Passive exposure to cannabidiol oil does not cause microbiome dysbiosis in larval zebrafish

2.2. CBD oil quality

We selected the CBDPure hemp oil 100 (Full Spectrum CBD oil, Nutra Pure LLC, Vancouver, WA) based on publicly available product quality reports. Indeed, the CBD oil extract used in this study was certified by a third-party for cannabinoid profiling, pesticide testing, microbiological screening, terpene analysis, and residual solvent. In addition to third-party certification from the manufacturer, we validate the CBD oil quality internally using gas chromatography/mass spectrometry analysis. More specifically, we extracted cannabinoids using methanol (MeOH, Sigma-Aldrich, HPLC grade) without further purification. Extractions of commercial CBDPure 100 hemp oil (1.00 g samples, 12 × 1 mL MeOH) were performed in triplicate, combined, concentrated in vacuo, and redissolved in 1 mL MeOH before GC/MS analysis. Cannabinoid analytical standards: cannabidiol solution (CBD, 1.0 mg/mL in methanol, certified reference material), cannabinol solution (CBN, 1.0 mg/mL in methanol, certified reference material), and Δ⁹-tetrahydrocannabinol solution (THC, 1.0 mg/mL in methanol, certified reference material) were purchased from Sigma-Aldrich. A calibration curve for the CBD solution standard was obtained in triplicate and was linear (R² > 0.992) over the measured concentration range (0.03 mg mL⁻¹ – 1 mg mL⁻¹).

We performed gas chromatography/mass spectrometry analysis using a Varian-450 GC equipped with a Varian FactorFour™ VF-5 (30 m x 0.25 mm, 0.25 µm) capillary column and a Varian-220 MS. Helium gas was used as the carrier gas in constant flow mode at 1 mL min⁻¹, and the solvent delay was set at 3 min. The injector was maintained in a 1:50 split ratio mode at 220 °C while the oven was programmed as follows: 100 °C (2 min) to 260 °C at a heating rate of 15 °C min⁻¹ (5 min). The G.C. oven program’s run time was 20 min, and the mass spectrometer was in full scan mode in the range of m/z 60–350. Of the standards analyzed, we detected only the cannabinoid CBD at a concentration of 0.56 ± 0.07 mg mL⁻¹ (Figure S1). Finally, the M.S. provides conclusive evidence that a large concentration of CBD was indeed detected in CBDPure 100 (Figure S2). While the concentration of CBD detected with our methodology is lower than that reported by NutraPure, the analyses provided by a certified lab did indeed report 1.6 mg mL⁻¹. We will hereafter use the concentration reported by the certified third-party analytical laboratory.

2.3. Experimental populations

For each independent trial, between 20 and 30 fertilized eggs were sampled approximately one hour after a single mating and bleached twice in 0.5% hypochlorite solution for four minutes and once in sterile 1x E3 media (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂, 0.33 mM MgSO₄, 0.5 mg/L methylene blue). Fertilized eggs were placed in sterile Petri dishes filled with 60 mL of 1x E3 medium, of which 80% was replaced with fresh medium every 24 h and monitored for possible contaminants and embryo mortality.

For each trial, we established two control populations and two experimental populations using 24 healthy larvae (i.e., 5-day-post-fertilization (dpf)) for a total of 96 animals. Each larva, randomly selected from the larger pool, was placed in a single well of a 24-well plate filled with 2 mL of sterile 1X E3 medium. One control population, referred to as the control group, was placed in sterile E3 medium. The second control group, referred to as the DMSO group, was exposed to 0.05% DMSO, the concentration of DMSO used as a solvent for cannabidiol oil. As for the experimental populations, one was placed in sterile E3 medium supplemented with 20 μg/L cannabidiol oil, and another was exposed to sterile E3 medium supplemented with 200 μg/L cannabidiol oil. The two concentrations were chosen to represent a range of toxicity for cannabidiol previously investigated in zebrafish larvae (Valim Brigante et al., 2018). Cannabidiol solution was made fresh before each independent trial using a commercially available CBD oil (Full Spectrum CBD oil, Nutra Pure LLC, Vancouver, WA) and dissolved in a 5% DMSO solution. While most cannabinoids are known to be adsorbed to plastics over time, recent work showed that plastic tubes and glass tubes did not shown significant difference in adsorption/degradation in laboratory condition similar to the one described above (Bruno et al., 2019) or at concentration above the one used in this study (Molnar et al., 2013). Furthermore, approximately 80% of the medium was exchanged daily for sterile 1x E3 media with the relevant cannabidiol or DMSO concentration when applicable to minimize the risk of contamination and to maintain the cannabidiol oil concentration as close as possible to the original concentrations.

