For this study, we isolated 5 E. coli strains and 4 P. aeruginosa from urine collected from healthy Gyr heifers at the Agricultural Research Company of Minas Gerais State (EPAMIG) in Brazil. In total urine samples were collected from 10 different heifers in this herd. The reproduction within the herd is controlled by fixed time artificial insemination and there is no presence of bulls. This herd has a medium milk production of 3,700 lg/lactation/cow and a rigorous sanitary control. None of the animals from this study presented clinical reproductive signs for 12 months before sample collection. The sample collection and following experiments were previously approved by the Ethics Committee in Animal Experimentation of the Universidade Federal de Minas Gerais, Brazil (CEUA/UFMG - 40/2019). Prior to the urine sampling, the external genitalia of the animals were cleaned with soap and water. The mid-stream urine was collected using a sterile 50 ml centrifuge tube, frozen (−20°C), transported to the lab and processed. All samples were processed within 48 h of sampling.

Four of the urine samples were processed using the same technique as follows. First, samples were aliquoted and spun down. 500 μL of the supernatant was spread on an LB agar plate, incubated overnight at 37°C and the individual colonies were picked. The colonies were regrown in LB agar overnight at 37°C and this process was repeated at least 3 times to obtain pure colonies. The pure single colonies were grown in liquid LB media overnight at 37°C. Samples were spun down and the DNA was extracted using the Qiagen DNeasy UltraClean Microbial Kit following the manufacturer’s protocol. All of the isolates were submitted to sequencing of the 16S rRNA region using the 63F/1387R primer pair to identification. Sanger sequencing was performed by Genewiz (New Brunswick, NJ, United States) using each primer individually, providing 2x coverage. Raw reads were then manually trimmed, assembled and queried against the NCBI 16S rRNA Sequences Database via the blastn algorithm. Based upon the 16S rRNA gene sequence, four isolates from four different animals were identified as P. aeruginosa and five isolates, from the same four animals, were identified as E. coli. Isolates of other taxa have been described elsewhere (Giannattasio-Ferraz et al., 2020a,b,c,d,e, 2021a,b).

Whole genome sequencing was performed for all 9 isolates. The extracted DNA was sequenced at the Microbial Genomic Sequencing Center (MiGS) (Pittsburgh, PA, United States). For sequencing, the libraries were prepared using the Illumina Nextera kit and the genomes were sequenced using the NextSeq 550 platform. Raw reads were trimmed using Sickle v1.33¹ with a quality threshold (Phred quality score = 20) and length threshold after trimming of 100 nucleotides. Trimmed reads were then assembled using SPAdes v3.13.0 with the “only assembler” option for k values of 55, 77, 99, and 127 (Bankevich et al., 2012). To calculate genome coverage for the assemblies, BBMap v38.47² was used. The raw reads and genome assemblies were deposited in GenBank, Table 1 shows the accession number for each BioSample deposited. The assembled genomes were annotated using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP) v4.11 (Tatusova et al., 2016).

Genome assemblies also were examined with the Center for Genomic Epidemiology’s (CGE) tools PlasmidFinder v2.1 (Carattoli et al., 2014), using the Enterobacteriaceae database and a 90% threshold for identity and 60% minimum coverage, and ResFinder v3.2 (Zankari et al., 2012) with default parameters. The serotype for E. coli strains was verified using CGE’s SerotypeFinder v2.0 (Joensen et al., 2015). P. aeruginosa serotypes were determined using CGE’s Past v1.0 (Thrane et al., 2016). E. coli phylotype were determined using EzClermont (Waters et al., 2020). The VFAnalyzer tool was used to predict virulence factors (Liu et al., 2019). All of the strains were also screened for the CRISPR/Cas system (Grissa et al., 2007). Genes for flagellar synthesis, flagellar rotation, chemotactic signal transduction, and chemotactic membrane receptors were identified within the genome sequences using reciprocal blasts and manual curation. Local blast databases were created with the PGAP annotation files of the bovine urinary strains; blastn queries were used to identify the presence of flg (flgA-flgN), flh (flhA-flhE), fli (fliA, fliC-fliT, and fliZ), mot (motA, motB) and che (cheA, cheB, cheR, cheW, cheY, and cheZ) genes as well as chemotactic receptors tap, tar, trg, tsr, and aer. These sequences were manually inspected. Unless previously noted, default parameters were used for each tool.

