Isolation and comparative genomic analysis of reuterin-producing Lactobacillus reuteri from poultry gastrointestinal tract

Lactobacillus reuteri is a natural inhabitant of selected animal and human gastrointestinal tract (GIT). Certain strains have the capacity to transform glycerol to 3-hydroxypropionaldehyde (3-HPA), further excreted to form reuterin, a potent antimicrobial system. Reuterin-producing strains may be applied as a natural antimicrobial in feed to prevent pathogen colonization of animals, such as in poultry, and replace added antimicrobials. To date, only seven L. reuteri strains isolated from poultry have been characterized which limits phylogenetic studies and host-microbes interactions characterization. This study aimed to isolate L. reuteri strains from poultry GIT and to characterize their reuterin production and antimicrobial resistance (AMR) profiles using phenotypic and genetic methods. Seventy reuterin-producing strains were isolated from poultry crop, faeces and caeca and twenty-five selected for further characterization. Draft genomes were generated for the new 25 isolates and integrated in a phylogenetic tree of 40 strains from different hosts. Phylogenetic analysis based on gene content as well as on core genomes showed grouping of the selected 25 L. reuteri poultry isolates within the poultry/human lineage VI. Strains harbouring pdu-cob-cbi-hem genes (23/25) produced between 156 mM ± 11 and 330 mM ± 14 3-HPA, from 600 mM of glycerol, in the conditions of the test. All 25 poultry strains were sensitive to cefotaxime (MIC between 0.016 and 1 μg/mL) and penicillin (MIC between 0.02 and 4 μg/mL). Akin to the reference strains DSM20016 and SD2112, the novel isolates were resistant to penicillin, possibly associated with identified point mutations in ponA, pbpX, pbpF and pbpB. All strains resistant to erythromycin (4/27) carried the ermB gene, and it was only present in poultry strains. All strains resistant to tetracycline (5/27) harbored tetW gene. This study confirms the evolutionary history of poultry/human lineage VI and identifies pdu-cob-cbi-hem as a frequent trait but not always present in this lineage. L. reuteri poultry strains producing high 3-HPA yield may have potential to prevent enteropathogen colonization of poultry.


Introduction
Lactobacillus reuteri inhabits the gastrointestinal tract (GIT) of selected animals where it forms biofilms on the non-glandular, squamous epithelium lining the upper GIT. In poultry, L. reuteri is the most abundant Lactobacillus species in the GIT, mainly found in the crop and the caecum [1]. Distinct phylogenetic lineages of L. reuteri are coherent with host origin, reflecting co-evolution of this species with the vertebrate hosts [2]. The evolutionary adaptation differentiates the species in host-adapted phylogenetic lineages comprised of isolates from rodents (lineages I and III), humans (lineage II), pigs (lineages IV and V) and poultry/human (lineage VI) [2,3]. Host adaption has been linked to the occurrence of specific functional traits, e.g. rodent L. reuteri isolates possess the genes responsible for synthesis of urease, as the strains are constantly exposed to urea in the forestomach of mice [4].
Genomes of poultry and human L. reuteri isolates (lineages II and VI) have been shown to harbour the pdu-cbi-cob-hem operon, as a lineage specific trait [5]. This operon contains genes for glycerol and propanediol utilization (pdu) and for cobalamin biosynthesis (cbi-cob), hem genes and some accessory genes. Cobalamin is a cofactor for glycerol/diol dehydratase PduCDE (EC 4.2.1.30). PduCDE catalyzes the conversion of 1,2-propanediol to propanal, which can be further metabolized by other enzymes of the pdu operon to propanol or propionate [6]. Glycerol, a second substrate of PduCDE, is transformed to the intermediate 3-hydroxypropionaldehyde  which can be further metabolized to 1,3-propanediol or 3-hydroxypropionate [7]. 3-HPA produced from glycerol is released from the cell forming the dynamic multicompound reuterin system, with broad antimicrobial spectrum and consisting of 3-HPA, its hydrate and dimer and acrolein [8,9]. Acrolein, a highly reactive toxicant, was recently shown to be the main component for the antimicrobial activity of reuterin [10,11].
Due to the high persistence of L. reuteri in the poultry GIT and the established antimicrobial activity of reuterin, L. reuteri has high potential to be applied as a natural antimicrobial in feed to prevent pathogen infection of animals [12]. L. reuteri strains isolated from poultry GIT were shown effective against Salmonella spp. and Escherichia coli resistant to various antibiotics [13]. Moreover, L. reuteri in the early post-hatching period had a delayed effect on ileum microbiota of poultry, which resulted in the enrichment of potentially beneficial lactobacilli and the suppression of Proteobacteria [14]. To select functional L. reuteri strains, a key trait is the determination of their antimicrobial resistance (AMR) profiles to identify intrinsic and extrinsic resistances that may be potentially transferred. Lactobacilli are known to be intrinsically resistant against vancomycin. However, the occurrence of tetracycline and erythromycin genes on mobile elements has been reported for different Lactobacillus spp. [15].
It was therefore the aim of this study to isolate and characterize L. reuteri strains from poultry and characterize their reuterin production and AMR profiles using phenotypic and genotypic methods. Draft genomes of the isolates were analyzed combined with 40 L. reuteri genomes of strains previously isolated from different hosts to assess genetic diversity and gain insight into distinguishing features related to poultry, and enrich previously phylogenetic characterization of L. reuteri.

