A universal oyster infection model demonstrates that Vibrio vulni ﬁ cus Type 6 secretion systems have antibacterial activity in vivo

With the rapid increase of aquaculture contributing to sustainable food security, comes the need to better understand seafood associated diseases. One of the major aquatic bacterial genera responsible for human infections from seafood is Vibrio , especially from oysters. Currently, in vivo study of bacterial interactions within oysters is limited by the inability to promote high-level uptake of bacteria by oysters. This study has therefore evolved current natural marine snow protocols to generate ‘ arti ﬁ cial ’ marine snow, into which bacteria can be incorporated to facilitate extensive uptake by oysters. This presents an adapt-able model for bacterial study within ﬁ lter-feeding shell ﬁ sh. Using this model, we demonstrate for the ﬁ rst time the antibacterial activity of Vibrio vulni ﬁ cus Type 6 secretion systems in vivo , revealing an impor-tant role for the T6SS in V . vulni ﬁ cus ecology.


Introduction
The State of the World Fisheries and Aquaculture (SOFIA) report (2018) describes 80 million tonnes of aquaculture produced in 2016, of which molluscs totalled 17.1 million tonnes (SOFIA, 2018). Comprising 47% of global fish and mollusc production, fisheries and aquaculture are paramount food security concerns for ensuring continued and consistent seafood production. However, aquaculture practices are becoming increasingly intensive to meet this demand, resulting in proliferation of marine pathogens and food safety issues (Bondad-Reantaso et al., 2005;Rico et al., 2012). Vibrio species are a major hurdle in bivalve production, in particular oysters, due to their mortality in early developmental stages and their deleterious impact on adult shellfish (Dubert et al., 2016;Bruto et al., 2017). Seasonal losses to the oyster pathogen, Vibrio splendidus, have consistently reached up to 80% in oyster farms in the Bay of Morlaix over a 10-year span (Lacoste et al., 2001). Furthermore, the Vibrionaceae family harbours numerous human pathogens, including Vibrio vulnificus, Vibrio cholerae and Vibrio parahaemolyticus, which are responsible for around 80,000 vibriosis cases annually in the United States (CDC, n.d.). The investigation of Vibrio ecology is therefore essential for understanding how to best manage the threat they pose to aquaculture and human health.
Studies exploring bacterial behaviour within shellfish have previously utilized oysters inoculated through exposure to water supplemented with bacterial culture (Srivastava et al., 2009;Pu et al., 2018). However, this approach is limited by inefficient uptake of bacteria as oysters filter and ingest particles based primarily on size . Planktonic V. vulnificus (2 μm) fall below the optimal size range for ingestion, therefore only achieving low levels of uptake in such experiments (Froelich and Noble, 2014). Incorporation of bacterial cultures into larger, naturally forming aggregates known as 'marine snow' has shown potential as a method for enhancing bacterial uptake by oysters . Marine snow is composed of diatoms, faecal matter, microorganisms, protists and other miscellaneous elements brought together at random that present a unique environment for bacterial colonization and survival. However, this stochastic process generates inherent variability (Kiørboe et al., 1990;Grossart et al., 2003;Kiørboe et al., 2003;Passow et al., 2012).
Vibrio vulnificus is the leading cause of seafoodrelated disease in the United States and is increasing in prevalence globally, correlating with rising seasurface temperatures (Jones and Oliver, 2009;Baker-Austin et al., 2012;Vezzulli et al., 2015;Baker-Austin et al., 2017;Deeb et al., 2018). Ingestion can result in severe gastroenteritis which may develop into primary septicaemia, with symptoms including vomiting, haemorrhagic bullae, secondary lesions on extremities and severe hypotension (Beatty et al., 2017). Progression to primary septicaemia results in 50% patient fatality within 48 h of hospital admission (Chiang et al., 2003). Exposure of open wounds to contaminated water results in erythema, haemorrhagic bullae, cellulitis and intense pain (Oliver, 2005). Failing immediate treatment, infected lesions frequently turn necrotic. While visually more aggressive, wound infections carry a lower mortality rate of 25% compared with ingestion whose rates are >50% if treated within 24 h and 100% if untreated for over 72 h (Jones and Oliver, 2009;Kim et al., 2011). Particularly at risk are patients with compromised immune systems, liver disease and haematological disorders, with 80% of patients already presenting at least one pre-existing medical condition (Menon et al., 2014).
Studies comparing the genetic makeup of V. vulnificus strains identified two Type 6 secretion systems (T6SSs) labelled T6SS1 and T6SS2 (Church et al., 2016). A molecular syringe composed of 13 essential proteins, the T6SS functions to inject cytotoxic effectors into neighbouring cells (Ho et al., 2014). Shown to have a wide range of activity in other bacterial species, the T6SS has been implicated in both anti-prokaryotic and anti-eukaryotic activity (Salomon et al., 2013;Ray et al., 2017;Trunk et al., 2018;Berni et al., 2019). Interestingly, whilst the T6SS2 was identified in all V. vulnificus strains sequenced to date, only some strains possessed the T6SS1. in vitro competition assays showed that T6SS1-positive (T6SS1 + ) strains were able to target and kill T6SS-negative (T6SS1 − ) strains as well as other bacterial species. T6SS-mediated antibacterial activity has been demonstrated by a range of bacteria, such as Klebsiella pneumoniae, V. cholerae and Vibrio fluvialis, in various in vivo models including mice, rabbits and squid (Fu et al., 2018;Speare et al., 2018;Hsieh et al., 2019). It was thus hypothesised that the T6SS1 may play a significant role in competition between Vibrios, and on bacterial population dynamics in vivo.
