane. This showed that the catalytic activity is not required for membrane localization. In contrast, YopOΔ20–77‐EGFP did not localize to membranes and cells transfected with pML2, encoding YopOΔ20–77‐EGFP, did not round up. We concluded that the CBD was required for the membrane targeting and that this membrane localization was necessary for YopO‐EGFP activity (Figure 6). To delineate more accurately the MLD, we fused residues 20–80 and 20–90 of YopO to EGFP and monitored the localization of the hybrid. Residues 20–90, but not 20–80, were sufficient to target EGFP to the membrane. There is thus a good correlation between the CBD (20–77) and the MLD (20–90).
Figure 6

The alternative text for this image may have been generated using AI.
The CBD of YopO is an MLD. The different constructs drawn on the right panel were transfected in HEK293T cells. Cells were grown for 24 h, stained with Texas Red‐conjugated wheat germ agglutinin, fixed and analyzed by confocal microscopy. The constructs were encoded by pML1 (YopO‐EGFP), pML2 (YopOΔ20–77‐EGFP), pML3 (YopOD267A‐EGFP), pML26 (YopO20–80‐EGFP) and pML27 (YopO20–90‐EGFP).
The CBDs of YopE and YopT are also aggregation‐prone MLDs
The experiments described above demonstrate that SycO is required to mask the aggregation‐prone MLD of YopO inside the bacterium. We wondered whether this hypothesis could apply to other class‐I chaperones.
Like the CBD of YopO, residues 50–77 of YopE create the need for SycE (Boyd et al, 2000). Even more, residues 54–75 have been shown to represent an MLD for YopE (Krall et al, 2004). As shown in Figure 7A, YopE90‐EGFP expressed in HEK293T cells showed a peri‐nuclear localization including the _cis_‐Golgi. This does not mean that when YopE is injected by Yersinia, it would be associated to the Golgi, but it confirms that the CBD confers some membrane targeting properties to YopE, as shown earlier (Krall et al, 2004). It is thus likely that the CBD causes intrabacterial insolubility. We thus tested the intrabacterial solubility of YopE in the presence and absence of SycE. As shown in Figure 7D, SycE clearly increased the solubility of YopE. This result is in perfect agreement with previous data from Birtalan et al (2002) showing that binding of SycE rescues purified YopE from aggregation.
Figure 7

The alternative text for this image may have been generated using AI.
Role of the CBD of YopE and YopT. The fusions YopE90‐EGFP (A) and YopT124‐EGFP (B, C) were expressed in HEK293T cells after transfection of plasmids pSAM10 and pSAM12, respectively. Cells were grown 24 h, fixed, permeabilized and immunolabelled with α‐Giantin (A) or α‐LAMP1 (C). Pictures were acquired by confocal microscopy. (D) YopEhis and YopThis encoded by pML33 and pML34 were expressed in E. coli BL21 with or without SycE or SycT encoded by plasmids pSAM13 or pSAM15, respectively. The cleared lysates (see Materials and methods) were centrifuged at 100 000 g to separate soluble (Sn) from insoluble (P) proteins and analyzed by Western blotting.
As YopT, like YopE, targets Rho GTPases (Zumbihl et al, 1999; Black and Bliska, 2000; Von Pawel‐Rammingen et al, 2000; Shao et al, 2002) and also associates with host cell membranes (Aepfelbacher et al, 2003), we would predict that the CBD of YopT is also an MLD. The CDB of YopT has been localized, by limited proteolysis, between residues 52 and 139 (Buttner et al, 2005). In order to test whether this CBD would also correspond to the MLD, we fused residues 1–124 of YopT to EGFP and expressed the hybrid in HEK293T cells. YopT124‐EGFP presented a punctuated distribution evoking either lysosomal association or aggregation (Figure 7B and C). Using LAMP1 as a lysosomal marker, we observed that YopT124‐EGFP did not colocalize with lysosomes, suggesting that YopT124 was causing the aggregation of EGFP. We then tested whether the SycT chaperone would not influence the intrabacterial solubility of YopT. As shown in Figure 7D, this was indeed the case.
