Natural Bred ε2-Phages Have an Improved Host Range and Virulence against Uropathogenic Escherichia coli over Their Ancestor Phages

Alternative treatments for Escherichia coli infections are urgently needed, and phage therapy is a promising option where antibiotics fail, especially for urinary tract infections (UTI). We used wastewater-isolated phages to test their lytic activity against a panel of 47 E. coli strains reflecting the diversity of strains found in UTI, including sequence type 131, 73 and 69. The plaquing host range (PHR) was between 13 and 63%. In contrast, the kinetic host range (KHR), describing the percentage of strains for which growth in suspension was suppressed for 24 h, was between 0% and 19%, substantially lower than the PHR. To improve the phage host range and their efficacy, we bred the phages by mixing and propagating cocktails on a subset of E. coli strains. The bred phages, which we termed evolution-squared ε2-phages, of a mixture of Myoviridae have KHRs up to 23% broader compared to their ancestors. Furthermore, using constant phage concentrations, Myoviridae ε2-phages suppressed the growth of higher bacterial inocula than their ancestors did. Thus, the ε2-phages were more virulent compared to their ancestors. Analysis of the genetic sequences of the ε2-phages with the broadest host range reveals that they are mosaic intercrossings of 2–3 ancestor phages. The recombination sites are distributed over the whole length of the genome. All ε2-phages are devoid of genes conferring lysogeny, antibiotic resistance, or virulence. Overall, this study shows that ε2-phages are remarkably more suitable than the wild-type phages for phage therapy.


Introduction
Uncomplicated urinary tract infections (UTI), which are caused predominantly by uropathogenic Escherichia coli (UPEC) are amongst the most common reasons for medical consultation [1]. UPEC strains show different constellations of virulence factors and diverse genetic backgrounds. Several international studies performing DNA profiling revealed the predominance of the sequence types (ST) 69, 73, 95, and 131, among others, in isolates from patients with UTI [2][3][4][5][6][7]. These STs differ markedly in their antibiotic resistance profiles. While STs 69, 73, and 95 are largely susceptible except rarely extended-spectrum β-lactamases (ESBLs), ST131 isolates often carry ESBLs together with fluoroquinolone and aminoglycoside resistance [8]. Oral antibiotics are preferred for the treatment of strains, have generated phage mutants that can switch to utilize alternative receptors when under selection pressure [37][38][39]. Nevertheless, the determination for an increased host range of these mutants is sparse. Gibson et al. (2019) also claimed a successful host range expansion for E. coli phages, although the study does not provide supporting data [27]. Kuniska & Tanji (2010) showed the recombination of phage genomes after infecting one E. coli strain with a cocktail of three phages but did not compare the efficacy of these with their ancestors [40].
In this study, we isolated new E. coli phages from wastewater and showed their improvement towards therapeutic use in UTI treatment by intercrossing different wildtype phages by superinfection, and by selecting the best progeny (ε 2 -phages), which showed an improved host range and increased virulence.

