This study involves methods to reveal effects on a model fish host following alteration of the skin and gut microbiome communities composition by an antibiotic.
The commonality of antibiotic usage in medicine means that understanding the resulting consequences to the host is vital. Antibiotics often decrease host microbiome community diversity and alter the microbial community composition. Many diseases such as antibiotic-associated enterocolitis, inflammatory bowel disease, and metabolic disorders have been linked to a disrupted microbiota. The complex interplay between host, microbiome, and antibiotics needs a tractable model for studying host-microbiome interactions. Our freshwater vertebrate fish serves as a useful model for investigating the universal aspects of mucosal microbiome structure and function as well as analyzing consequential host effects from altering the microbial community. Methods include host challenges such as infection by a known fish pathogen, exposure to fecal or soil microbes, osmotic stress, nitrate toxicity, growth analysis, and measurement of gut motility. These techniques demonstrate a flexible and useful model system for rapid determination of host phenotypes.
It has been established that antibiotics can disrupt the human microbiome leading to dysbiosis, meaning a microbial community imbalance. The microbiota's compositional alteration after antibiotic treatments has been shown to lower the community's diversity, reduce key members, and alter community metabolism, especially in the gut1,2. Antibiotic disturbance of the gut microbiome can reduce colonization resistance to Clostridium difficile3,4 and Salmonella5.
Additionally, the disruption of the microbiota has been linked to the development of many syndromes and diseases in humans (e.g., antibiotic-associated enterocolitis, inflammatory bowel disease, metabolic disorders, etc.). Antibiotics are also widely implemented in agriculture as growth promoters in livestock and poultry production6. The usage of these powerful tools is not without collateral effects, which is evident in the rapid rise of antibiotic resistance, as well as the effects of a disrupted microbiome has with its inhabited host. Many studies have shown that broad-spectrum antibiotic usage has long lasting consequences to the structure and function of the microbiota, yet the side effects from an antibiotic-disrupted microbiome impacting host physiology are only speculations which have yet to be supported.
The interplay between host, microbiota, and antibiotics is far from being understood in a concise manner. Therefore, a simple and more tractable model is advantageous to shedding light on the highly complex mammalian system. Mucosal surfaces in humans, including the gut, harbor the highest density and diversity of microbes, and also the most intimate microbe-host interactions. The mucosal skin microbiome of fish offers several advantages as a model system. The Teleostei (bony fish) is one of the earliest lineages to diverge within the Vertebrata meaning that teleosts have both innate and acquired immune systems that have co-evolved a relationship with commensal bacterial communities7. Fish skin shares many characteristics with type 1 mucosal surfaces of mammals, such as physiological functions, immunity components, and arrangement of mucus-producing cells8. The external location of the fish skin mucosal surface offers a microbiome easy to experimentally manipulate and sample.
The Western mosquitofish, Gambusia affinis (G. affinis), is a model fish that has been used in the past for studying mating and toxicology9,10,11. Given the small size, population abundance in the wild as an invasive species, minimal care cost, and hardy nature, we have developed G. affinis as a mucosal microbiome model. Further, Gambusia share the physiology of giving birth to live young with viviparous mammals, which is uncommon in fish species. We completed the most extensive study at the time of fish skin normal microbiota using 16S profiling with Gambusia12. Further work demonstrated three negative effects on the host following disruption of the skin and gut microbiota using a broad-spectrum antibiotic13.
Five different effects were examined in the fish following antibiotic exposure. The most well established host benefit of the microbiome is competitive exclusion of pathogens. The fish pathogen Edwardsiella ictaluri is known to cause outbreaks of enteric septicemia in commercial catfish farms14. E. ictaluri has also been shown to lethally infect zebrafish15,16 and Gambusia17. A challenge with this pathogen from the water column can serve as a measure of exclusion. As a comparison to susceptibility to an individual pathogen, survival during exposure to a high density of mixed organisms was also carried out. Feces and organic-rich soil were used as commonly-encountered sources of microbial communities.
