In Vitro Evolution to Increase the Titers of Difficult Bacteriophages: RAMP-UP Protocol

Abstract

Background:

Bacteriophages are becoming increasingly important in the race to find alternatives to antibiotics. Unfortunately, bacteriophages that might otherwise be useful are sometimes discarded due to low titers making them unsuitable for downstream applications.

Methods:

Here, we present two distinct approaches used to experimentally evolve novel New Zealand Paenibacillus larvae bacteriophages. The first approach uses the traditional agar-overlay method, whereas the other was a 96-well plate liquid infection protocol that improved phage titers in as little as four days. We also used a mathematical model to probe the parameters and limits of the RAMP-UP approach to rapidly select mutants that improve bacteriophage titers.

Results:

Both experimental approaches resulted in an increase in plaque-forming units (PFU/mL). The liquid infection approach developed here, which we call RAMP-UP for Rapid Adaptive Mutation of Phage - UP, was significantly faster and simpler, and allowed us to evolve high titer bacteriophages in as little as four days. Titers were increased from 100-100,000-fold relative to their ancestors. The resultant titers were sufficient to extract and sequence DNA from these bacteriophages. An analysis of these phage genomes is provided.

Conclusion:

The RAMP-UP protocol is an effective method for experimentally evolving previously intractable bacteriophages in a high-throughput and expeditious manner.

Introduction

Bacteriophages (phages) are the most abundant entities on the planet, with an estimated number of 1 × 10³¹ virus particles, this number has been named the Hendrix product, a homage to Roger Hendrix, who was the first to describe it.^1,2 The discovery, purification, and sequencing of Actinobacteriophages in undergraduate classrooms through the adoption of the Howard Hughes Medical Institute-sponsored SEA-PHAGES (Science Education Alliance Phage Hunters Advancing Genomics and Evolutionary Science) program have brought personal empowerment through discovery to undergraduate classrooms around the globe.^3,4

This experience has also given some the impression that phages are universally discovered and isolated with ease. The facile and frequent discovery of new mycobacteriophages belies the lived experience of researchers who have embarked on phage hunts, only to discover that phages, even some mycobacteriophages, can be challenging to work with.⁵ For example, one of the steps in novel phage isolation is to achieve titers of over ∼5 × 10⁹ PFU/mL, a commonly used threshold concentration of particles for electron microscopy and DNA extraction.

Historically, physiological studies of phages have often been conducted at concentrations of at least 1 × 10⁸ PFU/mL,^6,7 although these concentrations come with their own challenges.⁸ Achieving these high concentrations of phage particles can prove difficult for reasons that are not clear, and can subsequently derail the discovery, description, and application of newly discovered phages.

DNA sequencing is a key step in the evaluation of phages for practical use. DNA sequencing reveals aspects of the life cycle and gene content of novel phages that were not obvious by other means, including their genetic diversity and gene content, genes associated with lysogeny, host toxins, or antibiotic resistance. Genome sequences can and should influence the selection of phages for application.^9,10 Although genetic sequencing cannot replace in vitro or in vivo testing, it is an invaluable tool for eliminating inappropriate candidates.

Complete genome sequencing of novel phages also contributes to the discovery and development of useful phage-encoded enzymes and our understanding of their biology. Finally, sequencing and publishing phages further our understanding of the evolution of the most diverse and undersampled entities on the planet.¹¹

In an effort to discover phages for practical use in the apiculture industry in New Zealand, we conducted a phage hunt using the honeybee pathogen Paenibacillus larvae. P. larvae is a spore-forming bacterium that leads to the honey bee disease known as American Foulbrood (AFB). AFB is deleterious at both the larval and whole-hive level and can cause irreversible damage to the beehive.¹² There have already been efforts from several laboratories around the world to try and find phages to prophylactically treat AFB, with promising results.^13,14 We encountered a set of plaque-forming P. larvae phages that could, with effort, be brought to a titer of 3 × 10⁷ PFU/mL but these proved intractable to efforts to further increase their concentrations.

A search of the literature led us to the Appelmans protocol, a 96-well plate-based procedure that has been used to expand the host range of phages by allowing strains to become simultaneously infected by multiple phages and allowing natural selection to screen recombinant phages for the most successful new chimaeras.¹⁵ Herein, we report two separate approaches we used to experimentally evolve our P. larvae phages, which allowed us to significantly increase their effective titer. The first approach was the experimental evolution and propagation of a phage in four parallel lineages for 25 days using a relatively low multiplicity of infection (MOI) of 0.05 in solid medium and an agar-overlay method.

The second approach was a protocol inspired by the Appelmans protocol¹⁵ that was initially performed for 30 days. We modified this approach by allowing pure phages to adapt to single hosts rather than propagating mixed populations of phages, thereby allowing natural selection to screen mutant phages for those most able to infect the strain in liquid conditions. Subsequently, we employed this to great effect after only 4 days (rapid adaptive mutation of phage-UP or RAMP-UP).

We describe both these approaches and the quality and speed of the titer improvements achieved. We report on the genomic traits of six separate phages that were evolved and subsequently sequenced using the RAMP-UP protocol. In addition, we modeled the population biology parameters operating in this brief experiment to understand the conditions under which mutation and selection act in this high-speed adaptive evolution protocol.

Materials and Methods

Bacterial strains and phage isolates

All bacterial strains and phage isolates are listed in Table 1 and Supplementary Table S1. Four strains of P. larvae were used for the 30-day protocol and 4-day RAMP-UP protocol described herein, these were isolated from beehives in New Zealand with symptoms consistent with AFB disease. P. larvae-PaFR-2017 and P. larvae-PaFR-2006 were both isolated from infected honeybee larvae provided by Plant and Food Research. P. larvae-TP was isolated from a symptomatic brood comb provided by Barry Foster (beekeeper) in February 2018.

