Development of a Robust and Quantitative High-Throughput Screening Method for Antibiotic Production in Bacterial Libraries.

Over the past 30 years, there has been a dramatic rise in the number of infections caused by multidrug-resistant bacteria, which have proliferated due to the misuse and overuse of antibiotics. Over this same time period, however, there has also been a decline in the number of antibiotics with novel mechanisms of action coming to market. Therefore, there is a growing need for an increase in the speed at which new antibiotics are discovered and developed. Natural products produced by bacteria have been and continue to be a robust source of novel antibiotics; however, new and complementary methods for screening large bacterial libraries for novel antibiotic production are needed due to the current agar methods being limited in scope, time consuming, and prone to error. Herein, we describe a rapid, robust, and quantitative high-throughput liquid culture screening method for antibiotic production by bacteria. This method has the ability to screen both mono- and coculture mixtures of bacteria in vitro and be adapted to other phenotypic natural product analyses. Over 260 bacterial species were screened in monoculture, and 38 and 34% were found to produce antibiotics capable of inhibition of Staphylococcus aureus or Escherichia coli, respectively, with 8 and 4% being classified as strong producers (≥30% growth inhibition), respectively. Bacteria found to not produce antibiotics in monoculture were also screened in coculture using an adaptation of this method. Of the more than 270 cocultures screened, 14 and 30% were found to produce antibiotics capable of inhibition of S. aureus or E. coli, respectively. Of those bacteria found to produce antibiotics in monoculture, 43 bacteria were subjected to 16S rRNA sequencing and found to be majority Pseudomonas (37%), Serratia (19%), and Bacillus (14%) bacteria, but two novel producers, Herbaspirillum and Kluyvera, were also found.

Introduction

Multidrug-resistant bacterial infections are one of the top three threats to global public health and are projected to cause up to 10 million deaths worldwide by the year 2050 if the current trend continues.[1,2] There are many factors that contribute to the upsurge of multidrug-resistant bacterial infections with the largest being the overuse and misuse of antibiotics both in medicine, as they are among the most commonly prescribed drugs, and in agriculture.[3,4] Resistance to a particular antibiotic can occur rapidly and broadly, where bacteria can become resistant to an entire class of drugs with the same mechanism of action. Therefore, there is a dire need for the constant development and production of novel and efficacious antibiotics to keep pace with the ever-evolving bacteria. Natural products, which have been fine-tuned by nature to elicit a specific biological activity, continue to be an excellent source of novel molecules that can be used as blueprints for antibacterial drug design; however, many challenges with natural product isolation and characterization remain.

Between 1981 and 2014, natural products and their derivatives, produced by a wide array of organisms including plants, fungi, and bacteria, accounted for 65% of all approved small molecule drugs and 73% of approved antibacterial agents.[5] Despite these compelling numbers, natural product discovery programs, and specifically those for antibiotic development, have been largely abandoned over the past 30 years by the pharmaceutical industry for more efficient, and therefore profitable, high-throughput screening technologies and a combinatorial chemistry approach to drug discovery. It has been estimated that it takes approximately 3 months of work and $50 000 to isolate and characterize a single novel natural product due to the time-consuming nature of natural product discovery, from finding a hit to cultivating or culturing the natural source to extracting the bioactive compound, and the to the issue of dereplication, which is the process of identifying known compounds with bioactivity prior to full isolation.[6,7] Over the past two decades, many strategies utilizing a variety of spectroscopic[7−10] methods have been developed to facilitate dereplication for all types of natural product screening. However, recent focus has shifted to improve the process of natural product identification and isolation so that less dereplication is needed to begin with. Bacteria, which were a robust source of bioactive (especially antibacterial) natural products beginning in the early 1900s, saw diminished interest by scientists in the past 30 years due to the high frequency of reisolation of known compounds. Recently, bacteria have re-emerged as organisms of interest due to many new methods that have been developed to specifically target novel natural product production.[11] These methods include screening unique bacterial species such as extremophiles, endophytes (kakadumycins from an endophytic Streptomycete associated with the fern-leafed grevillea[12]), and marine bacteria (marinomycins A–D from Marinispora(13)), developing methods for culturing “unculturable” bacteria (teixobactin from Eleftheria terrae via iChip technology[14] and malacidins A and B from Streptomyces albus(15)), and mixed microbial culture to mimic the natural, competitive environment and cross talk between species (pyocyanin produced by a Pseudomonas–Enterobacter dual culture[16]), which may utilize the bacteria’s thousands of uncharacterized, silent genes as shown by whole-genome sequencing.[17] For example, Garbeva et al. screened a 146 phylogenically diverse bacterial library in mono- and cocultures and found that 6% (154/2798 pairs) of interactions induced antibiotic activity against either Staphylococcus aureus or Escherichia coli.[18]