Here we demonstrate that V. vulnificus T6SS1 and T6SS2 mediate antibacterial activity in an oyster in vivo model. Killing was shown to be thermoregulated with only T6SS1 active at 30 C, but both T6SS1 and T6SS2 active at 21 C. This is the first study to present a role for the T6SS2 of V. vulnificus. For in vivo studies, we have adapted current marine snow methodologies to develop a defined, highly controllable 'artificial marine snow' (AMS) based on aggregation of the diatom species, Thalassiosira pseudonana. Using inoculated AMS, we achieved higher uptake of V. vulnificus and Salmonella enterica serovar Enteritidis by oysters than any study to date. We believe this methodology may have much broader application within research and aquaculture. Co-culture of strains within oysters showed that T6SS1 and T6SS2 mediate intra-and inter-species competition in vivo and may play a pivotal role in the establishment of V. vulnificus strains within filter-feeding shellfish and subsequent prevalence of virulent strains.

Results
Vibrio vulnificus T6SSs display no anti-eukaryotic activity against Dictyostelium discoideum The T6SS of V. cholerae demonstrates a range of antieukaryotic functionality, such as actin cross-linking to inhibit phagocytosis by macrophages (Pukatzki et al., 2007). We hypothesised that V. vulnificus T6SSs may also possess anti-eukaryotic activity. To explore this hypothesis, wild-type V. vulnificus 106-2A (T6SS1 + / T6SS2 + ) and mutants of both T6SSs were employed in Dictyostelium discoideum plaque-forming assays. T6SS1 and T6SS2 mutants were generated through inframe deletion of hcp1 and hcp2 genes encoding effector-transporting needle components of the T6SS (Fig. 1A) (Cascales and Cambillau, 2012). Mutants were labelled ΔT6SS1 (hcp1 − ) and ΔT6SS2 (hcp2 − ) respectively. Deletion of hcp1 and hcp2 was confirmed via PCR and expression of downstream genes verified by reverse-transcriptase-PCR (RT-PCR) ( Fig. 1B and C). For D. discoideum plaque-forming assays, wild-type V. vulnificus 106-2A and T6SS deletion mutants were combined with predatory D. discoideum AX3 and plated as a lawn on SM agar for 3-5 days at 21 C. If the T6SSs possess anti-eukaryotic activity, larger plaques of predation should be observed relative to wild-type V. vulnificus 106-2A following their deletion. No plaques were observed on plates containing wild-type V. cholerae V52, a strain with previously characterized anti-eukaryotic activity (Fig. 1D) (Miyata et al., 2011). However, visible plaques were recorded on V. vulnificus 106-2A, ΔT6SS1 (hcp1 − ) and ΔT6SS2 (hcp2 − ) lawns, indicative of V. vulnificus predation by D. discoideum ( Fig. 1E-G). Measurement of plaque diameter showed no significant difference in predation efficiency between V. vulnificus strains (Fig. 1H). These results indicate that the T6SSs of V. vulnificus are unable to target D. discoideum and have no observable anti-eukaryotic activity at 21 C in this model of infection. A similar lack of antieukaryotic activity has also been previously demonstrated in a Galleria mellonella model of virulence at 37 C (Church et al., 2016 Both V. vulnificus T6SS1 and T6SS2 demonstrate in vitro antibacterial activity at environmental temperatures Given the apparent lack of anti-eukaryotic phenotype from either of the T6SSs of V. vulnificus against D. discoideum, we sought to further investigate previous observations of V. vulnificus intra-and inter-species competition mediated by the T6SS (Church et al., 2016). As T6SS activation in V. cholerae has been identified at 23 C, we aimed to explore whether V. vulnificus also exhibits T6SS activity at environmental temperatures (Ishikawa et al., 2012). Competition assays were thus conducted at 21 C, an optimal temperature for V. vulnificus growth in sea water (Kaspar and Tamplin, 1993). To facilitate enumeration of strains in competition, attacker strains V. vulnificus 106-2A (wildtype), ΔT6SS1 (hcp1 − ) and ΔT6SS2 (hcp2 − ) were transformed with pVv3-Kan to confer kanamycin resistance (Klevanskaa et al., 2014). Likewise, an environmental 'prey' V. vulnificus strain, 99-743, (T6SS1 − , T6SS2 + ), was transformed with pVv3-Tmp to confer trimethoprim resistance.