Material and methods
Bacterial strains and growth conditions L. reuteri DSM20016 (Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany) and L. reuteri SD2112 (BioGaia AB, Stockholm, Sweden) were used as reference strains. Reference strains as well as all L. reuteri isolated in this study were propagated anaerobically (Oxoid, AnaeroGenTM, Basingstoke, UK) at 37°C in de Man, Rogosa and Sharpe (MRS) broth medium (Biolife, Milan, Italy).

Bacterial isolation
Six Lohmann brown poultry (13 weeks old) were obtained from six poultry farms in Switzerland. Crop, caeca and faeces were aseptically collected. In parallel, 10 whole gut of Cobb 500 broiler poultry were obtained from Schönholzer Werner abattoir in Wädenswil (Zurich), and transported to the lab within 1 hour.
L. reuteri strains were isolated from poultry crop using the protocol previously described [4], with some modifications. Briefly, one gram of crop, caecal or faecal content was added to 10 mL of sterile phosphate buffer saline (PBS) (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4, pH 7.4) and homogenized in a stomacher (BagMixer® 400 P, Interscience, Saint Nom, France) at high speed for 1 min. Suspensions from samples were serially diluted and spread on mMRS plates [16].
The agar plates were incubated overnight at 42 o C under anaerobic condition using AnaeroGen 2.5 L (Thermo Fisher Diagnostics AG, Pratteln, Switzerland). Replica plates were prepared using Scienceware replica plater and velveteen squares (Sigma-Aldrich, Buchs, Switzerland), and incubated overnight as presented above. After incubation, one plate was overlaid with 500 mM glycerol agar (1% agar) and incubated at 37 °C for 30 min for testing reuterin production of colonies. A colorimetric method was used with the addition of 5 mL 2,4-dinitrophenylhydrazine (0.1% in 2 M HCl), 3 min incubation, removal of the solution, addition of 5 mL 5M KOH. L. reuteri colonies showing purple zones indicating reuterin synthesis were streaked on MRS agar plates and single colony were sub cultured 3 times in MRS broth (1% inoculum, 18 h at 42 o C). Few negative colonies which show a colony morphology of L. reuteri but no purple zone were picked as negative controls. Species confirmation and reuterin production quantification was performed for both positive and selected negative selected colonies.