In addition to developing methods to produce novel antibiotics from microorganisms, there is also a need for improving the screening techniques for these producers to reduce false positives, which lead to loss of time and resources. New techniques for screening bacteria libraries using tools such as microfluidics[19,20] and mass spectrometry[21,22] have been developed that allow for rapid and large-scale analysis of producing bacteria, but these require highly specialized equipment and instrumentation that may be inaccessible for smaller institutions, limiting their ability to contribute to the field of antibiotic discovery. For laboratory-culturable bacteria, the standard “low-tech” method for assaying bacteria for antibacterial production has been an agar-based method, similar to the disk diffusion agar method, where the bacteria to be assayed is spotted onto a solid media agar plate via pipette or pin replication that has been overlaid or spread with the pathogen target of interest on top[23−25] (Figure ). Bacteria that produce antibiotics under these conditions will develop a “halo” or zone of inhibition around them (Figure , D5), where the pathogen has been cleared, which can be observed without the need for any specialized equipment or instrumentation. While this method allows for the evaluation of a large number of potential producers and is simple enough that even minimally trained scientists such as undergraduates can perform it, it is also highly prone to error, time consuming (typically taking 2 weeks for a single assay with growth times), and restricted to organisms that grow well and have low motility on solid media (Figure , B1 shows a high motility bacteria). The agar method is able to be quantified manually by measuring the size of the zone of inhibition, but this limits throughput.[25] Additionally, since the analysis of antibiotic production is conducted visually, bacteria that produce only weakly active compounds or produce active compounds but in low concentrations may be missed. Herein, we report the development of a new, time-efficient, high-throughput, and quantitative method for screening bacteria in a liquid culture that compliments the use of novel organisms and mixed microbial culture for antibiotic production to further improve the drug discovery process. We have validated this process by screening >260 bacteria isolated from a variety of environments including bulk and rhizosphere soil, the phytotelmata of Sarracenia pitcher plants, desert soil, and freshwater sediment for antibiotic production against S. aureus and E. coli in monoculture. Further, this method was adapted to screen coculture mixtures of bacteria, as well.

Figure 1

Agar overlay method against S. aureus. D5 represents a hit.

Results and Discussion

Assay Development

The goals for developing a new bacterial, phenotypic (specifically antibiotic) screening method were to

have a reduced assay time compared to traditional agar method;

be robust enough that even minimally trained scientists can get reliable results;

allow for quantification with error analysis via standard deviation (SD) of antibiotic activity;

be amenable to mono- and coculture analyses.

With these goals in mind, the assay shown in Figure was developed. For monoculture evaluation, a single bacterial colony is inoculated in 10% tryptic soy broth (TSB) and incubated at 25 °C for 24 h (day 1). These overnight cultures are then pipetted into a 96-well filter bottom plate (n = 4), diluted 15-fold in fresh media (10% TSB), and incubated at 25 °C for 48 h to allow for antibiotic production (days 2–3). After the 48 h production period, the plate is centrifuged into a new 96-well round-bottom plate containing the test pathogen in fresh media (full-strength TSB), allowing for analysis of the cell-free media containing any excreted secondary metabolites (day 4). After a final 24 h incubation period at 37 °C, the absorbance is taken at OD590 to assess pathogen growth in the presence and absence of culture filtrate (day 5). For coculture evaluation, the assay is extended to include an additional growth day of the monocultures in a 96-well assay plate before they are mixed in coculture and allowed to incubate for 48 h prior to filtration.

