Unusual, Virulence Plasmid-Dependent Growth Behavior of Yersinia enterocolitica in Three-Dimensional Collagen Gels

Abstract

As a first approach to establishing a three-dimensional culture infection model, we studied the growth behavior of the extracellular pathogen Yersinia enterocolitica in three-dimensional collagen gels (3D-CoG). Surprisingly, we observed that plasmidless Y. enterocolitica was motile in the 3D-CoG in contrast to its growth in traditional motility agar at 37°C. Motility at 37°C was abrogated in the presence of the virulence plasmid pYV or the exclusive expression of the pYV-located Yersinia adhesion gene yadA. YadA-producing yersiniae formed densely packed (dp) microcolonies, whereas pYVΔyadA-carrying yersiniae formed loosely packed microcolonies at 37°C in 3D-CoG. Furthermore, we demonstrated that the packing density of the microcolonies was dependent on the head domain of YadA. Moreover, dp microcolony formation did not depend on the capacity of YadA to bind to collagen fibers, as demonstrated by the use of yersiniae producing collagen nonbinding YadA. By using a yopE-gfp reporter, we demonstrated Ca²⁺-dependent expression of this pYV-localized virulence gene by yersiniae in 3D-CoG. In conclusion, this study revealed unique plasmid-dependent growth behavior of yersiniae in a three-dimensional matrix environment that resembles the behavior of yersiniae (e.g., formation of microcolonies) in infected mouse tissue. Thus, this 3D-CoG model may be a first step to a more complex level of in vitro infection models that mimic living tissue, enabling us to study the dynamics of pathogen-host cell interactions.

For decades, cell culture monolayers have successfully been used to study mechanisms of microbial adherence, invasion, and intracellular survival/multiplication. For in vitro infection studies, eukaryotic cells are grown on solid supports as monolayers and then are challenged with the respective pathogen. In spite of the liquid culture medium (third dimension) covering the adherent cell monolayer, this infection model can be considered a two-dimensional system which obviously does not reflect the environment of host tissue and dynamic events, such as cell migration, happening during infection. Host tissue (i) is vascularized, allowing the extravasation and migration of host immune cells toward the invading microbe, and (ii) consists of a network of extracellular matrix (ECM) proteins with diverse resident cells (e.g., fibroblasts, macrophages, etc.). Thus, there are many reasons to leave the two-dimensional system and establish an in vitro three-dimensional infection model by approaching in vivo conditions. In order to simulate a tissue-like environment, immunology and cell biology as well as tumor biology make use of three-dimensional collagen matrices to study cell migration, cell-cell interactions, and cell-matrix interactions (reviewed in reference 13). To develop and assess such a three-dimensional infection model, we started with the well-established, three-dimensional collagen gel (3D-CoG) on microscopic glass supports and with Yersinia enterocolitica as the prototype of an extracellular pathogen.

Y. enterocolitica causes food-borne gastrointestinal diseases by invading the intestinal mucosa and multiplying predominantly extracellularly in Peyer's patches, mesenteric lymph nodes, spleens, and livers. The pathogenicity of this enteric pathogen is controlled by the virulence plasmid pYV, which encodes a type III secretion system (TTSS), about six secreted antihost Yersinia outer proteins (YopE, YopH, YopM, YopO, YopP, and YopT) and the adhesin YadA (5) as well as diverse regulators of gene expression and protein secretion (e.g., VirF [6], reviewed in reference 7). Yops are microinjected into host cells by the TTSS and interfere with a variety of cell functions, including cytoskeletal regulation, cytokine production, and the control of apoptosis, to suppress the immune response to infection, resulting in predominantly extracellular survival of the pathogen in the lymphoid tissue of the host (reviewed in references 1, 16, 20, and 42).

Besides the virulence plasmid, essential for virulence are chromosomally located genes, such as the invasin-encoding gene inv and the yersiniabactin (ybt) determinant, located on the high pathogenicity island (HPI) which is responsible for ferric iron provision (reviewed in reference 37).

As reported recently, the extracellular environment plays a critical role in defining Yersinia-host cell interactions (11, 22). The ECM content is one of the determining factors for the phagocytic response of host macrophages. Fibronectin and laminin as well as diverse types of collagen enhance bacterial adhesion and internalization via YadA-mediated signaling and phagocytosis. The Yersinia invasin (Inv) and adhesin (YadA) both mediate binding to β1 integrins, which are also receptors for some of the ECM proteins. While Inv binds directly to host β1 integrin receptors, YadA presumably binds indirectly via an ECM bridging mechanism (10-12). It has previously been shown that YadA densely covers the bacterial surface by forming a fibrillar matrix of lollipop-shaped surface projections, which might mask the lipopolysaccharides as well as the Inv protein and protect Yersinia from the complement system (31) and defensin lysis (43). YadA belongs to the family of oligomeric, coiled-coil, lollipop-shaped adhesins (Oca family) with C-terminal outer membrane anchor domains, intermediate segments forming pillar-like coiled-coil stalks, and globular N-terminal head domains that promote tight adherence to host cells and ECM proteins (21, 32). Moreover, YadA is also involved in the autoaggregation/autoagglutination of Yersinia cells growing in cell culture medium at 37°C (39). The role of these diverse functions of YadA for pathogenicity is still unclear, but is expected to determine the growth behavior of extracellular Y. enterocolitica in infected mouse tissue as well as in 3D-CoG.