The 5 bovine urinary E. coli and 4 bovine urinary P. aeruginosa strains were first compared to bovine isolates from other niches (Supplementary Table 1). In the case of E. coli, 3049 bovine genomes are publicly available; the vast majority are from fecal samples (n = 1199). A subsample of these genomes were selected, limiting genomes to those isolated (regardless of sample type) from South America or vaginal samples [given prior evidence that uropathogens are frequently introduced from the vaginal tract (Yeruham et al., 2006)]. All bovine P. aeruginosa genomes were considered (n = 20). Genomes were retrieved from NCBI and compared to the bovine urinary genomes by ANI (Average Nucleotide Identity) using pyANI and the ANIm measure (Pritchard et al., 2016). Shared genic content was determined via reciprocal blastn queries. 95% nucleotide sequence identity was used as the threshold for these blasts, which were conducted locally using the blast + executable (v2.9.0). These other bovine isolates were also characterized using CGE’s SerotypeFinder v2.0 (Joensen et al., 2015), CGE’s Past v1.0 (Thrane et al., 2016), and EzClermont (Waters et al., 2020), as described above.

A comparative analysis was performed using publicly available genomes and genome assemblies of isolates from human urine. Urinary isolates were identified by referencing GenBank sequencing metadata (isolation source). Only quality assemblies were considered, based upon their completeness and checkM contamination score through PATRIC (Davis et al., 2020). The 5 bovine E. coli isolates were compared to 900 human urinary E. coli genomes/assemblies and the 4 bovine P. aeruginosa isolates were compared to 221 human urinary P. aeruginosa genomes/assemblies. The strains retrieved and their accession numbers are listed in Supplementary Table 2. The genomes were used to identify the pangenome of both species using anvi’o v6.2 (Eren et al., 2015). The pangenome was created using the anvi-pan-genome with mcl-inflation 10 and the other parameters were default. The concatenated core gene sequences were found using the anvi-get-sequences-for-gene-clusters command. Whole genome comparisons were conducted using the Mash program (Ondov et al., 2016) with default k and sample size (s) of 1,000,000.

The phylogenetic trees were derived using FastTree v2.1.11 (Price et al., 2010) plug-in through Geneious Prime using default parameters. Resulting trees were then visualized using iTOL (Letunic and Bork, 2016).

All E. coli and P. aeruginosa urinary genomes/assemblies included in our comparative genomics analysis (Supplementary Table 2) were uploaded to the tool PHASTER for prophage prediction (Arndt et al., 2016). PHASTER predicts incomplete, questionable and intact prophage regions. Only intact regions were used for our analysis. The intact nucleotide sequences were queried against NCBI nr/nt database’s viral sequences (TaxID: 10239) to predict the taxonomies of the prophages. GraphPad Prism v8.8.1 was used to illustrate taxonomic predictions of prophages found. Homologous gene sequences between predicted phages were identified anvi’o v6.2 (Eren et al., 2015). Gene calls were made using anvi’o and homologous proteins were identified with a minbit of 0.35. This choice of threshold is informed by prior work identifying homologous phage proteins [see (Shapiro and Putonti, 2021)]. These homologous gene clusters were then processed using a python script written to generate a gene presence/absence matrix and an edge list. This script is available through our GitHub repository.³ The resulting edge list was visualized using Cytoscape v3.8.2 (Shannon et al., 2003).

