Direct Injection of Functional Single-Domain Antibodies from E. coli into Human Cells

Intracellular proteins have a great potential as targets for therapeutic antibodies (Abs) but the plasma membrane prevents access to these antigens. Ab fragments and IgGs are selected and engineered in E. coli and this microorganism may be also an ideal vector for their intracellular delivery. In this work we demonstrate that single-domain Ab (sdAbs) can be engineered to be injected into human cells by E. coli bacteria carrying molecular syringes assembled by a type III protein secretion system (T3SS). The injected sdAbs accumulate in the cytoplasm of HeLa cells at levels ca. 105–106 molecules per cell and their functionality is shown by the isolation of sdAb-antigen complexes. Injection of sdAbs does not require bacterial invasion or the transfer of genetic material. These results are proof-of-principle for the capacity of E. coli bacteria to directly deliver intracellular sdAbs (intrabodies) into human cells for analytical and therapeutic purposes.


Introduction
The ability to express antibody (Ab) fragments in Escherichia coli has an enormous biotechnological and therapeutic potential [1]. The smallest Ab fragments (,12-15 kDa) are the so-called singledomain antibodies (sdAbs), which are composed of a single variable (V) immunoglobulin (Ig) domain [2,3]. The sdAbs are generated by engineering conventional Igs (e.g. human or murine) [4] or obtained from natural heavy-chain-only Igs expressed by certain animals like camelids [5]. The sdAbs from camelid heavychain-only Igs are known as V HH domains or Nanobodies. Importantly, the absence of a paired V domain in V HH s does not hinder their affinity for their cognate antigens, which is in the same range of conventional Abs with paired V H /V L domains (K D ,10 28 -10 210 M). Targets for therapeutic Abs are extracellular including cytokines, matrix proteins, and extracellular domains of membrane receptors [6]. Intracellular proteins (e.g. components of cell signaling cascades) are excellent therapeutic targets but plasma membrane prevents the access of Abs to them. Nonetheless, Ab fragments against different antigens have been expressed intracellularly (intrabodies) as inhibitors of proteins involved, for instance, in carcinogenesis and viral replication [7,8]. Intrabody expression requires transfer of the encoding gene into the cell, either using transfection with naked DNA, liposomes, or infection with recombinant viral vectors, which raises concerns given its possible integration into the host cell genome. Therefore direct transfer of antibody polypeptides into target cells constitute an attractive alternative. Since E. coli is employed for selection, engineering and production of IgGs and Ab fragments [9,10] this microorganism is an excellent candidate for delivery of intrabodies. Preferably, the delivery system should avoid the use of invasive E. coli strains that release their cell content after lysis in the phagosome [11]. Interestingly, intestinal pathogenic E. coli strains, such as the enteropathogenic E. coli (EPEC) O127:H6 [12] and enterohaemorragic E. coli (EHEC) O157:H7 [13], remain extracellular while using a type III protein secretion system (T3SS) to inject specific bacterial proteins, referred to as ''effectors'', into mammalian cells [14,15].
EPEC and EHEC adhere to enterocytes in the gastrointestinal tract while inducing characteristic ''attachment and effacement'' (A/E) lesions [16]. A chromosomal pathogenicity island of 35-40 kb, called the locus of enterocyte effacement (LEE) [17], encodes the proteins responsible of A/E lesion formation, including the outer membrane adhesin intimin (eae), the T3SS, and six effectors [18]. T3SSs are supramolecular protein assemblies embedded in the bacterial envelope (called the needle complex or injectisome). They are composed of a cytosolic ATPase, inner and outer membrane rings, a periplasmic shaft and an extracellular needle [19,20]. Protein translocation is dependent on insertion of hydrophilic translocation pore into the plasma membrane of the eukaryotic cell (the translocon) [19]. In EPEC and EHEC the needle complex is extended up to 700 nm by long and flexible EspA filament of polymerized EspA and the translocon is composed of translocated proteins EspB and EspD [21].
Effectors and translocators, the substrates of T3SSs, contain a non-cleavable N-terminal translocation signal usually comprising the first ,15-30 residues [20,22]. N-terminal fusions of effectors with viral antigens and certain enzymes have been secreted through T3SSs for the generation of live vaccines [23,24] or as translocation reporters [25,26]. The aim of this study was to determine whether non-invasive E. coli bacteria carrying a T3SS can be used to translocate Ab fragments into human cells.