Isolation of Phages from Waste Water Reveales 28 Novel Phages
Several lytic phages were isolated from wastewater samples. Twenty-eight phages showing clear plaques with activity against E. coli strains were selected for further characterization. Sequence analysis revealed that all these phages belonged to the order Caudovirales, tailed phages having double stranded DNA genomes [41]. The most predominant family among the sequenced phages was Autographiviridae (n = 11), followed by Myoviridae (n = 7), Siphoviridae (n = 7), and Podoviridae (n = 3). At the subfamily and genus level, Autographiviridae phages belonged to the subfamily Molineuxvirinae (Vectrevirus or unspecified genus), and Studiervirinae (Kayfunavirus genus). Phages from the Myoviridae family belonged to the Tevenvirinae subfamily, the majority characterized as members of the genus Tequatrovirus and one belonging to the genus Mosigvirus. Siphoviridae phages were classified into subfamily Guernseyvirinae (Kagunavirus genus) and genus Dhillonvirus, while all podoviruses belonged to the Kuravirus genus (Table 1). [x]-no subfamily assigned. # -all strains, except for UTI89 (ST95, O18) and 172859 (ST unknown, O25b), are O25b-antigen positive (identified by PCR as described by [42]), belonging to the 131 sequence type (DSM-strains-ST131 affiliation is assumed on the basis of the O25b antigen, but has not been proven). *-plaquing host range percentage measured against a panel of 79 E. coli strains originating from UTI and representing different serotypes (see Table S1).
To validate the classification achieved by sequence comparison, selected phages were additionally analyzed by transmission electron microscopy (TEM). The phages CHD94UKE2, 101117BS1, and CHD16UKE1, classified as Tequatrovirus, showed a long contractile tail with short spikes characteristic for the family Myoviridae ( Figure 1). The morphology of phage 101120B2 showed an isometric phage with a short, noncontractile tail as expected for the family Autographiviridae. Phage 101114BS3 was assigned to Kuravirus of the Podoviridae by sequence comparison, and its TEM picture revealed an elongated capsid and a short tail, previously seen in other Podoviridae [30]. The TEM picture of phage 22664BS2, belonging to the genus Kagunavirus of the Siphoviridae family, showed the characteristic long, non-contractile tail ( Figure 1).

The Plaquing Host Range Does Not Correlate with the Ability of Phages to Control Growth in Suspension
To gain a first insight into the efficiency of the isolated phages, the susceptibility of a panel of 79 E. coli strains, isolated from urinary tract infections (Table S1), was tested. Here, the formation of clear plaques was defined as phage susceptibility of the host strain. Overall, the plaquing host range (PHR) ranged from 61% (Autographiviridae 22664UKE3-2) to 1% (Siphoviridae UTI89UKE1) ( Table 1).
For further characterization of the host range of phages intended for the treatment of UTIs, we compiled a panel of 47 E. coli strains reflecting the most prominent sequence types of uropathogenic E. coli (UPEC), including antibiotic resistant and multiresistant strains ( Table 2 and Table S2). We selected phages with a promising host range for further characterization on the UPEC panel including the myoviruses CHD16UKE1, CHD94UKE2, G3G7, and 101117BS1, as well as members of the Autographiviridae, 22664UKE3-2, 101120B2, and 101118B1. The efficiency of plaquing (EOP) of each of the seven phages was determined against the UPEC panel. The plaquing host range (PHR) was defined as the % of strains where clear plaques were formed. Strains on which the phage spot locally impaired growth ("opaque lysis"), but no clear plaque was formed, were not counted for the PHR. The myovirus G3G7, as well as the Autographiviridae 22664UKE3-2, showed the highest PHR with each 49% followed by CHD16UKE1 (43%), CHD94UKE2 (34%), 101118B1 (16%), 101120B2 (13%), and 101117BS1 (9%) (Figure 2). In addition, we tested to which extent the bacterial growth in suspension could be impaired by the phages. Therefore, we measured what we termed, kinetic host range (KHR). The KHR is defined as the number of strains showing less than 10% growth in suspension after phage treatment compared with their untreated control. The percentage of growth was determined by OD600. The KHR values were remarkably lower than the PHRs for every phage, ranging between 0% (101118B1) and 20% (CHD16UKE1) (Figure 1). Phages G3G7 and 22664UKE3-2 showed a particularly large difference between the PHR and KHR, with a reduction of the host range by approximately 94 percentage points and 76 percentage points, respectively, in the KHR compared to PHR ( Figure 2).
To analyze the correlation between the EOP and the ability to inhibit the growth of bacteria in suspension after 24 h, we defined the OD ratio as the ratio of optical densities of the phage-treated and untreated bacterial suspensions. Anticorrelation is expected in this analysis, i.e., that growth in suspension (quantified as OD ratio) is inhibited best on strains where the EOP is high. However, this was the case only for a few bacterial strains ( Figure 2). As depicted in Figure 2b,c, the correlation was poor. All phages failed to inhibit bacterial growth in suspension for 24 h even on strains where they have a high EOP.