We isolated 5 E. coli strains and 4 P. aeruginosa from the urine of 4 healthy Gyr heifers and sequenced their genomes (Table 1). For each animal, both E. coli and P. aeruginosa were isolated; 2 E. coli strains were isolated from the same individual (UFMG-H7A and UFMG-H7C). All 5 E. coli genomes are characterized by the H34 flagellar antigen and are representatives of the phylotype B1, and all 4 P. aeruginosa genomes belong to the O5 serotype. Sequencing did not reveal the presence of plasmids within any of the strains. The CRISPR/Cas system was identified in all 5 of the E. coli genomes (type I-E). Furthermore, all 5 E. coli genome assemblies were found to include fimbriae (Supplementary Table 3) as well as genes for flagellar synthesis, flagellar rotation, chemotactic signal transduction, and chemotactic membrane receptors. Antibiotic resistance genes were only detected in the P. aeruginosa genome assemblies; all 4 P. aeruginosa strains are predicted to have the same resistances Phenicol (catB7), Beta-lactam (bla_OXA–50 and bla_PAO), Fosfomycin (fosA), and Aminoglycoside [aph(3′)-IIb].

The bovine urinary E. coli and P. aeruginosa genomes were first compared to bovine isolates via ANI. These genomes include isolates from fecal, milk, nasopharynx, vaginal, and internal organs. As expected, high ANI values were observed for these different strains of the same species (Supplementary Figure 1). Of note, the bovine urinary E. coli and P. aeruginosa genomes clustered distinctly from genomes from other isolation sites. Shared genic content between the bovine urinary genomes ranged from 63.46 to 93.25% for the bovine E. coli isolates and from 78.51 to 96.17% for the bovine P. aeruginosa isolates (Supplementary Table 1). The bovine urinary E. coli strains are representatives of the B1 phylotype; strains from bovine fecal and vaginal samples are also associated with this phylotype (Supplementary Table 1). In addition to the bovine urinary P. aeruginosa strains, genomes from bovine nasopharynx and milk isolates are also representatives of the O5 serotype.

Next, we compared the bovine urinary E. coli and P. aeruginosa genomes to all publicly available genomes of human urinary isolates. The pangenome and single copy number core genome was identified for 905 urinary E. coli strains (inclusive of the 5 bovine E. coli strains sequenced here) and for the 225 urinary P. aeruginosa strains (inclusive of the 4 bovine P. aeruginosa strains sequenced here).

The analysis of the 900 human and 5 bovine urinary E. coli isolates identified 23,678 homologous genes. The majority (75.47%) of the E. coli genomes contained one or more genes that were unique to their genome, i.e., they were not present in any of the other urinary E. coli genomes examined here. UFMG-H7C included 201 genes that were unique to its genome, while the other 4 bovine isolates had no unique genes. Given the diversity in gene content observed among the 905 urinary E. coli genomes in this pangenome, a small single copy gene core genome was identified, which had 346 genes. Using this core, a phylogenomic tree for the E. coli strains was derived (Figure 1). This tree places the 5 bovine urinary isolates (shown in bold red) in a single clade (red), within the B1 phylotype. These bovine urinary strains have a core genome most similar to human isolates from individuals with either asymptomatic bacteriuria (strains: ABU_1) or UTI (strains: 1724, NGE5, UMEA 3065-1, UMEA 3292-1), per GenBank metadata.

The pangenome and core genome also was computed for the 225 urinary P. aeruginosa genomes. 19,257 homologous genes were identified in the genomes, with 96.89% of these genomes containing at least one gene sequence unique amongst the group. Nevertheless, 1,546 of the homologous genes are single copy core gene sequences, shared amongst all urinary P. aeruginosa genomes. The phylogenomic tree of these core genes places the 4 bovine P. aeruginosa isolates together (Figure 2, red). The bovine clade’s core genome is most similar to that of strain TC4411 (Accession No. GCF_008033805), collected from human urine in Besancon, France in 2017; however, the donor’s symptom status is unknown. Other closely related genomes include those isolated from individuals with UTI (strains: AZPAE15053 and AUS430) or symptom status unknown (strain: MRSN1688).