We wondered whether such three-dimensional in vitro models would be more suitable for study of the growth behavior of bacterial pathogens and their interactions with host cells (in particular, professional phagocytes) than are the traditionally used two-dimensional cell culture monolayers. As a first step in this direction, we report for the first time on the characterization of the growth behavior of Y. enterocolitica in a 3D-CoG environment, with a focus on virulence plasmid-controlled functions, microcolony formation, and Yersinia motility.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

The bacterial strains used in this study are listed in Table 1 and characteristics are summarized in Table 2. Bacteria (Escherichia coli at 37°C and Yersinia strains at 27°C) were grown in Luria-Bertani (LB) or brain heart infusion medium. For analysis of secreted proteins and for 3D-CoG, stationary overnight Yersinia cultures grown at 27°C were diluted 1:20 or 1:40 in fresh medium and incubated for 2 h at 37°C to induce the expression of virulence factors. For YadA production and analysis, overnight cultures were inoculated into RPMI tissue culture medium (1:500) and grown overnight at 37°C without shaking. The absence or presence of YadA on the cell surface was checked by slide agglutination using anti-YadA(O:8) rabbit serum and by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) of total cell lysates (33).

TABLE 1.

Bacterial strains

Strain	Description	Reference
WA(pYVO8)	Y. enterocolitica serotype O:8 (WA-314), clinical isolate, pYVO8+	19
WA-C	Plasmidless derivative of WA(pYVO8)	19
WA(pYVO8ΔyadA)	WA-314 harboring pYVO8-A-0 with yadA inactivated by a kanamycin cassette	33
WA(pYVO8ΔyadA::yadA)	WA-314 harboring pYVO8-A-1 with integrated pGPS-A-1, complemented by wild-type yadA	33
WA(pYVO8-YadAH156/H159Y)	WA-314 harboring pYVO8-A-2 producing collagen nonbinding YadA because of replacement of His-156 and His-159 by Tyr	33
WA(pYVO8-YadAΔ29-81)	WA-314 harboring pYVO8-AΔ29-81 producing neutrophil nonbinding YadA (deletion of amino acids 29 to 81)	34
WA(pYVO8-YadAΔH)	WA-314 harboring pYVO8-A-H producing YadA with deleted head structure (deletion of amino acids 30 to 188)	32
WA(pYVO8-YadAΔS3)	WA-314 harboring pYVO8-A-S3 producing YadA with deleted four 15-mer repeats of the stalk (deletion of amino acids 229 to 289)	32
WAΔmotAB(pYVO8)	motAB mutant of WA-314 producing nonmotile flagellae, pYVO8+	44
WA-CΔmotAB	motAB mutant of WA-C producing nonmotile flagellae	This study
WA(pYVO8, pYopE138-GFP)	WA-C(pYVO8, pCJYE138-G3), WA-314 with a YopE138-GFP reporter (pCJYE138-G3), pYVO8+	26
WA(pYVO8, pFyuA-GFP)	WA-C(pYVO8, pCJFY-GL), WA-314 with a FyuA-GFP-Luc reporter (pCJFY-GL), pYVO8+	25
WAΔinv(pYVO8)	inv-negative mutant of WA-314, inactivation of the invasin gene by insertion of a kanamycin resistance cassette, pYVO8+	36
WA-C Δinv	inv-negative mutant of WA-C	36
WA(pYVO8ΔvirF)	virF-negative mutant of WA-314, pYVO8+	This study
WA-C(pBR322 EH-5)	WA-C harboring plasmid pBR322 with a 5-kb EcoRI-HindIII fragment of pYVO8 carrying yadA	35
WA-C(pLCR)	WA-C harboring plasmid pLCR with a 30-kb SalI/XbaI fragment of pVY carrying the genes of the TTSS on a low-copy vector	35
WAΔyenIR(pYVO8)	yenIR-negative mutant of WA-314, inactivation of the quorum-sensing system, pYVO8+	24
WA-CΔyenIR	yenIR-negative mutant of WA-C, inactivation of the quorum-sensing system	This study
WAΔmyf(pYVO8)	myfA-negative mutant of WA-314, inactivation of the myfA gene by insertion of a spectinomycin resistance cassette, pYVO8+	This study

TABLE 2.

Summary of strain characteristics

Strain	Autoagglutination	YadA expression	Yop expression/secretiona	Microcolony formationb
WA(pYVO8)	+	+	+	dp
WA-C	−	−	−	−
WA(pYVO8ΔyadA)	−	−	+	lp
WA(pYVO8 ΔyadA::yadA)	+	+	+	dp
WA(pYVO8-YadAH156/H159Y)	+	+	+	dp
WA(pYVO8-YadAΔ29-81)	+	+	+	dp
WA(pYVO8-YadAΔH)	−	+	+	lp
WA(pYVO8-YadAΔS3)	+	+	+	dp
WAΔmotAB(pYVO8)	+	+	+	dp
WA-CΔmotAB	−	−	−	lp
WA(pYVO8, pYopE138-GFP)	+	+	+	dp
WA(pYVO8, pFyuA-GFP)	+	+	+	dp
WAΔinv(pYVO8)	+	+	+	dp
WA-CΔinv	−	−	−	−
WA(pYVO8ΔvirF)	−	(+)	−	dp
WA-C(pBR322 EH-5)	+	+	−	dp
WA-C(pLCR)	−	−	#	−
WAΔyenIR(pYVO8)	+	+	+	dp
WA-CΔyenIR	−	−	−	lp
WAΔmyf(pYVO8)	+	+	+	dp

^a#, YopB, YopD, and YopN, but no effector Yops.

^b−, no microcolony formation.

The following antibiotics were used at the indicated concentrations: ampicillin, 100 μg/ml; chloramphenicol, 20 μg/ml; kanamycin, 50 μg/ml; spectinomycin, 100 μg/ml; and tetracycline, 20 μg/ml.

Nucleic acid manipulations.

A virF mutant of Y. enterocolitica WA(pYVO8) was constructed by the phage lambda Red recombinase cloning procedure (8). The entire coding region of the virF gene was replaced by a kanamycin resistance cassette. This replacement was achieved by homologous recombination mediated by λ phage Redα and Redβ recombinases, which were expressed in Yersinia from a curable plasmid pKD46 as described previously (41). For recombination, a PCR product harboring a kanamycin cassette with 50-nucleotide-homology arms was amplified from plasmid pACYC177 (New England Biolabs) with the primers VirF_ET_for (5′-ATGGTTGTACATCGCACGCATAATAACTCAATACACCTCATTAGATAAATTGATCACTGACACCCTCATCAGTG-3′) and VirF_ET_rev (5′-ATTTTACTTTATAGTCCAAAAGTGTTTTTTTAAAAAAAAAACAGATAATTCGTCAAGTCAGCGTAATGCTC-3′).