Competition between attacker and prey V. vulnificus confirms T6SS1 activity at 30 C, demonstrated by a clear and significant 1000-fold reduction of prey 99-743 following co-culture with 106-2A but no reduction when cultured with ΔT6SS1 (hcp1 − ) ( Fig. 2A). We therefore sought to see if a more environmentally representative temperature had an effect on T6SS activity. Competition assays were thus conducted at 21 C for both T6SS deletion strains (Fig. 2B). We observed a significant 10-fold reduction in 99-743 prey counts following individual co-culture with 106-2A and the ΔT6SS2 (hcp2 − ) mutant at 21 C. In addition, at 21 C we also observed for the first time a T6SS2 antibacterial activity, manifested as a reduction in prey numbers comparable with that observed for the T6SS1 (p < 0.05) (Fig. 2B). To confirm that the observed reduction in prey numbers were indeed T6SS-dependent and not the result of alternative antibacterial mechanisms active at 21 C, a 106-2A Δhcp1Δhcp2 double mutant was generated, named ΔT6SS1/2. No reduction in recovery of the prey strain 99-743 was observed following coculture with ΔT6SS1/2 (hcp1 − , hcp2 − ). These data confirm that killing by ΔT6SS1 (hcp1 − ) and ΔT6SS2 (hcp2 − ) was due to the T6SS2 and T6SS1, respectively, and that both systems are capable of intra-species targeting at 21 C in vitro. Given the broad nature of T6SS activity, we investigated whether the T6SS competent strains of V. vulnificus were able to demonstrate inter-species antibacterial activity. To address this question, we utilized Salmonella Enteritidis CC012 as a prey strain in competition with our panel of attacker V. vulnificus T6SS mutants. Similar to our intra-species data, we observed a T6SS1-driven 1000-fold reduction of prey CC012 at 30 C but no reduction with the ΔT6SS1 (hcp1 − ) attacker strain (Fig. 2C). Of interest, the competition assays at 21 C demonstrated that CC012 was targeted by 106-2A, ΔT6SS1 (hcp1 − ) and ΔT6SS2 (hcp2 − ), resulting in significantly reduced prey recovery (p < 0.01) (Fig. 2D). This reduction was not observed when co-cultured with the double mutant, ΔT6SS1/2 (hcp1 − , hcp2 − ), reinforcing the role of V. vulnificus' T6SS2 in bacterial killing at environmental temperatures.

Development of a delivery model to facilitate reproducible uptake of bacteria by oysters
Having demonstrated V. vulnificus T6SS activity in vitro, we sought to investigate whether such activity occurred in vivo by developing an oyster model for studying bacterial interactions. Previous in vivo models of bacterial uptake utilizing oysters have either employed directly inoculated sea water or 'natural marine snow' (NMS) (Srivastava et al., 2009;Pu et al., 2018). Our attempt at incorporating V. vulnificus into NMS, formed by aggregating seawater particulate matter, showed some success but did not achieve incorporation >3 × 10 4 CFU/g marine snow (Fig. 3A). This may be due to the competition from the naturally occurring bacteria that pre-colonize the particulate matter preventing V. vulnificus establishment. We showed that NMS contained diverse bacterial populations, including Vibrio species, as determined by growth on both LB and Thiosulfate-Citrate-Bile Salts-sucrose (TCBS) agar plates (Fig. 3A). Moreover, our NMS also varied by size and composition (Fig. S1). Given that the levels of were ligated for construction of in-frame deletion cassettes. Deletion cassettes were introduced to V. vulnificus 106-2A using pDM4 to generate ΔT6SS1 (hcp1 − ) and ΔT6SS2 (hcp2 − ). Truncated hcp1 and hcp2 genes from V. vulnificus ΔT6SS1 and V. vulnificus ΔT6SS2 are shown below with confirmatory sequencing data. Start and stop codons for hcp1 and hcp2 are highlighted with red boxes. B. PCR confirmation of hcp1 and hcp2 deletion. Mmolecular marker (bp), 1no template control, 2 -hcp1 positive control, 3 -ΔT6SS1 mutant, 4 -hcp2 control, 5 -ΔT6SS2 mutant. Lanes 2-5 also exhibit a species-specific vvhA band at 500 bps to confirm mutants are V. vulnificus. C. Expression of genes downstream of hcp1 and hcp2 (vgrG1 and clpV respectively) was confirmed using RT-PCR. Mmolecular marker (bp), 1no template control, 2no reverse-transcriptase gDNA control, 3 -106-2A wild-type positive vgrG1 control, 4 -106-2A wildtype positive clpV control, 5 -ΔT6SS1 vgrG1, 6 -ΔT6SS2 clpV. D. Dictyostelium discoideum cultures were combined with Vibrio cholerae V52, V. vulnificus 106-2A, V. vulnificus ΔT6SS1 and V. vulnificus ΔT6SS2. Example plaques are indicated with white arrows. E. Measurement of D. discoideum plaques (n = 4) following growth on V. vulnificus 106-2A, ΔT6SS1, ΔT6SS2 and V. cholerae V52. Nsnot significant. incorporation into NMS were <3 × 10 4 CFU/g marine snow and Vibrio levels in oysters can be up to 5 × 10 4 CFU/g tissue in areas where V. vulnificus infections are prevalent, it was clear that uptake material containing far higher numbers of Vibrio was required (Froelich et al., 2017).