Another established role the bacterial gut community performs is nutrient processing and energy harvest, thus affecting the overall nutritional uptake for the host. As a gross measurement of nutrition, fish body weight was compared before and after one month of being fed a standard diet. Antibiotic-treated fish as an average lost weight while control fish on average gained weight over the month. The mechanism for this lack of weight gain is unclear. One possible contributing factor is transit time of food in the gut. A GI motility method was adapted from zebrafish (Adam Rich, SUNY Brockport, personal communication) to determine transit time. It has not yet been determined if antibiotic-treated fish have an altered transit time.
A common challenge experienced in the natural environment by all organisms, especially fish, is osmotic stress. Gambusia have been shown to quickly adapt when acutely stressed in high concentrations of salinity18. Surprisingly, fish with an antibiotic-altered microbiome exhibited lowered survival to a high salt stress. The mechanism for this novel phenotype is under investigation. Another common stress on aquatic animals, especially in aquaria, is toxic forms of nitrogen (ammonia, nitrate, and nitrite). Survival against nitrate was not significantly different between antibiotic-treated and control fish. The methods presented in this manuscript can be used with Gambusia or similar fish model organisms, such as zebrafish and medaka, to measure phenotypes in the fish following experimental manipulation.
All animal experiments were conducted under approval of IACUC protocols, numbered 14-05-05-1018-3-01, 13-04-29-1018-3-01, and 14-04-17-1018-3-01.
1. Animal Collection, Handling, and Ethical Care
2. Initial Antibiotic Exposure for All Experiments
3. Microbiome Extraction
4. Infection Model Preparation and Bath of a Specific Pathogen
5. Polymicrobial Challenge with Feces & Soil
6. Osmotic Stress Challenge
7. Nitrate Toxicity Challenge
8. Growth Analysis of Individualized or Grouped Fish
9. Gut Transit Time
An overall schematic diagram of the experimental system used to study fish host effects from antibiotic exposure13 is represented in Figure 1A and includes the technique for extracting the skin (Figure 1B) and gut (Figure 1C) microbiomes from the fish. Three days was selected as the antibiotic period of exposure because previous data reveals that while the total skin culturable number drops early in treatment, it has returned to pre-treatment levels after 3 d. Meanwhile, the community composition, as determined by 16S profiling, has been strongly altered. Therefore, the 3-day period should be optimal for analyzing the effects from a changed community of approximately the same density. Note that culture analysis, using colony numbers on agar plates, may leave out a number of species. Efficiency of the skin microbiome dispersion method (Figure 1B) was analyzed by comparing colony number on plates from suspension buffer (typically in range of 104 – 105 CFU/g fish weight) to colony number from several fish subjected to the same suspension procedure a second time (to quantify any remaining bacteria). Counts from this second suspension were lower than 100 CFU/g, suggesting the suspension method is effective (unpublished).
Fish with an antibiotic-altered microbiome appeared to be more susceptible to an E. ictaluri infection than control fish (Figure 2). The difference in mean time to death of 56.1 ±15 h for the treated fish and 98.5 ±48 h for control fish was not statistically significant (two-tailed Student's t-test, p = 0.12). This is likely due to the small group size (treated n=6, control n=5). Group sizes of at least a dozen are therefore recommended. An advantage of this infection model is the bath protocol, which does not require needles for infection or tracking of food intake. E. ictaluri naturally invades catfish from the water column. Safety is high because E. ictaluri is temperature sensitive, and thus poorly infectious to humans. Other studies have revealed that the time to death in G. affinis correlates with the initial dose of bacteria. The incubation temperature of 27 °C was selected as optimal for both bacteria and fish.