Figure 2

Workflow of the monoculture liquid high-throughput screening method. The 5 day process starts after the colonies are quadrant-streaked from agar plates and allowed to grow for 36–48 h. Hits are determined as any inhibition 2 standard deviations from the pathogen in TSB. Strong producers are categorized as those hits that have ≥30% inhibition.

For our specific library of bacteria, which is composed of bacteria isolated from both aqueous and rhizosphere soil samples found in western North Carolina and the southwestern United States, we found that the general optimal growth conditions were nutrient-poor media (diluted TSB to 10% of the manufacturer’s recommendation) and 25 °C. Additionally, due to the small volume in each well, a 48 h incubation allowed for us to observe antibiotic production activity without issues associated with longer incubation times such as cell death due to the lack of nutrients or evaporation. These conditions can be optimized for any specific bacterial library, and larger culture volumes can be accommodated with deep-well filter plates. Pathogens, S. aureus and E. coli, were grown under optimal growth conditions of full-strength TSB and 37 °C.

In total, this assay takes 5–6 days to complete depending on whether mono- or cocultures are used. The 96-well plate format allows for an appropriate number replicates to be run in parallel so that the “hits” are reliable and reproducible. For analysis of antibiotic production by bacteria in mono- or coculture, an n = 4 was deemed sufficient to be able to statistically differentiate (2 standard deviations (SDs), 95% confidence) a producing culture compared to the uninhibited pathogen control, which is detailed in the Supporting Information. Additionally, we set a threshold of ≥30% inhibition of the pathogen compared to the uninhibited pathogen control for bacteria to be deemed as “strong producers” or bacteria that produce either a large quantity of an antibiotic or a highly potent antibiotic. In a practical sense, this is a threshold that we found to be amenable to natural product isolation in our laboratory using standard large-scale culture methods (6–12 L), which was validated in the isolation pseudopyronine B from a Pseudomonas soil bacteria,[26] but other thresholds could be established based on the specific goals of an individual screen.

One additional utility of this screening method is that it is modular and can be readily adapted for specific growth requirements of the bacteria being screened (i.e., extended growth time or varying media conditions) and for other phenotypic screens (i.e., cytotoxic where the filtrate is centrifuged into mammalian cells). Limitations of this method are that only bacteria that can be cultured in liquid are able to be screened, bacteria with erratic growth under standard conditions can lead to confounding results, and multiculture is limited by the volume of the 96-well plate (typically <200 μL).

Screening in Monoculture

To validate the screening method, a bacterial library containing bacteria isolated from a variety of aqueous and rhizosphere soil environments was screened for antibiotic production using S. aureus and E. coli as representative Gram-positive and -negative pathogens. 302 bacterial samples were tested for single production against S. aureus, and 38% (n = 115) were found to produce antibiotic activity, with 23 of those 115 being classified as strong producers (≥30% inhibition) (Figure ). There were 267 bacterial samples tested for single production against E. coli, and 34% (n = 90) were found to produce antibiotic activity, with 11 of those 90 being classified as strong producers (≥30% inhibition) (Figure ). Bacteria that had irreproducible growth in liquid media at any stage of the protocol were removed from the final analysis. A representation of the data output with 95% confidence intervals for the screening of 21 bacteria is shown in Figure . Nonproducers are shown in blue, which were found to not significantly inhibit the growth of the S. aureus (SA in TSB) compared to that of the TBS-only control. Producers are shown in yellow and orange, with orange being those that inhibited the pathogen by ≥30%.

Figure 3

Total hits in monoculture against S. aureus and E. coli.

Figure 4

Representative data for 21 bacteria screened for antibiotic production against S. aureus. Each individual data point has been background-subtracted to remove absorbance from 32.5% TSB (“TSB only”) and normalized to the growth of the pathogen (“SA in TSB”), which has been set to 1. Black dots show individually normalized well values for each bacterial species (n = 4), colored bars show the mean of the individually normalized well, and 95% confidence intervals (±2 standard deviations) are shown as error bars. Blue bars indicated no inhibition of pathogen, yellow indicate weak inhibition (<30%), and orange indicate strong inhibition (≥30%) of pathogen compared to the pathogen-only (−)-control SA in TSB. Bacteria 400, a Pseudomonas sp., which was found to strongly inhibit both S. aureus and E. coli, was used as a (+)-control for each individual assay.