Table 1.

Paenibacillus larvae Strains and Phage Isolates Used in This Study

Bacterial/phage strain	Source	Isolation source	Geographical region	Notes	Accession no.
P. larvae-PaFR-2006	Plant & Food Research	Infected larvae	Hamilton		JARDAI000000000
P. larvae-PaFR-2017	Plant & Food Research	Infected larvae	Auckland		JARDRG000000000
P. larvae-TP	AsureQuality	Symptomatic Brood Frame	Rotorua		JARDRJ000000000
P. larvae-F1A	Beekeeper	Symptomatic Brood Frame	North Canterbury		JARDRL000000000
Phage Callan	This study	Soil	Wellington	Underwent 30-day RAMP-UP	OP503989
Phage Dash	This study	Wax	Wellington	Underwent 30-day RAMP-UP	OP503990
Phage Lilo	This study	Soil	Waikato	Underwent 30-day RAMP-UP	OP503991
Phage Logan	This study	Soil	Gisborne	Underwent 30-day RAMP-UP	OP503980
Phage AJG77	This study	Soil	Otago	Underwent 30-day RAMP-UP	OP503969
Phage ABAtENZ	This study	Soil	Waikato	Underwent 30-day RAMP-UP	OP503968
Phage Wildcape	This study	Soil	Gisborne	Underwent 4-day RAMP-UP	OP503988
Phage Carlos	This study	Soil	Wellington	Underwent 4-day RAMP-UP	OP503973
Phage ApiWellbeing	This study	Soil	Wellington	Underwent 4-day RAMP-UP	OP503970
Phage LunBun	This study	Soil	Gisborne	Underwent 4-day RAMP-UP	OP494865
Phage FutureBee	This study	Soil	Waikato	Underwent 4-day RAMP-UP	OP503975
Phage Rae2Bee1	This study	Soil	Ashburton	Underwent 4-day RAMP-UP	OP503983

P. larvae-F1A was isolated from a symptomatic brood comb provided by AsureQuality in December 2018. The phages chosen for the 30-day protocol were all isolated from samples of soil from around healthy beehives. Phages Callan and Dash, from an apiary in the lower North Island, Phage Lilo from an apiary in the Greater Auckland region, Phage Logan from an apiary on the East Coast of North Island, Phage AJG77 from an apiary in the lower South Island, and Phage ABAtENZ from an apiary in the central North Island. The phages chosen for the 4-day RAMP-UP protocol were also isolated from around healthy beehives: Phage Wildcape and LunBun from an apiary in Gisborne, Phage Carlos, and ApiWellbeing from an apiary in the Wellington region, Phage FutureBee from an apiary in Hamilton, and Phage Rae2Bee1 from an apiary in Ashburton.

Bacterial strains were grown in brain heart infusion (BHI) broth (Oxoid CM1135). Phages were grown by infecting the bacterial strain they were isolated on (see Table 1), using the agar-overlay method. Phage titers were established by plating serial dilutions of the phage lysates using the agar-overlay method.

Experimental evolution to increase the infectivity of P. larvae phage Lilo

MOI was established by plating serial dilutions of both colony-forming units (CFUs) and PFUs. Specifically, a single colony isolate of the P. larvae strain was inoculated in 5 mL of BHI broth, incubated at 37°C, and shaken at 100 rpm for 2 days. This P. larvae liquid culture was serially diluted to a total dilution factor of 10⁻⁸, and 5 μL of each dilution was applied to 1.5% agar BHI plates. Plates were incubated without shaking at 37°C for 2 days. Colonies were counted to determine the number of CFUs/mL of culture.

PFU determination was similar, the P. larvae phage Lilo lysate was serially diluted to a factor of 10⁻⁸, and 5 μL of each dilution was applied as a spot test to a P. larvae bacterial lawn plated in 0.5% BHI top agar. The plate was incubated without shaking at 37°C for 2 days. The number of plaques observed for each dilution was counted to estimate the titer of the lysate.

Four biological replicates of phage Lilo were serially passaged at a low MOI. At each serial passage, 500 μL of bacterial culture was inoculated with lysate for an estimated MOI of 0.05 (2 × 10⁶ PFUs were plated with ∼4 × 10⁷ CFUs). These were incubated without shaking at room temperature for 30 min to facilitate adhesion before they were plated in 3 mL of 0.5% top agar BHI with 1 mM CaCl₂, 1 mM MgCl₂, and 1 mM thiamine hydrochloride, onto 1.5% agar BHI plates. This was our standard plaque plating approach and was used throughout.

Plates were incubated without shaking at 37°C for 2 days. The confluent or cleared plates were flooded with 8 mL BHI broth and incubated without shaking at room temperature for 2–4 h. The resulting lysate was collected and passed through a 0.45 μm syringe filter, in preparation for use in the next passage. Lysates from each passage were titered by spot plate assay. Plaque counts were used to determine lysate titers from each passage and used to adjust the volumes plated in subsequent passages to maintain the desired MOI of 0.05. Plaque diameters were measured using ImageJ software.¹⁶

RAMP-UP protocol in 96-well plates

We implemented a modified Appelmans protocol as in Burrowes et al. with changes to the media and by restricting the horizontal wells in the 96-well plate to a single strain of host and a single strain of phage.¹⁵ Using a 96-well plate, we seeded 100 μL of serially diluted phage lysate in 100 μL double-strength BHI broth containing 4 μL culture of a single strain (Fig. 2A). Plates were incubated at 37°C on a shaking platform at 100 rpm. Wells showing complete lysis, plus the first turbid well, were pooled. If no lysis was observed, wells containing undiluted phage were harvested. The pooled lysates were sterilized by vortex mixing with 1:100 CHCl_3, left for 10 min and subsequently centrifuged at 15,000 g for 15 min.