Last, we investigated if the bovine urinary genomes were more similar to other bovine isolates or to human urinary isolates. The 5 bovine urinary E. coli genomes were compared to all of the bovine genomes previously considered in our ANI analysis (Supplementary Table 1) and the human urinary isolates most closely related to the bovine urinary genomes (genomes in the insert of Figure 1). The most closely related genome sequence for the 5 bovine urinary isolates was the human urinary isolate E. coli ABU 1. On average, the bovine urinary isolates were as similar to the human urinary isolates (n = 900) as the bovine vaginal isolates (n = 9) (Supplementary Table 4). The bovine urinary P. aeruginosa genomes were most similar to the human urinary genome strain TC4411 and more similar to the human urinary genomes than the genomes isolated from other bovine samples (Supplementary Table 4).

Prophage prediction was also conducted for all of the human urinary genomes and the bovine urinary genomes (Supplementary Table 3). Only high confidence (predicted to be intact) prophages were considered for downstream analysis. A total of 2,663 prophages were identified. Sequence similarity to characterized phage genomes enables us to assign all but 14 of these prophage sequences to one of 6 different viral families (Figure 3C); 14 predicted prophages exhibited no sequence similarity to any characterized phages (“Unclassified”).

The 5 E. coli UFMG strains include one intact prophage sequence each, which are identical at the nucleotide level and are 40.1 kb long, encoding for 51 viral proteins including tail and head associated proteins, repressor proteins and integrase. The bovine prophages exhibit greatest sequence homology (61% query coverage and 97% sequence identity) to the myovirus Shigella phage SfII (Accession No. NC_021857). The bovine prophages exhibited sequence similarity to prophages identified within the human urinary E. coli genomes. They shared genes, and thus are connected in Figure 3A, to 1,767 human urinary prophages. The number of shared genes ranged from 1 to 46. We further investigated human urinary prophages that shared over 80% of their genes with the bovine urinary prophages. Seven prophages met this threshold, harbored by the following E. coli strains: 21_fCAUTI, HM46, UMB1346, UMB1347, UMB1354, UMB1359, and UMB5337. Sequence alignment found that these sequences had 72-84% nucleotide sequence identity to the bovine urinary prophages; the five UMB strains are identical. Referring to the core phylogeny of the urinary core genome identified the bacterial strains containing these prophages were phylogenetically distant to the 5 bovine urinary E. coli strains (Supplementary Figure 2).

Among the P. aeruginosa strains analyzed, 495 phages were predicted belonging to 4 different viral families; nine predicted prophages exhibited no sequence homology to characterized phages (“Unclassified”) (Figure 3B). Each of the bovine P. aeruginosa strains contained one predicted intact prophage, which were identical (100% nucleotide sequence identity). This prophage is 34.9 kb long (63.96% GC content) with 42 viral proteins, including an excisionase, two hydrolases and tail associated proteins. While identical across the bovine urinary strains, these prophage sequences had very little sequence similarity to a previously characterized phage [3% query coverage and 76% sequence identity to Pseudomonas phage vB_Pae_BR141a (Accession No. MK510991.1)]. Given the placement of the bovine P. aeruginosa prophages in Figure 3B, the bovine phages are predicted to belong to the order Caudovirales.

Homologs to the bovine urinary prophage also were identified in the human urinary P. aeruginosa genomes. The bovine urinary prophage encoded genes homologous with 224 prophages in human urinary strains, ranging from a single similar gene to all 42 annotated genes. In fact, the bovine urinary prophage sequence was identical to a prophage harbored by P. aeruginosa TC4411 (Accession No. GCF_008033800), the same strain that the bovine urinary core genomes claded with in Figure 2. It also exhibited over 98% sequence identity with 10 other human urinary P. aeruginosa strains. These prophages were found in P. aeruginosa strains closely related to the bovine urinary isolates, as well as more distantly related P. aeruginosa strains (Supplementary Figure 3).

Most of the prophages identified for the two species are predicted to be tailed phages (order: Caudovirales) (Figure 3C). The predicted prophages are most frequently predicted to be representatives of Siphoviridae and Myoviridae, 85.96% and 79.96%, respectively. For the two bacterial hosts, however, the Siphoviridae presented higher percentage numbers in the urinary P. aeruginosa strains in comparison to Myoviridae in E. coli. Furthermore, Inoviridae (order: Tubuvirales) were detected more frequently in the urinary P. aeruginosa strains and Podoviridae in the E. coli strains.