Secretion of functional sdAbs into E. coli culture media
The N-terminal 20 amino acids of the effector EspF, which are fully conserved in EPEC strain E2389/69 and EHEC strain EDL933stx (Table 1), were selected to drive the T3 secretion of the sdAb fragments. We chose V HH s as sdAb fragments due to their favorable biophysical properties and ability to function as potent enzyme inhibitors [5,27]. Two characterized V HH s, named Vamy and Vgfp, recognizing amylase (Amy) and the green fluorescent protein (GFP) respectively, were employed as models [28,29]. We used the IPTG-inducible bacterial expresion vector pSA10 ( Table 2) to express EspF 20 T3 signal (T3s) fused to Vamy (T3sVamy) or Vgfp (T3sVgfp) (Fig. S1). The V HH s were tagged with His and E-tag epitopes at their C-termini to allow metalaffinity purification and detection with monoclonal antibodies (mAbs). Although T3SS are cell-contact dependent, secretion can be triggered in vitro under by growing bacteria under certain growth conditions [20]. In EPEC this is achieved by growth in DMEM 5% CO 2 at 37uC. Thus, we used this growth conditions to analyze whether T3sV HH s are secreted by the T3SS of EPEC (Fig. 1). EPEC wild-type (wt) strain and DescN strain (Table 1), which lacks the T3SS ATPase EscN [30], transformed with pSA10 (empty vector), pT3sVamy, or pT3sVgfp, were grown in DMEM and induced with IPTG for 3 h. Induction of T3sV HH fusions did not affect the growth of EPEC strains, which reached the same final optical density (OD 600 nm ,1.2) as cultures with the empty vector. Coomassie staining of proteins secreted from wt EPEC revealed discrete bands corresponding to the T3SS substrates (e.g. EspA, EspB, EspD) and protein bands of ca. 21-23 kDa, corresponding to the expected size of T3sV HH s, when bacteria carried pT3sVamy or pT3sVgfp ( Fig. 1B; lanes 1-3). No secreted T3sV HH s were seen from wt EPEC carrying the empty vector. Neither protein bands corresponding to the T3SS substrates nor the T3sV HH s were present in the supernatant of DescN strain ( Fig. 1B; lanes 4-6). A band corresponding to the autotransporter EspC [31], which is secreted by the Sec-pathway, was present in the supernatants of both the wt and DescN strains.
We next evaluated expression and secretion of the T3sV HH s by Western blot (WB). The anti-E-tag mAb detected the T3sV HH s (ca. 21-23 kDa) in the culture supernatants of wt EPEC but not in the DescN strain containing pT3sVamy and pT3sVgfp ( Fig. 1C; top panel; lanes 1-6). In contrast, T3sV HH s were detected in the bacterial lysates of both the wt and DescN strains with these plasmids ( Fig. 1C; lanes 7-12). The membranes were also probed with anti-GroEL mAb ( Fig. 1C; bottom panels) to control for non-specific release of cytoplasmic EPEC proteins. The result showed no reactivity with anti-GroEL in culture supernatants (lanes 1-6) whereas strong signals were detected in bacterial lysates (lanes 7-12). Secretion of the T3sV HH s to culture supernatants was also observed in EHEC wt strain but not in the EHEC DescN strain ( Fig. S2; Table 1). As an additional proof of the T3-dependent secretion, a derivative of pT3sVgfp was constructed in which the T3s was deleted (pDsVgfp). WB of induced wt EPEC and EHEC carrying pDsVgfp resulted in intracellular accumulation of DsVgfp but not in its secretion (Fig. S3).
To determine whether secreted T3sV HH s remained as soluble proteins, and to rule out their aggregation or association with outer membrane vesicles (OMVs), the culture supernatants from induced EPEC were ultracentrifuged at 100,000 g. The presence of T3sV HH s in the soluble and pellet fractions was analyzed by WB ( Fig. S4) showing that $95% of T3sV HH s are found as soluble proteins. Taking advantage of the His-tag present at their C-termini, both T3sV HH s were purified ( Fig. S5) with standard yields between 0.5-1 mg/L of culture supernatant. Binding activities of the purified T3sV HH s were tested by ELISA using antigens Amy, GFP or BSA (as negative control). Bound T3sV HH s were developed with anti-Etag mAb demonstrating the specific antigen-binding activity of the T3sV HH s over a range of concentrations (Fig. 1D).