To facilitate reproducible, high-level uptake of bacteria by oysters, we developed an AMS model, exploiting a natural food source for oysters. Commercial shellfish are often reared on microalgae concentrates and we therefore investigated whether bacteria could be rationally incorporated into this material. Essentially, our AMS model is the aggregation of the diatom species T. pseudonana CCMP 1335, with the subsequent incorporation of bacteria as detailed in the Experimental procedures section. This model reliably and reproducibly demonstrated levels of bacterial incorporation of >10 7 CFU/g AMS (Fig. 3B). These levels were observed for V vulnificus 106-2A, 99-743 and S. Enteritidis CC012.
To control for the presence of contaminating bacteria in the T. pseudonana, AMS was plated on LB, Vibrioselective TCBS and Salmonella-selective ChromoSelect agar. The remarkably high levels of bacterial incorporation into AMS, observed by plate count enumeration, contrasted starkly to the levels of incorporation into the NMS, the latter being of four orders of magnitude less than the >10 7 CFU/g AMS levels. Both V. vulnificus 99-743 and S. Enteritidis CC012 demonstrated significantly greater incorporation into AMS than V. vulnificus 106-2A (p < 0.05 and p < 0.0001 respectively). Given that we want to explore intra and inter-species competition, it was necessary to generate AMS with similar levels of bacterial strains. This was achieved simply by inoculating the starting material with appropriately higher numbers of bacterial cultures, resulting in AMS of consistent bacterial loads (Fig. 3C). Co-incorporation of attacker and prey bacteria into AMS confirmed that there was no significant killing of prey bacteria occurring in AMS that may impact . Prey counts following coculture with attacking strains were statistically compared with prey counts following co-culture with just LB using a one-way analysis of variance with Dunnett's corrections for multiple comparisons. * p < 0.05, ** p < 0.01.   = 3), error bars show the standard deviation. Control oysters exposed to bacteria-free AMS were plated on LB, TCBS and ChromoSelect agar. Samples in which bacterial load fell below the limit of detection of 100 CFU/g oyster tissue are indicated by a downward arrow. I. Bacterial ingestion was measured over 48 h. Oysters were enumerated at T 0 , T 4 , T 8 , T 24 and T 48 h (n = 3). Error bars represent the standard deviation. Statistics were performed using either a one-way analysis of variance (ANOVA) with Dunnett's corrections for multiple comparisons ( Fig. 1A-C) or a nested one-way ANOVA with Dunnett's corrections for multiple comparisons (Fig. 1H). Nsnot significant, * p < 0.05, ** p < 0.01, **** p < 0.0001.
downstream assays (Fig. S2). AMS-incorporating bacteria were also imaged via scanning electron microscopy (SEM). Vibrio vulnificus 106-2A (blue), V. vulnificus 99-743 (pink) and S. Enteritidis CC012 (green) are clearly incorporated into T. pseudonana AMS (Fig. 3D-G). Given the effectiveness of the developed AMS model, the next step was to investigate the suitability of this material for uptake by oysters. Oysters were exposed to inoculated AMS for 24 h at 21C. Following exposure, oyster stomachs were removed and homogenized for bacterial enumeration by plate counts. Figure 3H presents viable cell counts of V. vulnificus 99-743, 106-2A and S. Enteritidis CC012 in CFU/g tissue. To account for oyster variability, six oysters were utilized per replicate (n = 3). Control oysters were exposed to bacteria-free AMS and plated on LB, TCBS and ChromoSelect agar. Results showed low levels of naturally occurring microorganisms in all tested oysters (10 2 -10 4 CFU/g tissue).
Colony counts on TCBS and ChromoSelect agar determined that natural Vibrio and Salmonella levels were typically <1000 CFU/g and <100 CFU/g tissue respectively. Reassuringly, colony counts following exposure to inoculated AMS demonstrated highly successful uptake of both V. vulnificus and S. Enteritidis with levels of between 10 6 and 10 8 CFU/g tissue consistently recovered. Furthermore, no strain demonstrated significantly greater uptake than other strains being tested. Oyster uptake of AMS was assessed over 48 h (Fig. 3I). Rapid initial uptake was observed for all strains, with oysters showing 10 5 -10 7 CFU/g tissue at T 4 . Ingested concentrations rose to 10 7 -10 8 CFU/g tissue at T 8 ; however, between T 8 and T 48 there was no increase in bacterial uptake. While we observed 10 2 CFU/g tissue at T 0 consisting of naturally occurring bacteria in the negative controls, this was not deemed a problem due to the significantly higher volumes resulting from exposure to inoculated aggregates.