Bacterial identification
Genomic DNA was isolated using a lysozyme-based cell wall digestion followed by the To confirm the identity of the isolates, 1.6 kbp full region of 16S rRNA gene was amplified by PCR using universal primers bak4 (5'-AGGAGGTGATCCARCCGCA-3') and bak11w (5'-AGTTTGATCMTGGCTCAG -3') [17,18]. The 16S rRNA PCR assay consisted of 5 min at 95 °C, followed by 35 cycles of 15 s at 95 °C , 30 s at 60 °C and 2 min at 72 °C and final extension for 7 min at 72 °C . Sanger sequencing of the PCR amplicon was performed at GATC (Konstanz, Germany). To identify the closest homologs, DNA sequences obtained were aligned using Basic Local Alignment Search Tool (BLAST) [19]. Sequence homology greater than 97% was used to identify L. reuteri.

Generation and annotation of draft genomes
Genomic DNA from the isolated strains was obtained as described above and standardized to 100 ng/µL. The whole genomes of 25 L. reuteri isolates were sequenced with the standard set of 96 Illumina paired-end barcodes on a HiSeq 2500 Illumina Technology (Illumina Inc., San Diego, USA) with 2x 125 high output mode.
The genomic library was generated using reagents from NEBNext Illumina preparation kit. Raw paired-end reads were quality trimmed using the default settings of Trimmomatic [23]. Read pairs were merged by FLASH [24]. De novo assembly was performed with SPAdes assembler (version 3.12) with -careful option [25]. The quality of the assembly based on evolutionarily-informed expectations of gene content from near-universal single-copy orthologs selected from OrthoDB v9 was assessed by BUSCO [26] and QUAST [27]. Reference guided ordering of scaffolds based on iterative alignment steps was performed by QUAST using L. reuteri DSM20016 genome as reference. SeqKit [28] and QUAST were used to retrieve genome features.
The complete genome of L. reuteri DSM20116 type strain was compared individually with each of the 25 draft genomes of L. reuteri poultry isolates. Average nucleotide identity (ANI) values between two genomic datasets were calculated using JSpecies [29]. Genomes with ANI values above 95 % were considered as belonging to the same species [30].

Comparative genomics
Comparative genome analysis was based on gene content tree, core genome phylogenetic tree and a nucleotide-content similarity matrix (ANI matrix). For the generation of the gene content tree, a matrix based on gene content (binary data for presence or absence of each annotated gene) was generated comprising all annotated genes of 25 L. reuteri draft genomes plus the 40 L. reuteri NCBI genomes. The gene content tree was then constructed using the hierarchical cluster analysis (hcust) on R 3.4.4 while a core genome phylogenetic tree was calculated using EDGAR 2.3 [32].
Concatenated sequences were used to calculate a distance matrix which provided the input for the neighbour-joining method with PHYLIP implementation. EDGAR 2.3 calculated the core genome of each identified clusters as a set of orthologous genes present in all strains belonging to each cluster. For the generation of the similarity matrix, ANI was calculated on EDGAR 2.3 with an all-against-all comparisons at nucleotide level for all 65 L. reuteri strains.
Identified lineages were named based on host origin of the strains, and lineages specific features were determined using "Define metacontigs" function in EDGAR 2.3.
Core groups were created for the sets of strains derived from the phylogenetic analysis. A Venn diagram was then designed in EDGAR to identify shared and unique genes among the group of strains. Unique genes of poultry/human lineage VI were manually categorized based on UniProt protein description into the following groups: The presence of AMR genes in the assembled genomes was also manually checked and integrated with the results from ResFinder 3.0 tool [33].
For penicillin resistant L. reuteri strains, point mutations (SNPs) in penicillin-binding proteins genes ponA, pbpX, pbpF and pbpB were checked by aligning deduced amino acid sequences of resistant and sensitive strains with NCBI-deposited sequences for Lactobacillus in Geneious 9.1.8.
PCR for tetW and ermB detection PCR for tetW and ermB genes was performed to confirm the genomic data. TetW was amplified with tetw-rev and tetw-fw primers [15,34]