We fused the first 90 amino acids of YopH to EGFP, but no specific distribution was observed. EGFP was homogeneously distributed in the whole cell (data not shown). In addition, there was no clear difference in the solubility of YopH in the presence or absence of SycH.
Finally, we analyzed the cellular localization of YopP, for which no chaperone was described. The hybrid YopP‐EGFP transfected into HEK293T cells appeared uniformly distributed to the cytosol (not shown).
In conclusion, we suggest that SycO, SycE and SycT cover an aggregation‐prone MLD.
Discussion
In Yersinia, the chaperones of effector proteins are SycE (Wattiau and Cornelis, 1993; Birtalan and Ghosh, 2001), SycH (Wattiau et al, 1994; Phan et al, 2004) and SycT (Iriarte and Cornelis, 1998; Buttner et al, 2005; Locher et al, 2005). SycO described in this paper shares all the properties of these Syc proteins and clearly represents a new member of this group. However, its phenotype is somewhat more subtle and this explains why it is discovered more than 10 years later than the others. At first sight, SycO does not appear to be required for in vitro secretion of YopO but it is required for translocation of YopO into cells. If chaperones are necessary for export of effector molecules that are stored in the cytoplasm, but not for export of those synthesized while the secretion apparatus is active, as demonstrated by Parsot and co‐workers (Page et al, 2002) for Spa15, then our results indicate that translocation of YopO into cells is essentially post‐translational. In contrast, in vitro secretion of YopO, induced by Ca2+ chelation, could be also co‐translational.
Having shown that SycO is a new member of the group of Yersinia effector chaperones, we tried to address its function, taking into consideration the different functions already proposed for the proteins of this group. We observed that binding of SycO to YopO does not prevent the activity of the kinase domain of YopO, implying that the catalytic domain (residues 160–400) and the actin‐binding domain (residues 709–729) (Juris et al, 2000) are folded. Thus, SycO does not maintain YopO in a completely unfolded secretion‐competent state. This observation is in perfect agreement with the observation of Birtalan et al (2002) that SycE does not prevent the GAP activity of YopE and with the observation of Akeda and Galan (2005) that SicP does not prevent the tyrosine phosphatase activity of SptP, revising an earlier interpretation of Stebbins and Galan (2001). For SycO, we thus rule out that its function is to maintain YopO in an unfolded state, which also implies that the injectisome can unfold its substrates during export, as was clearly demonstrated for SptP by Akeda and Galan (2005).
However, we observed that SycO keeps YopO soluble in the cytosol of Yersinia before export and that it is precisely the SycO‐binding domain that makes YopO insoluble. Thus, the presence of the CBD creates the need for the chaperone. As pointed out by Birtalan et al (2002), it is unlikely that chaperones exist simply to mask an aggregation‐prone effector region whose only function is chaperone binding. We thus looked for a function of this aggregation‐prone domain and we found that it was an MLD. A function of SycO is thus to hide the aggregation‐prone MLD of YopO while YopO is in the bacterium.
Would this new function apply to other class‐I chaperones? Previous data from Krall et al (2004) and data presented here show that it also applies to YopE. It probably also applies to YopT because the CBD causes aggregation of transfected hybrid proteins in the host cell. However, it should still be demonstrated that this domain targets YopT to the membrane when YopT is injected rather than transfected. We also tried to extend the hypothesis to YopH, which targets focal adhesions (Persson et al, 1999) and another membrane‐associated complex (Black et al, 2000). We did not observe any membrane localization of transfected YopH90‐EGFP and any change in the intrabacterial solubility of YopH with or without SycH. However, for YopH, the situation could be more complex as the CBD already partially overlaps with the phospho‐tyrosine recognition domain (Montagna et al, 2001). Moreover, different domains of YopH including its CBD have been shown to interact with the Fyn binding protein in macrophages (Yuan et al, 2005). Thus, the CBD of YopH might not be a membrane targeting domain, it is nevertheless an intracellular targeting domain. In good agreement with our hypothesis about class‐I chaperones, YopM (Benabdillah et al, 2004) and YopP, which do not have a chaperone, do not localize to membranes. In Yersinia, there is thus a correlation between the presence of a chaperone and an activity of the effector at the membrane of the target cell.