Treated or control fish did not have significant differences in survival in water contaminated with high counts of mixed microbes. No mortality was observed over 4 days for control (n=8) or treated (n=9) fish when soil was used as the microbial source. While some mortality did occur with human feces as the source of microbes (Table 1), antibiotic treatment made no difference. When the concentration of feces in APW was at 20 mg/mL, 40 – 50% of the fish died. However, the dissolved oxygen concentration was only 10% (compared to >80% in buckets or aquaria), thus the hypoxic conditions confounded the interpretation. When lower levels of 16 or 10 mg/mL were used, the survival rates were matched (two-sided Fisher's exact test, p=0.95 or greater for a difference between groups). Dissolved oxygen levels in these two trials were 65% or above. Gambusia are a very hardy invasive species that can live in low-quality water. It would be interesting to use these assays of natural microbial exposure to measure survival of other species against a challenge of contaminated water. Samples of water from specific environmental sites, especially those subjected to eutrophication, could be used in a similar assay, although water quality (dissolved O2, nitrate, salinity, etc.) would need to be determined, as a potentially confounding factor.
Fish exposed to rifampicin were more susceptible to osmotic stress than control fish (Figure 3). The log-rank test observed a significant difference (p = 0.049) in survival rate (43% death in control group and 88% death in treated group, n = 9 for both groups) when challenged with elevated salinity. Results from this assay agreed with a determination of 18.1 mg/mL as the LC50 for NaCl with G. affinis at 24 h in freshwater20. Signs of saline stress that fish exhibit include reddening around the gills and decreased swimming movement. With this assay, salinity levels above 18 mg/mL resulted in rapid fish death.
When exposed to the toxin nitrate in the water column, pre-exposure to antibiotic did not affect survival (Table 2). For each trial, death was recorded at both a short and long time point, representing acute and more chronic effects. The concentration of 10 mg/mL was selected based on it being the LD50 for G. affinis in freshwater at 48 h21. Results from this assay were consistent with this LD50 value, with a 50% lethality in both treated and control groups after 90 h. A higher challenge of nitrate (17.5 mg/mL) was also examined, leading to more rapid death. Again there was not difference in treatment groups. Nitrite is dramatically more toxic than nitrate to G. affinis, with an LD50 of 0.0015 mg/mL at 48 h22. Community biochemical analysis using the analytical profile index system (unpublished) shows that the fish skin microbiota has the potential to reduce nitrate. However, nitrite levels in the APW during both trials remained below the detection limit (0.001 mg/mL). This challenge method could be used with any small soluble chemicals.
When individualized in cups and fed the same amount to each fish for a month, a trend was observed for the antibiotic-treated fish to not gain weight as well as control fish, without a statistically significant difference (data not shown). To avoid the stress of individualization, fish were housed together as a group for treated and untreated fish in buckets for one month and given matched amounts of food. The limitation of this setup is that fish may not be receiving the same food on an individual basis. However, in the group model, treated fish on average lost weight and control fish on average gained weight (Table 3). The amount of food the fish were consuming did not seem to be affected by prior antibiotic exposure, so appetite suppression is an unlikely candidate explanation for lack of weight gain. Numerous other factors could contribute, including changes in inflammation in the gut, levels of mucus production, gut permeability, and/or gut motility. A major advantage of this assay to measure nutritional effects is simplicity. It is inexpensive, and only requires a laboratory balance as instrumentation. It is suitable to screen for an effect, leading to other involved experimentation to determine mechanism.
One example factor to examine related to fish weight gain is food transit time, which is related to gut motility. FITC-labeled dextran can be incorporated into gelatinized food and is nonlethal to the fish. Measuring fluorescence in the surrounding water over time gives a measurement of how fast the FITC-dextran is passing through the gut. Fluorescence above background can be detected as soon as 2 hours, with a maximum reached after 16 hours post-feeding (Figure 4). This result is with control fish from an aquarium. Reliable results from antibiotic-treated fish have not yet been obtained. One limitation of this procedure is that a 2-d starvation period is required for fish to eat the food. An advantage of the protocol is high sensitivity (low background fluorescence), as fish eating only one food section can give results, although data is more consistent when two sections are eaten. This protocol is less complicated than a similar and lethal method for mice23.