Demonstration of controlled intra and inter-bacterial species killing in vivo facilitated by AMS
The aim of this study was to replicate in vitro competition assay data in an in vivo oyster model utilizing our AMS. Oysters were exposed to bacteria-free AMS and AMS containing individual attacker and prey bacterial cultures at 21 C for 24 h. These controls were enumerated by plating on LB and TCBS agar ( Fig. 4A and B). Single culture controls demonstrated no significant difference in uptake between the attacker and prey strains. For the killing assays, oysters were exposed to separate AMS suspensions inoculated with either attacker or prey strains at final attacker to prey ratios of 5:1 and 10:1. Bacterial incorporation into AMS was adjusted to ensure oysters were exposed to the required attacker to prey ratios. in vivo co-culture of the prey strain 99-743 with either ΔT6SS1 or ΔT6SS2 resulted in significantly reduced recovery of 99-743 compared with the 99-743-only control oyster cohort (p < 0.001) (Fig. 4A). Typically, there was a reduction in viable counts from >10 7 to <10 6 CFU/g tissue. These reductions in numbers of prey are similar to those obtained with wild-type 106-2A as the attacker strain. To confirm that these in vivo levels of reduction of prey bacteria manifested by ΔT6SS1 and ΔT6SS2 were not attributable to alternative killing mechanisms, the assay was conducted using the double T6SS mutant, ΔT6SS1/2. Excitingly, we saw no significant impact on 99-743 viability when challenged with the ΔT6SS1/2 attacker strain compared with the wild-type 106-2A attacker strain. This striking observation demonstrates that the killing observed by attacker strains ΔT6SS1 and ΔT6SS2 is a consequence of the remaining T6SS. These data strongly support the hypothesis that both the T6SSs of V. vulnificus play a role in intra-species targeting in vivo. To further explore these results, we repeated the assay with an increased ratio of attacker to prey of 10:1 (Fig. 4B). This assay resulted in an even greater reduction of 99-743 when challenged with 106-2A, ΔT6SS1 and ΔT6SS2, typically a >1000-fold reduction from >10 7 to <10 4 CFU/g tissue (p < 0.001).
As we had previously observed that V. vulnificus is able to kill S. Enteritidis in addition to other species of Vibrio in a T6SS-dependent manner in vitro, we aimed to determine whether such a phenotype could be observed in vivo (Church et al., 2016). Thus, to test inter-species targeting of the V. vulnificus T6SSs, we conducted in vivo co-culture assays employing S. Enteritidis CC012 as the prey strain at an attacker to prey ratio of 5:1 (Fig. 4C). Similar to the killing of V. vulnificus 99-743 seen previously, the attacker strain V. vulnificus 106-2A was able to reduce S. Enteritidis levels from >10 7 to <10 6 CFU/g tissue following in vivo coculture. Both attacking strains, ΔT6SS1 and ΔT6SS2, also killed S. Enteritidis to comparable levels as the wild-type attacker strain 106-2A. Fascinatingly, the ΔT6SS1/2 attacker strain had no observable killing phenotype towards the S. Enteritidis prey strain. As previously noted for 99-743, increasing the attacker to prey ratio from 5:1 to 10:1 resulted in a significant, dose-dependent reduction in S. Enteritidis within oyster stomach tissue, with levels being reduced from >10 7 to <10 4 CFU/g oyster tissue (p < 0.05) (Fig. 4D). In this study, we have demonstrated that V. vulnificus strains are capable of killing neighbouring bacterial species using both T6SS1 and T6SS2 at 21 C in vivo as demonstrated in vitro. Competition with a ΔT6SS1/2 mutant confirmed that this killing was T6SS-dependent.

Discussion
Relevant in vivo models are the idyll of studying hostbacteria interactions. While previous studies have attempted to generate oyster models of bacteria uptake, these models have suffered from reproducibility issues, along with an inability to achieve sufficiently adequate levels of uptake (Morrison et al., 2011;Morrison et al., 2012;). Our unique model of AMS, used here as a vector for delivery of bacteria to oysters, provides a robust and relevant model for studies requiring bacteria-oyster interaction. We utilized this model to demonstrate its suitability for investigating intra and inter-species bacterial killing, reporting for the first time an in vivo phenotype for the T6SSs of V. vulnificus.
The roles of T6SSs of bacteria are multifaceted, providing a selective advantage to bacteria possessing these systems over those without, through the secretion of 'effector' proteins involved in disabling target cells. Several bacterial pathogens, such as V. cholerae, also secrete effectors that are involved in virulence, although no one factor is uniquely correlated with virulence (Pukatzki et al., 2006;Burtnick et al., 2011;Baker-Austin and Oliver, 2018). Given that the T6SS2 is present in all sequenced strains of V. vulnificus to date, we assessed the potential role of this T6SS in virulence in a D. discoideum eukaryotic model at 21 C (Solomon et al., 2000). As shellfish digestive glands are highly polymicrobial, we hypothesised that T6SS1 and T6SS2 may target eukaryotic microorganisms at this lower temperature. Anti-eukaryotic T6SS activity has been observed in a number of microorganisms, such as the antifungal T6SS effectors, Tfe1 and Tfe2, of Serratia marcescens (Trunk et al., 2018). However, our data demonstrated that unlike V. cholerae, neither the wild-type nor the T6SS mutants of V. vulnificus exhibited any anti-eukaryotic for 24 h at 21 C at attacker to prey ratios of 5:1 and 10:1. In parallel to co-culture oysters, single strain uptake and bacteria-free controls were assessed to ensure bacteria were being ingested as expected. Each point represents the mean of six technical oyster replicates (n = 3). Horizontal bars represent the mean, error bars represent the standard