Would this new hypothesis about effector chaperones and MLDs also apply to effectors of other T3S systems? The hypothesis predicts that the effectors that have a chaperone would act at the membrane of the host cell. Among the best‐characterized class‐I chaperones, we find SicP (Fu and Galan, 1998) and SigE (Darwin et al, 2001) from Salmonella enterica, CesT from enteropathogenic E. coli (Elliott et al, 1999) and Spa15 from Shigella (Page et al, 2002). SptP, the partner of SicP, is a GAP for Rho (Fu and Galan, 1999) and has been shown to localize at the plasma membrane of infected cells (Cain et al, 2004); SigD/SopB, the partner of SigE, is an inositol phosphatase (Norris et al, 1998), which was shown to be membrane‐associated (Marcus et al, 2002; Cain et al, 2004); Tir, the partner of CesT, is the membrane‐associated intimin receptor (Kenny et al, 1997); and finally, Spa15 is, among others, the chaperone of IpaA, which contributes to Shigella entry by binding the focal adhesion protein vinculin (Tran Van Nhieu et al, 2000). Thus, all these effectors have a chaperone of the SycE family, and are targeted to the membrane. Furthermore, the Salmonella effectors SopE, SopE2 and SipA, which all share the InvB chaperone (Bronstein et al, 2000; Ehrbar et al, 2003), have also been shown to localize at the membrane of infected cells (Cain et al, 2004).
Does this new function for class‐I chaperones exclude any targeting function? Certainly not, as the CBD and SycO are required for efficient delivery of YopO into cells by wt Yersinia bacteria. This situation is reminiscent of YopE. Indeed, YopE deprived of its CBD cannot be delivered by wt bacteria but it can be delivered by mutant bacteria deprived of all the other effectors (Boyd et al, 2000). Chaperones bound to the effector thus improve targeting to the secretion apparatus, either by acting as a three‐dimensional secretion signal, as suggested for SycE (Birtalan et al, 2002), or simply by presenting the secretion signal. The presence of the chaperone could also favor the interaction with the ATPase, and so facilitate the unfolding of the effector (Akeda and Galan, 2005). Our data do not allow to discriminate between these hypotheses.
We thus suggest that the class‐I chaperones have two functions: (i) cover an aggregation‐prone MLD and (ii) facilitate export. Which one of these two functions would have appeared first? As some effectors seem not to have a chaperone (YopM and YopP in Yersinia), we would speculate that the primary function of effector chaperones is to cover an MLD and that targeting of effector chaperone complexes has been gradually optimized by evolution. In support of this view, YopO and YopE deprived of their CBD can be translocated by multi‐effector knockout mutant bacteria. CBDs and chaperones thus evolved to serve two different essential functions, one inside the bacterium and one inside the host cell.
Materials and methods
Bacterial strains, plasmids and genetic constructions
Y. enterocolitica E40 (Sory and Cornelis, 1994) and W22703 (Cornelis et al, 1986), both from serotype 0:9, were used for T3S experiments. E. coli BL21 was used for protein expression and E. coli Top10 was used for plasmid amplification and cloning. Plasmids are listed in Table I. Oligonucleotides labelled with * in Table I were used to delete or mutate domains by inverse polymerase chain reaction, using the Pfu turbo polymerase (Stratagene). Every construct was sequenced using 3100‐Avant genetic analyzer (ABI Prism). Oligonucleotides are listed in Supplementary Table S1.
Table I List of the plasmids used in this work
Induction of type III secretion by low Ca2+
This was carried out as described earlier (Agrain et al, 2005).
Standard protein purification and analysis protocols
See Supplementary data 3.