We maintained the zebrafish larvae in experimental conditions for five days without feeding and assessed survival daily. Based on our previous work (Dahan et al., 2018; Pindling et al., 2018a), this timeline is sufficient for the host-microbiome to develop to a size fit for investigation without affecting survival. At the end of the experiment, three larvae from each treatment, for a total of 12 animals, were randomly selected and anesthetized according to established euthanasia techniques (Matthews and Varga, 2012; NIH, 2009). Following euthanasia, fish were washed three times with nuclease-free water to minimize the presence of free-living bacteria and stored at −80 °C. The procedure was repeated over three trials using independent broods from the same mating pair for a total of 36 preserved animals (i.e., 12 animals per trial).

2.4. DNA extraction and processing

We extracted and purified microbial DNA from each fish larvae using a modified protocol for the DNeasy Blood and Tissue Kit (QIAGEN, Germantown, MD) as implemented in (Pindling et al., 2018b, 2018a). Purified gDNA samples were stored in nuclease-free H₂O at −20 °C. Fish microbiomes were characterized via targeted gene amplification of the 16S rRNA V4 region using Golay-barcoded primers 515F and 806R (Caporaso et al. 2012). PCR and sequencing reactions were performed at the Wright Labs (Huntingdon, PA) using the MiSeq paired-end Illumina platform adapted for 250-bp paired-end reads as previously described in (Dahan et al., 2018; Pindling et al., 2018a). All sequence reads are available at the Sequence Read Archive of the NCBI (PRJNA692159).

2.5. Processing of 16S rRNA sequence data

Microbiomes were characterized using the DADA2 pipeline version 1.8.1 (Callahan et al. 2016) available at (https://github.com/benjjneb/dada2) and implemented in R version 3.2.3 (http://www.r-project.org). In brief, forward and reverse reads were filtered, de-replicated, and de-noised using DADA2′s default parameters. After removing chimeras and all sequences identified as either mitochondrial or chloroplast DNA, each unique sequence is defined as an amplicon sequence variant (or ASV). A full list of ASV and their abundance in each sample can be found in Table S1. For each ASV, we then assigned taxonomy using IDTAXA (Murali et al., 2018) available via the DECIPHER Bioconductor package (DOI: 10.18129/B9.bioc.DECIPHER) trained on the SILVA ribosomal RNA gene database version 138 (Quast et al., 2013) as well as the RDP trainset 16 (McDonald et al., 2012; Werner et al., 2012). Full taxonomic assignment, along with DNA sequences, can be found in Table S2. Lastly, we build a maximum likelihood phylogenetic tree based on a multiple alignment of all the ASVs using the phangorn package version 2.1.3 (Schliep 2011).

3.2. 16S rRNA V4 reads using DADA2

After removing non-desired sequences, we retained 1056,920 (or 90.0% of the total) sequence reads from 35 larval microbiomes, which resulted in 782 unique amplicon sequence variants (ASVs). Using 16S rRNA profiling, the ASVs were assigned to 17 phyla, 79 orders, and 192 known genera (see Table S2 for full listing). While 256 (or 32.7% of all) ASVs could not be identified at the genus level, the taxonomic assignment was consistent when using different reference databases (Figure S3). Further results will hereafter be presented for taxonomic profiling based on the SILVA database release 138.