To date, 43 antibiotic producers have been identified by polymerase chain reaction (PCR) amplification and DNA sequencing of the full-length 16S rRNA genes (see the Supporting Information). Of those identified, the highest abundant genera were Pseudomonas (37%), Serratia (19%), and Bacillus (14%). Bacillus, Brevibacterium, Brevibacillus, Chromobacterium, Collimonas, Herbaspirillum, Kluyvera, Microbacterium, Staphylococcus, and Streptomyces encompass the remaining identified isolates. Out of these identified genera, Pseudomonas,[26−28]Serratia,[29]Bacillus,[30]Brevibacterium,[31]Brevibacillus,[32]Chromobacterium,[33]Collimonas,[34]Staphylococcus,[35] and Streptomyces(36) are all known antibiotic producers. Isolates, which have not yet been found to produce antibiotic compounds, include Herbaspirillum and Kluyvera.

Screening in Coculture

As has been discussed, one way to increase the chances of isolating novel compounds produced by bacteria, and reduce the need for dereplication, is to use co- or multiculture techniques to elicit the production of small molecules from silent genes that are not active in monoculture. To this end, bacteria from the library that were found to not produce an antibiotic in monoculture were then screened pairwise in coculture, selected at random, using the adapted protocol described above. Three hundred and three bacterial pairs were screened against S. aureus, and 14% (n = 43) were found to have activity with 3.3% (n = 10) showing ≥30% inhibition (Figure ). Two hundred and seventy-eight bacterial pairs were screened against E. coli, and 30% (n = 84) were found to have activity with 0% (n = 0) showing ≥30% inhibition (Figure ). These numbers include single bacteria that in coculture with any other bacteria caused antibiotic production. An example of the coculture data is shown in Figure .

Figure 5

Total hits in coculture against S. aureus and E. coli.

Figure 6

Representative data for 20 bacteria pairs found to be nonproducers in monoculture screened for antibiotic production against S. aureus. Each individual data point has been background-subtracted to remove absorbance from 32.5% TSB (TSB only) and normalized to the growth of the pathogen (SA in TSB), which has been set to 1. Black dots show individually normalized well values for each bacteria pair (n = 4), colored bars show the mean of the individually normalized well, and 95% confidence intervals (±2 standard deviations) are shown as error bars. Blue bars indicated no inhibition of pathogen, yellow indicate weak inhibition (<30%), and orange indicate strong inhibition (≥30%) of pathogen compared to the pathogen-only (-)-control SA in TSB. Bacteria 400, a Pseudomonas sp., which was found to strongly inhibit both S. aureus and E. coli, was used as a (+)-control for each individual assay.

Conclusions

In summary, the discovery of novel antibiotics is and will continue to be a worldwide need as antibiotic resistance continues to rise. To help address this need, we have described the development of a new method for screening liquid bacteria in mono- and cocultures for antibiotic production that is rapid, quantitative, and robust enough that even minimally trained scientists, such as undergraduates, can perform, which is complimentary to and can be used as an initial screening tool for other more “high-tech” methods for antibiotic discovery and analysis in the field. This method not only allows for the identification of bacteria that produce molecules in large quantity or ones that are highly active but also allows for the identification of weakly producing bacteria that may be read as false negatives in standard agar methods used in the field today. Mono- and cocultures that were found to be active in our screens are being scaled for antibiotic isolation using methods previously described.[26] Finally, this method can be easily adapted to other phenotypic screens, such as mechanism of action assays and LC-MSMS analysis of metabolites, since only the cell-free media with excreted, bioactive secondary metabolites are being evaluated.

