Assessing the Microbial Quality of Commercial Aquaculture Probiotics by Next Generation Sequencing Revealed Faecal Bacteria and Bacteria of Public Health Signicance

There is an increasing trend in the application of probiotics in aquaculture as an environment friendly alternative to antibiotics for the control and treatment of diseases. The microbiome of commercial aquaculture probiotics were evaluated using NGS to understand the microbial species composition vis-à-vis the label on the probiotic product. Bacteria belonging to 12 bacterial species stated on the labels of ve probiotics were absent. The sequence reads of the microorganisms declared on the label ranged from 0.28-79 % for the probiotics tested while the rest were contaminants. The number of contaminant bacteria varied from 2 to 15 for the tested products. All products were found to have at least one bacterial species that were potentially pathogenic or of fecal origin. To our knowledge this is the rst study that utilized the potential of NGS to assess the microbial quality of commercial aquaculture probiotics providing novel insights in understanding the microbial composition of probiotic products which are otherwise dicult to identify. It is proposed that probiotic products shall be approved for use in aquaculture after determining the performance eciency by in vivo trails in farms, analysis of bacterial composition and concentration which shall be displayed on product labels.


Introduction
During the last few decades, worldwide usage of commercial probiotics in shrimp aquaculture has shown tremendous growth 1 . Probiotic products with various bene ts that include controlling diseases, promoting growth, enhancing feed digestion, and improving pond water quality have been introduced in aquaculture 2,3 . Easy availability of aquaculture probiotics in the market re ect the increasing usage of these products in the industry and the expansion of global marketplace for probiotic products. The market of aquaculture probiotics used for shrimp farming in India is estimated to be more than $100 million. In a recent study, 100% of the P. vannamei farmers surveyed in the four districts of Andhra Pradesh, India used probiotics 4 .
Even though the aquaculture probiotics are highly commercialized and widely used, effective implementation of regulations to assess the quality of probiotics is still lacking in most countries.
Microbial composition displayed on label of many commercial probiotics do not accurately match with their actual content 5 and the recent trend in aquaculture probiotic industry of formulating products with complex blend of microbes call out for methods ideal for effective analysis of the relative numbers of mixed bacterial species. The probiotic may not contain all the bacterial species claimed on the product label or contain contaminant microorganisms with undesirable effects on both animal and aquaculture environment. The reliability of traditional microbiological methods for enumeration of multiple strains of probionts present in a single probiotic is doubtful, due to low selectivity and reciprocal inhibition in plate 6 .
Even microbial enumeration with different culture media may not be able to detect all groups of microorganisms present in a probiotic product 6 . Molecular biology techniques like real-time quantitative PCR (RT-qPCR or qPCR) 8, 9,10 , uorescent in-situ hybridization (FISH) 11,12 , ow cytometry 13,14,15 , MALDI-TOF mass spectrometry 16 etc. were used to determine the microbial quality of commercial probiotics. Even though these techniques are able to detect and/or quantify the microbial species stated on the label, they have limited ability to detect microbes that are present as contaminants in the probiotic.
The variable and conserved regions of 16S rDNA makes it an ideal indicator for taxonomic classi cation of bacteria. The 16S rDNA sequencing of cultures isolated from the probiotic can be employed for identifying the microbial content of probiotic products. However, this method is laborious, time consuming and requires isolation of cultures from different media. High throughput sequencing using shorter regions of 16S rDNA produce more read depth and allow an effective analysis of complex bacterial species bypassing the need of culturing and limitations thereof. Next generation sequencing is being widely used for analyzing microbial diversity of great variety of environments, including soil 17,18 , sea 19,20 , air 21 and other agriculture food samples 22 . In this study Illumina sequencing based NGS using single 16S phylogenetic marker was employed to identify the microorganisms present in commercial aquaculture probiotics. This method could successfully be used to ascertain the veracity of the bacterial species declared on the label and also to determine the contaminant microorganisms present in the product.