Breeding of Myoviridae Phages Improves Their Kinetic Host Range against Uropathogenic E. coli
To improve the host range of the wild-type phages, mixtures of two different sets of phages were bred by incubation with 24 E. coli strains (Table S2) in artificial urine for 20 rounds, essentially as described [29]. The first breeding cocktail contained the myoviruses CHD16UKE1, CHD94UKE2, 101117BS1, and G3G7. A second breeding cocktail included the three Autographiviridae 22664UKE3-2, 101120B2, and 101118B1. The host range of the single phages isolated after breeding (ε 2 , or evolution squared phages) was determined by plaquing and by the OD ratio on the panel of 47 UPEC strains described above ( Table 2). The PHR of the ε 2 myoviruses ranged from 47-51%, which was not improved over the best ancestor G3G7, but higher than the ancestors with the best KHR, CHD16UKE1 (PHR = 43%) and CHD94UKE2 (PHR = 34%). Strikingly, four of the six ε 2 phages with morphology of the Myoviridae tested showed an improvement of 18-23% in KHR compared to their most efficient ancestor phage, CHD16UKE1 ( Figure 3). Therefore, the ε 2 Myoviridae seem to have acquired the best combination of properties with regards to maximizing both the PHR and KHR. This effect is further visualized in Figure 3b, showing that the bred phages had both a higher KHR and PHR compared to the ancestors. In addition, for these four ε 2 Myoviridae phages, we analyzed the correlation between the EOP and OD ratio after 24 h. As depicted in Figure 3c, the correlation of the EOP and OD ratio was still limited, but low OD ratios were reached for more strains compared to the ancestors. This directly results in a higher KHR, as shown in Figure 3a,b.
In the breeding group of Autographiviridae, ε 2 -phages showed no improvement at either plaquing or kinetic host range over their ancestors ( Figure S1).
Next, we analyzed whether the activity of the phages could be linked to the phylogenetic relationships of the UPEC strains. All Myoviridae ancestor phages were isolated from wastewater by propagation on various E. coli strains of the sequence type 131, one of the most prevalent UPECs. The host range data of the individual tested E. coli strains revealed that the phages were most active on strains and sequence types within the B2 phylogenetic group to which the ST131 belongs ( Figure 4). Furthermore, multiple E. coli of the phylogenetic groups A and D, less so B1, were susceptible to the Myoviridae phages. Here, not all representative strains tested of a phylogenetic group or sequence type showed the same susceptibility ( Figure 4). Comparison of the kinetic host range data after 6 h and 24 h incubation revealed that the phages could initially inhibit the growth of up to 60% of the bacteria tested, comparable to the plaquing host range. Nevertheless, after 24 h this number dropped distinctly suggesting a putative resistance evolution of the bacteria (Figure 4).

Myoviridae ε 2 -Phages Are More Virulent than Their Ancestors
We further analyzed whether the breeding process, in addition to expanding the host range, also improved the virulence, or killing efficacy, of the ε 2 -phages over their ancestors. We examined the growth of a subset of 22 of the 47 UPEC strains (Table S2) in the presence of selected ε 2 -phages at different multiplicities of infection (MOI). This subset panel included 12 strains of the most prevalent phylogenetic group B2 and is further described in Table S1. The phage titer was kept constant at 1 × 10 7 PFU/mL and was used to infect bacterial suspensions with MOI 100-0.1. At MOI 100, the highest MOI tested, the Myoviridae ancestors could only inhibit the growth of 30-50% of the strains used. In contrast, the ε 2 -phages restrained the growth of 55-79% of the strains at MOI 100 ( Figure 5).
Reducing the MOI decreased the overall number of bacteria whose growth could be restricted. Nevertheless, the ε 2 -phages, especially PM123, were more efficient compared to their ancestors with the same MOI. Furthermore, the efficacy of PM123 at MOI 1 (31% of strains) was comparable to the efficacy of the ancestors at MOI 10 (range 17-33% of strains). At MOI 10, PM123 was even more efficient in inhibiting bacterial growth (62% of strains) than any of the ancestors at MOI 100 (range 30-50% of strains) ( Figure 5). These results show an increase of phage virulence after the breeding, in the panel of bacterium tested.