FIG. 2.

Experimental evolution using RAMP-UP protocol. (A) Schematic of 96-well format RAMP-UP protocol. (B) Increase in phage titer (PFU/mL) over the 30-day experimental evolution for six phages Logan, AJG77, ABAtENZ, Callan, Dash, and Lilo. Error bars = standard error (N = 3). (C) Increase in plaque size from the initial phage lysate until day 30 of the evolution experiment, (*p < 0.05; paired t-test). Error bars = standard error (N = 9). (D) Change in plaque size over 30 days for phages Dash and Lilo. Scale bar = 1 mm.

Lysates were removed, leaving the CHCl₃, and transferred to fresh tubes for storage. These pooled lysates were used to initiate the next round of directed evolution in a similar manner. Pooled lysates were tested for titer after every second round using the agar-overlay spot titer method, three technical replicates were performed. Each phage was only put through the RAMP-UP protocol once. Six phages underwent a 30-day RAMP-UP protocol and 20 phages underwent the faster 4-day RAMP-UP protocol (Table 1 and Supplementary Table S1).

Genomic DNA extraction

Evolved phages were purified by selecting a single plaque from a plate and using this to infect in a standard agar overlay; this was repeated three times. Phages from the final round of purification were used to inoculate a 60 mL culture of the bacterial strain they were isolated on at a MOI of 0.1. Cultures were incubated at 37°C with shaking at 100 rpm overnight. This is consistent with previously published methods.¹³ Honeybee brood is maintained at high temperatures.¹⁷ Phage lysates were purified as described previously.¹⁸

In brief, an aliquot of phage lysate was transferred into a falcon tube and centrifuged at 4000 g for 20 min, the supernatant was collected and transferred into a new falcon tube. The lysate was filter-sterilized through a 0.22 μm filter, and chloroform was added to bring the total volume of the lysate to 1.10 (0.1 of volume in chloroform). The lysate was then vortexed and incubated at room temperature for 10 min. The lysate was then centrifuged at 4000 g for 5 min and transferred into Nalgene Oak Ridge High-Speed PPCO Centrifuge Tubes, leaving the chloroform behind. The sterilized phage lysate was then concentrated by centrifugation at 38,724 g for 45 min.

Subsequently, 50 mL of this pelleted lysate was resuspended into a 5 mL volume of BHI. The phage DNA was extracted using a modified zinc chloride precipitation method.¹⁹ Modifications included the addition of 1 μL Proteinase K that was incubated at 37°C for 10 min after the TES buffer (0.1 M Tris-HCl, pH 8; 0.1 M EDTA and 0.3% SDS) step. The tubes were left on ice overnight after isopropanol was added. On day 2 of the protocol, 1 μL of pure glycogen was added to each tube before centrifugation to aid in pelleting and visualization of the DNA. DNA pellets were resuspended in 50 μL nuclease-free water.

Library preparation and sequencing

A total amount of 1 μg DNA per sample was used as input material for the DNA sample preparations. Sequencing libraries were generated using NEBNext^® Ultra™ DNA Library Prep Kit for Illumina (NEB, USA) following the manufacturer's recommendations, and index codes were added to attribute sequences to each sample. In brief, the DNA sample was fragmented by sonication to a size of 300 bp. DNA fragments were end-polished, A-tailed, and ligated with the full-length adaptor for Illumina sequencing with further PCR amplification.

At last, PCR products were purified (AMPure XP System) and libraries were analyzed for size distribution by Agilent2100 Bioanalyzer and quantified using real-time PCR. The clustering of the index-coded samples was performed on a cBot Cluster Generation System according to the manufacturer's instructions. After cluster generation, the library preparations were sequenced on an Illumina NovaSeq 6000 platform and paired-end reads were generated.

Genome assembly

Phage genome assembly was carried out using Geneious 9.05 (Auckland, New Zealand), using the built-in Geneious assembler, with either medium sensitivity/fast or medium–low sensitivity/fast, and circularize contigs with matching ends selected.

The genome ends and DNA packaging strategy were identified by sequence similarity to already published P. larvae phages.²⁰ Phages were searched for the two known 3′ overhang sequences “CGACGGACC” or “CGACTGCCC” near the terminase genes.²⁰ Phages AJG77, ABAtENZ, and Logan were found to have the “CGACGGACC” sequence, whereas Dash, Callan, and Lilo contained the “CGACTGCCC” sequence. Genomes were rearranged so these 3′ overhang sequences were at the end of the genome. This resulted in the small terminase gene starting at either 50 or 51 base pairs downstream of base 1.

Genes were identified by running the rearranged files through Phage Commander²¹ with all gene identification programs selected. The GenBank-formatted output files were entered in DNA Master (cobamide2.bio.pitt.edu) to manually check for false positives, missing genes, and identify start codons as described in detail in Ref.²² Putative protein functions were assigned as described in previous study.²⁰

Host range assays

The ability of phages to infect an isolate was assessed by 3 μL spots of a phage lysate onto a double-layer agar containing 500 μL of bacterial lawn. Each P. larvae bacterial isolate was tested separately.

Simulation to model mutation frequency in phages

We constructed a mathematical model to estimate the number of mutations that could develop in this RAMP-UP experiment. The number of available cells, phages, and time have been input from the experimental protocol. The model to simulate the number of phage mutants at the end of 4 days of experimental evolution assumes the following kinetic equations.

The first equation represents the infection of a bacterium by a phage:

where B is a bacterium, P₀ is a phage with phenotype 0, β₀ is the infectivity of P₀, and C₀ is a bacterium infected by P₀.