To investigate the stability of the zebrafish larval microbiome in our study, we then compared the microbiomes of eight control larvae sampled randomly from the tree trials we conducted; one of the microbiomes from trial T1 failed to be sequenced. Among the eight microbiomes extracted from control larvae, we recovered all of the 782 ASVs sampled from our whole study, demonstrating the zebrafish larval microbiome variability. Comparable to previous work done of the core microbiome of zebrafish living in different environments (Roeselers et al., 2011), we found that a majority of ASV belonged predominantly to the phylum Proteobacteria, 80.2(5.0)%, and, to a lesser extend, Firmicutes, 9.6(0.1)%, (Fig. 1A). While some of the most common ASVs we identified as part of the Pseudomonas genus, totalizing up to 17.1(0.07)% of the total abundance, a large number of ASVs, 39.8(0.05)%, could not be matched to a known genus. Also, while the composition of the different control microbiomes differed somewhat between samples, with 62.7% similarity as measured with the Morisita-Horn Similarity Index, we found no overall significant dissimilarity between the three independent trials (F_{(2, 5)} = 2.69; R² = 0.52; P = 0.07). Possible dissimilarities between control microbiomes were even less significant when considering phylogenetic distance within the communities measured as weighted UniFrac (F_{(2, 5)} = 1.56; R² = 0.38; P = 0.14), meaning that the ASVs that differed between trials were closely related genetically and predicted to fill similar ecological niches.

Lastly, we tested whether the presence of DMSO, the solvent used to dissolve cannabidiol in the growth medium, affected zebrafish larval microbiomes by comparing the microbiome of larvae exposed to DMSO at a concertation similar to that used in the main experiment to that of the control microbiomes. We found that DMSO did not significantly alter the zebrafish larval microbiomes (F_{(1, 10)} = 1.08; R² = 0.06; P = 0.31). Once again, the effect was not significant when considering the phylogenetic distance among the communities (F_{(1, 10)} = 1.03; R² = 0.07; P = 0.38; Fig. 1B). Interestingly, when combining the control microbiomes to the microbiomes of larvae exposed to DMSO as a single control group, we found a small difference between the three trials when considering the similarity between the two control groups (F_{(2, 5)} = 3.42; R² = 0.38; P = 0.033) as well as the phylogenetic distance between the two control groups (F_{(2, 5)} = 2.39; R² = 0.30; P = 0.03). While this difference is not significant when applying a correction for multiple statistical testing, we will test the possible effect of trials on CBD treatment in subsequent analysis.

3.3. Cannabidiol does not affect the diversity of zebrafish larval microbiome

To investigate whether cannabidiol could impact zebrafish larval microbiome, we first compared the overall microbial diversity in microbiomes sampled from control larvae to that of microbiomes sampled from larvae exposed to either 20 µg/L or 200 µg/L. We found no significant differences in average richness, estimated as the total number of ASVs, among the different groups (F_{(2, 19)} = 1.26; P = 0.31; Fig. 2A) and this was true in all three trials (F_{(2, 5)} = 0.48; P = 0.63). On average, we found a total of 53.86(7.7) ASVs in the control microbiomes, 56.86(16.5) ASVS in the fish exposed to 20 µg/L, and 46.86(10.3) in the fish exposed to 200 µg/L. Interestingly, we found similar richness when considering the Chao1 index, which estimates the number of predicted taxa expected if the population was infinite (Table 1), indicating that our sampling was exhaustive.

We then look at measures of diversity that include evenness. The latter is important in defining community diversity as it indicates whether one or a few species dominate the community. Again, we found no significant difference between the three treatments estimated as Simpson’s D Index (F_{(2, 19)} = 0.56; P = 0.58; Fig. 1A). On average, we found a Simpson Index of 0.87(0.08) in the control larvae microbiome, 0.84(0.15) in the 20 µg/L cannabidiol oil treatment, and 0.83(0.5) in the 200 µg/L cannabidiol oil treatment. These results indicate that, on average, two ASVs taken at random in any group have an ~85% chance of being different from each other, demonstrating that no ASVs tend to dominate the larval microbiome even in the presence of higher concentrations of cannabidiol.

The work conducted in this study was supported by the Biology Department of Bard College, which had no role in study design, data collection and interpretation, or the decision to submit the work for publication. We also thank Heather Bennett and Frank Scalzo for guidance with experimental design and Maureen O’Callaghan-Scholl and Daniela Azulai for assistance with zebrafish husbandry. Finally, we thank Prof. Swapan Jain for supervising the preparation of the CBD-supplemented medium.