Bacterial Quorum-Sensing Regulation Induces Morphological Change in a Key Host Tissue during the Euprymna scolopes-Vibrio fischeri Symbiosis

ABSTRACT Microbes colonize the apical surfaces of polarized epithelia in nearly all animal taxa. In one example, the luminous bacterium Vibrio fischeri enters, grows to a dense population within, and persists for months inside, the light-emitting organ of the squid Euprymna scolopes. Crucial to the symbiont’s success after entry is the ability to trigger the constriction of a host tissue region (the “bottleneck”) at the entrance to the colonization site. Bottleneck constriction begins at about the same time as bioluminescence, which is induced in V. fischeri through an autoinduction process called quorum sensing. Here, we asked the following questions: (i) Are the quorum signals that induce symbiont bioluminescence also involved in triggering the constriction? (ii) Does improper signaling of constriction affect the normal maintenance of the symbiont population? We manipulated the presence of three factors, the two V. fischeri quorum signal synthases, AinS and LuxI, the transcriptional regulator LuxR, and light emission itself, and found that the major factor triggering and maintaining bottleneck constriction is an as yet unknown effector(s) regulated by LuxIR. Treating the animal with chemical inhibitors of actin polymerization reopened the bottlenecks, recapitulating the host’s response to quorum-sensing defective symbionts, as well as suggesting that actin polymerization is the primary mechanism underlying constriction. Finally, we found that these host responses to the presence of symbionts changed as a function of tissue maturation. Taken together, this work broadens our concept of how quorum sensing can regulate host development, thereby allowing bacteria to maintain long-term tissue associations.

Plasmid and mutant construction. Primers used to create V. fischeri expression and gene deletion plasmids are listed in Table 1. Genomic insertion of ainS or luxI under control of their respective endogenous promoters at the genomic attTn7 site was performed as previously described using a mini-Tn7 vector (4). Briefly, ainSp-ainS or luxIp-luxI were cloned from the V. fischeri genome and inserted into pEVS107 via AvrII and SpeI sites. The transposition helper plasmid pUX-BF13 (5) was used to insert the pEVS107 construct at attTn7. The luxI expression vector was created by cloning luxI from the V. fischeri genome, and inserting it into pVSV105 (6) via XbaI and KpnI sites. Deletion of luxIR and replacement with the modified lac promoter PA1/O4/O3 upstream of luxCDABEG was performed similarly, as previously described (7). Briefly, ~1-kb fragments surrounding luxIR were cloned from the V. fischeri genome and fused via EcoRI and KpnI sites to either end of PA1/O4/O3 cloned from pAKD601 (8). The fused fragment was inserted via BamHI and SacI sites into pSMV3 (9), which carries kanamycin-resistance and sacB cassettes. Counter-selection to remove pSMV3 from the V. fischeri genome and replace luxIR with PA1/O4/O3 was performed on high-sucrose, low-salt LB plates (7) at room temperature.