The second equation represents the lysis of an infected bacterium:

where l is the rate of lysis, U is the mutation rate per replication per genome, b₀ is the burst size of P₀, and P₁ is a phage with phenotype 1. The value of b₀ is assumed to be constant.

There are two more equations, which describe the infection and lysis by P₁:

where β₁ is the infectivity of P₁ and b₁ is the burst size of P₁.

We assume that b₁ = b₀ (1 + s_b) and β₁ = β₀ (1 + s_β), where s_b and s_β are the selection coefficients. For example, if s_b = 1, the burst size of P₁ is twice that of P₀.

For simplicity, we do not assume the spontaneous decay of phages or that there are more than two phenotypes of phage. For the purposes of this simulation, we assume a single class of beneficial mutations that contribute to the phenotype. We made this assumption because the number of generations is too short for multiple distinct beneficial mutations to be fixed. The above kinetic equations are simulated with the Gillespie algorithm. The model assumes 10 tubes, each containing 40,000 B at the beginning as in the experimental evolution described above.

At the start of each simulation, the initial number of P₀ in different tubes is set to numbers drawn from Poisson distributions with the following means: 1000 (zeroth tube), 100 (first tube), 10 (second tube), etc. Thus, about three tubes will have any phage particles at the beginning of day 0. The initial number of P₁ is set to 0 in all tubes. The above initial condition is equivalent to assuming that the concentration of phage particles obtained from the solid medium before the experimental evolution is 10 phage particles per microliter.

All tubes are incubated until all bacteria (B) are lysed in the tubes that have at least one phage particle P₀ or P₁.

The tubes that have at least one phage particle are pooled, plus one tube that has no phage particles. All tubes are assumed to contain 200 μL of medium. Then, 100 μL of the pooled medium is added to the zeroth tube, 10 μL to the first tube, 1 μL to the second tube, etc. The number of phage particles in each tube is again determined by drawing a number from a Poisson distribution with the corresponding mean number of phage particles.

The tubes are incubated again until all bacteria are lysed in the tubes that have at least one phage particle. In a typical simulation, about five tubes contain phages from day 1. The above repeats up to three transfers, that is, 4 days.

One hundred simulations were run for each condition tested. Any arbitrary large value could be used, we also spot-checked running the model with 1000 simulations with similar results. Parameters were varied within experimentally reasonable values as described in the results.

Results

Using strains of P. larvae as hosts, 26 novel P. larvae phages were isolated using standard phage discovery methods. In brief, direct isolation and enrichment methods were employed and phages able to form plaques were discovered and isolated from hive materials, bee debris, and soil from around or within healthy noninfected hives. The complete process of the discovery of these P. larvae phages will be described elsewhere (Kok, Zhou, and Hendrickson, in preparation).

The plaques of the six phages used in the 30-day RAMP-UP protocol were pinprick size plaques in 0.5% top agar. We employed a suite of standard methods for increasing the titers of these phage lysates, but none was successful (webbed plates, liquid infections, large volume infections, concentrating picked plaques, and enhanced centrifugation). The titers of each lysate generally remained between 5 × 10⁵ PFU/mL or up to 3 × 10⁷ PFU/mL, 1000–10,000 times too low for DNA extraction, electron microscopy or downstream in vitro testing.

Unable to proceed further in our characterization, we chose to experimentally evolve these phages in the hopes that by selecting for increased efficiency in host infection, we would be able to increase the titers. We first employed a method that used an agar overlay, reasoning that we had observed plaques but had not had success in propagating these phages in liquid during attempts to perform burst size assays.

Agar-overlay method of evolution improves titer

We developed an experimental protocol in which phages were serially propagated in top agar overlays with an abundance of host bacteria. We chose phage Lilo and performed infections using a relatively low MOI of 0.05 (1 phage for 20 bacterial cells), reasoning that phages capable of rapid infection and higher burst sizes would be favored. Four low MOI Lilo lysates were continually harvested and propagated on the ancestral bacterial strains for 25 days (Fig. 1A). Phage titers were calculated after each overnight passage for all Lilo replicate lines (Lilo E, F, G, and H).

FIG. 1.

Experimental evolution of phage Lilo on solid medium. (A) Schematic of experimental method (created with BioRender.com). (B) Increase in phage titer (PFU/mL) over the 25-day experimental evolution for four phage Lilo lineages E, F, G, and H. (C) Increase in plaque diameter from the initial phage lysate until day 25 of the evolution experiment (*p < 0.05; paired t-test). Error bars = standard error. (D) Representative plaques for the ancestral Lilo and Lilo lineages E, F, G, and H after 25 days. Scale bar = 1 mm. PFUs, plaque-forming units.

From the initial titer of 3 × 10⁷ PFUs/mL for the starting lysate, after 15 passages an increase in titer of 4.1- to 14.5-fold was observed in three out of four lines (Fig. 1B, n = 1). Those three lines maintained an average fivefold to sevenfold increase from the starting lysate's titer. Lilo E maintained an average increase of threefold throughout the experiment.

Plaque improvements from agar-overlay method

Plaque size changes can indicate a mutation has taken place in a phage of interest.²³ The ancestral Lilo phages created plaque diameters that measured at 0.70 mm (±0.05) (mean, n = 10). To assess whether the experimental evolution was having an effect on the phages, the diameters of the evolved Lilo plaques were measured after each overnight passage. After 15 passages, a 56–121% increase in the mean diameter of plaques was observed in all four evolved Lilo lines, with lines E–H measuring an average of 1.09 (±0.06) mm, 1.52 (±0.04) mm, 1.55 (±0.11) mm, and 1.36 (±0.07) mm, respectively (Fig. 1C).