The Genomes of ε 2 -Phages Are Intercrossed from Up to 3 Ancestors
The ε 2 -phages resulting from the Myoviridae breeding were classified as Tequatrovirus by sequence similarity. Each of the Myoviridae ε 2 -phages showed extensive genomic recombinations, and genome stretches were inherited from up to four different ancestor phages. The backbone of all Myoviridae ε 2 -phages consists largely of the ancestors CHD16UKE1 (A in Figure 6a) and G3G7 (B in Figure 6), which were also the ancestor phages with the highest KHR. Phages are aligned against PM123 and genome sequence were divided in the predicted CDS (1 to 279), and each CDS is given a color according to sequence similarity with the ancestors. As indicated in the legend, the phage CHD16UKE1 (A) is represented in pink, G3G7 (B) in green, CHD94UKE2 (C) in blue and 101117BS1 (D) in orange. Mixed colors indicate two possible ancestors. The grey CDS represent a sequence similarity to more than two ancestors, while white CDS represents the lack of relation to any of the ancestors. CDS with a predicted role in tail or baseplate formation are marked in black in the last column. (b)~10 kbp genome region of PM123, which is a mosaic of ancestor phages CHD16UKE1 (A), G3G7 (B) and CHD94UKE2 (C). Dashed lines indicate similarity to both ancestors. Arrows indicate recombination sites. Annotation of predicted CDS is indicated, referring to its number. Arn-putative anti-restriction nuclease.
As the identity of CHD16UKE1 and G3G7 was 99.5%, several genomic stretches of the progeny were identical to both (represented as "AB" in Figure 6). Stretches of ancestors CHD94UKE2 and 101117BS1 were distributed differently along the genomes of the progeny. In PM123 and PM136, genomic stretches from the ancestor CHD94UKE2 comprised 11% and 7% of their total coding regions, respectively, and the ancestor 101117BS1 comprised only 1% in both cases. In contrast, the ancestor CHD94UKE2 made up 1 to 2% and 101117BS1 2-8% of the genomes of PM105, PM107, PM108, and PM133 ( Figure 6).
A more detailed analysis of an~10 kb genomic region (protein coding sequence (CDS) 117 to 133) of phage PM123, one of the phages with improved virulence, revealed at least eight recombination sites (Figure 6), where the PM123 genome switches from one ancestor to another. CDS 117 was identical to the ancestors CHD16UKE1 and G3G7, whereas CDS 118 to 123 were identical to the ancestor CHD94UKE2 and CDS 124 was inherited from the ancestor G3G7 ( Figure 6). CDS 130 encodes for a tail fiber protein and was inherited from CHD94UKE2. However, it also contained a mutation (highlighted by a green box in Figure 6b), not presented in any of the ancestors, which translated into an amino acid change (Q166R). This confirms that the breeding process not only promotes recombination, but also allows for the occurrence of spontaneous mutations. Moreover, CDS 131, another tail fiber-encoding gene, was identical to ancestor CHD94UKE2 for the first 2.3 kbp, then switched to ancestors CHD16UKE1 or G3G7 (identical in this stretch) for approximately 400 bp, before switching back to CHD94UKE2 for the last 300 bp ( Figure 6).
All ε 2 -phages recovered from the Autographiviridae mix were classified as Kayfunavirus indicating the absence of the positive selection or the dilution of the Vectrevirus 101118B1 ancestor during the Autographiviridae breeding. The genomic analysis of the ε 2 -phages revealed a distinct mosaicism, where individual genome stretches are inherited from different ancestor phages. The individual Autographiviridae ε 2 -phages had a very similar genome arrangement with a backbone inherited from 101120B2. Only CDS 010 could be unambiguously identified as inherited from ancestor 22664UKE3-2, while several CDS might be inherited from either of both phages (101112B2 or 22664UKE3-2, Figure S2). The uniqueness of each Autographiviridae ε 2 -phage was defined by different single point mutations, the majority of them accumulated in CDS 001 (putative transglycosylase) and CDS 002 (K5 lyase).
We further analyzed the phage life cycle with PHACTS, screened the genome for antibiotic resistance genes with CARD and ResFinder 4.1, and for virulence factors in Virulence Finder 2.0. All ε 2 -phages (Myoviridae and Autographiviridae) were devoid of genes conferring lysogeny, antibiotic resistance, or virulence, and were therefore suitable for phage therapy in principle.