Colonization competition between strains carrying fluorescent labels on plasmids.
To determine whether carriage of either the plasmid backbone of gfp-carrying pVSV102 or the fluorescent markers CFP and YFP conferred a competitive disadvantage on V. fischeri strains during colonization, we competed the WT strain expressing each of the labels against each other, as well as the dark Dlux strain (DluxCDABEG) carrying the markers, and found no effect of the carriage of the plasmid with either label. Competition experiments were performed by exposing juveniles to two strains with a ~1:1 inoculum (unless otherwise stated), using at least identified in the plating assays by counting colonies under light passed through filter sets that revealed CFP or YFP using a dissecting microscope (NightSea LLC, Lexington, MA). To ensure that the competition defect of DluxIR lacZp-lux in colonization experiments was not due to interactions between strains in culture, we carried out the competitions (described in Fig. 5) in vitro as well. Sample fixation and microscopy. At the endpoint of colonization assays, juveniles were dropped into 4% paraformaldehyde in marine phosphate-buffered saline (mPBS; 0.45 M NaCl in a 50 mM sodium phosphate buffer, pH 7.4) and fixed overnight at 4 °C with rotation. Following three washes with mPBS, light organs were dissected and permeabilized and stained in 0.1% Triton-X 100 stained with rhodamine phalloidin for F-actin and TO-PRO-3 to label nuclei for two days at 4 °C with rotation. Briefly, washed samples were mounted and imaged on a Zeiss LSM 710 (Carl Zeiss AG, Jena, Germany) confocal microscope as well as a Leica SP8 X confocal microscope (Leica Camera AG, Wetzlar, Germany) as described previously (11). Each side of a light organ (crypts 1-3) was considered one set of measurements. For image analysis, Fiji (ImageJ) (12) was used to generate Z-projections of sub-stacks to generate an unobstructed view of the bottleneck portion of tissue, using the phalloidin channel. In the software, a straight line was drawn between the F-actin rich, terminal web at the narrowest point of the bottleneck structure and used as the estimation of the diameter. While the average wild-type bottleneck diameter at early stages of symbiotic development was reduced to 3 μm or less (Fig. 2B), we used a conservative cutoff of >5 μm to operationally define any response to a mutant as different from wild type when comparing across experiments; i.e., the mean response to wild type was always a constriction to less than 5 μm (Table 2).
For intensity measurements to estimate abundance of V. fischeri labeled with green fluorescent protein (GFP) in specific tissue regions, all light organs were imaged with the same laser intensity and gain on the detector, and done on the same day. In Fiji, projections of the GFP channel were used with the freeform tool to draw the regions of migration path (MP) 1 and crypt (C) 1 (regions shown in Fig. 7A). The final intensity measurement used was the mean intensity per unit area of each specified region with the background (the fluorescence of tissue away from the two regions of interest) subtracted.
Statistical analysis. Data were analyzed using GraphPad Prism software, version 7.0 (GraphPad Software, Inc., La Jolla, CA), and were first tested for normality using a D'Agostino-Pearson (omnibus K2) test. When data passed this criterion (P > 0.05), parametric tests including a one-way ANOVA or two-way ANOVA with time as a factor were used to examine differences between the groups. When the difference was considered significant, then a Tukey or Holm's post-hoc test with a one-way ANOVA or a Sidak's multiple comparison's test for a two-way ANOVA was performed. If data did not pass the normality test, a Kruskal-Wallis test and Dunn's post-hoc test were used to determine differences between groups. Significance indicated by asterisks as follows: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.
Mathematical modeling. Data were analyzed using the program R (13) and the package for linear mixed models (lme4) (14) as described previously (15). Three sets of data were analyzed using mathematical modeling approaches. To analyze the data from the co-inoculation with WT, DluxCDABEG (dark mutant), and DluxIR lacZp-lux (signaling but luminous mutant), the strains were scored in terms of their LuxI-LuxR functionality (Table S1; see data in Figs. 5, S4). Strains with LuxIR functionality included WT and DluxCDABEG, whereas DluxIR lacZp-lux was excluded from this group. We used a nearest neighbor analysis to determine if there was an effect of LuxIR functional strains in nearby crypts to a given focal crypt. We treated an adjacent crypt on the same side of the light organ and the opposing side's crypt 1 when evaluating a focal crypt 1 due to the proximity of the blind ends of each C1. Data were log-transformed and a linear mixed model fit by maximum likelihood was used to compare across all combinations of focal crypt function for neighborhood effect.
A third analysis tested differences in the luminescence production (light functionality) of strains to assess if the restoration of light in the DluxIR lacZp-luxCDABEG strain had the same impact on the bottleneck as did wild-type light (Table S1). Empty or DluxCDABEG-colonized crypts were scored as non-luminous, whereas wild-type light and the restored light of DluxIR lacZp-luxCDABEG were categorized separately. We used a mixed model to compare whether treating light from both strains as the same (only two classifications, absent or present) was a better characterization or if the model fit better when light was split into the three groups (absent, present, or restored). A linear mixed model was fit by maximum likelihood framework.
We assessed fit by AIC with a broad distinction where a DAIC of 2 indicated a significantly better fit.