Translocation of sdAbs into human cells
We tested whether T3sV HH s can be injected into human cells with EPEC strain (Fig. 2A). Toward this end, vector pCX340 (Table 2), which encodes the TEM-1 b-lactamase (Bla) reporter devoid of its natural N-terminal Sec-dependent signal peptide [26], was used to generate T3s-Bla, T3sVamy-Bla, and T3sVgfp-Bla fusions (Fig. S6A). Plasmids pT3s-Bla and pCX340 were used as positive and negative controls of translocation, respectively. Cultured HeLa cells, infected with wt and DescN EPEC carrying these Bla-expressing plasmids, were incubated with the nonfluorescent esterified CCF2/AM substrate. Upon passive entry into the eukaryotic cell, CCF2/AM is transformed by eukaryotic esterases to the fluorescent substrate of Bla CCF2, which is mostly contained inside the eukaryotic cell [26]. Injection of Bla into the cytoplasm of the eukaryotic cells leads to hydrolysis of CCF2 changing its fluorescence emission from 520 nm (green) to 447 nm (blue), which could be detected under the fluorescence microscope or quantified in a fluorimeter. Examination by fluorescence microscopy revealed a clear shift to blue fluorescence of HeLa cells infected with wt EPEC carrying pT3s-Bla, pT3sVamy-Bla, or pT3sVgfp-Bla (Fig. 2B). In contrast, green fluorescence was observed in HeLa cells infected either with wt EPEC/pCX340 or with any of the DescN strains (Fig. 2B), demonstrating that the hydrolysis of CCF2 is only due to the Bla translocated into the cytosol of the HeLa cells. The fluorescence intensity of infected cells was also quantified in a fluorimeter and is shown as the ratio between blue emission fluorescence (447 nm) and green emission fluorescence (520 nm) (Fig. 2C). Expression of Bla fusions in wt and DescN EPEC was confirmed by WB with anti-Bla antibodies (Fig. S6B). Thus, these results demonstrate that EPEC carrying a functional T3SS are able to inject T3sV HH s-Bla fusions into human cells.
Next, we investigated whether T3sV HH s could be detected in the cytoplasm of infected HeLa cells. Toward this end we employed a fractionation method of the infected cells based on Streptolysin-O (SLO), a pore-forming cytolysin from Streptococcus pyogenes that selectively bind to cholesterol groups in the eukaryotic plasma membrane [32,33]. HeLa cells were infected for 90 or 150 min with EPEC expressing T3sV HH s in the presence of IPTG, placed on ice and the monolayer washed with PBS. Following SLO treatment the cytoplasmic content of HeLa cells was collected (''Eukaryotic cytoplasm'' extract). Ghost HeLa cells and bound bacteria were lysed in a SDS-containing buffer (''SLOinsoluble'' extract). Protein extracts were subjected to WB revealing a time-dependent accumulation of T3sVgfp in the cytoplasm of HeLa cells infected with wt EPEC but not with EPEC DescN (T3sV HH top panel, Fig. 2D). No sign of bacterial contamination was detected in the eukaryotic cytoplasmic extracts using a mAb directed against DnaK (Fig. 2D). Efficiency of the SLO-treatment in the different samples was controlled by WB developed with anti-GFP mAb (Fig. 2D). In addition, WB of the ''SLO-insoluble'' extracts demonstrated similar expression levels of T3sVgfp in wt EPEC and DescN strains (T3sV HH bottom panel, Fig. 2D). The comparable signal from the ''SLO-insoluble'' extracts probed with anti-DnaK indicated similar cell attachment levels of wt EPEC and DescN (Fig. 2D). This was confirmed by fluorescence microscopy of the infected cultures ( Fig. 2E), in which microcolonies of wt EPEC and DescN bacteria (red) were stained adhered to HeLa cells at similar levels, although actin pedestals are only observed in cells infected by wt EPEC (Fig. 2E, staining of Factin in green).
Since wt EPEC also secretes T3sVgfp to the extracellular medium, we wanted to exclude the possibility that the T3sVgfp molecules detected in the cytoplasmic extracts entered the cells indirectly from the extracellular medium through the pores formed in the plasma membrane by the EspBD translocation pore or during the SLO-treatment (despite removal of the extracellular medium before addition of SLO). For this, HeLa cells were infected with wt EPEC harboring empty pSA10 (expressing a functional T3SS but not T3sVgfp) or EPEC DescN/pT3sVgfp (lacking a functional T3SS and thus unable to inject T3sVgfp). After 120 min of infection, the culture media was replaced by conditional, bacteria-free, medium obtained from an induced wt EPEC/pT3sVgfp culture (containing secreted T3sVgfp). After a further 30 min incubation, infected cells were fractionated with SLO into ''eukaryotic cytoplasm'' and ''SLO-insoluble'' extracts, as above. WB showed that T3sVgfp was not found in the ''eukaryotic cytoplasm'' (Fig. S7), hence ruling out the possibility that T3sVgfp molecules were entering HeLa cells from the extracellular medium. Taken together, the experiments with Bla reporter and biochemical fractions demonstrated that wt EPEC are able to directly inject T3sV HH s into human cells.
We also investigated whether the presence of their cognate antigen influence the stability of T3sV HH s. Toward this end HeLa cells, transfected with pEGFP-N1 or with the empty vector pCS2+MT (Table 2), were infected with wt EPEC and EPEC DescN harboring pT3sVgfp. ''Eukaryotic cytoplasm'' and ''SLOinsoluble'' extracts were analyzed by WB revealing that the amount of T3sVgfp found in the cytoplasm of HeLa cells was ,3fold higher when GFP was expressed (Fig. 3A, T3sV HH panel). Detection of the cytoplasmic enzyme glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as a loading control of the ''eukaryotic cytoplasm'' extracts ( Fig. 3A, GAPDH panel). WB with anti-DnaK demonstrated the absence of bacterial contami-nation in the ''eukaryotic cytoplasm'' extracts and similar signals in the ''SLO-insoluble'' extracts ( Fig. 3A, DnaK panels). Therefore, higher levels of injected T3sV HH s are found when HeLa cells express the relevant antigen. We estimated that after 150 min infection in the presence of IPTG, an average of ,7610 5 molecules of T3sVgfp per cell are found in pEGFPN1-transfected cells. This estimation was done by densitometry of the WB signals of T3sVgfp in ''eukaryotic cytoplasm'' extracts from three independent infections with the signals generated with known protein concentrations of purified T3sVgfp used as standard curve (Fig. S8).