The increase in plaque diameter was maintained throughout the remainder of the 25-day experiment, however, some of these apparent gains were not retained in the course of the selection. Lines F and G saw the greatest increase in plaque size (Fig. 1D), measuring 1.31 (±0.05) mm and 1.40 (±0.07) mm, respectively, at the end of the 25 days, whereas lines E and H measured 0.91 (±0.04) mm and 1.06 (±0.06) mm, respectively (Fig. 1C). Ultimately, the maximum increase in plaque size at the end of the experiment was Lilo G [1.40 (±0.07)], which doubled from the initial Lilo plaque size of 0.70 (±0.05) mm.

RAMP-UP to improve phage titers

Although the top agar-overlay evolution experiment showed signs that mutations had likely taken place, the increases in titer achieved (4–14-fold) were not sufficient to justify the effort of propagating phages in top agar overlays for this period of time. We, therefore, turned our attention to the literature and found the Appelmans protocol described by Burrowes et al.¹⁵ The Appelmans protocol is a small-volume high-throughput-directed evolution experiment administered to foster recombination between similar phages in liquid lysates over a period of 30 days. Historically, this method has been used to increase host range by passaging serial dilutions of mixed phages on several hosts in a 96-well plate format (Fig. 2A).

We adopted the basic protocol to use this passaging procedure to apply selection pressure on populations of a single phage on a single host to improve infection properties in liquid medium over a similar time period (Fig. 2A). This RAMP-UP protocol was performed on six novel P. larvae phages. Phage efficacy was checked by measuring the PFUs after every second transfer (Fig. 2B). Initial phage titers of our six phages ranged from 2 × 10⁵ PFU/mL to 3 × 10⁷ PFU/mL. After 30 days of the RAMP-UP protocol, the titers of these phages had increased to between 1 × 10⁷ PFU/mL and 2 × 10⁸ PFU/mL, a 100-fold increase on average (Fig. 2B).

Plaque improvements from RAMP-UP protocol

A significant increase in plaque size was observed in four out of the six phages. Plaque sizes were measured on days 0, 15, and 30 (Fig. 2C). AJG77 and Callan did not exhibit larger plaque sizes and measured ∼0.65 (±0.02) mm and 1.40 (±0.05) mm, respectively, for the duration of the experiment. ABAtENZ, Logan, Lilo, and Dash increased in their respective plaque sizes from 0.31 (±0.02) mm, 0.33 (±0.02) mm, 0.66 (±0.06) mm, and 0.57 (±0.04) mm to 0.50 (±0.01) mm, 0.90 (±0.03) mm, 1.27 (±0.04) mm, and 1.52 (±0.09) mm, respectively (p < 0.05; paired t-test).

ABAtENZ and Lilo (Fig. 2D) achieved 63% and 92% increases in plaque size, respectively, over 30 days, whereas Dash (Fig. 2D) and Logan saw the greatest increase in plaque size with an increase of ∼165% and 170%, respectively.

Similar results seen after 4 days

The original Appelmans protocol used a 30-day time period. The RAMP-UP protocol experiment described above appeared to yield significant increases in phage titers after the first few rounds of plating (Fig. 2B). This was particularly evident in Lilo. We, therefore, selected a new set of six novel P. larvae phages and subjected them to a 4-day RAMP-UP treatment (Fig. 3). The initial titers of the six phages selected for RAMP-UP ranged from 1 × 10⁴ PFU/mL to 1 × 10⁵ PFU/mL, with Rae2Bee1 having the lowest titer and Wildcape the highest titer.

FIG. 3.

RAMP-UP experimental evolution increases lysate titer in as little as 4 days. (A) Increase in phage titer (PFU/mL) in 4 days for six phages. Error bars = standard error (N = 3). (B) Colony measurements for the RAMP-UP evolved phages in A. (C) The RAMP-UP protocol appears to have a natural limit on the potential for phages that depends on their starting titer.

After 4 days of RAMP-UP, the titers increased from 2 × 10⁷ PFU/mL to 4 × 10⁸ PFU/mL, a 1000-fold increase on average (Fig. 3A). Rae2Bee1 showed the smallest increase in titer, with a 220-fold increase, ApiWellbeing, Carlos, and Wildcape all experienced a 1000-fold increase in titer, and LunBun and FutureBee saw the largest increases in titer with increases of 2700-fold, respectively.

An increase in plaque size was also seen in all six phages ranging an increase from 53% to 514%, FutureBee and Carlos, respectively (Fig. 3B). Interestingly, the average increase in plaque size did not correlate with increases in titer as FutureBee experienced the largest increase in titer but the smallest increase in plaque size. Ultimately, we applied either the 30- or 4-day RAMP-UP protocol to all 26 of our P. larvae phages.

The largest improvement observed was phage NHScienceFair that had a starting titer of 2.2 × 10² PFU/mL and achieved a final concentration of 4.1 × 10⁷ PFU/mL after 4 days of RAMP-UP, a 185,000-fold improvement in titer. We observed a negative correlation between the starting titer and the titer improvement, which suggests that there is a limit to the potential titer enhancement that can be achieved in the RAMP-UP protocol (Fig. 3C).

RAMP-UP leads to high phage titers and high-quality DNA

The resulting lysate concentrations after the RAMP-UP protocol were sufficient to extract high-quality DNA for Illumina sequencing. Sequencing results are shown for phages that underwent the longer 30-day protocol (Table 2 and Fig. 4). The genomes of these phages ranged between 40 and 44 kbp in length and have 70–82 genes each. Pairwise comparisons of these genomes were carried out using Phamerator, a bioinformatics tool that uses the “Align Two Sequences” program contained within BLAST.²⁴

Table 2.