The transformation of this PCR product into the wild type resulted in WA(pYVO8ΔvirF), which was verified by PCR and analysis of secreted proteins (Yops) and YadA expression (the loss of the transcriptional activator VirF resulted in the attenuation of yop and yadA gene expression).

Using the same technique, we constructed a yenI yenR double mutant (24). The corresponding PCR fragment for recombination was amplified with the primers YenR_ET_for (5′-TAATTGACTATTTTGATAACGAAAGTATTAATGAAGATATAAAGAACTATCTCTGATGTTACATTGCACA-3′) and YenI_ET_rev (5′-ACTCTTTAACGTAAATTTTAATAATATGCCGGAAAGGAAGTTAGACGAGACGTAATGCTCTGCCAGTG-3′).

The transformation of this PCR product into WA(pYVO8) resulted in WAΔyenIR(pYVO8), which was verified by PCR and analysis of N-(3-oxohexanoyl)-l-homoserine lactone production (24).

The myf mutant WAΔmyf(pYVO8) was constructed by conjugation as described previously (40). The primers used for amplification of the conjugation fragment were myf_sacI_f (5′-GAGCTCACTTGCCAGTATATGCCTCG-3′) and myf_sphI_r (5′-GCATGCCTCTCAGTTAGGCTATACGT-3′). A spectinomycin cassette was inserted into the myf fragment at the KpnI cleavage site.

The inv mutant WAΔinv(pYVO8) (invasin negative) and the motAB mutant WAΔmotAB(pYVO8) (nonmotile, but flagellum positive) have previously been described (36, 44).

For pYV plasmid curing, Y. enterocolitica strains were grown on brain heart infusion agar supplemented with 5 mM EGTA at 37°C as described previously (19). The selected plasmid-cured strains were WA-CΔyenIR, WA-CΔinv, and WA-CΔmotAB.

Constitutively red fluorescent protein (RFP)-expressing and green fluorescent protein (GFP)-expressing bacteria were constructed as described previously; red-fluorescing Yersinia strains carry pLACRFP and green-fluorescing Yersinia strains pLACGFP (29).

PCR was performed with the high-fidelity PCR system of Roche (Germany), and the preparations of plasmid DNA and the isolation of DNA fragments from agarose gels were carried out with OLS kits (Omni Life Science OLS, Germany), as recommended by the manufacturer.

Analysis of YadA and Yop production.

All tested strains were checked for YadA production by slide agglutination and SDS-PAGE as described previously (33). Alternatively, YadA was detected on the surfaces of Yersinia cells by indirect immunofluorescence using anti-YadA(O:8) rabbit serum as primary antiserum (34).

The secretion of Yop proteins was induced as described previously (18). Secreted proteins were precipitated from the culture supernatant with 10% trichloroacetic acid for 1 h on ice. After washing with ice-cold acetone, released proteins were separated by SDS-PAGE on a 12% polyacrylamide gel, followed by Coomassie staining.

3D-CoG model.

For growth studies, 10⁵ Yersinia cells were suspended in liquid collagen solution (Nutacon BV, The Netherlands) containing bovine type I collagen at a final concentration of 1.7 mg/ml in RPMI 1640 medium adjusted to pH 7.4 (total volume, 66 μl). This solution was placed into a small self-constructed chamber (a tracking chamber) built by a hollowed coverslip on a glass slide and allowed to polymerize for 45 min (37°C; 5% CO₂) as described previously (17). The depth of the resulting collagen gel was about 400 μm. The remaining space in the tracking chamber was filled with RPMI 1640 medium, and thereafter, the chamber was sealed with wax (1:2, paraffin to vaseline) and incubated at 37°C in a cell culture incubator.

Growth was checked microscopically (phase contrast) after different time points (with a conventional microscope) or followed by time-lapse video microscopy.

To determine bacterial growth rates, yersiniae were released from the 3D-CoG by using collagenase (Clostridium histolyticum collagenase; Sigma) at a concentration of 1,000 U/ml in phosphate-buffered saline at 37°C. After digestion of the collagen gel, bacteria were washed with phosphate-buffered saline, and serial dilutions of the cell suspension (bacterial clusters and aggregates were dissociated, as checked by microscopy) were plated onto LB agar plates. CFU were counted after 48 h of incubation at 27°C. Collagenase treatment of yersiniae did not impair viability, as was checked by our ability to grow Yersinia in liquid medium.

To measure diameters of Yersinia microcolonies in the 3D-CoG by laser scanning microscopy (Leica TCS NT), green-fluorescing Yersinia strains carrying pLACGFP (29) were cultured overnight within the 3D-CoG.

In addition, for motility and growth studies, Yersinia strains were grown in the above-described tracking chamber filled with LB agar (1.5 and 0.5% agar), RPMI 1640 agar (0.5% agar), or motility agar (3 g/liter meat extract, 10 g/liter peptone, 5 g/liter NaCl, 8% gelatin, 0.4% agar).

If not otherwise mentioned, all collagen and agar experiments were performed in the above-described tracking chamber.

Time-lapse video microscopy.

Growth of Yersinia strains in 3D-CoG was monitored by time-lapse video microscopy using the PerkinElmer UltraVIEW LCI system with the temporal software module. For growth studies, Yersinia cells were imaged for a sample period of 12 to 15 h at a constant frame rate of 1 frame/10 min and for motility tracking studies for a sample period of 11 min at a constant frame rate of 1 frame/10 s. Tracking was carried out with MetaMorph (offline version 6.2r6).