Antigen binding activity of injected sdAbs
To investigate formation of intracellular antigen-V HH complexes, ''eukaryotic cytoplasm'' extracts from pEGFP-N1-transfected cells, infected with wt EPEC carrying pT3sVgfp or pT3sVamy, were immunoprecipitated (IP) with anti-E-tag mAb bound to protein G-beads. WB of the input extracts and IP proteins (Fig. 3B) revealed that GFP was specifically co-IP from the cytoplasmic extracts of cells infected with EPEC/pT3sVgfp demonstrating the formation of intracellular GFP-T3sVgfp complexes. Since T3sVamy does not accumulate at detectable levels in the cytoplasm of HeLa cells (likely caused by the absence of its antigen in the cytoplasm; see Discussion), purified T3sVamy was added to the input extract of cells infected with EPEC/pT3sVamy to provide a control of the specific co-IP of GFP with T3sVgfp (Fig. 3B). Similar amounts of GFP and GAPDH proteins were detected in the input extracts (Fig. 3B). These data provide direct evidence that the injected T3sV HH s have the capacity to bind their specific antigen in the cytoplasm of human cells.

Translocation of sdAbs by attenuated EPEC
The EPEC wt strain is a pathogen causing strong cytopathic effects due to the injection of its natural repertoire of T3 effectors [14,18]. Therefore most biotechnological applications will require using attenuated strains deficient in the major or all of the T3 effectors. To obtain a proof-of-principle of the injection of V HH s with attenuated bacteria, we employed quadruple (''quad'') mutant strain of EPEC that assembles functional injectisomes but carries four deletions in genes encoding the adhesin intimin and the effectors Tir, EspF and Map [34]. Bla translocation assays in HeLa cells infected with the ''quad'' mutant or with EPEC DescN (as a control) demonstrated the injection of T3s-Bla, T3sVamy-Bla and T3sVgfp-Bla by the ''quad'' mutant strain (Fig. 4A). The emission ratio at 447/550 nm obtained with the ''quad'' mutant were ,60% those reached with wt EPEC (compare Fig. 2C and Fig.4A) suggesting a lower infection efficacy with the attenuated mutant. T3sVgfp was also detected in the cytoplasmic extracts of HeLa cells transfected with pEGFPN1 and infected with ''quad'' mutant'' (Fig. 4B). Infection with wt EPEC and DescN bacteria, both carrying pT3sVgfp, were used as positive and negative controls, respectively. The level of translocated T3sVgfp by the ''quad mutant'' strain was found to be ,60% that reached by the wt EPEC strain (Fig 4B, T3sV HH top panel). Absence of bacterial contamination in the cytoplasmic extracts was confirmed with anti-DnaK and detection of GAPDH and GFP was used as a control of gel loading and of the efficacy of the SLOtreatment. ''SLO-insoluble'' extracts showed similar levels of T3sVgfp and DnaK in infections with EPEC wt and ''quad'' strains (Fig. 4B). Fluorescence microscopy of infected cells showed microcolonies of the ''quad'' strain on the surface of HeLa cells and the absence of F-actin pedestals (Fig. 4C). Altogether these data demonstrate an attenuated E. coli strain with a functional T3SS is capable of translocation of T3sV HH s into HeLa cells.