The Genomic Characteristics of the First Six Phage Genomes That We Sequenced Using the RAMP-UP Protocol to Increase Titers

	Genome length (bp)	DNA packaging strategy	GC content (%)	No. of genes	Genes per 1000 bp	Percent coding (%)
Logan	44,419	3′ cos	43.0	82	1.85	94.01
AJG77	44,417	3′ cos	43.0	82	1.85	93.46
ABAtENZ	44,419	3′ cos	43.0	82	1.85	94.01
Lilo	40,941	3′ cos	40.3	70	1.71	91.78
Callan	44,768	3′ cos	39.6	77	1.72	91.56
Dash	44,599	3′ cos	39.4	79	1.77	93.40

FIG. 4.

Pairwise genome maps of phages evolved for 30 days. Maps generated by Phamerator. Shading indicates high sequence similarity between sequences as determined by BLASTN, with purple being the highest (E-value = 0). Genes are represented by boxes, boxes with the same color indicate genes that belong to the same families (“phams”).

The maps show how related or divergent the phages are, showing that Lilo, Callan, and Dash are highly related, whereas ABAtENZ, AJG77, and Logan are highly related to each other but diverge from the other three phages apart from two regions of the structural genes (Fig. 4). The details of these genomes and their functional genes will be discussed elsewhere (Kok, Tsourkas, Gosselin, and Hendrickson in preparation).

Changes to host range of New Zealand P. larvae phages after RAMP-UP protocol

The host range of all 26 P. larvae phages was assessed against 30 New Zealand P. larvae bacterial strains before they underwent the RAMP-UP protocol. The host range was then checked again after the phages had been evolved. An expansion in the host range was seen in 32 instances (red asterisks in Fig. 5) in 11 phages, and there were no instances of a reduction in the host range. Eight instances of phages that expanded their host range to include P. larvae strain W19_08100, whereas five RAMP-UP evolved phages gained the ability to infect P. larvae strain W19_08099.

FIG. 5.

Host range of 26 Paenibacillus larvae phages on 30 P. larvae bacterial isolates from New Zealand. Dark gray boxes indicate complete cell lysis, light gray boxes indicate some cell lysis and white boxes indicate no cell lysis has occurred. Red asterisks show where host range expansion has occurred. White asterisks show the original P. larvae strain the phage was isolated on.

Both of these P. larvae isolates previously had fewer potential infecting phages than the majority of bacterial strains in this project. Phage ABAtENZ gained the ability to infect an additional five bacterial strains, which were previously only capable of being infected by phages Callan, Dash, and Lilo before the RAMP-UP protocol.

Phages TonyLawson77, Bob, BarryFoster_Benicio, Rosalind, Bloomfield, and NHScienceFair all had low starting titers (10² PFU/mL to 10³ PFU/mL, or 1–10 phages in 3 μL). It is possible that the observation of increased host range in these six phages could actually be due to these low titers in the initial lysates. In other words, the increase in titer may have made a previously unobservable infection on the host observable using a 3 μL aliquot of lysate. That said, in the phages that had an initial titer of 20–150 phage particles in 3 μL, we consider the initial titer to have been sufficient to show positive or negative signs of lysate in this small volume.

We, therefore, consider the expanded host range of phage ABAtENZ, and the capability of phages GIW2016, ApiWellbeing, Carlos, and GaryLarson to lyse W19_08100 after evolution to be due to mutations that arose during evolution and had the effect of expanding the host range of these phages.

As we were unable to sequence these phages before evolution, we could not compare their genomes before and after evolution to ascertain what may have caused the expansion of the host range.

We investigated the possibility that mutations might have occurred that allowed escape from CRISPR recognition. Previously sequenced P. larvae strains are known to have CRISPR-Cas systems, and P. larvae phages may have evolved point mutations to escape their hosts' CRISPR systems.²⁵ We subjected preliminary sequences (in contigs) of our 8 P. larvae strains to CRISPRFinder²⁶ to identify 38 unique CRISPR spacers.

We searched the phage genomes for these P. larvae CRISPR spacers, allowing up to 80% nucleotide divergence (∼7 bp changes allowed). Callan, Dash, and Lilo all contain spacers that range from 81% to 100% nucleotide identity to at least one P. larvae CRISPR spacer found in our sequenced isolates. However, these three phages did not experience any expansion in their host range after the RAMP-UP protocol. Based on this analysis, we cannot determine what the mechanism of host range expansion was, but it does not appear to be caused by the evasion of CRISPR spacers.

Modeling of RAMP-UP protocol

To model the possibility of a beneficial mutation becoming dominant after as little as 4 days of evolution in the RAMP-UP protocol, we assembled a simulation. The simulation was designed to answer two questions. The first of which is given the number of bacteria in this small volume format was ∼40,000, how many mutant phages would we expect after 4 days? Second, how great a fitness advantage would such a mutation needs to provide to be observed as larger plaques? In this format, mutations that increase the effective burst size or phage infectivity or both are likely responsible for the increased titer observed.

We began our simulation study by evaluating the reasonable parameters that might occur in our experiment. The literature suggests that P. larvae phage burst sizes can be as low as 8.34.²⁷ The maximum burst sizes of any phages on record appears to be those of Hafnia phages that have been recorded to have burst sizes of up to 10,000+ (±1097).²⁸ Based on these observations, we chose our basal burst size to be 10 and the maximum burst size included in our simulation was 600 (basal 10 × 60-fold increase). We also estimated the maximum burst size that might explain the lysis observed in our wells and it was 500. Last but not least, the genome size was chosen retrospectively as 40,000 bp, consistent with the P. larvae phages.²⁹

The initial parameters for the simulation were as follows:

• Rate of lysis: l = 1/30 (/min), meaning that lysis takes, on average, 30 min.