Ultracentrifugation of crude extracts
YopOhis, YopOΔ20–77his and SycO coexpressed with YopOhis were produced in E. coli BL21 from the pET22 vector using plasmids pML7, pML8 and pML9. Bacteria were grown overnight at 37°C, diluted to an OD600 of 0.1 in LB with ampicillin 100 μg ml−1 and incubated at 37°C until they reached an OD600 of 0.6. Then, the expression of the proteins was induced by adding 0.125 mM IPTG and the cultures were incubated at room temperature (RT) with shaking for 3 h. Cells were harvested by centrifugation, resuspended in PBS containing Triton X‐100 0.1% (v/v), protease inhibitors (Complete mini, Roche) and lysed by using a French press. The lysate was spun for 30 min at 6000 g to eliminate unbroken cells and debris. The supernatant was centrifuged for 1 h 30 min at 100 000 g to separate soluble compounds (supernatant) from insoluble ones (pellet).
Kinase assay
Proteins were purified as described in Supplementary data 4. YopO and YopOD267A were eluted together with GST in order to improve their solubility. The purified YopO‐(SycO)2 complex used for the kinase assay was previously run on gel filtration in order to use only proteins associated in a complex. Kinase assays were performed using 5 μg of purified proteins in the kinase reaction buffer: PBS, 20 mM MgCl2, 1 mM DTT, 0.5 mM ATP and 2 μCi [γ‐32P]ATP, ±0.1 μg of purified G‐actin (Pierce). The reaction was incubated at 30°C for 30 min and stopped by addition of SDS‐PAGE loading buffer. The kinase reactions were loaded on SDS‐PAGE gel. The autoradiography was obtained on Phosphor Screen (Molecular Dynamics).
Adenylate cyclase reporter translocation assay
J774 macrophages were grown in RPMI 1640 medium (Gibco BRL), supplemented with 2 mM L‐glutamine (Invitrogen), 1 mM sodium pyruvate (Invitrogen), 0.0004% (v/v) β‐mercaptoethanol (Sigma) and 10% (v/v) fetal bovine serum (Invitrogen). Adenylate cyclase reporter translocation assays were carried out as described earlier (Sory and Cornelis, 1994). All experiments were performed in triplicate.
Transfection and fluorescence microscopy
HEK293T cells were grown in Dulbecco's modified Eagle's medium (Gibco BRL), supplemented with 10% (v/v) fetal bovine serum (Invitrogen) and 1 mM sodium pyruvate (Invitrogen). Coverslips (Huber & Co) were coated with poly‐lysine (BD Diagnostic System), and 104 cells well−1 were seeded in 500 μl of media. The next day, cells were transfected following the calcium phosphate procedure. Cells were grown for 24 h, stained with 10 μg ml−1 wheat germ agglutinin Texas Red®‐X conjugate (Molecular Probes) for 7 min at RT and fixed for 20 min in 3% paraformaldehyde in PBS. For Giantin and LAMP1 staining, cells were permeabilized with 0.1% saponin in PBS containing 3% BSA, incubated during 40 min with α‐Giantin (G1/133) or α‐LAMP1 (G1/139) (Axxora), washed and incubated for 30 min with a Texas Red®‐coupled secondary antibody (Molecular Probes). Slides were examined with a Leica TCS SP confocal microscope. Pictures were processed with the Leica confocal software version 2.5.