Discussion
Expression of Ab fragments in the cytoplasm of human cells allows targeting of intracellular proteins that participate in disease and infection processes [7]. In this work we have shown that noninvasive E. coli bacteria carrying a functional T3SS are able to secrete and translocate to the cytoplasm of human cells sdAb fragments with full capacity to bind their cognate antigens. Singledomain V HH s appeared specially suited for this application given their potential as enzyme inhibitors, monomeric nature, stability, and size (2-3 nm diameter) [5] that fit in the protein channels of T3 needles and EspA filament [19,35]. We have shown that T3secreted V HH s present in the extracellular medium do not enter in the cytoplasm of HeLa cells via EspBD pores, and that translocation only occurs directly from bacteria to the mammalian cell. In addition, we obtained evidence of the formation of antigen-V HH complexes in the cytoplasm of the infected cells and quantified that ,7610 5 V HH molecules accumulated per cell in the presence of antigen. Interestingly, we found that expression of the antigen in the cytoplasm of the mammalian cell increases the level of V HH s in the cytoplasm, which suggests that formation of antigen-V HH complex could stabilize the intracellular V HH . Although the actual reason of this stabilization is unclear, it is possible that the higher molecular weight of the antigen-V HH complex might reduce its susceptibility to proteolyic degradation in vivo. In this regard, detection of translocated T3sV HH -Bla in cells that do not express their cognate antigens could also be explained by their higher molecular weight of these fusions.
To establish this proof-of-principle we took advantage of noninvasive intestinal pathogenic E. coli strains that harbor a T3SS, such as EPEC and EHEC [12,13]. Although in vivo infection by these strains is restricted to humans and certain animals (e.g. cattle), in vitro they can infect many mammalian cell lines from distinct cell types and species, including human and murine epithelial cells, fibroblasts, and macrophages among others [36][37][38]. Interestingly, the mouse pathogen Citrobacter rodentium also carries a LEE pathogenicity island encoding a T3SS almost identical to that found in EPEC and EHEC strains [39].
Applications of intrabody delivery by bacterial injection require the use non-pathogenic bacterial strains. In this study we employed an attenuated strain lacking the adhesin Intimin and three major T3-effectors (EspF, Map, Tir) of EPEC. In the absence of intimin and Tir EPEC looses its intimate adhesion and actin pedestal formation capabilities, while the lack of EspF and Map diminishes major cytopathic effects due to disruption of the mitochondria and activation of Rho GTPases [14,34]. A study with human volunteers who ingested an EPEC null mutant in the intimin gene (eae) demonstrated a strong attenuation of this strain [40]. The EPEC strain E2348/69 used in this study encodes 21 effectors of the T3SS [12]. Accordingly, it would be possible to engineer an attenuated E. coli strain lacking all these effectors following a genome minimization approach [41]. Such bacterial strain will be extremely useful for antibody injection into mammalian cells. Another appealing strategy for bacterial injection of sdAbs is the use of commensal E. coli strains endowed with a functional T3SS. It has been reported that E. coli K-12 strain carrying the complete LEE from EPEC on a cosmid is able to induce actin polymerization in human cells in vitro [42]. However, the LEE is expressed weakly in E. coli K-12 [43] and, therefore, additional engineering is needed for efficient injection from E. coli K-12.
The use of extracellular bacteria with a T3SS as vectors for delivery of proteins (including intrabodies) into mammalian cells differs from other approaches that need the transfer of the protein-encoding gene by viral infection or transfection [8]. Invasive E. coli expressing the Invasin (Inv) from Yersinia and the Listeriolysin O (LLO) from Listeria have been employed previously to deliver proteins, DNA, and interfering shRNA into mammalian cells [11]. Upon cell invasion, invasive E. coli cells (Inv+ LLO+) are lysed in the phagosomes releasing their total cellular content. In contrast to this situation, non-invasive E. coli cells carrying a T3SS remain extracellular and could inject specifically the desired sdAb. In addition, extracellular E. coli bacteria are more sensitive to antibiotic treatment (e.g. gentamycin) facilitating their elimination from in vitro cultures as well as from whole animals in vivo.
The use of live bacteria has a great potential for delivery of therapeutic proteins in vivo, in specific organs or tissues of the animal, where they can produce a continuous supply of the polypeptide. For instance, probiotic strains of lactic acid bacteria have been used in vivo for the extracellular secretion of cytokines, enzymes, antibody fragments, etc. against infectious and inflammatory diseases [44][45][46]. Also mucosal and systemic infections with live attenuated invasive bacterial strains (e.g. Salmonella, Listeria) have been employed for intracellular delivery of antigens for vaccination and cytotoxins for tumor therapy [47][48][49][50][51][52][53][54]. Interestingly, probiotic E. coli strains are currently used as therapeutic agents in humans. For instance, colonization of the gastrointestinal tract by E. coli Nissle 1917 [55] is used to treat some inflammatory bowel diseases such as Crohn's disease [56,57]. In addition, deliberate colonization of the urinary bladder with probiotic E. coli strain ABU83972, isolated from an asymptomatic bacteriuria patient [58], is being used to treat recurrent urinary tract infections by uropathogenic E. coli strains [59,60]. Therefore, an attractive possibility is to engineer probiotic E. coli strains to carry a functional T3SS to deliver intracellularly therapeutic sdAbs targeting proteins involved in diseases such as inflammation and cancer in the gastrointestinal and urinary tracts. Systemic infections to treat other organs and solid tumors are not excluded [61,62]. Importantly, sdAbs interfering the function of relevant intracellular targets involved in cell proliferation (e.g. Ras), apoptosis (e.g. Caspase-3), cell migration (e.g. Gelsolin), and HIV-replication proteins (e.g. Rev) have already been selected [63][64][65][66]. The levels of intracellular sdAbs reported here with the E. coli T3SS (10 5 -10 6 molecules per cell) seem appropriated to modulate the activity of regulatory and cell-signaling proteins, which often have intracellular levels below 10 6 molecules per cell [67,68] and, in addition, could trigger a downstream signaling cascade that would amplify the initial effect of the intrabody.
In conclusion, we believe that injection of sdAbs into mammalian cells using non-pathogenic bacterial strains carrying a T3SS is a promising technology for in vitro and in vivo intrabody applications that target host cell functions and signaling pathways.