Experimental Section

General Procedures

All steps were completed with aseptic techniques. All media and glassware were sterilized via an autoclave at 121 °C for 60 min. All agitation occurred at 160 rpm in a temperature-controlled console shaker (Excella E25) at 25 °C. Bacteria in the library were stored at −80 °C as 25% glycerol stocks in 10% tryptic soy broth (TSB, 3 g BD Bacto TSB powder per 1 L deionized water). From these freezer stocks, bacteria were quadrant-streaked and grown on 10% tryptic soy agar (TSA, 3 g BD Bacto TSB powder, and 20 g Bacto agar per 1 L deionized water) for 36–48 h at 25 °C. Purchased bacteria strains used were S. aureus (ATCC 29213) and E. coli (ATCC 15022).

Agar Spread Plate Method

The overnight target pathogen suspension (70 μL) was pipetted onto a 10% TSA plate. A sterilized plastic spreader was used to evenly spread the pathogen across the plate, and the plates were left to dry. The desired cell suspension (1.7 μL) was spotted onto dried plates and monitored up to a week at room temperature for zones of inhibition against the target pathogen.

Agar Overlay Method

TSA (10%, 4 mL) was melted via microwave and cooled to 50 °C in a hot water bath. The target pathogen overnight suspension (100 μL) was added to the soft agar, inverted, and poured onto a prewarmed, 37 °C, 10% TSA (1.5% agar) plate. Plates were slightly rocked for even coverage and left to dry. The desired cell suspension (1.7 μL) was spotted onto dried overlay plates and monitored at room temperature up to a week for zones of inhibition against the target pathogen.

High-Throughput Liquid Assay Method

Growth Parameters and Controls

Screened bacteria were grown in 10% TSB at 25 °C. Pathogens were grown in full-strength TSB (30 g BD Bacto TSB per 1 L deionized water) at 37 °C. For all liquid assays, the (-)-control was a pathogen in 32.5% TSB (150 μL 10% TSB + 50 μL full-strength TSB), and the blank was 32.5% TSB to confirm no contamination present and to allow for background subtraction in data analysis (see Figures S2–S4). Additionally, a confirmed producer bacteria from our library (400, Pseudomonas sp.) was used as a (+)-control for all mono- and coculture assays.

Monoculture

As shown in Figure , the 5 day screening process began with creating liquid bacterial cultures by inoculating a selected colony into 4 mL of 10% TSB, which was shaken at 25 °C for 24 h. On the second day, the liquid cultures from day 1 were used to make 96-well filtration plates (a 0.22 μM hydrophilic low protein binding Durapore membrane). The filter plates were prepared with 140 μL of 10% TSB and 10 μL of the appropriate sample in each well (n = 4). Filter plates were shaken for 48 h. On the third day during the filter plate incubation period, overnight cultures of S. aureus and E. coli were grown in 4 mL of full-strength TSB from a selected colony and shaken for 24 h at 37 °C. On the fourth day, 50 μL of the overnight cultures of S. aureus or E. coli were pipetted into all wells of a sterile 96-well VWR tissue culture plate except for the final column, which served as a (+)-positive control and was filled with 50 μL of full-strength TSB only. Then, the filter plates were placed directly on top of the newly made plates and were centrifuged for 10 min at 1000g at 25 °C. This allowed all cellular material to stay in the filter membrane, while all excreted secondary metabolites were transported into the new plate. The final TSB concentration for all wells was 32.5% TSB. The plates were shaken for 24 h at 37 °C.

On the fifth and final day, plates were analyzed for absorbance at 590 nm using a Gen5 All-In-One Microplate Reader software with BioTek Synergy HTX Multimode Microplate Reader. Each plate’s data was individually analyzed using Microsoft Excel. The average of TSB without pathogen was subtracted from each well of the 96-well plate to subtract the media background and create a standard baseline of “0” (TSB only, Figures –6). The new average of TSB without pathogen represents the absorbance of a well, which contains 0% of the pathogen or 100% inhibition of bacterial growth. Each well was compared to the pathogen absorbance in TSB (SA in TSB, Figures –6) as a ratio of the well/pathogen in TSB to normalize all values to the pathogen = 1 (“limit of acceptance”, Figures –6). The new average of the pathogen in TSB represents the absorbance of a well, which contains 100% of the pathogen or 0% inhibition of bacterial growth. The average and two standard deviations (SDs) of all ratios per bacteria were used to create a 95% confidence interval of the total pathogen inhibition. If the bacterial sample had an average of +2 SD below 1, then it qualified to have antibiotic activity. The bacterial sample qualified to be a strong producer when at least 30% inhibition occurred against the pathogen.