Confocal reflection contrast imaging of 3D-CoG.

For visualization of collagen fibers within an unstained, nonfixed 3D-CoG, confocal reflection contrast was used on an inverted confocal laser scanning microscope (Leica TCS) with TCSNT software as described previously (14). For excitation, the beam splitter RT30/70 was chosen. The reflection signal was combined with fluorescence microscopy to detect RFP-expressing WA(pYVO8, pLACRFP) and WA-C(pLACRFP), respectively. Three-dimensional reconstructions of sequential xy sections (20 to 30; step-size approximately 1 μm) were performed as projection images.

Electron microscopy.

Samples were fixed in 3.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4) for 3 h at 4°C, followed by transfer to 1% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4). After washing with phosphate buffer, samples were incubated with 2% osmium tetroxide in aqua bidest for 2 h at 4°C and washed with phosphate buffer at room temperature before being dehydrated in a graded series of ethanol (30, 50, 70, 90, and 100%). We performed embedding in araldite with polymerization for 48 h at 60°C and thin sectioning and staining with 2% unbuffered uranyl acetate according to standard procedures. The stained samples were imaged in a Philips CM-10 electron microscope.

Nucleotide sequence accession numbers.

For primer construction, gene sequences with the following accession numbers in GenBank GeneID were used: 4715922 (inv), 4715936 (motA), 4715935 (motB), 4713029 (myfA), 4711913 (virF), 4711901 (yadA), 4714977 (yenI), 4714976 (yenR), and 4711874 (yopE).

RESULTS

Growth characteristics of Yersinia in the 3D-CoG.

As Y. enterocolitica is known to multiply within the ECM of infected tissue (Peyer's patch, spleen, and liver) of mammals and its growth and interaction with the ECM are controlled by pYV, this pathogen should be particularly suitable for the study of its growth behavior in a 3D-CoG model. Therefore, we started growth studies in 3D-CoG by comparing growth of the isogenic pair WA(pYVO8) and WA-C (the plasmidless derivative).

In the 3D-CoG model, strain WA-C grew as single cells or in small clusters of only a few cells (Fig. 1A1 and B1 and 2A to C), while the pYV-positive strain WA(pYVO8) formed spherical microcolonies with diameters of 40 to 50 μm within 20 h (Fig. 1A2 and B2). This result exemplifies that the formation of microcolonies by yersiniae in 3D-CoG is pYV dependent. Comparing the two different growth phenotypes within the 3D-CoG by electron microscopy, we observed different ultrastructural characteristics: in contrast to the case for WA-C, the densely packed (dp) microcolony forming WA(pYVO8) was covered by a fibrillar layer on top of the bacterial surface (Fig. 1C1 and C2).

FIG. 1.

Growth characteristics of Yersina in 3D-CoG. Imaging of the growth characteristics of Yersinia WA(pYVO8) and its plasmidless derivative WA-C in a 3D-CoG environment at 37°C after 20 h by different microscopic techniques. (A) Conventional phase contrast microscopy. Scale bar, 10 μm. (B) Confocal reflection contrast microscopy (an illustration of collagen fibers is shown in green) combined with fluorescence microscopy (an illustration of Yersinia cells expressing RFP is shown in red). Scale bar, 10 μm. (C) Electron microscopy. (C1a) Original magnification, ×1,650. Scale bar, 1 μm. (C1b) Original magnification, ×21,000. Scale bar, 100 nm. (C1c) Original magnification, ×66,740. Scale bar, 100 nm. (C1d) Excerpt of panel C1c. Scale bar, 100 nm. (C2a) Original magnification, ×20,080. Scale bar, 1 μm. (C2b) Excerpt of panel C2a. Scale bar, 100 nm. (C2c) Original magnification, ×37,000. Scale bar, 500 nm. (C2d and C2e) Excerpts of panel C2c. Scale bars, 100 nm. The plasmidless strain WA-C (A1, B1, and C1) grows singly or in small clusters of only a few cells, while the pYV-positive strain WA(pYVO8) (A2, B2, and C2) forms spherical dp microcolonies. An analysis of the cellular ultrastructure by electron microscopy reveals a fibrillar layer (red arrows) on top of WA(pYVO8) (C2), which forms dp microcolonies, in contrast to the case for WA-C (C1).

FIG. 2.

Growth progress of Yersinia in a 3D-CoG environment at 37°C. WA(pYVO8) and its plasmidless derivative WA-C were studied in 3D-CoG by time-lapse video microscopy (phase contrast) for 20 h, and growth characteristics were compared. WA-C (A to C) grows in small clusters of only a few cells or singly, while WA(pYVO8) forms chains of rods (which start to convolute) and establishes dp spherical microcolonies within 20 h (D to F). Scale bars, 10 μm.

The observed fibrillar layer is evocative of Myf fibrillae. Analyzation of the surface structure of a myf deletion mutant [WAΔmyf(pYVO8)] within the 3D-CoG by electron microscopy revealed that the fibrillar layer still existed (data not shown).

Further studies with Y. enterocolitica of other human pathogenic serotypes (O:3, O:8, and O:9) pointed out that the different growth behaviors of pYV-positive and pYV-negative strains are independent of the serotype (data not shown).

By observing the bacterial replication for a certain time interval using time-lapse microscopy, we found that WA(pYVO8) formed chains of rods without separation of the descendants. After a few hours, the chains formed by WA(pYVO8) rods started to convolute and to establish dp spherical microcolonies (Fig. 2D to F). In contrast, WA-C dissociated after replication (Fig. 2A to C).

To determine the growth rates of strains WA-C and WA(pYVO8), the 3D-CoG was liquefied with collagenase at fixed time points to release yersiniae for plating on solid medium. From colony counts of three independent experiments, we calculated average doubling times of 37 min for WA(pYVO8) and 29 min for WA-C (Fig. 3 shows the growth curves). It was impossible to determine exact cell numbers of WA(pYVO8) at later time points (for example, after overnight incubation to determine the maximal cell number of one microcolony), due to stable microcolony formation which did not allow entire disruption.