Bacterial strains, growth, induction and infection conditions
Escherichia coli strains used in this work are listed in Table 1 [12,13,34,37,69,70]. Bacteria were grown at 37uC in Luria-Bertani (LB) agar plates, in liquid LB medium or in Dubelcco's modified Eagle's medium (DMEM), as indicated. Media were supplemented with appropriated antibiotics for selection. Antibiotics were used at the following concentrations: Ampicillin (Ap) 150 mg/ml; Chloramphenicol (Cm) 30 mg/ml; Kanamycin (Km) 40 mg/ml; Tetracycline (Tc) 10 mg/ml. For secretion into the extracellular media of sdAbs, 5 ml cultures were grown overnight in LB at 37uC under static conditions. Next day, these cultures were used to inoculate 15 ml of DMEM (initial OD 600 ,0.05) in a capped Falcon tube (Beckton Dickinson) and incubation continued at 37uC with shaking until OD 600 ,0.4. At this point, 0.1 mM isopropyl-1-thio-b-D-galactoside (IPTG) was added for 4 h. For infection experiments, overnight LB cultures (as above) were used to inoculate 15 ml of DMEM (initial OD 600 ,0.1) and the cultures were incubated under static conditions at 37uC with 5% CO 2 for 2 h, as a pre-activation step. Bacteria from these cultures were used for infection of HeLa cell cultures (,10 5 cells/well in 24-well tissue culture plates; Falcon) at a multiplicity of infection (MOI) 300:1 (bacterial CFUs: HeLa cells) and 0.1 mM IPTG was added. Infection continued at 37uC with 5% CO 2 for the indicated time (90-150 min) and were stopped on ice.

Plasmids and DNA constructs
Plasmids employed in this study are summarized in Table 2. Standard methods of DNA manipulation were used [71]. All DNA constructs were sequenced using an automated DNA sequencer (Perkin Elmer). Oligonucleotides were synthesized by Sigma Genosys (Table 3). Plasmid pSA10 is a vector that contains lacI q repressor and a multiple cloning site under the control of Ptac promoter [72]. Plasmid pT3sVamy contains a DNA fragment of 549 bp, cloned at the EcoRI site of pSA10 under the Ptac promoter, which encodes the V HH anti-amylase (Vamy) fused to the T3-signal EspF 20 at its 59-end and a six-histidine (His) tag and the 12-amino acid epitope E-tag (GAPVPYPDPLEP) at its 39-end. This DNA fragment was obtained by EcoRI digestion of a 560 bp DNA product generated by homology-driven PCR, fusing two PCR subfragments with a final amplification of the fused product with oligonucleotides R1-Xb-SD-espF y RI-stop-E (Table 3). Subfragment 1 (119 bp), containing EspF 20 signal, was amplified from genomic DNA from EDL933stx with primers R1-Xb-SD-espF and SfiI-espF. Subfragment 2 (517 bp), corresponding to Vamy with His and E-tag epitopes, was amplified using plasmid pEHLYA4SDVamy as a template and oligonucleotides SfiIVamy and RI-stop-E as primers. Plasmid pEHLYA4SDVamy is a derivative of pEHLYA2SDVamy [28] with the His and the E-tag epitope at the 39end of Vamy. Plasmid pT3sVgfp was obtained by SfiI and NotI digestion of pT3sVamy, substituting Vamy coding sequence by Vgfp. The DNA encoding Vgfp was obtained by PCR using plasmid pcAbGFP4 [73] as template and oligonucleotides Vhh-sfiI2 and Vhh-NotI2 as primers. The amplified fragment was digested with SfiI and NotI and the resulting 358 bp DNA molecule was cloned in the backbone of vector pT3sVamy (,4.3 kb) digested with the same enzymes. Plasmid pDsVgfp was constructed by amplication of a DNA segment, encoding Vgfp-His-E-tag devoid of the T3-signal, from pT3sVgfp template with oligonucleotides Dsign-EcoRI and RI-stop-E. The ,0.5 kb PCR product was digested EcoRI and inserted in the same site of vector pSA10 under the control of Ptac promoter.
Plasmids pT3s-Bla and pT3sVamy-Bla are derivatives of pCX340 [26], a vector employed to make fusions with the TEM b-lactamase lacking its Sec-dependent-signal peptide ('blaM). To construct pT3s-Bla, first an 83 bp DNA segment encoding the T3signal of EspF was amplified from genomic DNA from EDL933stx with the oligonucleotides NdeI-espF and EcoRI-espF. Next, this DNA segment was fused, by homology-driven PCR, with a 1.2 kb DNA fragment encoding ('blaM), that had been amplified from pCX340 with oligonucleotides EcoRI-TEM and BamHI-tetra. The resulting 1.3 kb fragment, amplified with oligonucleotides NdeI-espF and BamHI-tetra, was digested with NdeI and BamHI and ligated with backbone fragment of pCX340 digested with the same enzymes. To construct pT3sVamy-Bla, a DNA fragment encoding T3sVamy was amplified from plasmid pT3sVamy with oligonucleotides NdeI-espFVamy and EcoRIVamy-espF, digested with NdeI and EcoRI and ligated in the same sites of pCX340.

In vitro cell culture and plasmid transfection
The human epithelial cell line HeLa clone HtTA1 was grown as monolayer in DMEM, supplemented with 10% fetal bovine serum (FBS) and 2 mM glutamine, at 37uC with 5% CO 2 . For transfection, HeLa cells were seed in tissue 24-well culture plates (,10 5 cells/well), grown for 20 h at 37uC with 5% CO 2 . Plasmid pEGFPN1 (Clontech) or pCS2+MT [74] was added (0.6 mg DNA/well) to the cultures in calcium phosphate [75]. After 22 h incubation, the medium was removed, wells were washed three times with PBS, filled with 1 ml of complete medium and incubated for 1 h at 37uC with 5% CO 2 . This medium was replaced by serum-free medium before cell cultures were infected as described above.