Taxonomic Classi cation of Quality Reads
Based on the 16S amplicon sequencing of the V3 region for product codes 1, 2, 3, 4 and 5, quality reads OTUs respectively for product codes 1 to 5. The 16S rRNA gene is present among the members of the paraphyletic prokaryotic domains, Bacteria and Archaea 23 and sequencing of hypervariable regions of the gene can be used for identi cation of bacteria in a complex sample. The aim of our work was to identify bacterial species present in the products, hence processed reads falling under the bacterial domain were only considered for further taxonomic classi cation. The graphical depiction of taxonomic distribution of processed reads from each product is given in Fig. 1. From each hierarchy rank, sequences categorized as unknown were eliminated before calculating percentage of reads. Quality reads were found to be low at the kingdom level itself for products 3 and 4 which may be either due to sequences that were not categorised to any bacteria or the sequences were unknown. Due to the high sequence similarity between cyanobacteria and chloroplast 24 , some sequences under the phylum cyanobacteria could not be further resolved to any bacterial class and were recorded as chloroplast and considered as unknown from thereon. The high number of chloroplast reads in product 2 and 4 contributed to the reduced number of total quality reads of these products from class level onwards. The claim made by probiotic manufacturers on the microbial species present in the product and the percentage of reads that each species contributes in each probiotic as revealed by next generation sequencing is given in Table 1. The sequence reads of microorganisms that were declared on the label ranged from 0.28-79% for the probiotics tested. All the microorganisms claimed by the manufacturers were present in products 1 and 4 contributing 60.77 and 0.28% of the sequence reads respectively while the rest were contaminant microorganisms.

Qualitative Analysis of Metagenomic Sequences
The metagenomic sequences of all the products revealed discrepancies against product declarations and revealed incongruity between the label claims and the actual bacterial content of products (Supplementary table, S1). Microorganisms with sequence abundance greater than 0.01% were included in our results. All products contained several contaminant microorganisms when evaluated at both genus and species level. The ratio of number of bacteria actually present to the total number of bacteria claimed on the label for products 1 to 5 were 2:2, 1:5, 2:4, 1:1 and 3:9 respectively. The number of contaminant bacteria ranged from 5 for product 3 to 30 for product 4 at the genus level (Supplemetary Table, S1A).
The presence of Bacillus spp. was con rmed in all the screened probiotic products with Bacillus subtilis in four products, B. licheniformis in two products and B. coagulans and B. pumilus in one product each as expected. None of the products contained any species of Bacillus not speci ed in the labels, however products 2, 3 and 5 did not contain many of the Bacillus species mentioned in the labels. Pediococcus acidilactici was present in product 1 as per the label declaration, while in products 2 and 5 it was present as a contaminant. Two products that claimed to contain nitrifying bacteria; Nitrobacter and Nitrosomonas could not be detected even at the genus level. Propionibacterium acne was observed as the most occurring contaminant with incidence in three products. All tested products had at least one bacterial species which are either potentially pathogenic or of faecal origin; Acinetobacter baumannii, A. junii, Enterobacter cloacae, Enterococcus faecium, E. coli, P. acne and Pseudomonas aeruginosa.
It is evident that some organisms that were categorized till genus level were not grouped to its respective species (Supplemetary Tables, S1A and S1B). The entire quality reads sorted to a particular taxonomic rank may not be always resolved to its sub-categories in next level of hierarchy due to the inabilities of databases 25 . In product 2, 3 and 5, the Bacillus genera were not classi ed to its species level for 13, 12.1 and 0.45 % of the sequences, respectively.