• Basal infectivity: β₀ = 10⁻⁶ (per virion per min).

• Basal burst size: b = 10.

Under these parameters, it would take ∼10 h for all 40,000 bacteria to be lysed. The parameters U (mutation opportunity), s_b (burst size coefficient), and s_β (infectivity coefficient) were varied as follows:

Mutation opportunity (U) was set between 0.0 × 10⁰ and 1.0 × 10⁻⁵. This is equivalent to multiplying the mutation rate, 5 × 10⁻⁷ per base pair (or Drakes rule³⁰), by the number of possible sites that could be mutated (out of ∼40,000 bp) that would result in an increase in the burst size of that phage. We set this number of possible sites for mutation within the U parameter to between 0 and 20 nucleotide sites (e.g., 20 × 5 × 10⁻⁷ = 1.0 × 10⁻⁵). The selection coefficient of burst size (s_b) was set between 1 and 60 (meaning that a mutation increases the actual burst size from 10 up to 600).

The selection coefficient of infectivity (s_β) was set between 1- and 10-fold increase in infectivity. This means that depending on the parameters set for any given simulation, the largest potential increase in total fitness due to a single mutation would result if that mutation increases burst size by 60-fold and also increases infectivity by 10-fold, for example, 60 × 10 = 600-fold increase in phage fitness.

One hundred simulations were run for each chosen set of parameters, and a frequency histogram of the fraction of the mutant phage after 4 days was obtained. Histograms showed a bimodal distribution, showing peaks at 0 and 1 (Fig. 6). This bimodality comes from the fact that the limiting step for the mutant phage to emerge in a population is the occurrence of a fitness-improving mutation. If the mutant occurs, it can spread within a small amount of time.

FIG. 6.

Histogram of three representative simulations when infectivity is set to 1. Black shows when parameters are set so the simulation results in 0 mutant phages (U = 0, s_b = 1, s_β = 1). Gray shows the transition point or at least 50% of the experiment results in 70% or more of the phages having the fitness-increasing mutation (U = 5.0 × 10⁻⁶, s_b = 5, s_β = 1). White represents the upper limit of the simulation or when all experiments have at least 70% fitness-increasing mutant phage after 4 days (U = 1.0 × 10⁻⁵, s_b = 30, s_β = 1).

The percentage out of 100 simulations that resulted in populations in which 70% or more of the phages were mutants with a fitness advantage after 4 days of RAMP-UP varied depending on the parameters (Fig. 7). Setting the mutation opportunity (U) to 0 resulted in zero mutant phages regardless of parameters affecting increased burst size or infectivity (Fig. 7A).

FIG. 7.

Heat maps of simulations of RAMP-UP. One hundred simulations were run for each change in parameters. Mutation opportunity, increase in burst size, and increase in infectivity coefficient were varied. (A) Infectivity fold increase set to 1. (B) Infectivity fold increase set to 5. (C) Infectivity fold increase set to 10. Blue indicates the lowest number in the series and orange indicates the highest number. The boxes that are highlighted are represented in Figure 6.

One way to consider these results is to address the transition point, or the point at which at least 50% of the experiments resulted in 70% or more of the phages having fitness-increasing mutations. If the increase in infectivity was set to 1 (s_β = 1), the transition point appears when the mutation opportunity was set to 5.0 × 10⁻⁶ (10 possible mutation sites) and the burst size increase for a mutation was set to fivefold (Fig. 7A). The upper limit (all experiments having 70% mutations after 4 days) was reached when the mutation opportunity (U) was set to 1.0 × 10⁻⁵ (20 nucleotide sites) and burst size increased by 30-fold.

If the increase in infectivity in the event of a mutation was set to fivefold, the transition point appears when the mutation opportunity (U) was set to 5.0 × 10⁻⁷ (1 nucleotide site) and burst size for a mutation was set to 10-fold (Fig. 7B). In this case, a single mutation would increase fitness by 50-fold. The upper limit was reached when the mutation opportunity (U) was set to 7.5 × 10⁻⁶ (15 nucleotide sites) and the burst size increase for a mutation was set to 5-fold, these settings designated an increase in fitness by 25-fold in the event of a single mutation.

The last set of simulations had the increase in infectivity in the event of a mutation set to 10-fold; the transition point in this instance appears when the mutation opportunity (U) was set to 5.0 × 10⁻⁷ (1 nucleotide site) and the burst size increase for the same mutation was set to 5-fold increase above the basal burst size of 10 (Fig. 7C). This effectively increased fitness by 50-fold if the mutation of that single available site occurred. The upper limit was reached when the mutation opportunity (U) was set to 5.0 × 10⁻⁶ (10 nucleotide sites) and the burst size increase for a mutation in 1 of these 10 sites was set to 5-fold increase in burst size; these mutations would increase overall fitness by 50-fold.

Discussion

Like many other phage biologists before us, we discovered P. larvae phages isolated from nature and found them to be intractable, meaning they had persistently low titers that could not be increased by standard laboratory procedures. Other Paenibacillus phage biologists have had similar issues. Yost et al., working in the United States, described concentrating 100 mL volumes to 3 mL to obtain lysates for DNA extraction and EM.¹³

Similarly, a Paenibacillus polymyxa phage isolated in Slovakia required the concentration of a 200 mL volume of lysate.³¹ More recently a similar study from Poland described 5 P. larvae phages, 4 of which were at a titer of 8 × 10⁶ or lower and required additional centrifugation steps before DNA sequencing.³² Similar protocols for tractable phages often require volumes of 1–10 mL and are easily brought to titers >5 × 10⁹ PFU/mL.