Discussion
Increasing antibiotic resistance rates in UPEC requires alternative treatment options. In this study, we isolated new phages infecting UPEC and improved their therapeutic potential by cross-breeding. We describe for the first time the phage breeding by the principles of the Appleman's protocol [23] for improving E. coli phages towards therapeutic use.
In order to direct the phage activity towards UPEC, we isolated the phages on strains of sequence type 131, one of the most prevalent in UTI. After an initial characterization, the most promising ones according to host range were selected for breeding. Superinfections of phage cocktails of 3-4 ancestors on a defined set of UPEC strains were used to induce genomic recombination and create bred phages, which we termed ε 2 -phages (evolution squared). The breeding process increased the host range of the ε 2 -phages by 23% compared to the best ancestor and increased their virulence.
Comparing the plaque formation and the killing of planktonic bacteria in suspension revealed significant differences in the determined host range with less efficacy of phages in the kinetic measurements. Recently, we observed this phenomenon for S. aureus phages as well [29]. The KHR after 6 h was comparable to the PHR, indicating that initial phage efficiency is comparable in both assays. However, the 24 h KHR determination showed resistance formation of the bacteria against the phages. This rapid regrowth was previously described for E. coli and for other Gram-negative bacteria such as Salmonella and Pseudomonas [43][44][45]. Since the ability of plaque formation does not reflect the phage capability of controlling or reducing bacterial growth, which might be related to the formation of resistance, the use of the "phagogram" [33,46] as the sole predictor of the therapeutic success could be questioned. In fact, some specialized centers are already testing activity in broth, in addition to the traditional spot testing before phage therapy [47].
The Myoviridae ancestor and ε 2 -phages were more efficient against ST131 strains or other sequence types belonging to the same phylogenetic group, B2, such as ST73. Other phylogenetic groups such as A and D showed less susceptibility towards the phages. Nevertheless, the pathogenic E. coli strains causing extraintestinal infections mainly belong to group B2 and to a lesser extent to group D, whereas commensal strains belong to group A and B1 [48]. Hence, the breeding is successfully driving phage capabilities towards the treatment of UTI infections. The fact that the host range of the ε 2 -phages was increased and not limited to ST131could confer an advantage upon treatment over the narrower host-range of phages previously described [43].
Although the panel used for breeding was diverse, the E. coli B2-phylogenetic group was predominant, and the selective pressure of this group might have influenced the enhanced virulency towards these strains of Tequatrovirus ε 2 -phages, which showed an increase in host range and virulence. Previous studies have pointed to gp37 as the driver of host range expansion in a T4-related phage [49]; however, the number of mutations and recombination found in our ε 2 -phages makes it virtually impossible to identify which genes caused this improvement. Contrarily, Holtzman et al. (2020) showed that phage T7 specializes and contracts its host range towards certain types of LPS if the other LPStypes do not allow productive infection [37]. Our breeding could have had a similar impact in Autographiviridae phages, which did not improve their host range. The diverging outcome of the breeding protocol for the different phage families could be due to i) the fact that genetically more similar Myoviridae-phages were used in the initial ancestor cocktail, allowing more chances of recombinations or ii) due to the Myoviridae larger genome, which might allow more recombinations, compared to the Autographiviridae phages.
Recently, using a similar method, we created natural ε 2 -phages of S. aureus showing a host range twice as broad as their ancestors [29]. The fact that the host range improvement in the ε 2 -phages of E. coli is not as large could be related to the fact that E. coli strains are more diverse. The core genome of E. coli comprises only up to 25% of the genes, while the remaining are considered as accessory genomes [50][51][52]. In contrast, approximately 75% of S. aureus genes were classified into the core genome [53].
Although phage cocktails were not formulated in this study, the genomic differences of the ε 2 -phages could allow the creation of complementary cocktails. Since adsorption inhibition seems to be the main phage-resistance mechanism identified in E. coli [54], the identification of specific phage receptors and subsequently creating cocktails targeting different receptors has the potential to reduce resistance formation and therefore could drive therapeutic success [55]. Furthermore, the breeding was carried out in artificial urine, which might confer an advantage to the ε 2 -phages during UTI treatment.
In conclusion, this study describes ε 2 -phages lysing E. coli, bred from newly isolated wild-type phages. Homologous recombination and selection led to phages with superior killing in suspension (KHR) and higher virulence than the wild-type phages. The specialization of these phages towards certain ST types of E. coli, independent of its antibiotic resistance, could confer an advantage for the therapeutic treatment of UTI. We show that it is possible to substantially improve key phage properties required for therapeutic success by applying directed evolution with limited knowledge of the underlying functions of the altered parts of the genomes. The ε 2 -phages described in this study may not only have increased therapeutic efficacy compared to the ancestors, but, due to the increased kinetic host range, may have also reduced the number of phages required in a fixed cocktail or in a phage bank for personalized phage therapy.