Relative Abundance of Bacterial Species in the Probiotic Products
Relative percentage of abundant bacterial species (with ≥ 0.1% of total quality reads) present in the ve products is graphically represented in Fig. 2. The proportion of bacterial species present as contaminants ranged between 14.82 to 91.1% for the products tested. Among the bacteria declared on the label, Bacillus species constituted major proportion of three products with a share of 84.76, 83.02 and 80.72% in products 1, 3 and 5, respectively whereas the corresponding value were 9.27% in product 2 and 8.90% in product 4. Although we were able to con rm the presence of P. acidilactici in product 1, Bacillus subtilis in product 2 and 4, and Bacillus licheniformis and Bacillus pumilus in product 5 which were anticipated as per the labels, their relative abundance was negligible compared to the contaminants. In product 2 and 4, the ratio of contaminant organisms was almost 10 times higher than the declared bacteria on label.
The presence of fecal indicator organisms was observed in three products; E. cloacae (6.83%) in product 3, E. faecium (9.83%) in product 4 and E. coli (18.96%) in product 5. Incidence of multiple numbers of potential human infectious bacteria (A. baumannii, A. junii, P. acne and P. aeruginosa) with a total share of 22.39% in product 2 raises concern on health safety aspect of the probiotic. The presence of potential pathogenic microorganisms such as P. acne (9.59% in product 1 and 4.04% in product 4) and P. aeruginosa (0.88% in product 3) were also observed. The bacteria under genera Acidovorax, Staphylococcus and Streptococcus contributed 27.5 relative percentage in product 2. The relative abundance of contaminant bacteria; Streptococcus thermophiles (74.64%) was higher than that of its only declared organism, Bacillus subtilis in product 4.