SDS-PAGE and Western blot analysis
Sodium Dodecy Sulfate-Polyacrylamide gel electrophoresis (SDS-PAGE) and Western blot was performed following standard methods [71] using the Miniprotean III system (Bio-Rad). Proteins Table 3. Oligonucleotides used in this study.

Analysis of secreted and cellular proteins in E. coli strains
Whole-cell protein extracts from induced E. coli cultures were obtained from cells harvested by centrifugation (30006g, 5 min) from 1 ml aliquot of liquid cultures (OD 600 ,1.2), resuspended in 100 ml of phosphate-buffered saline (PBS), and mixed with the same volume of 2X SDS-PAGE sample buffer. Samples were boiled for 10 min, briefly sonicated (5 sec; Labsonic B Braun), and centrifuged (140006g, 5 min) to remove insoluble material before loading on SDS-PAGE. For analysis of proteins secreted in the culture medium, culture supernatants obtained after centrifugation (30006g, 15 min) were filtered utilizing a 0.22-mm sterile lowprotein binding PVDF filter unit (Millex GV, Millipore) and the serine-protease inhibitor phenyl-methyl-sulphonyl-fluoride (PMSF) was added to 1 mM final concentration. The proteins in the filtered-culture supernatants were mixed with 1/5th of the volume of SDS-PAGE sample buffer (5X) for WB or were precipitated with trichloroacetic acid (TCA 20% w/v; Merck) for Coomassie staining. After centrifugation (140006g, 15 min), TCA-precipitated protein pellets were rinsed with cold acetone (220uC) and resuspended in SDS-PAGE sample buffer (1/10th of the precipitated volume). Purification of His-tagged T3-secreted sdAbs with Talon resin (Clontech) is described below. To test the solubility of the T3-secreted sdAbs, the filtered-culture supernantants (see above) were centrifuged at 100.0006g in a Beckman TL-100 ultracentrifuge for 1 h at 4uC. The supernatant (Soluble) and the pellet (insoluble) fractions obtained after this centrifugation were adjusted to the same final volume in SDS-PAGE sample buffer, boiled and analyzed by WB.

Purification of T3-secreted sdAbs
Cultures of E. coli EPEC with the indicated plasmids (pT3sVamy or pT3sVgfp) were grown at 37uC with shaking in 200 ml of DMEM with ampicillin (using a capped bottle). IPTG was added at 0.1 mM final concentration when the OD 600 reached ,0.4 and incubation continued for 4 h. Culture supernatants were filtered and PMSF was added as described above. Next, supernatants were equilibrated to PBS 1X and incubated overnight at 4uC with 2 ml of Cobalt-containing chromatography resin (Talon, Clontech) for binding of His-tagged sdAbs. Resin was packed in a column, washed 4 times with 10 ml of PBS containing 5 mM imidazole, and eluted in 1 ml aliquots with PBS containing 100 mM imidazole. Eluted fractions were stored at 4uC.

Enzyme-linked immunosorbent assays (ELISA)
ELISA conditions were based on those described previously [76]. Briefly, 96-well immunoplates (Maxisorp, Nunc) were coated for 2 h at room temperature with purified antigens (10 mg/ml) in PBS. Antigens employed: alpha-amylase (Amy; Sigma), the green fluorescent protein (GFP; Upstate), bovine serum albumin (BSA, Roche). Plates were washed with PBS and blocked in PBS buffer containing 3% (w/v) non-fat milk, before incubation with purified T3-secreted sdAbs or filtered-culture supernatants obtained after IPTG induction (at the indicated concentration or dilution in the same buffer). After PBS-wash, bound E-tagged sdAbs were revealed with anti-E-tag mAb-POD (1:2000) and developed with o-phenylenediamine (OPD, Sigma) and H 2 O 2 (Sigma). The OD at 490 nm of the plates was determined in a microplate reader (Bio-Rad).

b-lactamase translocation assay
We followed the method described by [26]. Briefly, the indicated bacterial strains were used to infect HeLa cells grown in vitro in 8-well Falcon culture slides (Beckton Dickinson), IPTG was added for induction and incubation continued for 90 min. The medium was removed and cells were washed three times with Hank's balance salt solution (HBSS). Next, 200 ml of HBSS and 40 ml of the b-lactamase substrate CCF2/AM mix (K1024, Invitrogen) were added. Cells were incubated for additional 90 min at room temperature in the dark, washed three times with HBSS and analyzed by fluorescence microscopy (Nikon Eclipse E600, excitation UV light 330-380 nm). For quantitative analysis of Bla translocation, HeLa cells were seeded in a 96-well opaque plate (Nunc) at aprox. 85% confluence (,2610 4 cells/well). After 16 h incubation at 37uC with 5% CO 2 , infection were done with pre-activated EPEC strains in serum-free DMEM and the cultures were further incubated for 30 min before addition of IPTG, and 60 min after this addition. Infections were washed three times with HBSS, and 200 ml/well of HBSS were added plus 20 ml of CCF2/ AM substrate mix. Samples were incubated for 90 min in the dark, washed three times with HBSS and finally 100 ml/well of HBSS were added. Plates were read in a FLUOstar Optima Microplate Fluorometer with a filter set 450/520 nm.