FIG. 3.

Growth curve of WA(pYVO8) (A) and WA-C (B) within the 3D-CoG at 37°C. For the determination of the growth rate, the 3D-CoG was liquefied with collagenase at fixed time points to release yersiniae for plating on solid medium. From colony counts, doubling times were calculated. *, because of microcolony formation of WA(pYVO8), which did not allow disintegration into single bacterial cells, later time points were canceled. Data were collected from three independent experiments. Error bars indicate standard deviations.

In contrast to their growth in 3D-CoG, WA-C and WA(pYVO8) showed comparable growth behaviors in three-dimensional motility agar with respect to spatial distribution, as shown in Fig. 4A and B, respectively. The two strains formed tiny aggregates, but no microcolony formation could be observed in three-dimensional motility agar of 37°C. Similar results were obtained after growth in tracking chambers filled with 1.5 or 0.5% LB agar, or 0.5% RPMI agar (data not shown). This exemplifies that the pYV-dependent growth of WA(pYVO8) in microcolonies compared to that of WA-C, which grows in small clusters or singly, is unique for the three-dimensional collagen environment.

FIG. 4.

Growth of Yersinia in three-dimensional motility agar. For comparison purposes, WA-C (A) and WA(pYVO8) (B) were analyzed in an agar environment instead of a collagen environment. After growing for 20 h at 37°C, both strains formed tiny aggregates. Scale bars, 10 μm.

YadA-dependent packing density of microcolonies.

The Yersinia adhesin YadA is known to bind diverse ECM proteins, such as collagen and laminin as well as fibronectin, and mediate attachment to the surfaces of host cells (reviewed in reference 12). We reasoned whether the binding properties of YadA might influence the growth behavior of Yersinia in the 3D-CoG environment.

By studying different yadA mutant strains, we demonstrated that the pYV-encoded adhesin YadA is crucial for the packing density of the observed microcolonies. YadA-positive strains grew in dp spherical microcolonies (Fig. 1A2). In contrast, YadA-negative strains formed net-like, loosely packed (lp) structures (Fig. 5A). In the 3D-CoG model, strains producing the collagen nonbinding mutant YadA_H156/H159Y [strain WA(pYVO8-A-2)] or the neutrophil nonbinding truncated YadA_Δ29-81 [strain WA(pYVO8-YadA_Δ29-81)] all featured dp spherical microcolonies (Fig. 5B and C). YadA with deletion of the stalk domain [WA(pYVO8-YadA_ΔS3)] also mediated dp microcolonies (Fig. 5E). In contrast, YadA with deletion of the head region [WA(pYVO8-YadA_ΔH)] mediated lp structures (Fig. 5D) comparable to the YadA-negative strain [WA(pYVO8ΔyadA)] (Fig. 5A). As expected, complementation of the YadA-negative stain with the yadA gene [resulting in WA(pYVO8 ΔyadA::yadA)] restored the YadA-positive phenotype of the dp microcolonies (Fig. 5F). The presence of YadA in Yersinia microcolonies within 3D-CoG was verified by immunofluorescence staining (Fig. 6A) and SDS-PAGE, followed by Coomassie staining (Fig. 6B).

FIG. 5.

Growth behavior of yadA as well as inv mutant strains in 3D-CoG after 20 h at 37°C. The growth phenotype of different yadA mutant strains (A to F) was studied with respect to the diverse functions of YadA. YadA-negative yersiniae (A) formed net-like, lp microcolonies, while collagen nonbinding YadA yersiniae (B) as well as neutrophil nonbinding YadA yersiniae (C) and stalk-deleted YadA yersiniae (E) featured spherical dp microcolonies. Only the deletion of the head domain of YadA (D) mediates lp microcolonies comparable to those of the yadA-negative strain (A). Complementation of the yadA-negative strain with wild-type YadA (F) restores the YadA-positive phenotype of dp microcolonies. In contrast to the pYV- encoded adhesin YadA, the chromosomally encoded invasin Inv of Yersinia has no impact on the growth behavior in the 3D-CoG (G to H): WAΔinv(pYVO8) and WA-CΔinv show no differences in the growth behavior within the 3D-CoG model compared to WA(pYVO8) and WA-C, respectively. Scale bars, 10 μm.

FIG. 6.

YadA expression in the 3D-CoG. (A) Verification of the presence of YadA in Yersinia WA(pYVO8) microcolonies by immunofluorescence using a YadA-specific rabbit antiserum. The left panel shows phase contrast; the right panel shows fluorescence. Scale bars, 10 μm. (B) Verification of the presence of YadA in Yersinia WA(pYVO8) microcolonies in cell lysates by SDS-PAGE. The arrow indicates multimeric YadA (180 to 200 kDa). Lane 1, WA(pYVO8); lane 2, WA(pYVO8ΔvirF).

These findings demonstrate that it is neither the collagen binding function of YadA nor the absence of the stalk domain which influences the growth behavior of Yersinia within the 3D-CoG. Only the deletion of the complete head domain or the deletion of the whole yadA gene converted the growth behavior from dp to lp microcolonies, which was actually distinct from that of the pYV-cured derivative (WA-C). Thus, the dp micocolony phenotype is related to the YadA-mediated autoagglutination trait. In contrast to the pYV-encoded adhesin YadA, the presence of a functional invasin gene (inv) of Yersinia had no impact on the growth behavior in the 3D-CoG: WAΔinv(pYVO8) and WA-CΔinv showed no differences in the growth behavior within the 3D-CoG model relative to the behavior of WA(pYVO8) and WA-C, respectively (Fig. 5G to H).

Motility of Yersinia strains in the 3D-CoG.

The motility of Y. enterocolitica is regulated in a temperature-sensitive manner (27): Y. enterocolitica is motile at 27°C but immotile at 37°C under in vitro growth conditions (liquid culture medium or motility agar) due to a shut down of flagellum synthesis.