References
- Aepfelbacher M, Trasak C, Wilharm G, Wiedemann A, Trulzsch K, Krauss K, Gierschik P, Heesemann J (2003) Characterization of YopT effects on Rho GTPases in _Yersinia enterocolitica_‐infected cells. J Biol Chem 278: 33217–33223
Google Scholar - Agrain C, Sorg I, Paroz C, Cornelis GR (2005) Secretion of YscP from Yersinia enterocolitica is essential to control the length of the injectisome needle but not to change the type III secretion substrate specificity. Mol Microbiol 57: 1415–1427
Google Scholar - Akeda Y, Galan JE (2005) Chaperone release and unfolding of substrates in type III secretion. Nature 437: 911–915
Google Scholar - Barz C, Abahji TN, Trulzsch K, Heesemann J (2000) The Yersinia Ser/Thr protein kinase YpkA/YopO directly interacts with the small GTPases RhoA and Rac‐1. FEBS Lett 482: 139–143
Google Scholar - Benabdillah R, Mota LJ, Lutzelschwab S, Demoinet E, Cornelis GR (2004) Identification of a nuclear targeting signal in YopM from Yersinia spp. Microb Pathog 36: 247–261
Google Scholar - Birtalan S, Ghosh P (2001) Structure of the Yersinia type III secretory system chaperone SycE. Nat Struct Biol 8: 974–978
Google Scholar - Birtalan SC, Phillips RM, Ghosh P (2002) Three‐dimensional secretion signals in chaperone‐effector complexes of bacterial pathogens. Mol Cell 9: 971–980
Google Scholar - Black DS, Bliska JB (2000) The RhoGAP activity of the Yersinia pseudotuberculosis cytotoxin YopE is required for antiphagocytic function and virulence. Mol Microbiol 37: 515–527
Google Scholar - Black DS, Marie‐Cardine A, Schraven B, Bliska JB (2000) The Yersinia tyrosine phosphatase YopH targets a novel adhesion‐ regulated signalling complex in macrophages. Cell Microbiol 2: 401–414
Google Scholar - Boyd AP, Lambermont I, Cornelis GR (2000) Competition between the Yops of Yersinia enterocolitica for delivery into eukaryotic cells: role of the SycE chaperone binding domain of YopE. J Bacteriol 182: 4811–4821
Google Scholar - Bronstein PA, Miao EA, Miller SI (2000) InvB is a type III secretion chaperone specific for SspA. J Bacteriol 182: 6638–6644
Google Scholar - Buttner CR, Cornelis GR, Heinz DW, Niemann HH (2005) Crystal structure of Yersinia enterocolitica type III secretion chaperone SycT. Protein Sci 14: 1993–2002
Google Scholar - Cain RJ, Hayward RD, Koronakis V (2004) The target cell plasma membrane is a critical interface for Salmonella cell entry effector‐host interplay. Mol Microbiol 54: 887–904
Google Scholar - Cornelis G, Sory MP, Laroche Y, Derclaye I (1986) Genetic analysis of the plasmid region controlling virulence in Yersinia enterocolitica 0:9 by Mini‐Mu insertions and lac gene fusions. Microb Pathog 1: 349–359
Google Scholar - Cornelis G, Vanootegem JC, Sluiters C (1987) Transcription of the yop regulon from Y. enterocolitica requires trans acting pYV and chromosomal genes. Microb Pathog 2: 367–379
Google Scholar - Cornelis GR (2002) The Yersinia Ysc‐Yop ‘type III’ weaponry. Nat Rev Mol Cell Biol 3: 742–752
Google Scholar - Cornelis GR, Wolf‐Watz H (1997) The Yersinia Yop virulon: a bacterial system for subverting eukaryotic cells. Mol Microbiol 23: 861–867
Google Scholar - Darwin KH, Robinson LS, Miller VL (2001) SigE is a chaperone for the Salmonella enterica serovar Typhimurium invasion protein SigD. J Bacteriol 183: 1452–1454
Google Scholar - Dukuzumuremyi JM, Rosqvist R, Hallberg B, Akerstrom B, Wolf‐Watz H, Schesser K (2000) The Yersinia protein kinase A is a host factor inducible RhoA/Rac‐ binding virulence factor. J Biol Chem 275: 35281–35290