Reuterin production
Reuterin production was determined using a two-step process [35].  Considering the small sampling size, a high diversity of L. reuteri strains was found in this study. The colorimetric method applied allowed an easy phenotypic isolation of reuterin-positive colonies, that accounted for approximately 50 % of the bacterial colonies obtained on MRS isolation plates. This result indicated the high frequency of reuterin-producing L. reuteri strains in the poultry GIT, compared to the absence of reuterin production for rodent isolated strains [4].
Isolates from caecum did not appear to have a unique genetic profile when compared to isolates from faeces, and this could be due to the transition of strains from caecum to faeces, when not colonizers.
For the first time here, several L. reuteri strains were isolated from poultry and this will significantly enrich the database which will now account for 32 poultry strains making it 19 % of the total number of deposited L. reuteri strains, compared to 5 % before.
Genomes analysis of L. reuteri poultry strains of this study The whole genomes of 25 L. reuteri strains were sequenced and draft genomes were characterized. Strains had an average genome size of 2.159  0.17 Mbp (Table 1).
BUSCO assembly assessment showed good quality of assembly with 431 to 433 single-copy orthologous on a total of 433 for all assembled genomes (Supplementary Figure S2). Shotgun reads were assembled into contigs higher than 500 bp, ranging from 68 (PTA4_C4) to 657 (PTA5_C6B  Figure S3).

Comparative phylogenetic analysis of L. reuteri strains from different hosts
The apparent relatedness between microbial community composition in the gut and host phylogeny has been interpreted as evidence of coevolution [39]. Symbiotic gut microbes associated with the host are predicted to evolve host-specific traits and, as a result, display enhanced ecological performances in their host [36,40]. To assess evolution and adaptation of L. reuteri strains to different hosts, the gene content of poultry isolates was analyzed together with that of 40 L. reuteri strains available by NCBI, obtained from different hosts: human (6), rat (1), mouse (3), pig (4), sourdough (4), goat (5), sheep (4), cow (4), horse (3) and poultry (6). The gene content tree, in which strains sharing more genes clustered together, identified three main clusters namely cluster I, cluster II and cluster III, that contained previously observed L. reuteri lineages [2]. Those host-adapted lineages were first described after the characterization of the genetic structure of L. reuteri strains isolated from human, mouse, rat, pig, chicken and turkey, and the same lineage names were also applied in our study for coherency [2]. Cluster I, corresponding to the previously defined poultry/human lineage IV, comprised all 25 L. reuteri poultry isolates of this study and all, except one (P43), poultry NCBI isolates. The same cluster also included two humans strains (SD2112 and CF48-3A) and this was also the case in all previous phylogenetic analysis of L. reuteri isolates from different hosts [2,36,38]. The fact that even with a much higher number of poultry isolates the two human isolates, respectively isolated from human breast milk in Peru (SD2112) and from the child faeces in Finland (CF48-3A), still cluster together with 29 poultry isolates, suggests that these strains could be of poultry origin. In a previous study, administering human isolates of lineage VI were shown to colonize the poultry GIT and therefore may be a necessary colonization factors indicating rather co-evolution events of human and poultry isolates [36]. In our study, cluster II included the majority of herbivorous isolates (new defined herbivorous lineage VII of our study) in addition to four human (DSM20016, MM2_3, IRT and JCM1112), one sourdough (CRL1098), one rodent (mlc3) and one porcine (20-02) strains, belonging to lineages II, III and IV previously defined [2]. Cluster III was composed of pig, herbivorous, sourdough and rodent isolates ( Figure 1) corresponding to lineages I, III, V and newly defined in this study: herbivorous VIII. ANI analysis of the 65 L. reuteri isolates (Supplementary Figure S4) identified the same three clusters, with exception of published strains 20_02 and mlc3 that were assigned to cluster III instead of cluster II.
The phylogenetic tree based on core genomes of the 65 genomes covered a core of 1152 genes per genome, for a total of 74880 genes. In agreement with the gene content tree, all isolated L. reuteri poultry strains clustered together with NCBI poultry isolates (strains JCM1081, CSF8, An71 and An166), except P43, forming poultry/human lineage VI [2]. As indicated above, this cluster also included the two human isolates CF48-3A and SD2112 (cluster I, Figure 2