deviation. Attacking strains are represented with circles, prey strains with diamonds. Vertical dotted lines separate attacker, prey and control values into their respective groups. Downward arrows indicate that some oysters were below the 100 CFU/g tissue limit of detection. A. Vibrio vulnificus 106-2A, ΔT6SS1, ΔT6SS2 and ΔT6SS1/2 co-culture with V. vulnificus 99-743 at a 5:1 ratio. B. Vibrio vulnificus 106-2A, ΔT6SS1, ΔT6SS2 and ΔT6SS1/2 coculture with V. vulnificus 99-743 at a 10:1 ratio. C. Vibrio vulnificus 106-2A, ΔT6SS1, ΔT6SS2 and ΔT6SS1/2 co-culture with S. Enteritidis CC012 at a 5:1 ratio. D. Vibrio vulnificus 106-2A, ΔT6SS1, ΔT6SS2 and ΔT6SS1/2 co-culture with S. Enteritidis CC012 at a 10:1 ratio. Statistical significant reduction in prey strain CFU/g tissue was determined through comparing prey counts from co-culture oysters to prey counts from the single-strain prey cohort using a nested one-way analysis of variance with Dunnett's multiple comparisons test. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. properties in this model, reinforcing the hypothesis that these T6SSs are predominantly antibacterial. Examination of V. vulnificus genomes has identified two T6SSs, termed T6SS1 and T6SS2 (Church et al., 2016). While all sequenced V. vulnificus strains possess a T6SS2, only a subset possesses a T6SS1. in vitro competition assays have previously demonstrated that T6SS1 + strains exhibit antibacterial activity against T6SS1 − strains, suggesting a role for the T6SS1 in bacterial competition (Church et al., 2016). In this paper, we demonstrate that both the T6SS1 and T6SS2 are active at environmental temperatures and that both facilitate intra-and inter-species killing. Interestingly, our data show that V. vulnificus 106-2A is able to target and kill V. vulnificus 99-743 in a T6SS2-dependent manner, despite 99-743 also being T6SS2 + . This may be due to 106-2A encoding T6SS2 effectors not found in 99-743, against which it does not express cognate immunity proteins. This has been demonstrated previously in V. cholerae where genomic analysis of a panel of V. cholerae strains revealed that strains possessed a variety of effectors not always identical to one another (Unterweger et al., 2014). Strains with dissimilar effector panels were labelled 'incompatible' and shown to compete with one another. Similarly, V. alginolyticus contains an array of 'orphan' effectors neighbouring mobile elements obtained from other marine bacteria through horizontal gene transfer (Salomon et al., 2015). It is thus possible that V. vulnificus 106-2A has horizontally acquired T6SS2 effector proteins to which V. vulnificus 99-743 does not possess the cognate immunity protein, rendering 99-743 vulnerable to T6SS2-mediated killing. Future work will be required to confirm this is the case in V. vulnificus.
Although the large majority of seafood-related deaths in the United States are attributed to V. vulnificus with several reported cases having a fatality rate exceeding 50%, the clinical incidence is low given the prevalence of V. vulnificus in shellfish coupled with the quantity consumed. Of between 18.5 and 26.5 million Americans at risk, fewer than 100 cases are reported annually (Warner and Oliver, 2008). Historical reports demonstrate a varied population structure of V. vulnificus within oysters, and it may well be that the T6SS1 and T6SS2 systems contribute to a reduction in numbers of strains that are clinically isolated. We further demonstrated in vitro inter-species killing potential of both T6SSs by co-culture assays utilizing S. Enteritidis CC012 as prey. Salmonella species have been identified in up to 8.6% of domestic oysters within the United States, although there are few Salmonella outbreaks associated with fish or shellfish documented in the literature (Morrison et al., 2012). Whether this is a reflection on detection techniques or in part due to inter-species bacterial competition, or both, remains to be resolved.
While the in vitro data presented in this paper empirically demonstrate the competitive ability of V. vulnificus, the experimental approach of using AMS to replicate an in vivo scenario supports this competitive behaviour in vivo. The AMS model provides a means for reproducible and efficient uptake of high numbers of bacteria by oysters. Previous studies into the uptake of V. vulnificus or S. Enteritidis by oysters have relied on the filtration of planktonic bacteria added to oyster tank water. Such research resulted in levels no higher than 10 5 CFU/g tissue (Srivastava et al., 2009;Froelich et al., 2010;Morrison et al., 2011;Morrison et al., 2012;Pu et al., 2018). NMS as a vehicle to promote V. vulnificus uptake by oysters was first documented by , where bacteria were incorporated into NMS generated from seawater. Despite successful incorporation of bacteria into NMS, a number of factors limiting incorporation and subsequent oyster uptake levels were proposed. To resolve these issues, we developed an AMS model using the diatom, T. pseudonana CCMP 1335, as a substrate for aggregate formation as it had previously been shown that marine snow can be generated using a range of diatom species (Grossart and Ploug, 2001;Flintrop et al., 2018). Utilizing T. pseudonana in AMS at specific concentrations led to consistent aggregate formation, with reproducible incorporation levels of V. vulnificus and S. Enteritidis, 10,000-fold greater than those observed with NMS. Exposure of oysters to AMS resulted in extremely rapid ingestion of bacterial cultures to levels of 10 6 -10 8 CFU/g tissue, >1000-fold higher than other oyster uptake experiments published to date . This study has demonstrated that AMS is a versatile and highly reproducible model for facilitating rapid and highlevel uptake of bacteria into oysters compared with NMS or simple tank-inoculation practices.