Figure 1: Schematic Overview of the Common Experimental Protocol. A is a flow chart illustrating, left to right, fish transferred from aquarium tank, separated into antibiotic-treated (represented by red water) or control (blue) groups, and then placed into individual cups to track phenotypes. B and C depict the process of extracting the fish skin and gut microbiomes. After vortexing, bacteria are suspended into the solution for analysis of the microbial community. Please click here to view a larger version of this figure.
Figure 2: Pathogen Susceptibility. A survival curve during exposure to E. ictaluri for fish pre-treated with rifampicin (red line) or an untreated control group (black line) fish. Please click here to view a larger version of this figure.
Figure 3: Susceptibility to Osmotic Stress. A survival curve during exposure to high salinity for fish in both antibiotic-treated and untreated groups. Please click here to view a larger version of this figure.
Figure 4: Food Transit Time. Fluorescence over time in the water from two fish fed FITC-dextran. Fish A ate two gelatin food sections and fish B ate one. Please click here to view a larger version of this figure.
Table 1: Susceptibility to High Microbial Environments. Exposure to human fecal matter was assessed at 3 different concentrations. Death ratio of x:y shows the total number of fish dead at the indicated time point x compared to the total number of fish in the experimental group y.
Table 2: Nitrate Toxicity Challenge. Recorded time of fish mortality in treated and control groups throughout exposure to a toxic nitrate concentration. Death ratio of x:y shows the total number of fish dead at the indicated time point x compared to the total number of fish in the experimental group y.
Table 3: Growth Analysis. Changes in total body weight after one month following antibiotic or control treatment. The percent difference in initial mean body weight (for the fish in that trial group) compared to final mean body weight is the column Δweight. N is the number of fish in that group. The mean weight per fish at the end of the trial is underweight/fish.
Some challenges require a rest period in clean APW after antibiotic treatment for the drug to be depleted in fish tissues. If the rest period is skipped then antibiotic presence can confound the results, especially when the assay involves exposure to bacteria. In order to examine effects from an altered microbiome composition without large changes in the total number of microbes on the host, preliminary experiments monitoring microbiome composition (16S profiling or whole genome sequencing) and population density (16S quantitation via qPCR) during antibiotic exposure would be required. While 3 d is optimal in this system, changing the host and/or antibiotic would require recalibration.
Typical of biomedical research, the mouse model is commonly utilized for microbiome studies. The most common fish model is zebrafish. In order to get a better understanding of the universal properties of the structure and function of mucosal microbiomes and host interactions, new and atypical models are a necessity. Our WT fish model serves as an authentic source for studying host-microbiome interactions by including natural host genetic variability that other models have lost due to generations of lab-raised animal inbreeding24. A major experimental advantage of Gambusia is their hardy nature, tolerating a wide range of conditions. High survival rates have been observed in water that varies in temperature (4 – 38 °C), salinity (0 – 17 mg/mL), dissolved oxygen (110 - 25%), and pH (4 – 8). This allows not only examining many environmental conditions, but also for multi-step experimental procedures that are often too stressful for other fish species.
The procedures presented here are suited for rapid screening to discover host effects from a particular treatment. Follow-up studies are required to determine direct causality and mechanism. Example extension studies for the fish microbiome host effects include: measuring intestinal inflammation, examination of fish health by quantifying fat stores, and measurement of mucus levels on the skin and in the gut. A gnotobiotic system is being developed to study in detail links of specific microbes to particular host phenotypes.
The authors have nothing to disclose.
This project was partially funded by a FAST (Faculty and Student Team) Award to TPP and JMC from EURECA (Center for Enhancing Undergraduate Research Experiences and Creative Activities) at Sam Houston State University.
Rifampicin | Calbiochem | 557303-1GM | |
Sodium Nitrate | Sigma Aldrich | S5506 | |
Fluorescein-labeled 70 kDa anionic dextran | ThermoFisher Scientific | D1823 | |
PBS tablets | Calbiochem | 6500-OP | tablets dissolve in water to make phosphate-buffered saline |