Genes unique for the poultry/human lineage VI
Presence of host specific lineages in itself does not necessary provide evidence for natural selection, as a cluster can arise by neutral processes, such as genetic diversion [2]. It has been demonstrated how strains from rodent display elevated fitness in mice, and biofilm formation in the forestomach is restricted to strains from rodent lineages.
Moreover, L. reuteri rodents strains were able to effectively colonize rodent host in vivo [2,5,36]. However, this was not the case for pig isolates [36,37]. Here, forty unique genes of the poultry/human lineage VI were identified (Supplementary Table S3) and were mainly categorized as transport proteins DNA-binding proteins and transferase proteins. Such genes could not be directly linked with adaptation to poultry physiology or feeding (Figure 3). Further studies are needed to elucidate the specific role (s) of those unique genes that may be linked to poultry adaptation.

Reuterin synthesis
The presence and composition of reuterin operon genes (pdu-cbi-cob-hem) was investigated in all 65 genomes ( Figure 4) and reuterin production was determined as a marker for PduCDE activity. All L. reuteri strains isolated in this study that possessed the complete pdu-cbi-cob-hem operon produced 3-HPA when incubated in 600 mM glycerol. In contrast strains PTA5_11 and PTA8_1 only possessed hemH, hemA, cobC and cobB but lacked all the others operon genes ( Figure 4) and therefore did not produce reuterin. Under the conditions of the test, 3-HPA yield ranged from 156.9 mM ± 11.0 (PT6_F1) to 330.2 mM ± 14.9 (PTA4_C4) starting from 600 mM glycerol ( Figure   2). Reference strains DSM 20016 (human lineage II) produced 132.8 ± 4.3 mM while SD2112 (human lineage VI) produced 432.9 ± 9.0 mM.
Cluster I strains (corresponding to poultry/human lineage VI) all harbored pdu-cob-cbihem genes, except for a small sub-cluster of three by reuterin-negative poultry isolates CSF8, PTA5_11 and PTA8_1. Isolates assigned to Cluster II possessed pdu-cob-cbihem and have been shown to mostly form reuterin on MRS agar plates overlaid with 500mM glycerol agar [4]. Isolates of this cluster lacked pduW and hemN genes, which therefore seem not essential for reuterin production. These genes were also not detected in the vast majority of reuterin-positive strains isolated in our study (Figure 4).
In contrast, the prevalence of pdu-cob-cbi-hem scattered in Cluster III comprising isolates of rodent lineages I and III, herbivorous and porcine lineages VIII and V, respectively. Only strains ATCC53608 and ZLR003 possessed a complete functional pdu-cob-cbi-hem operon while cluster III rodent isolates lpuph and 100_23 lack the majority of the operon genes.
Interestingly, herbivore isolates LR6, LR7, LR12 and the rodent isolate I49 possessed the pdu but not the cbi and cob genes while cobalamin is a cofactor for 3-HPA production. Therefore these strains are likely not able to form 3-HPA from glycerol (or propanal from 1,2-propanediol) and are reuterin-negative (indicated with red branches in Figure 1 and Figure 2) unless they acquire the vitamin from other sources or microbes. Rodent strains, which are mostly reuterin-negative based on the analysis of operon genes in this study, have been previously suggested as being at the root of the evolutionary history of L. reuteri-host associates [36]. This data suggests that the pducbi-cob-hem operon and thus reuterin production was acquired later during evolution by L. reuteri strains in rodents and also in poultry/human lineage VI strains.
Antimicrobial sensitivity profiles of L. reuteri strains The horizontal transfer of AMR genes is a rising risk concern, and the absence of transferable AMR genes must be demonstrated for application of new strains in food and feed (WHO, 2017). Antimicrobials used in farmed animals for diseases prevention have been associated with an increase of frequency of resistant bacteria in chickens, swine, and other food-producing animals GIT [41]. The high use of antimicrobials in animal production is likely to accelerate the development of antimicrobial resistance in pathogens, as well as in commensal organisms, resulting in treatment failures, economic losses and source of gene pool for transmission to humans [41]. Poultry is one of the most widespread food industries worldwide and various antimicrobials are used to treat infections mainly in young poultry [42,43].