Phage Collection and Propagation
Phages were isolated from wastewater samples collected in 2018 from the Charité (Berlin, Germany), the DSMZ (Braunschweig, Germany) and the University Medical Center Hamburg-Eppendorf (Hamburg, Germany) at the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures.
Briefly, wastewater samples were filtered through a 0.22 µm-pore filter and mixed with several E. coli strains grown to log phase (OD 600 nm = 0.4). The mixture was then added to liquid soft agar (0.4% w/v) and poured onto a solid agar plate. After overnight incubation at 37 • C, clear plaques indicating the presence of lytic phages were harvested using a pipette tip and resuspended in an SM-buffer (100 mM NaCl, 50 mM Tris-HCl, 8 mM MgSO 4 and 0.01% gelatin). The suspension was mixed with the log-phase growing E. coli propagating strain in liquid soft agar and poured onto a solid agar plate. Clear plaques were suspended in an SM buffer after overnight incubation followed by streaking on a double agar plate. At least 4 consecutive single plaque isolates were processed for a pure phage preparation, which was used for further propagation. In general, logarithmic growing cultures were infected with phages and incubated at 37 • C 180 rpm until visible lysis or overnight. After centrifugation and filtration (ULTRAFREE-CL GV 0.22 µm Centrifugal Devices (Merck Millipore, Germany)), the titer of the lysate was determined using the double agar method and the lysate was stored at 4 • C.

Transmission Electron Microscopy (TEM)
For TEM analysis, phages were treated as previously described [57]. Briefly, phages were allowed to adsorb onto thin carbon support films. Afterwards, they were negatively stained with 2% (w/v) aqueous uranyl acetate, pH 5.0. Samples were examined in a Zeiss EM-910 or Zeiss Libra120Plus transmission electron microscope (Carl Zeiss, Oberkochen, Germany) at an acceleration voltage of 80/120 kV at calibrated magnifications using a crossed line grating replica. Size determination of heads and tails was performed using the ITEM Software (Olympus Soft Imaging Solutions, Münster, Germany) from at least 3-10 different phage particles and further classification was conducted according to Ackermann [58].