Google Scholar - Ehrbar K, Friebel A, Miller SI, Hardt WD (2003) Role of the Salmonella pathogenicity island 1 (SPI‐1) protein InvB in type III secretion of SopE and SopE2, two Salmonella effector proteins encoded outside of SPI‐1. J Bacteriol 185: 6950–6967
Google Scholar - Elliott SJ, Hutcheson SW, Dubois MS, Mellies JL, Wainwright LA, Batchelor M, Frankel G, Knutton S, Kaper JB (1999) Identification of CesT, a chaperone for the type III secretion of Tir in enteropathogenic Escherichia coli. Mol Microbiol 33: 1176–1189
Google Scholar - Evdokimov AG, Tropea JE, Routzahn KM, Waugh DS (2002) Three‐dimensional structure of the type III secretion chaperone SycE from Yersinia pestis. Acta Crystallogr D 58: 398–406
Google Scholar - Feldman MF, Cornelis GR (2003) The multitalented type III chaperones: all you can do with 15 kDa. FEMS Microbiol Lett 219: 151–158
Google Scholar - Fu Y, Galan JE (1998) Identification of a specific chaperone for SptP, a substrate of the centisome 63 type III secretion system of Salmonella typhimurium. J Bacteriol 180: 3393–3399
Google Scholar - Fu Y, Galan JE (1999) A Salmonella protein antagonizes Rac‐1 and Cdc42 to mediate host‐cell recovery after bacterial invasion. Nature 401: 293–297
Google Scholar - Galyov EE, Hakansson S, Forsberg A, Wolf‐Watz H (1993) A secreted protein kinase of Yersinia pseudotuberculosis is an indispensable virulence determinant. Nature 361: 730–732
Google Scholar - Galyov EE, Hakansson S, Wolf‐Watz H (1994) Characterization of the operon encoding the YpkA Ser/Thr protein kinase and the YopJ protein of Yersinia pseudotuberculosis. J Bacteriol 176: 4543–4548
Google Scholar - Ghosh P (2004) Process of protein transport by the type III secretion system. Microbiol Mol Biol Rev 68: 771–795
Google Scholar - Hakansson S, Galyov EE, Rosqvist R, Wolf‐Watz H (1996) The Yersinia YpkA Ser/Thr kinase is translocated and subsequently targeted to the inner surface of the HeLa cell plasma membrane. Mol Microbiol 20: 593–603
Google Scholar - Iriarte M, Cornelis GR (1998) YopT, a new Yersinia Yop effector protein, affects the cytoskeleton of host cells. Mol Microbiol 29: 915–929
Google Scholar - Iriarte M, Cornelis GR (1999) The 70‐kilobase virulence plasmid of Yersiniae. In Pathogenicity Islands and Other Mobile Virulence Elements, Kaper JB, Hacker J (eds) pp 91–126. Washington, DC: American Society for Microbiology
- Juris SJ, Rudolph AE, Huddler D, Orth K, Dixon JE (2000) A distinctive role for the Yersinia protein kinase: actin binding, kinase activation, and cytoskeleton disruption. Proc Natl Acad Sci USA 97: 9431–9436
Google Scholar - Kaniga K, Delor I, Cornelis GR (1991) A wide‐host‐range suicide vector for improving reverse genetics in gram‐negative bacteria: inactivation of the blaA gene of Yersinia enterocolitica. Gene 109: 137–141
Google Scholar - Kovach ME, Elzer PH, Hill DS, Robertson GT, Farris MA, Roop II RM, Peterson KM (1995) Four new derivatives of the broad‐host‐range cloning vector pBBR1MCS, carrying different antibiotic‐resistance cassettes. Gene 166: 175–176
Google Scholar - Kenny B, DeVinney R, Stein M, Reinscheid DJ, Frey EA, Finlay BB (1997) Enteropathogenic E. coli (EPEC) transfers its receptor for intimate adherence into mammalian cells. Cell 91: 511–520
Google Scholar - Krall R, Zhang Y, Barbieri JT (2004) Intracellular membrane localization of pseudomonas ExoS and Yersinia YopE in mammalian cells. J Biol Chem 279: 2747–2753
