AbstractHatcheries nationwide suffer from unexplained acute production failures, termed crashes. The microbiota of oysters relates to larval health with previous studies showing that some bacterial species have positive and others negative effects on oyster health. To investigate microbial correlates of crashes, we collected samples from every batch of oyster larvae produced by the Horn Point Laboratory Oyster Hatchery since 2021 and analyzed the microbiota of 15 of those batches over their duration in the hatchery, from age of 3 to 5 days until either harvest or complete die off of the batch. Across events, die-offs generally became evident at or after six days of age. We found that the microbiota of oyster larvae appears to respond to die-off events with crashed batches having fundamentally different microbiota than good batches at age 7 to 9 and 9 to 12 days. Crashed batches were often taken over by microeukaryotes and bacterial taxa from the Protobacteria and Bacteroidetes phyla. However, this presumably opportunistic community differed between batches. Observed Vibrio species level groups did not appear to be oyster pathogens and appeared to respond to, rather than precede, crashes. The microbiota of 3 to 5 day old larvae were statistically related to whether a die-off occured later in the larval batches’ life, only when the taxa were first agglomerated to family level. The detection of two microbial species not previously known to associate with oysters, along with an increased presence of Dinophyceae, predominantly the toxin-producing Gyrodinium jinhaense, in 3 to 5 day old oyster larvae was statistically linked with subsequent batch crashes.This study suggests that the health of larval oysters shapes their microbiome. Conversely, it provides hints that the microbiome of larvae, and perhaps harmful algae, may drive hatchery crashes.IntroductionPrivate and government run shellfish hatcheries support oyster aquaculture and restoration efforts by providing farmers and managers with broodstock oyster larvae and spat on shell (Wallace et al., 2008). Although hatchery production has improved immensely over the past several decades (Helm and Millican, 1977; Elston et al., 1981; Urban and Langdon, 1984; Lewis et al., 1988; Robert and Gérard, 1999), hatcheries regularly experience massive unexplained die-off of stock (termed “crashes”) without clear causative factors (Walker, 2017; Gray et al., 2022). The Horn Point Laboratory Oyster Hatchery (HPLOH) is the largest producer of oyster seed (Crassostrea virginica ) on the Atlantic Coast of the United States, and experiences slowdowns and halts to its production every year (Gray et al. 2022). Crashes at HPLOH have common characteristics, with most to all of the larvae in a feeding tank ceasing to feed and then dying at the age of 6 to 21 days. Preliminary observations associate low salinity in the oysters’ overwintering location with lower spat yield (Gray et al., 2022). However, a causal relationship is not established. Changes to precipitation patterns driven by a warming climate may drive both changes in salinity and concurrent changes in chemistry and microbial community structure (Gibson and Najjar, 2000; Wang et al., 2021).Microbial communities are known to modulate the health of a range of animal hosts (Peixoto et al., 2021), including many aquaculture species (Infante-Villamil et al., 2021), and both adult and larval oysters (Yeh et al., 2020). Some bacteria are pathogenic, such as many species of the bacterial genusVibrio, which have been shown to sicken oyster larvae (Elston and Leibovitz, 1980; Richards et al., 2015). Contrastingly, other bacteria can help their hosts. Additions of probiotic species, includingBacillus pumilus RI06-95, and Phaobacter inhibens S4 have been shown to protect their hosts from Vibrio coralliilyticusinfection, by shaping the host’s innate immune response (Stevick et al., 2019; Modak and Gomez-Chiarri, 2020). Furthermore, bacteria can serve as indicators of changes in the host’s environment and health. For instance, oysters fed compromised microalgae showed both decreased health and a concurrent change in their microbiome (Vignier et al., 2021). Specifically, many species from the Rhodobacteraceae family positively associated with multiple measures of host fitness, while two species of Flavobacteriaceae associated with low fitness (Vignier et al., 2021). Preliminary investigation of larval crashes suggests that microbial communities differed between one crashed and one non-crashed larval brood at HPLOH (Cram et al., 2022; Gray et al., 2022). However, identifying a pattern requires surveying more than two batches of samples.Microbial communities, both in the waters of the Chesapeake Bay that feed the HPLOH hatchery (Arora‐Williams et al., 2022), and in other animal and environmental systems are in a state of constant change (Fauci and Dick, 1994; Fuhrman et al., 2015; Bashan et al., 2016). Time-series observations offer an advantage over those taken at a single point in time by identifying statistical associations in which one factor changes before the other factor, because leading indicators are valuable for prediction and can provide clues about causality (Fuhrman et al., 2015). Specifically, if one variable leads to another, the lagging variable cannot predict the leading variable. In hatchery crashes, this is important because crashes and microbes both shape each other – microbial changes that happen before crash events are unlikely to be caused by the crash itself and are therefore more likely to shape crashes, or else are associated with another variable that leads to crashes.While some studies have explored the community composition in oyster hatcheries (Le Deuff et al., 1996; Ramachandran et al., 2018; Arfken et al., 2021; Gray et al., 2022) and dynamics of those communities (Stevick et al., 2019), none to our knowledge have explored the microbial dynamics of normal production hatcheries over multiple crash and non-crash events. Therefore, we have been collecting samples of larvae and their microbiota from every water change of the HPLOH since the beginning of the 2021 growing season. In this study, we performed amplicon sequencing of the microbiota from every time point of selected batches from the 2021 season, including some that were healthy and some that experienced die-offs. Our goal was to explore whether microbiota from the first time-point were predictive of crashes, and whether the microbiota over the life of healthy and underperforming batches changed in predictable ways over the development or die-off of the larvae batches.