Plaquing Host Range (PHR) and Efficiency of Plaquing (EOP)
Logarithmic growing bacterial cultures were diluted in liquid soft agar (0.4%) to a concentration of~1 × 10 7 CFU/mL. The soft agar was poured onto rectangular solid agar plates. After solidification, 3 µL of SM buffer serial-diluted phage lysates were dropped onto the bacteria-containing layer. Plates were kept at 37 • C and plaques were counted after several hours or overnight incubation. EOP was calculated as the ratio of the PFU measured on the strain under investigation and the host strain was used for propagation. The PHR was defined as the percentage of strains showing clear plaque formation. Strains with no clear plaque formation but impairment of growth at the site of the phage spot ("opaque lysis") were not counted as plaque forming. PHR and EOP determinations were performed twice.

Kinetic Host Range (KHR)
Bacterial overnight cultures and phage lysates were diluted in fresh artificial urine to approximately 2 × 10 7 CFU/mL and 2 × 10 8 PFU/mL, respectively, and mixed 1:1 (= MOI 10) in a 96-well plate with a final reaction volume of 150 µL. Bacteria without added phages were used as untreated control. Plates were incubated overnight at 37 • C in a humidified atmosphere and OD 600 was measured after 6 and 24 h incubation using the Multiskan ™ FC microplate photometer (Thermo Scientific, Waltham, MA, USA). The KHR was defined as the percentage of strains showing no growth after 24 h. Therefore, OD-ratios between untreated control and phage-treated wells were calculated and an OD ratio of < 0.1 was defined as no growth. KHR was determined twice.

Phage DNA Isolation
Phage DNA isolation was carried out with the Phage DNA Isolation Kit (Norgen Biotek, Thorold, ON, Canada) or with DNeasy Blood & Tissue Kit (Qiagen, Hilden, Ger-many) based on the instructions provided by the manufacturer, with slight modifications. Phage lysates were concentrated, if necessary, prior to DNA extraction by precipitation with 10% PEG 8000 and 1M NaCl as described [59] to ensure concentrations of ≥ 10 8 PFU/mL. Prior to DNA isolation, phages lysates of at least 10 8 PFU/mL were DNase and RNase treated to avoid contaminants from the bacterial host. The DNase was then inactivated by adding EDTA and heating up to 65 • C. For the Phage DNA Isolation Kit, Lysis Buffer B was added, and the sample was vortexed vigorously for 10 s, followed by 15 min incubation at 65 • C, mixing the tube by inversion every 5 min. After adding isopropanol, the sample was flown through the provided column and allowed to bind. The column was washed by adding Wash Solution A and centrifuging at 6000× g for 1 min. To elute the DNA, the column was incubated for 1 min with Elution Buffer B and centrifuged for 1 min at 6000× g. For the DNeasy Blood & Tissue Kit, samples were treated with Proteinase K for 30 min at 55 • C before adding buffer AL, mixing thoroughly, and incubating for 10 min at 70 • C. After adding ethanol, the sample was flown through the provided column and allowed to bind. The column was washed by adding buffer AW1 and subsequent AW2 and centrifuging at 6000× g for 1 min. To elute the DNA, the column was incubated for 1 min with nuclease-free water and centrifuged for 1 min at 6000× g. Samples were stored at 4 • C. The concentration was determined using the Qubit ® dsDNA HS Assay Kit (Thermo Fisher Scientific) following the manufacturer's instructions.