Google Scholar - Kubori T, Matsushima Y, Nakamura D, Uralil J, Lara‐Tejero M, Sukhan A, Galan JE, Aizawa SI (1998) Supramolecular structure of the Salmonella typhimurium type III protein secretion system. Science 280: 602–605
Google Scholar - Lee SH, Galan JE (2004) Salmonella type III secretion‐associated chaperones confer secretion‐pathway specificity. Mol Microbiol 51: 483–495
Google Scholar - Locher M, Lehnert B, Krauss K, Heesemann J, Groll M, Wilharm G (2005) Crystal structure of the Yersinia enterocolitica Type III secretion chaperone SycT. J Biol Chem 280: 31149–31155
Google Scholar - Luo Y, Bertero MG, Frey EA, Pfuetzner RA, Wenk MR, Creagh L, Marcus SL, Lim D, Sicheri F, Kay C, Haynes C, Finlay BB, Strynadka NC (2001) Structural and biochemical characterization of the type III secretion chaperones CesT and SigE. Nat Struct Biol 8: 1031–1036
Google Scholar - Marcus SL, Knodler LA, Finlay BB (2002) Salmonella enterica serovar Typhimurium effector SigD/SopB is membrane‐associated and ubiquitinated inside host cells. Cell Microbiol 4: 435–446
Google Scholar - Menard R, Sansonetti P, Parsot C, Vasselon T (1994) Extracellular association and cytoplasmic partitioning of the IpaB and IpaC invasins of S. flexneri. Cell 79: 515–525
Google Scholar - Michiels T, Cornelis GR (1991) Secretion of hybrid proteins by the Yersinia Yop export system. J Bacteriol 173: 1677–1685
Google Scholar - Mills SD, Boland A, Sory MP, van der Smissen P, Kerbourch C, Finlay BB, Cornelis GR (1997) Yersinia enterocolitica induces apoptosis in macrophages by a process requiring functional type III secretion and translocation mechanisms and involving YopP, presumably acting as an effector protein. Proc Natl Acad Sci USA 94: 12638–12643
Google Scholar - Montagna LG, Ivanov MI, Bliska JB (2001) Identification of residues in the N‐terminal domain of the Yersinia tyrosine phosphatase that are critical for substrate recognition. J Biol Chem 276: 5005–5011
Google Scholar - Mueller CA, Broz P, Muller SA, Ringler P, Erne‐Brand F, Sorg I, Kuhn M, Engel A, Cornelis GR (2005) The V‐antigen of Yersinia forms a distinct structure at the tip of injectisome needles. Science 310: 674–676
Google Scholar - Neyt C, Cornelis GR (1999) Insertion of a Yop translocation pore into the macrophage plasma membrane by Y_ersinia enterocolitica_: requirement for translocators YopB and YopD, but not LcrG. Mol Microbiol 33: 971–981
Google Scholar - Norris FA, Wilson MP, Wallis TS, Galyov EE, Majerus PW (1998) SopB, a protein required for virulence of Salmonella dublin, is an inositol phosphate phosphatase. Proc Natl Acad Sci USA 95: 14057–14059
Google Scholar - Page AL, Sansonetti P, Parsot C (2002) Spa15 of Shigella flexneri, a third type of chaperone in the type III secretion pathway. Mol Microbiol 43: 1533–1542
Google Scholar - Parsot C, Hamiaux C, Page AL (2003) The various and varying roles of specific chaperones in type III secretion systems. Curr Opin Microbiol 6: 7–14
Google Scholar - Persson C, Nordfelth R, Andersson K, Forsberg A, Wolf‐Watz H, Fallman M (1999) Localization of the Yersinia PTPase to focal complexes is an important virulence mechanism. Mol Microbiol 33: 828–838
Google Scholar - Pettersson J, Nordfelth R, Dubinina E, Bergman T, Gustafsson M, Magnusson KE, Wolf‐Watz H (1996) Modulation of virulence factor expression by pathogen target cell contact. Science 273: 1231–1233
Google Scholar - Phan J, Tropea JE, Waugh DS (2004) Structure of the Yersinia pestis type III secretion chaperone SycH in complex with a stable fragment of YscM2. Acta Crystallogr D 60: 1591–1599