Surprisingly, by live-cell imaging, we found that the pYV-negative strain WA-C, which grows singly or in small clusters, is motile at 27°C as well as at 37°C in the three-dimensional collagen environment. In contrast, the pYV-positive strain WA(pYVO8) is immotile at 37°C. Studies with a motAB mutant (WA-CΔmotAB), which produces flagella but is immotile because it lacks motor components, revealed growth of yersiniae in large irregular clusters, dissimilar to dp microcolonies (Fig. 7B). These results demonstrate that the flagellar motor components MotA and MotB are necessary for WA-C motility in the three-dimensional collagen environment at 37°C. The deletion of motAB in the pYV-positive strain [WAΔmotAB(pYVO8)] had no effect on the growth phenotype in the 3D-CoG (Fig. 7A).

FIG. 7.

Growth behavior and motility of different mutant strains in the 3D-CoG after 20 h at 37°C. Studies with motAB mutant strains [WAΔmotAB(pYVO8) and WA-CΔmotAB] revealed immotile yersiniae and the growth of WAΔmotAB(pYVO8) yersiniae in spherical dp microcolonies (A) and that of WA-CΔmotAB yersiniae in large irregular clusters (B), dissimilar to dp microcolonies. WA(pYVO8ΔvirF) showed the same growth behavior as that of the parent strain WA(pYVO8), with dp spherical microcolonies (C) and no motile cells. Strain WA-C(pTTSS), carrying a minivirulence plasmid with the genes of the TTSS and yadA, formed spherical dp microcolonies and no motile cells could be detected (D). WA-C(pBR322 EH-5), expressing only YadA, was also able to form dp microcolonies with immotile yersiniae (E). Furthermore, strain WA-C(pLCR), with a 30-kb SalI/XbaI fragment of pYVO8, which harbors the genes of the TTSS needle complex, but neither yadA nor genes encoding one of the six Yop effector proteins nor YscM2, showed a growth behavior similar to that of WA-C (that is, the formation of small clusters and single motile bacteria) (F). Scale bars, 10 μm.

Recently, evidence suggesting a cross-control between the Yop virulon and the flagellum regulon was provided from transcriptome analysis (4, 30). For example, the master regulator FlhDC is suggested to downregulate the Yop virulon, presumably through FliC, the alternative sigma factor which probably competes with the regulator of the Yop virulon VirF. For this reason, we checked the growth behavior of a virF mutant [WA(pYVO8ΔvirF)], which lacks the major positive transcriptional regulator VirF of pYV genes, in the 3D-CoG environment. WA(pYVO8ΔvirF) showed the same growth behavior as that of the parent strain WA(pYVO8), with dp spherical microcolonies (Fig. 7C) and no motile cells. By growing the VirF-negative strain in liquid medium at 37°C, weak production of YadA could be detected by SDS-PAGE or anti-YadA serum agglutination, which could be sufficient for bacterial autoaggregation (Fig. 6B, lane 2). This phenotype of dp microcolonies could be confirmed by the growth behavior of WA-C(pBR322 EH-5), harboring a 5-kb EcoRI-HindIII fragment of pYVO8, which carries yadA, but no virF, and thus also produces low quantities of YadA as the virF mutant does. Figure 7E shows that this strain was also able to form dp microcolonies. Obviously, even the presence of small amounts of YadA on the cell surfaces made Yersinia cells sticky, prevented motility, and led to dp microcolony formation.

To challenge the different behaviors of the yadA deletion mutant (lp microcolony formation is shown in Fig. 5A) and the pYVO8-negative strain WA-C (no microcolony formation, motile cells [Fig. 1A1]), a possible role of the TTSS in microcolony formation was investigated. Therefore, strain WA-C(pLCR), with a 30-kb SalI/XbaI fragment of pYVO8, was analyzed. pLCR harbors the genes of the TTSS needle complex, but not yadA or genes encoding one of the six Yop effector proteins or YscM2, a regulator of TTSS (35). The growth of WA-C(pLCR) in the 3D-CoG was similar to that of WA-C, with the formation of small clusters (the clusters seem to be a little enlarged compared to those of WA-C) and single motile bacteria (Fig. 7F). These results indicate that besides YadA, another pYV-encoded factor that is not located within the TTSS is involved in motility regulation.

To analyze the impact of the quorum-sensing system on Yersinia motility in the 3D-CoG, we finally checked the growth behavior of a yenIR double mutant (WA-CΔyenIR) (24) in the 3D-CoG at 37°C. YenI and YenR of Y. enterocolitica are homologous to the quorum-sensing LuxI (homoserine lactone synthase) and LuxR (response regulator) protein families, respectively. In contrast to the case for WA-C, WA-CΔyenIR was immotile in the 3D-CoG and formed large irregular clusters (lp microcolonies [data not shown]), comparable to those of WA-CΔmotAB (Fig. 7B).

To document the motility phenotype of strain WA-C, 0.1 or 1% anti-flagellum antiserum was added to the 3D-CoG and the growth phenotype was checked microscopically. As shown in Fig. 8A1, anti-flagellum antiserum within the 3D-CoG resulted in immotile WA-C yersiniae and the formation of large irregular clusters comparable to those of WA-CΔmotAB. Control antiserum (Fig. 8A1) or anti-YadA antiserum (data not shown) did not inhibit the motile phenotype of strain WA-C. Furthermore, dp microcolony formation of WA(pYVO8) was not affected by the addition of anti-flagellum or other antisera (Fig. 8A2). To further document the motility of strain WA-C, tracking experiments were performed with strains WA-C and WA-CΔmotAB (Fig. 8B).

FIG. 8.