Streptolysin-O (SLO) fractionation of infected cell cultures
Conditions of SLO treatment were based on those described by [77] with some modifications. Briefly, 24-well plates containing infected HeLa cell cultures (as described above) were placed on ice and washed three times with 1 ml/well of freshly prepared cold SLO-buffer (150 mM sucrose, 25 mM Hepes pH 7.4, 150 mM K-acetate, 2.5 mM MgCl 2 , 4 mM EGTA, 2 mM DTT). Next, 175 ml/well of SLO-buffer containing 10 mg/ml of Streptolysin O (purchase from Prof. Sucharit Bhakdi's laboratory, Institute of Medical Microbiology and Hygiene, Hochhaus am Augustusplatz, Mainz, Germany) were added followed by 15 min of incubation on ice to allow binding of SLO to cells. After this incubation, unbound SLO was removed by washing three times with 1 ml/ well of cold SLO buffer and 175 ml/well of this buffer was added. Plates were incubated at 37uC for 15 min for SLO-pore formation to allow release of the cytosolic content of HeLa cells (which was confirmed by observation in an inverted light microscope; Carl Zeiss). Extracellular media containing released cytoplasmic proteins were collected from plates. Protease inhibitors were added (Complete EDTA-free Protease Inhibitors Cocktail, Roche) to these extracts and centrifuged at 45006g (15 min, 4uC) in order to eliminate any bacteria and cells detached from plates. This supernatant was collected and centrifuged at 160006g (15 min, 4uC). The final resulting supernatant was referred to as ''eukaryotic cytoplasm'' protein extract. Eukaryotic cells debris and bacteria that remained attached to plates after SLO treatment were collected in 175 ml/well of a SDS-PAGE sample buffer (1x) and this extract was referred to as ''SLO-insoluble'' protein extract.

Immunoprecipitation assays
Eukaryotic cytoplasm protein extracts (350 ml obtained as described above from two infected tissue culture wells) were incubated with 40 ml of anti-E-tag mAb bound to protein G-Sepharose resin. Anti-E-tag mAb (1 mg, Phadia) was previously crosslinked to protein G-Sepharose resin (1 ml, Sigma) with dimethyl pimelimidate dihydrochloride (DMP; Sigma). After 16 h incubation at 4uC in an orbital shaker, the resin was collected by centrifugation (8006g, 1 min) and washed three times with 1.5 ml of 200 mM sodium phosphate buffer (pH 8.2). Bound proteins were eluted from resin by incubation with 60 ml of 0.1 M glycine pH 2.8 (10 min at room temperature) followed by centrifugation (8006g, 1 min). Supernatants were collected and 30 ml of 200 mM sodium phosphate buffer (pH 8.2) was added for pH neutralization, referred to as immunoprecipitated (IP) proteins. Routinely, 12 ml of these IP proteins were mixed with 3 ml of 5X SDS-PAGE sample buffer, boiled and gel-loaded for Western blot analysis.

Immunofluorescence microscopy
Infected HeLa cell cultures, grown on coverslips in 24-well plates, were washed three times with 1 ml/well of PBS, fixed with 3% (w/v) paraformaldehyde (in PBS) for 20 min at room temperature, and washed again with PBS three times. Cells were permeabilized by incubation with PBS containing 0.1% (v/v) of Triton X-100 (Sigma) for 5 min. To label EPEC, a rabbit polyclonal anti-O127 serum was diluted 1:200 in PBS with 10% donkey serum (Jackson Immunoresearch) and incubated 1 h at room temperature. Coverslips were washed three times with PBS and goat anti-rabbit IgG-Texas Red conjugated secondary antibody (1:500 in PBS with 10% donkey serum; Molecular Probes) was added along with Oregon-Green conjugated phalloidin (1:100; Invitrogen) and DAPI (1:1000; Sigma) to label F-actin and DNA, respectively. Coverslips were washed three times with PBS after incubation and 4 ml of mounting medium (DAKO) was added. Coverslips were analyzed by conventional epifluorescence microscopy using a Zeiss Axio imager microscope.

Supporting Information
Figure S1 Scheme of plasmid vectors used for expression of T3sV HH fusions. The sequence encoding the first 20 amino acids of EspF effector (T3s) is fused to the corresponding V HH (Vamy or Vgfp in pT3sVamy or pT3sVgfp, respectively). Epitope tags (His and E-tag) at the C-termini of fusions and unique restriction sites SfiI, NcoI and NotI flanking V HH domain are indicated. Gene constructs are under the control of the IPTGinducible Ptac promoter. The presence of lacI q repressor, transcriptional terminators (T1,T2) from 5S ribosomal RNA gene, ampicillin-resistance (amp r ) gene, and origin of replication (ori) are also shown.