Discussion
In the present study, metagenomic sequencing was employed to evaluate claims regarding bacterial composition on the labels of ve commercial aquaculture probiotics sold in India. A probiotic product is said to be reliable if bacterial strains are present in right composition and concentration as claimed on the label. Culture-based isolation and identi cation of probiotic bacterial strains may result in missing out of few bacterial species 26 . The traditional method of culture-based enumeration of speci c organisms will only identify bacteria that are able to grow and reproduce on synthetic media under de nite conditions 27 . High Throughput DNA sequencing can offer resolution required for examining all microbial communities including the contaminant and unculturable organisms in mixed samples to the preferred taxonomy up to strain level 28 . In this study, the microbial quality of probiotic products were evaluated through NGS technique by means of the discriminative power of V3 region sequencing of 16S rDNA. Identi cation of bacteria were done only up-to species level as the disclosure on labels were up to that level. NGS sequencing of the V2 and V3 regions of 16SrRNA were found to be most suitable for distinguishing most bacterial species up to the genus level 29 . NGS has the potential to detect unexpected microbes 30 with the advantage of being high-throughput having ability to generate large sequence data.
It was evident that all the probiotic products tested contained several contaminant bacteria and the microbial composition of the product were not in agreement with the label information of the product. Aquaculture probiotic products approved in Vietnam were found to contain bacterial strains not declared on the product labels. Also, many bacterial species including Bacillus spp. declared on the label were not isolated 26 . In our study, only 2 products (1 and 4) contained all bacterial species claimed by the manufacturer. The relative abundance of the sole declared organism, B. subtilis was just 8.9 % in Product 4. Bacillus is one of the most widely used bacteria incorporated in majority of the aquaculture probiotics. When Bacillus spp., is used in aquaculture pond, it can easily become part of the culture environment by competing with pathogenic or non-pathogenic bacteria 31 . Bacillus spp. was reported to act as a growth promoter 32 , inhibit pathogen 33 , improve nutrient digestibility 34 , enhance water quality 35 and ameliorate reproduction 36 . Bacillus spores are able to survive under various conditions including desiccation, freezing, elevated temperatures, chemicals, increased pressure and even radiation 37 . The ability of Bacillus spp. to form spores will help in the long-term storage of probiotic products.
In the present study, absence of B. licheniformis and B. coagulans in product 2, B. subtilis and Lactobacillus acidophilus in product 3, Alkaligenes faecalis, B. amyloliquefaciens, B. megaterium and B. polymyxa in product 5 and Nitrobacter sp., and Nitrosomonas sp. in products 2 and 5 highlighted the poor correlation between the label claims and actual content of the products. The nitrifying bacteria are bene cial as they increase the number of microbial species in water and improve the water quality by eliminating ammonia and nitrate toxicity 38, 39 . The accumulation of ammonia nitrogen and nitrite are two major toxic factors that directly affect the economic bene ts of aquaculture 40,41,42 .
One of the foremost concerns to be considered while evaluating the quality of commercial probiotics is the existence of unspeci ed organisms within the product. P. acne, an anaerobic bacterium causing acne and one of the most abundant organisms found on human skin 43 , is a common taxonomic group found in three out of the ve probiotic products analyzed. This organism was totally unexpected in the products and might have got incorporated into the product due to lack of good manufacturing practices and proper quality control during production process. Based on sequence analysis it was found that there were many microorganisms that were not declared in the labels but were found in the product. E. faecium was present in product 5 as a contaminant and this bacteria was found to decrease the mortality of European eel due to edwardsiellosis 44 . Even though Enterococci are Lactic Acid Bacteria, their application in animal and human health are hindered by occurrence of virulence and antibiotic resistant genes and pathogenicity 45,46 . Enterococcus spp. is neither recommended for the QPS status by the European Food Safety Authority (EFSA) 47  faecium isolates in Germany exhibiting resistance to vancomycin was 26.1% in 2017 51 . Arthospira platensis was present in product 3, a lamentous cyanobacterium widely used as a prebiotic in aquaculture because of its best-known nutritional qualities 52 . Interestingly in Product 4, an organism that was not claimed on the label, Streptococcus thermophiles, was found to be present. S. thermophiles is a moderate thermophile and a widely used human probiotic 53 . The presence of E. faecium, A. platensis and S. thermophilus in the products cannot be considered as unintentional. It is suggested that the product label should specify all the organisms that are incorporated in a product and their intended use. Several strains of S. thermophilus have also been shown to have several in vitro and in vivo probiotic properties ranging from the production of a range of short-chain fatty acids to moderately high survivability levels under simulated gastrointestinal tract conditions 54,55,56 .
The existence of human infectious bacteria; A. baumannii, A. junii and P. acne, human and sh pathogen; Pseudomonas aeruginosa and faecal contaminants; E. coli, E. cloacae and E. faecium in probiotic was unanticipated. The use of probiotics with contaminant bacteria carrying potential risks and of unknown provenance in aquaculture farms could cause a major shift in natural microbial community 57 . There is a high risk due to human exposure to the pathogenic contaminant bacteria present in aquaculture probiotics, during handling of such probiotics or probiotic fed aquatic animals and by dissemination to surrounding environment and terrestrial animals.
To our knowledge, this is the rst study that has employed NGS for quality analysis of commercial aquaculture probiotics. Inconsistencies in the bacterial composition against the label declarations and occurrence of Bacillus as the integral constituent of products observed in the current study were also observed in similar reports on the culture dependent evaluation of aquaculture probiotics 5,26 . The microbiological analysis of seven probiotics commonly used in Vietnamese shrimp culture, revealed that all the bacteria as stated on the label could not be detected while few bacteria that were not declared; Aerococcus sp., Klebsiella spp., and Bacillus spp. were present in the products 26 .
Since knowledge about the heterogeneity and growth requirements of contaminant bacteria in a probiotic are limited, isolation and identi cation of these bacteria will be di cult with traditional microbiological techniques or even with molecular methods. High-throughput based metagenomics studies can identify entire range of organisms that were not declared on the label. However, some closely related species would have little divergence among the variable regions of 16S rDNA. Therefore, discrimination of such species that belongs to the same group as Bacillus subtilis group will be di cult and the categorization will be terminated at the genus level 25 . Since 16S database is incomplete, bacterial species present in the tested products which were unclassi ed and recorded as unknown could not be identi ed. However, this weakness of the technique has not affected the results since the major focus of our work was to isolate contaminant microorganisms.
It is mandatory that all aquaculture probiotics sold in the Indian market have to be registered with Coastal Aquaculture Authority (CAA). The approval is solely based on the absence of antibiotics in the product. Many probiotics sold in the Indian market donot carry information on the bacterial composition and concentration of each species, the method or the frequency of application. The claims on the effectiveness of the products is often highly exaggerated. Aquaculture farmers purchase the probiotic products based on the claims of the merchandise, and hence misleading, incorrect, or deceptive labeling on products need to be avoided. The ndings from the current study raises concerns regarding the quality of aquaculture probiotics in the interest of the stakeholders including the farmers and the nal consumer. It is evident that all the probiotics screened in this study do not ful ll the basic requirement of labeling the exact content of the probiotic product. Besides, the occurrence of potential pathogenic organisms in the product is an added public health concern. This underlines the need for stringent quality control and regular screening by competent authorities to monitor the manufacture of aquaculture probiotics and implementing a legislation system for such an emerging and growing industry. The effect due to application of probiotic product shall clearly be mentioned on the label and the information provided on the label shall be evaluated based on on-farm trials before the product is approved for use.