Documentation of the motility phenotype of strain WA-C within the 3D-CoG at 37°C. (A) Strains WA-C and WA(pYVO8) were grown with 1% anti-flagellum antiserum within the 3D-CoG and checked microscopically in regard to the growth phenotype. WA-C yersiniae were found to be immotile and formed large irregular clusters (A1a). Control antiserum did not inhibit the motile phenotype of strain WA-C (A1b). Furthermore, dp microcolony formation of WA(pYVO8) was not affected by the addition of anti-flagellum (A2a) or control antisera (A2b). Scale bars, 10 μm. (B) Documentation of the motility of strain WA-C (B1) in contrast to that of WA-CΔmotAB (B2) by tracking experiments.

For comparison purposes, we performed motility experiments with WA(pYVO8) and WA-C in a three-dimensional agar environment (motility agar, 1.5 or 0.5% LB agar, or 0.5% RPMI agar). As expected, both strains were found to be immotile and formed small aggregates at 37°C as mentioned above (Fig. 4A and B), but were motile at a 27°C growth temperature (data not shown).

These results demonstrate that the collagen environment is responsible for flagellum-mediated motility of pYV-negative Y. enterocolitica at 37°C. On one hand, the weak expression of YadA is obviously sufficient to prevent the motile phenotype. On the other, the deletion of yadA alone is not sufficient to restore the motile phenotype. Thus, besides YadA, another factor encoded by the virulence plasmid is involved in the restoration of motility. Furthermore, the functional quorum-sensing system YenR/YenI is required for motility.

Expression of virulence genes of Yersinia in the three-dimensional collagen environment.

By using a yopE-gfp reporter plasmid [WA(pYVO8, pYopE₁₃₈-GFP)], we demonstrated that yopE was expressed in the 3D-CoG at 37°C (Fig. 9), suggesting an intact Yop virulon. As a control, WA(pYVO8ΔvirF, pYopE₁₃₈-GFP) lacking virF showed no fluorescent yersiniae when grown at 37°C in 3D-CoG (data not shown).

FIG. 9.

Expression of yop effector genes of Yersinia in the 3D-CoG. To check the influence of Ca²⁺ on growth behavior and on yopE expression, WA(pYVO8, pYopE₁₃₈-GFP) was grown in normal 3D-CoG medium (A and C1) and in parallel in Ca²⁺-depleted 3D-CoG medium (B and C2). (A) By using a yopE-gfp reporter plasmid [WA(pYVO8, pYopE₁₃₈-GFP)], we demonstrated that yopE is expressed in the 3D-CoG at 37°C. Scale bars, 10 μm. The left panel shows fluorescence; the right panel shows phase contrast. (B) Ca²⁺ depletion in the 3D-CoG medium led to a drastic growth inhibition. Scale bars, 10 μm. The left panel shows fluorescence; the right panel shows phase contrast. (C) Intensity plot of GFP fluorescence of WA(pYVO8, pYopE₁₃₈-GFP) in 3D-CoG medium (C1) and in Ca²⁺-depleted 3D-CoG medium (C2) after 20 h at 37°C. Images were recorded with the same parameter settings to allow quantitative evaluation.

When we grew the yopE-gfp reporter strain in the 3D-CoG at 37°C for 20 h, dp microcolony formation could be observed (Fig. 9A and 9C1). The depletion of Ca²⁺ in the 3D-CoG medium by the addition of 10 mM EGTA and 10 mM MgCl₂ led, on the one hand, to an increase of the fluorescence intensity, indicating an upregulation of yopE expression, and on the other, to a drastic growth inhibition (Fig. 9B and 9C2). Instead of dp microcolonies with 30- to 40-μm diameters, only short chains (Fig. 9B) or very small microcolonies (Fig. 9C2) could be detected. The growth of the pYV-negative WA-C was completely insensitive to EGTA when analyzed in a manner similar to that described above (data not shown). These results demonstrate the well-known features of calcium-dependent growth restriction as well as TTSS activation at 37°C of pYV-positive yersiniae (28, 40) within our model system.

To check the expression state of the yersiniabactin determinant, which is located chromosomally on the HPI and is repressed in the presence of iron, we used strain WA(pYVO8, pFyuA-GFP) carrying a fyuA-gfp reporter plasmid (fyuA encodes the yersiniabactin siderophore outer membrane receptor). As expected, in contrast to the case for the yop genes of pYV, fyuA of the HPI appears not to be derepressed under 3D-CoG growth conditions, indicating a sufficient iron supply (data not shown).

DISCUSSION

The advantage of 3D-CoG for studying the migration and interaction of phagocytic cells with Aspergillus fumigatus conidia and Candida albicans has recently been reported (3). To establish such a tissue-like, three-dimensional infection model for the well-characterized, extracellularly multiplying pathogen Yersinia, we studied the growth behavior and expression of virulence-associated genes of the virulent Y. enterocolitica serotype O8 in 3D-CoG, an in vitro infection model which enables the simulation of in vivo situations. To our surprise, the growth of yersiniae in 3D-CoG revealed novel phenotypic characteristics controlled by the virulence plasmid pYV (a summary is shown in Table 2). These characteristics have not been described before and are probably important for improving our understanding of Yersinia pathogenicity.

First, in contrast to the repression of motility of Y. enterocolitica at 37°C on agar-solidified medium, we found motile yersiniae in 3D-CoG at 37°C in the absence of pYV. The production of YadA or YadA variants abrogated motility and provoked microcolony formation (dp microcolonies are shown in Fig. 1 and 5), suggesting a strong correlation between yadA expression and dp microcolonies. Furthermore, we demonstrated that motility is dependent on a functional flagellar motor, as shown by growth analysis of a motAB mutant in the 3D-CoG (WA-CΔmotAB [Fig. 7B]). Contrary to the case for WA-C, WA-CΔmotAB was immotile at 37°C within the 3D-CoG (Fig. 8B). Moreover, we could also demonstrate indirectly a functional quorum-sensing system, YenR/YenI, because WA-CΔyenIR grown in the 3D-CoG was found to be immotile as expected (2).