We endeavored to evolve one of our New Zealand P. larvae phages Lilo using traditional agar-overlay methods, a 4.1–14.5-fold increase in titer was observed across the different replicates in the first 15 days; this increase was maintained for the rest of the duration of the experiment. We also saw increases in plaque size between days 0 and 15 and then saw decreases in plaque size between days 15 and 25. Increases in plaque size are complicated and can be due to a number of different factors. For example, shorter phage latent periods and faster virion diffusion can both lead to larger plaque sizes as well as a larger phage burst size, especially if initial burst size is small.³³ The increases in titer observed were not, however, sufficient to obtain enough DNA to progress to complete sequenced genomes.

Spatial structure could be a contributing factor to our difficulty in evolution of phages on agar overlay. There are many consequences of spatial structure including resource concentration, barriers and gradients, superinfection, and altered gene expression.³⁴ It is possible that gene expression changes in the host may determine which method works best for a particular phage.^35,36 In addition, the limits to diffusion that are in play when a phage is infecting in an agar overlay are significant.

We subsequently attempted the evolution of six New Zealand P. larvae phages using the RAMP-UP protocol, naturally evolving each phage independently on the bacterial strain on which they were isolated.¹⁵ This method resulted in a 100-fold increase in titer on average across the different phages. We also saw an increase in plaque size of between 63% and 170% in four of our phages.

Interestingly, we saw the greatest increase in titer at ∼4–5 days after commencing evolution. We, therefore, modified the protocol again and evolved a further six P. larvae phages using a shorter RAMP-UP protocol. From this protocol, we saw an average 1000-fold increase in titer and increases in plaque size ranging from 53% to 514%. The 4-day RAMP-UP protocol highlighted the possibility of evolving phages in even shorter timeframes and still obtaining desirable results.

The RAMP-UP protocol described appears to facilitate mutations and the increase of these mutants in the population, as indicated by changes in titer and plaque dimensions. We sequenced and analyzed the genomes of these experimentally evolved phages but could not analyze their ancestors to determine what mutations led to the observed improvements. We did look at possible changes in CRISPR spacers that might have driven large increases in phage fitness by avoiding bacterial defence,³⁷ but we could not find evidence that mutational escape from CRISPR drove these mutants.

Our experience of the RAMP-UP protocol and the increased titers of the phages did make us question whether it was possible that mutations could arise so quickly. We have, therefore, constructed a computer simulation of the population parameters of the RAMP-UP protocol to determine whether advantageous mutations (and not some other phenomenon) could drive new phages to become dominant in the population in as little as 4 days. This simulation focused on the likelihood of mutations occurring that increased either the infectivity or the burst size.

One of the factors that we considered including in our simulations but did not include was that of the timing from infection to lysis, or in other words, the duration of the intracellular progeny accumulation phase. In the experimental setup we are attempting to model, 40,000 cells are present per well. In these conditions, wells that are inoculated with very few phages may have the opportunity for rapid secondary infections that would contribute to the competitive fitness of mutant phages, but the number of mutant phages present in these wells will also be very low.

Similarly, the majority of mutant phages will be competing against one another in one-step infection processes wherein all of the available hosts are infected and lysed at once. We reasoned that the wells with the highest phage populations will host the majority of advantageous mutations and natural selection will, therefore, be acting in these one-step infections.

We varied the target size for a random mutational event by changing the mutation opportunity (U) from 0.0 × 10⁰ to 1.0 × 10⁻⁵. We allowed single mutations to increase the infectivity from 1- to 10-fold and allowed the burst size to increase from 1- to 60-fold. Ultimately, we observed that with a mutation that would increase fitness by only 5-fold, as many as 50% of experiments would have at least 70% mutant phages. This phenomenon is likely similar to that of the mutational jackpot events, where beneficial mutations may occur early on in growth experiments resulting in these mutants dominating the population.^38,39 These results lead us to believe that this method may be highly relevant for increasing the utility of a wide range of difficult phages.

Serial passage experiments have previously been used to evolve phages to increase titer. In one experiment, phage FCV-1 was coevolved with Flavobacterium columnare in lake water over the course of 29 days. In the first 4 days, the phage titer increased from 10⁴ to 10⁷ PFU/mL (a three-log increase).⁴⁰ We believe that this is the method that we have put forward here is the first generalizable and high-throughput method that can be attempted rapidly with phages that are recalcitrant to other methods for increasing titer.

Conclusions

Communication with phage-hunting colleagues suggests that many phages have been discovered only to be abandoned due to issues with low titer. Herein, we have presented a novel method to experimentally evolve difficult phages to increase their titer and in some cases their host range in as little as 4 days. We have based this rapid method on the previously published Appelmans protocol.¹⁵

We have used this RAMP-UP protocol to evolve 26 New Zealand P. larvae phages to rapidly improve stocks that are recalcitrant to high-titer lysate creation by normal means. These high-titer lysates enable the extraction of high-quality DNA for sequencing as well as other downstream applications such as large-scale phage production. RAMP-UP presents a fast and effective way to experimentally evolve previously intractable phages allowing researchers to study entities that might otherwise be lost to science.

Authors' Contributions

Conceptualization of this study was done by H.L.H. and D.K.; methodology, formal analysis, visualization, and writing—original draft preparation were by D.K.; software was carried out by P.T. and N.T.; validation and data curation were by P.T.; investigation was taken care by J.T. and D.K.; writing—review and editing was by D.K., J.T., H.L.H., N.T., and P.T.; supervision, project administration, and funding acquisition were by H.L.H. All authors have read and agreed to the published version of the article.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This research was funded by AGMARDT, grant number AIGITINQ-000301; The Sustainable Food & Fibre Futures Fund, grant number 405604; and D.N.K. received funding from NZPPS.

Supplementary Material

Supplementary Table S1