Google Scholar - Sarker MR, Neyt C, Stainier I, Cornelis GR (1998) The Yersinia Yop virulon: LcrV is required for extrusion of the translocators YopB and YopD. J Bacteriol 180: 1207–1214
Google Scholar - Schlumberger MC, Muller AJ, Ehrbar K, Winnen B, Duss I, Stecher B, Hardt WD (2005) Real‐time imaging of type III secretion: Salmonella SipA injection into host cells. Proc Natl Acad Sci USA 102: 12548–12553
Google Scholar - Shao F, Merritt PM, Bao Z, Innes RW, Dixon JE (2002) A yersinia effector and a pseudomonas avirulence protein define a family of cysteine proteases functioning in bacterial pathogenesis. Cell 109: 575–588
Google Scholar - Sory MP, Boland A, Lambermont I, Cornelis GR (1995) Identification of the YopE and YopH domains required for secretion and internalization into the cytosol of macrophages, using the cyaA gene fusion approach. Proc Natl Acad Sci USA 92: 11998–12002
Google Scholar - Sory MP, Cornelis GR (1994) Translocation of a hybrid YopE‐adenylate cyclase from Yersinia enterocolitica into HeLa cells. Mol Microbiol 14: 583–594
Google Scholar - Stebbins CE, Galan JE (2001) Maintenance of an unfolded polypeptide by a cognate chaperone in bacterial type III secretion. Nature 414: 77–81
Google Scholar - Trame CB, McKay DB (2003) Structure of the Yersinia enterocolitica molecular‐chaperone protein SycE. Acta Crystallogr D 59: 389–392
Google Scholar - Tran Van Nhieu G, Bourdet‐Sicard R, Dumenil G, Blocker A, Sansonetti PJ (2000) Bacterial signals and cell responses during Shigella entry into epithelial cells. Cell Microbiol 2: 187–193
Google Scholar - Trulzsch K, Roggenkamp A, Aepfelbacher M, Wilharm G, Ruckdeschel K, Heesemann J (2003) Analysis of chaperone‐dependent Yop secretion/translocation and effector function using a mini‐virulence plasmid of Yersinia enterocolitica. Int J Med Microbiol 293: 167–177
Google Scholar - van Eerde A, Hamiaux C, Perez J, Parsot C, Dijkstra BW (2004) Structure of Spa15, a type III secretion chaperone from Shigella flexneri with broad specificity. EMBO Rep 5: 477–483
Google Scholar - Von Pawel‐Rammingen U, Telepnev MV, Schmidt G, Aktories K, Wolf‐Watz H, Rosqvist R (2000) GAP activity of the Yersinia YopE cytotoxin specifically targets the Rho pathway: a mechanism for disruption of actin microfilament structure. Mol Microbiol 36: 737–748
Google Scholar - Wattiau P, Bernier B, Deslee P, Michiels T, Cornelis GR (1994) Individual chaperones required for Yop secretion by Yersinia. Proc Natl Acad Sci USA 91: 10493–10497
Google Scholar - Wattiau P, Cornelis GR (1993) SycE, a chaperone‐like protein of Yersinia enterocolitica involved in the secretion of YopE. Mol Microbiol 8: 123–131
Google Scholar - Woestyn S, Allaoui A, Wattiau P, Cornelis GR (1994) YscN, the putative energizer of the Yersinia Yop secretion machinery. J Bacteriol 176: 1561–1569
Google Scholar - Woestyn S, Sory MP, Boland A, Lequenne O, Cornelis GR (1996) The cytosolic SycE and SycH chaperones of Yersinia protect the region of YopE and YopH involved in translocation across eukaryotic cell membranes. Mol Microbiol 20: 1261–1271
Google Scholar - Zhou H, Monack DM, Kayagaki N, Wertz I, Yin J, Wolf B, Dixit VM (2005) Yersinia virulence factor YopJ acts as a deubiquitinase to inhibit NF‐kappaB activation. J Exp Med 202: 1327–1332
Google Scholar - Zumbihl R, Aepfelbacher M, Andor A, Jacobi CA, Ruckdeschel K, Rouot B, Heesemann J (1999) The cytotoxin YopT of Yersinia enterocolitica induces modification and cellular redistribution of the small GTP‐binding protein RhoA. J Biol Chem 274: 29289–29293
Google Scholar