Second, we demonstrated that the packing density of microcolonies in the 3D-CoG is YadA dependent. YadA-producing yersiniae formed dp microcolonies, in contrast to the case for pYVΔyadA-carrying Yersinia mutants, which showed growth in lp microcolonies (Fig. 5). Further studies with strains producing different YadA variants point out that the head domain of YadA obviously is crucial for the packing density. Up to now, we have known that the N-terminal head domain of YadA is involved in binding to proteins of the ECM (fibronectin, laminin, and collagen) or to epithelial cells. In addition, mutants lacking the head domain lost the capacity to mediate autoaggregation/autoagglutination (21, 32). The aggregation of bacteria in stationary cultures due to YadA-mediated hydrophobic interactions is specified as autoagglutination (39). In our experiments, only the complete deletion of the head domain resulted in the reduced packing density of the microcolonies in the 3D-CoG. Neither the substitution of the two histidyl residues (H156 and H159) by tyrosine residues in YadA (abrogation of collagen binding [Fig. 5B]) nor the deletion of amino acids 29 to 81 of YadA (responsible for adherence to neutrophils [Fig. 5C]) influenced the packing density of microcolonies in the 3D-CoG. These findings suggest a strong correlation between the autoagglutination phenotype and dp microcolonies. Furthermore, we conclude that the collagen binding property of YadA is not required for dp microcolony formation.

Third, YadA-producing yersiniae formed dp microcolonies, contrary to the case for pYVΔyadA-carrying Yersinia mutants, which showed growth in lp microcolonies. This finding suggests the possibility of a further pYV-encoded determinant, which abolishes motility in the absence of yadA expression. As WA-C(pLCR) carrying the minivirulence plasmid with a functional TTSS (but lacking yadA and yop genes) is motile in 3D-CoG at 37°C, this pYV-encoded determinant is presumably not located within the plasmid region encoding the TTSS needle structure. At the moment, we are in search of this factor regulating motility.

Fourth, when we used a yopE-gfp reporter, at least one of the Yersinia TTSS-encoded effector genes, yopE, was expressed under the terms of growth in the 3D-CoG (Fig. 9), suggesting an intact yop regulon. The depletion of Ca²⁺ during growth of yersiniae in the 3D-CoG led to a drastic growth inhibition and to an induction of YopE, indicating a successfully operating TTSS. It is well known that growth restriction as well as TTSS activation at 37°C of pYV-positive yersiniae is calcium dependent (28, 40), and we demonstrated these essential features within our model system.

Interestingly, contrary to in vivo mouse experiments (25), the yersiniabactin system appeared to be down-regulated in the 3D-CoG, as demonstrated by the lack of green fluorescence of yersiniae carrying the fyuA-gfp reporter. The latter result is not surprising, because under the supplied growth conditions in the 3D-CoG, yersiniae do not starve for iron and there is obviously no need for the expression of the iron uptake system.

Furthermore, the growth of wild-type yersiniae in chains, well known from attached cell monolayer experiments, is also seen here in our newly established in vitro 3D-CoG infection model. The growth in chains seems to be responsible for microcolony formation. The dp microcolony formation within the 3D-CoG reminds us of the microabscess formation in the mouse oral infection model, where yersiniae are known to replicate predominately extracellularly and to form microcolonies/microabcesses (38, 42). In addition, the 3D-CoG in vitro infection model enables us to study Yersinia-host interactions and Yersinia growth phenotypes (lp-microcolony-forming as well as singly growing strains), which cannot be studied within the oral mouse infection model, because these strains do not survive within the host (34).

Like biofilms, dp microcolonies can possibly be taken as a defense against phagocytes, which try to attack Yersinia clusters, but this possibility has to be proven by further experiments. Concerning the defense against phagocytes, the fibrillar layers on the surfaces of the yersiniae within the dp microcolonies (Fig. 1C2) could play an important role as well. The observed layer is evocative of the Myf fibrillae which appeared as a layer of extracellular material extending locally 2 μm from the bacterial surface (23). But analyzing the surface structure of a myf deletion mutant [WAΔmyf(pYVO8)] within the 3D-CoG by electron microscopy, we have to conclude that the fibrillar layer is not the Myf fibrillae. Thus, the fibrillar structures are probably novel appendages not described yet. Future studies are necessary to clarify their identity.

To demonstrate the three-dimensional organization within our model system (3D-CoG), we used confocal reflection contrast imaging combined with fluorescence microscopy to visualize collagen fibers and Yersinia cells (Fig. 1B1 and B2). The collagen fibers in the 3D-CoG appeared to be ignored by YadA-expressing yersiniae (Fig. 1B2). Most strikingly, we could not observe any influence of the YadA collagen binding capacity on growth behavior.

Bacterial pathogens that invade host tissue are recognized by the host defense system and elicit the migration of professional phagocytes toward the infectious agent. In contrast to the case for two-dimensional cell monolayer infection models, where microorganisms are moving toward target cells, 3D-CoG provides us with the possibility to simulate the host tissue environment with host cells migrating toward the pathogen. The 3D-CoG model was chosen because it is regarded as useful in migration studies and analysis of cell-cell contact in immunology as well as tumor biology (3, 9, 15, 17). The collagen matrix simulates the ECM of the tissue and provides a three-dimensional environment for the cells. As a result, a three-dimensional organization of cells and the formation of cell-cell contacts is possible. Therefore, physiological processes can approximately be compared to in vivo situations. Future studies will deal with the impact of different Yersinia effector proteins on migration and the means of phagocyte-pathogen interaction within the 3D-CoG as well as the fate of the phagocytes after pathogen contact.

In conclusion, this study revealed for the first time unique plasmid-dependent growth behavior of yersiniae in a three-dimensional matrix environment which appears to be closer to that of yersiniae in infected mouse tissue (29) than in two-dimensional cell culture systems. Thus, this three-dimensional culture system may be a first step to a more complex level of in vitro infection systems mimicking living tissue.