Coculture

The high-throughput liquid assay method for screening cocultures was modified to a 6 day process. On the first day, selected colonies from agar plates were inoculated into 4 mL of 10% TSB and shaken for 24 h at 25 °C. On the second day, 140 μL of 10% TSB and 10 μL of the appropriate liquid bacterial culture were pipetted into a 96-well VWR tissue culture plate and shaken for 24 h at 25 °C. This added step allowed each bacteria an additional 24 h to grow separately before growing in a coculture. The different bacteria were placed into pairs based on the order of the 96-well VWR tissue culture plate made on the second day. The plate was flipped on itself horizontally and vertically or rotated 180̊ to make the random pairings (Figure S1 in Supporting Information). The resulting coculture pairs were checked to ensure that no repeats were tested during an assay. On the third day, 130 μL of 10% TSB was pipetted into the 96-well filtration plates (a 0.22 μM hydrophilic low protein binding Durapore membrane), and 10 μL of each bacteria in a pair were pipetted into a well so that each well contained a total of 150 μL. The plates were then shaken for 48 h at 25 °C. The last 3 days of this procedure follow the same methods outlined in the monoculture screening process.

Microbial Characterization

Gram Staining

Samples were placed onto a microscope slide by pipetting 10 μL of sterile deionized water and smearing a colony. The slide was allowed to dry completely and then subjected to gentile heat fixation via flame. The smear was subjected to crystal violet (60 s), iodine (60 s), decolorizer (80% ethanol, 10–20 s), and safranin (60 s). Between each step, the smear was washed thoroughly with deionized water. Bacterial cells were visualized under bright-field microscopy at 40× magnification.

Genomic Purification

Bacterial DNA was acquired using a PureLink Genomic DNA Mini Kit and protocol. Samples were grown in 4 mL of 10% TSB overnight, and cells (2 × 109) were harvested via centrifugation. Protocol for bacterial cell lysate was followed with respect to Gram-staining results. All samples were subject to purifying genomic DNA using a spin column-based procedure. DNA elution from the spin column for each sample was completed twice with a 50 μL PureLink Genomic Elution Buffer to a total volume on 100 μL. Purified bacterial DNA was kept at −18 °C.

PCR and Identification

Genus-level identification was determined using 16S rRNA amplification and DNA sequence analysis. Universal 16S rRNA primers were 27F (AGA GTT TGA TCM TGG CTC AG) and 1492R (CGG TTA CCT TGT TAC GAC TT). Polymerase chain reactions (PCRs) were completed for 50 μL reactions containing a 5× Phusion green HF buffer (10 μL), 10 mM dNTP (1 μL), Phusion DNA polymerase (0.5 μL), 10 mM universal primers (2.5 μL), 100 ng/mL template, and nuclease-free water. Thermocycler was set for initial denaturing at 98 °C (3 min), followed by 29 cycles of 98 °C (30 s), 48 °C (30 s), and 72 °C (5 min). After 29 cycles were completed, the sample underwent a final extension at 72 °C (5 min). PCR products were visualized for size separation on 1% agarose gel containing 0.5 μg/mL ethidium bromide in 1× TAE. A Quick-Load Purple 2-log DNA ladder (0.1–10.0 kb) was used for size reference. Products (approximately 1460 bp) were visualized under UV light and purified using QiaQuick PCR purification kit (Qiagen). Purified PCR product send-off was prepared in a total volume of 15 μL for sequencing at 4 ng/μL products, 25 pmol 27F primer, and nuclease-free water. Samples were sent to GeneWiz (Cambridge, MA) for Sanger sequencing. Sanger sequencing results were cross-checked with the National Center for Biotechnology Information and Ribosomal Database Project for genus-level identification.