Commercial Aquaculture Probiotics
Five brands of aquaculture probiotics marketed in India that revealed count and content of the bacterial species on the label procured from retail aquaculture shops in Andhra Pradesh, India were selected for the study. The products were assigned codes in the sequential order from 1 to 5. The labels on products 1, 3 and 4 provided bacterial composition up to species level while products 2 and 5 revealed species information for all bacteria other than for two bacteria belonging to genera; Nitrobacter and Nitrosomonas. All products were in powder form and were within the expiry period at the time of analysis. After procurement, based on the instructions on the label, the products were stored appropriately in cool, dry place and away from light till sampling was done. All probiotics were sampled on the day of opening of the packets.

DNA extraction and PCR ampli cation
Total DNA was extracted from 250 mg of probiotic product using the HiPurA™ Soil DNA Puri cation Kit (Himedia, India) following manufacturer's instructions. During sampling, thorough mixing and pooling of each product powder was performed aseptically. The V3 region of the 16S rRNA gene was ampli ed using the primers V3F (5'-CCAGACTCCTACGGGAGGCAG-3') and V3R (5'-CGTATTACCGCGGCTGCTG-3'). PCR ampli cation was carried out with 2X Master mix (Thermo Scienti c, US) in a Veriti 96 well Thermal Cycler (Thermo Fisher Scienti c, US) with the following PCR conditions; initial denaturation at 95 ℃ for 5 min, 35 cycles of 95 ℃ for 45 s, 59 ℃ for 45 s and 72 ℃ for 45 s, followed by a nal extension step at 72 ℃ for 5 min 29 .

Library preparation and sequencing
Library of V3 region was prepared using NEBNext® Ultra™ DNA Library Prep Kit for Illumina (NEB, USA). Manufacturer's instructions were followed for preparation of the amplicon for illumina MiSeq sequencing platform. Adapter ligated PCR amplicons were sequenced at AgriGenome Labs Pvt Ltd, India.

Preprocessing, processing, and taxonomic classi cation of sequences
The initial raw reads generated from sequencing were subjected to sequential stages of ltration. Spacer and conserved region were trimmed from the paired-end data and consensus V3 region sequence was constructed using ClustalO program 58 (Sievers et al. 2011). Chimeras were removed using the de-novo chimera removal software UCHIME 59 (Edgar et al. 2011) implemented in the USEARCH tool. The preprocessed reads obtained were pooled and clustered into Operational Taxonomic Units (OTUs) based on their sequence similarity using Uclust program 60 (Edgar 2010) (similarity cutoff = 0.97). QIIME 61 (Caporaso et al. 2010) program was used for the entire downstream analysis. Representative sequences were identi ed for each OTU and aligned using PyNAST 62 (Caporaso et al. 2009) with GreenGenes 63 (DeSantis 2006) core set. The reads corresponding to the OTUs after singleton removal were taken as the nal processed reads. Further these quality reads were aligned against reference chimeric data sets. The sequences that did not have any alignment with GreenGenes database were categorized as Unknown. Taxonomy classi cation to phylum, class, order, family, genus and species was performed using RDP classi er 64 (Wang et al. 2007) against SILVA OTUs database 65 (Quast et al. 2012).
by Department of Biotechnology, India and carried out at ICAR-Central Institute of Fisheries Technology, Kerala, India. The authors are thankful to Dr. Ravishankar C. N, Director, ICAR-CIFT for continuous support and guidance.

Data availability
The datasets generated during the current study are available from the corresponding author on reasonable request Figure 1 Taxonomic distribution of bacteria in aquaculture probiotics that conform to label information: Percentage of quality reads categorized to the respective hierarchy level, except the unknown sequences are represented. The shaded region denotes the percentage of quality reads that match to product label claims at the respective taxonomic rank