The antimicrobial susceptibility profile of L. reuteri poultry isolates and reference strains DSM20016 and SD2112 showed that all strains were sensitive to cefotaxime (MIC values from 0.016 to 1 µg/mL). All poultry isolates were also sensitive to penicillin with MIC values from 0.02 and 3 µg/mL, in contrast to DSM20116 and SD2112 that showed resistance to this antibiotic (MIC> 256 ug/mL). Penicillin resistance were shown to result from point mutations of the chromosomally located genes encoding penicillinbinding proteins Pbp [44]. Penicillin-binding genes pbpX, pbpF, pbpB and ponA were identified in all 65 strains with both resistant and sensitive phenotype. Several SNPs were observed for DSM20016 (Table 3) showed resistance phenotype (MIC > 256 µg/mL) to erythromycin, confirmed by the presence of ermB, which is usually found on a plasmid [45,46]. The ermB gene encodes enzymes that modify the 23S rRNA by adding one or two methyl groups, reducing the binding to the ribosome of different classes of antibiotics [47]. The presence of ermB in the genome of resistant strains was confirmed by using PCR (data not shown). Among the 40 NCBI strains analyzed in this study, ermB was also detected in the genome of poultry isolate CSF8 (Figure 1 and Figure 2). Erythromycin resistance has been found in Bacillus cereus, Pseudomonas aeruginosa, C. jejuni, C. coli, Clostridium perfringens and Enterococcus poultry isolates [43].
Tetracycline resistance genes tetA and tetO were detected in all 65 L. reuteri genomes analyzed, but did not appear to be directly correlated with this resistance phenotype (  Table S2).
Tetracyclines are a family of compounds frequently employed due to their broad spectrum of activity as well as their low cost, compared with other antimicrobials [48].
However, in vivo transferability of tetW from L. reuteri ATCC55730 to other human gut microbes in a double-blind clinical study was not demonstrated [50].
Lactobacilli are suggested to be intrinsically resistant to vancomycin and ciprofloxacin [51]. In agreement with previous studies [15], all poultry L. reuteri strains of this study were resistant to vancomycin (MIC > 256 µg/mL). Vancomycin resistance in lactobacilli has been shown to be linked to the vanX gene encoding a d-Ala-d-Ala dipeptidase [52]. Other vancomycin resistance genes were described in the literature with vanA, vanB, vanC and vanE [34,52]. None of those genes were detected in the genomes of L. reuteri in agreement with previous studies [34,52]. However, changes in membrane composition have also been associated with intrinsic vancomycin resistance [53].
VanH, a D-lactate dehydrogenase gene was detected in one pig strain (ATCC 53608) [54]. The same gene had been previously associated with vancomycin resistance of L. reuteri has been affiliated to different hosts, which might be exposed to different levels and types of antimicrobials and is commercially used as probiotic in food and feed [11,12,58,59]. The results of this study indicated that L. reuteri poultry isolates harbour some AMR genes. In view of application L. reuteri in feed to prevent pathogen infection, strains without transferable AMR genes must be carefully selected. The first applied L. reuteri probiotic strains (SD2112) harbours the tetracycline resistant gene tetw on a plasmid [34]. This strain was however curated for the plasmid free daughter strain DSM17938 [44], which is used for commercially.

Conclusion
In conclusion, this study substantially enriched the pool of poultry L. reuteri strains for comparative genomic and evolutionary studies. The phylogenetic analysis confirmed and straightened the co-evolution of human isolates of lineage VI with poultry (poultry/human lineage VI). However, due to the high number of poultry isolates in this lineage compared to only two human isolates, we speculate a possible cross contamination during isolation of the two human strains belonging to lineage VI. The pool of L. reuteri poultry isolates of this study may be useful to select and characterize high potential strains exhibiting reuterin activity, and develop application in poultry, as a natural antimicrobial system to prevent pathogen infections and colonization of the poultry GIT.