We employed our AMS model to prove that V. vulnificus T6SS1 and T6SS2 have intra-and inter-species antibacterial activity in an oyster in vivo model. This is the first study to demonstrate T6SS in vivo activity in oysters facilitated by AMS. Our work suggests that both T6SS1 and T6SS2 are responsible for killing at environmental temperatures around 21 C. It will be of interest to determine the effective environmental ranges of the T6SS in vivo through further studies and the control that other regulators such as salinity have over the activity of the two systems. Interestingly, we observed more effective killing of prey bacteria in vivo compared with the in vitro platebased assays. While this may be simply explained by the greater ratio of attacker to prey, another reason for this may lie in the regulation of the T6SSs, of which little is still fully understood. As T6SSs are involved in microbial competition, it is possible that they are upregulated under conditions where increased competition is required, such as nutrient-limiting environments. Specifically, recently identified roles for bacterial T6SSs have been in the acquisition of essential micronutrients. Burkholderia thailandensis, Pseudomonas aeruginosa and Yersinia pseudotuberculosis T6SSs have all demonstrated secretion of effectors involved in nutrient acquisition, including manganese scavengers, zinc-binding proteins and iron-chelators (Liu et al., 2015;Si et al., 2017a;Si et al., 2017b).
In conclusion, this study has further characterized the activity of the T6SSs of V. vulnificus and has identified thermoregulatory conditions under which these are active. These are the first published findings confirming T6SS2 antibacterial activity of V. vulnificus. This study has also developed a widely applicable AMS model which facilitates incorporation of high numbers of bacterial that can subsequently be ingested by oysters for in vivo studies. Such studies demonstrated that both the T6SS1 and T6SS2 of V. vulnificus are active in vivo and capable of intra-and inter-species killing. Future work will look at expanding these competition assays to further understand T6SS regulation, activity and mode of action both in vitro and in vivo.
Dictyostelium discoideum (accession number: DBS0235542) was ordered via DictyBase (Fey et al., 2013) and received growing on a lawn of E. coli B/r. Axenic cultures were generated by inoculating 30 ml HL5 media (Formedium™) supplemented with ampicillin (100 μg/ml) and streptomycin (300 μg/ml) with an inoculating loop of D. discoideum fruiting bodies. Cultures were incubated at 21 C, 180 rpm for 2-3 days. Before cultures exceeded 10 6 cells/ml, 1 ml of culture was used to inoculate 30 ml of antibiotic-free HL5. Cultures were incubated at 21 C, 180 rpm for 2-3 days and routinely checked for contamination.
Generation of T6SS in-frame deletion mutants V. vulnificus ΔT6SS1, ΔT6SS2 and ΔT6SS1/2 were generated through in-frame deletion of hcp1 and hcp2.
Upstream and downstream flanking regions for both genes were PCR-amplified and cloned into pGEM-T Easy for sequencing (Eurofins). Once confirmed, regions were enzymatically excised, purified and ligated into allelic exchange vector, pDM4 (Milton et al., 1996). Constructs pDM4-Δhcp1 and pDM4-Δhcp2 were transformed into E. coli S17-1 λ pir and triparentally conjugated into V. vulnificus 106-2A using conjugal donor, E. coli HB101 pRK2013. First crossover integrants were selected for on TCBS agar supplemented with chloramphenicol. To promote a second crossover event, first crossovers were plated on LB agar supplemented with 10% sucrose and incubated at 37 C for 24 h. Second crossover mutants were confirmed through PCR and sequenced to confirm the insertion was in-frame. Expression of downstream genes was confirmed using RT-PCR.

RNA extraction and RT-PCR
RNA was extracted from bacterial cultures using the RiboPure™-Bacteria kit (ThermoFisher) according to the manufacturer's instructions. 50 ng random hexamers were mixed with 1 μg total RNA, 1 μl 10 mM dNTPs and MiliQ H 2 O to 13 μl. This solution was incubated at 65 C for 5 min and incubated on ice for a further 5 min. Following incubation, the following reagents were added to the microcentrifuge tube: 4 μl 5× First-Strand Buffer, 1 μl 0.1 M DTT, 1 μl RNaseOUT™, 1 μl SuperScript™ III RT (all reagents Invitrogen). The solution was incubated at 25 C for 5 min before being incubated for a further 60 min at 50 C for cDNA generation.

Generation of antibiotic-resistant V. vulnificus strains
To enable antibiotic selection of V. vulnificus attacker and prey strains following co-culture, plasmids conferring antibiotic resistance were electroporated into V. vulnificus and selected for on LB supplemented with either kanamycin or trimethoprim. pVv3-Kan was utilized to confer kanamycin resistance to attacker strains (Klevanskaa et al., 2014), and pVv3-Tmp was employed to confer resistance to trimethoprim for prey strains. Electroporation was performed according to the protocol presented in Klevanskaa et al. (2014). As selective chromogenic media was utilized for Salmonella selection, there was no need to generate antibiotic-resistant Salmonella strains.