Library Preparation, Whole Genome Sequencing and Genomic Analysis
An amount of 0.5 ng DNA was enzymatically fragmented using the Nextera DNA Library Preparation Kit (Illumina, FC-121-1030) for 10 min at 55 • C. After the addition of 11 µL KAPAmaster mix (Library Amplification Kit, Kapa Biosystems) and 4.4 µL of each Nextera index primer, the reaction mixture was incubated according to the following program: 3 min at 72 • C, 5 min at 98 • C, 13 × (10 s at 98 • C, 30 s at 62 • C, 30 s at 72 • C), and 5 min at 72 • C. For PCR clean-up and size selection, the PCR product was incubated with magnetic beads (Ampure XP, Beckman Coulter). The pool of DNA libraries was adjusted to 4 nM and sequenced using MiSeq technology (Reagent Kit v3, 600-cycle, MS-102-3003, Illumina).
Before assembly, the sequencing reads were trimmed using prinseq-lite 0.20.4 [60]. Quality checks of the reads were performed with FastQC 0.11.9 [61]. SPAdes genome assembler version 3.13.0 and 3.14.1 [62] was used for the assembly of Illumina reads. Read mapping was conducted with the map to reference feature of Geneious (Geneious Prime 2020.01.02.) with medium sensitivity. Phage genome annotation was conducted with Prokka version 1.14.5 [63].
Analysis of genome mosaicism was conducted by comparing each CDS sequence of the bred phages with the ancestor sequence, determining the best matching ancestor. This was conducted manually using the progressive Mauve algorithm in build into Geneious.
Taxonomic classification was inferred from the closest sequenced relatives identified by BLAST analysis of phage sequences to NCBI's RefSeq database.

Breeding of Phages
Phage breeding was conducted based on the principles of the Appleman's Protocol [23]. Briefly, 3 to 4 ancestor phages were pooled to create the input phage mixture which was used (diluted and undiluted) to infect each of the subpanels of 24 E. coli strains in a 96-well microtiter plate. Two separate phage combinations were used, a mixture of the Myoviruses CHD16UKE1, CHD94UKE2, 101117BS1, and G3G7, and a mixture of members of the Autographiviridae, 22664UKE3-2, 101118B1, and 101120B2. For the first round of breeding, single phages with similar concentrations were mixed and then serial 10-fold diluted in artificial urine. Starting concentrations were 2 × 10 7 PFU/mL for myoviruses and 2 × 10 6 PFU/mL for Autographiviridae. Bacterial overnight cultures were diluted in fresh artificial urine to~1 × 10 8 CFU/mL. Bacteria and phage dilution volumes were mixed 1:1. After overnight incubation at 37 • C, clear lysates and the first dilution of turbid lysates were pooled, sterile filtrated, and used to infect the next round of breeding. This was repeated for up to twenty rounds. Single individual phages were isolated from the bred mixture lysate by re-streaking on the host strain at least 4 times.

Determination of Phage Efficiency
Phage virulence was examined by determination of the bacterial growth after incubation of a constant phage concentration with different numbers of bacteria. In detail, 22 E. coli strains with a starting inoculum of 1 × 10 5 , 1 × 10 6 , 1 × 10 7 and 1 × 10 8 CFU/mL were mixed with 1 × 10 7 PFU/mL (MOI 100, 10, 1 & 0.1) of each ancestor and selected ε 2 -phages in artificial urine in a 96-well plate. Plates were incubated overnight at 37 • C and OD 600 was measured after 24 h incubation. OD-ratios between the untreated control and phage-treated wells were calculated and an OD ratio of <0.1 was defined as no growth. Determinations were performed in triplicate.

Statistics
One-and two-way Anova analysis were performed with GraphPad Prism 9.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/antibiotics10111337/s1, Figure S1: Host range of Autographiviridae ε2-phages. Comparison of plaquing (PHR) and kinetic (KHR) host range over the panel of 47 E. coli strains between ancestor (bold) and ε2-phages, Figure S2: Sequence analysis of ε 2 -Autographiviridae phages; Table S1: Composition of the 79 E. coli strains originating from urinary tract infections used for initial HR determination of wastewater isolated phages, Table S2: Overview of the 47 E. coli strains used for comparison of PHR and KHR and further characterization.