In vitro co-culture competition assays Attacker and prey strain cultures were subcultured in 30 ml fresh selective LB to an OD 590nm of 0.03 and grown to an OD 590nm of 1 before being adjusted to an OD 590nm of 0.8. Cultures were pelleted, washed twice with phosphatebuffered saline (PBS) to remove any antibiotics before being resuspended in PBS at an OD 590nm of 0.8. Attacker and prey strains were combined at a ratio of 3:1 and 25 μl spotted onto LB agar. Following incubation, bacterial spots were excised using a scalpel and resuspended in 1 ml PBS. Tenfold serial dilutions were performed and triplicate 10 μl spots for each dilution were plated on the required selective media. Plates were incubated at 37 C and the average CFU/ml was calculated the following day.
Dictyostelium discoideum plaque-forming assays Overnight bacterial cultures were pelleted by centrifugation and resuspended to an OD 590nm of 5.5 in KK2 buffer. Of note, 1 ml D. discoideum culture was pelleted by centrifugation at 2500 rpm for 5 min. Pelleted cells were washed once with KK2 buffer, centrifuged for a further 5 min and resuspended in 1 ml KK2 buffer. Resuspended D. discoideum AX3 were added to 1 ml of bacterial culture at a final concentration of 10 2 cells/ml. Of note, 200 μl of the resulting suspension was spread on SM agar and incubated at 21 C for 3-5 days to allow plaque formation (Pukatzki et al., 2006).

Generating natural and AMS
NMS was generated by combining 1 l collected seawater (Dawlish Warren) with 1 ml bacterial culture adjusted to an OD 590nm of 0.8 and 1 μg/L hyaluronic acid (Fisher). Tubes were rolled for 24 h at 8 rpm and 21 C. AMS was generated by combining 1 l artificial seawater (20 ppt salinity), with 1 ml TP 1800 (10 9 cells) (Reed Mariculture), 1 ml bacterial culture adjusted to an OD 590nm of 0.8 and 1 μg/L hyaluronic acid (Fisher). Tubes were rolled for 24 h at 8 rpm and 21 C.

Enumerating bacterial incorporation into marine snow
Marine snow was vacuum-filtered through qualitative filter paper No. 6 (Whatman) with a pore size of 3 μm. Nonaggregated bacteria are too small to be retained by the filter paper. Any cells incorporated into larger marine snow aggregates were retained on the filter paper. Filtered aggregates were resuspended in 50 ml PBS and vortexed vigorously to disrupt the aggregates. Samples were centrifuged to pellet the material, resuspended in 10 ml PBS and serially diluted 10-fold for spread-plate enumeration on LB, TCBS or ChromoSelect agar. Plates were incubated at 37 C overnight and CFU/g marine snow calculated the following day.

Scanning electron microscopy
Marine snow samples were fixed in 2% glutaraldehyde and 2% paraformaldehyde in 0.1 M Pipes buffer pH 7.
Fixed samples were transferred onto a 0.1 μm polycarbonate filter and washed for 5 min in Pipes buffer for a total of three times. Aggregates attached to the filter were dehydrated using an ethanol gradient. The filter was incubated for 3 min in hexamethyldisilazane before air-drying. Dried filters were mounted onto an aluminium sample pin using carbon conductive tabs. Samples were then sputter-coated with 10 nm gold/palladium (80/20) and imaged using a JEOL JSM 6390 LV scanning electron microscope operated at 5 kV. Images were contrastadjusted and false-coloured using Adobe Photoshop CC 2019 V20.0. Raw images can be seen in Figure S3A-D.

Oyster husbandry
Pacific oysters (Crassostrea gigas) were obtained from Lindisfarne oyster farm (Lindisfarne oysters, Northumberland). Upon arrival, oysters were scrubbed and placed in static holding tanks containing 20 ppt salinity artificial seawater at 21 C. Oysters were maintained at the University of Exeter in the Aquatic Resources Centre under their prescribed oyster husbandry protocols. Oysters were allowed to acclimatize for 5 days. Water was changed daily, and oysters were fed with 250 μl TP 1800® (Reed Mariculture) every other day.
In vivo oyster exposure assays Oysters were maintained in individual 5 l holding aquaria at 21 C. AMS was added to the holding tanks taking care to not disrupt the aggregates. Six oysters were utilized for each exposure. Oysters were left to filter and ingest the inoculated marine snow for 24 h. A circulatory current was generated by the presence of an air stone to prevent aggregate settling. Following exposure, oysters were removed from their tanks and briefly cleaned with 70% ethanol to remove non-ingested microorganisms. Using a flame-sterilized knife, oysters were shucked and the stomachs excised. Samples were weighed before being suspended in 10 ml PBS and homogenized for 3 min in a Stomacher 400 (Seward). The resulting homogenate was filtered through a 40-μm cell strainer (Merck) into a clean 50-ml falcon tube to remove tissue debris. The resulting filtrate was serially diluted 10-fold in PBS. Of note, 100 μl of each dilution was spread on LB agar and incubated overnight at 37 C. The following day, all colonies were replica plated onto Vibrio selective TCBS and incubated overnight at 37 C. Colonies were then enumerated based off colour and recorded for CFU/g tissue calculations. Assays involving Salmonella were not plated onto LB but instead straight onto Salmonella-selective ChromoSelect agar.