Background

Developing an understanding of organisms in their natural ecological niches requires the ability to measure the dynamic interaction with their environment, either at the level of the individual or at population scales. For metazoa, a number of approaches have been utilized to track individuals of a species, including simple bands or markings conferring unique identifiers, Passive Integrated Transponders (PITs), telemetry devices, and biologgers [1, 2]. These approaches are limited to large organisms, as they require either direct visual inspection or electronic devices that can be attached by physical means to the body of the organism in question. As the field of environmental microbiology continues to mature, novel tools to facilitate “tag and release” studies are critical to understanding microbial interactions within existing environmental niches or in the context of introduction into new environments. Early efforts to track environmental fate of genetically modified organisms in field releases utilized fluorescently or metabolically marked strains of Pseudomonas putida [3, 4] and Pseudomonas fluorescens [5, 6]. Likewise, spontaneous rifampicin-resistant mutants have been used to track establishment and persistence of introduced isolates in field trials [7]. However, conventional selectable, chromogenic, or fluorescent markers carry metabolic costs that can compromise the carrier strain’s fitness in resource-constrained environments [8], revealing the need for phenotypically neutral, non-coding, genomic insertions that can differentiate introduced strains from native flora.

The development of DNA synthesis chemistry, microarray technology, quantitative PCR (qPCR), and high-throughput sequencing resulted in the development of several important capabilities and tagging approaches. Early studies used transposons containing short synthetic barcodes to identify virulence factors in several organisms [9, 10]. As oligonucleotide synthesis technology became more sophisticated and costs decreased, longer tags could be produced, resulting in the use of tagged strains to study the spatiotemporal dispersion in systems otherwise unamenable to tracking. In particular, significant work has been done to understand the details of stochastic dynamics of Salmonella infections by monitoring the relative quantities of tagged strains in different locations within the host [11,12,13]. These tagged strains, known as wild-type isogenic tagged strains (WITS), contain short, unique sequences inserted into the genome to allow quantitation by qPCR [11]. Similar work has been done to study population dynamics during infection for several other bacterial and viral pathogens [13,14,15,16,17,18,19,20].

The ability to track the fate of microbes introduced into an environment is also of interest to the biodefense research community. Spores of Bacillus anthracis, the causative agent of anthrax, were used in the high-profile 2001 anthrax mail attacks and were historically weaponized by both the United States and Soviet Union on large scales [21]. An important angle for preparedness against a potential attack includes an understanding of how spores released into the environment might disperse, persist, and migrate. The release of live B. anthracis spores (and indeed, even of attenuated strains) in an outdoor test is impossible due to public health concerns. Instead, close biological relatives are used as simulants. In the case of B. anthracis, recent work has used Bacillus thuringiensis serovar kurstaki (Btk) due to its similar physiological and biochemical properties [22,23,24,25,26]. Yet, even with the use of an adequate simulant, repeated dispersion testing on the same test site is problematic due to a need to distinguish between past and present testing, especially for a ubiquitous environmental bacterium such as Btk that is also in widespread use as a commercial biopesticide [27,28,29,30]. In addition, the problem of “signature erosion” has diminished the utility of endogenous genomic signatures as detection tools as the diversity of sequence data in public databases has exploded [31, 32].

To overcome these challenges, we previously inserted unique artificial genetic “barcodes,” designed to enable rapid detection by qPCR, into the Btk genome [24] and subsequently tested the system in a field release [23]. We note that the concept of “barcoding” described in this context differs from that used in taxonomic identification of living organisms [33], which—in contrast to the synthetic signatures employed here—uses endogenous mitochondrial and genomic DNA sequences as molecular classification tools. The barcoded strains constructed for our field release [23] were successfully detected in field samples using qPCR assays, but, like the earlier WITS strains, did not exploit the full ability of bioinformatics and synthetic biology that has become available. Most notably, each of the tags required its own specific PCR assay conditions, which makes scaling up to larger numbers of barcodes prohibitive. In this work, we have built upon our previous work by developing an algorithm, called barCoder, to generate barcode sequences that are unique among a pool of barcoded strains and require minimal development of qPCR assays. The algorithm also provides numerous features to minimize experimental troubleshooting efforts and customize amplicon properties. Here, we present the algorithm, as well as experimental validation of its ability to generate a potentially unlimited pool of highly diverse DNA barcodes, each with its own specific qPCR assay.

Results

Barcode design

We envisioned the possibility of creating synthetic barcodes that would amplify specifically, optimally, and in a manner consistent with the genomic characteristics of the target organism. We separated the barcode module into five constituent segments: two primer binding sites, the probe sequence, and two spacer sequences (Fig. 1a). Our design takes advantage of commonalities between the two major types of qPCR strategies, which are based on intercalating dyes (e.g. SYBR Green) that generate a fluorescent signal only in response to the formation of double-stranded DNA, or based on the liberation of a 5ʹ-linked fluorophore from a probe dually labeled with a 3ʹ-linked quencher by the 5ʹ–3ʹ exonuclease activity of the polymerase (referred to herein as TaqMan® for the probes used). Both assay types require a forward and reverse primer with similar constraints such as amplicon length, melting temperature (Tm), G+C content, GC clamp, potential secondary structure, and primer dimer formation. The primary design difference between the qPCR assay types is presence of a third probe sequence that is required only for TaqMan® assays; the probe sequence carries its own recommended design guidelines. Thus, in general, the same primer set can be used for either approach with the same barcode. From the perspective of qPCR assay design, the remaining spacer sequences between primers/probe are largely immaterial other than to meet ideal amplicon size targets, and therefore can be generated randomly, with constraints (see Algorithm design).

Fig. 1
figure 1

Overview of barcode design and algorithm work flow. a Barcodes consist of two synthetic primer binding sites, a probe annealing site, and two spacer regions. Spacer regions can be adjusted to match the overall G+C content of the organism to be barcoded. b barCoder algorithm work flow

Full size image

Algorithm design

The barCoder algorithm workflow is depicted in Fig. 1b. The algorithm starts by generating random primer/probe sequences to meet PCR-related specifications. Unlike typical primer/probe design where sequences are constrained by an existing sequence of interest, here there is almost complete freedom to design primer/probe sequences that have ideal PCR properties. Using an approximate melting temperature (Tm) formula (Eq. 1):

$$T_{m} = 69.4 + \frac{{41*\left( {n_{GC} - 16.4} \right)}}{{n_{total} }}$$
(1)

where \(n_{GC}\) is the number of G and C bases and \(n_{total}\) is the length of the primer/probe. The number of A’s+T’s and G’s+C’s needed to satisfy the specified Tm value can be calculated for primer/probe sequences within user-adjustable constraints on length and G+C content. From this set, a sequence meeting these constraints is randomly generated. For probe sequences generated containing more G’s than C’s, the complement sequence is used. All primer/probe sequences are subsequently screened for several PCR-related properties, such as maximum homopolymer repeats and secondary structure (Table 1). If any requirements are not met, the sequence is rejected and a new sequence generated.

Table 1 Algorithm parameters and default values for primer/probe design
Full size table

A sequence that meets PCR restrictions is then tested for uniqueness. First, the sequence is compared to a list of other primer/probe sequences, which includes any sequences already generated locally by the algorithm and an optional user-provided list of other primer/probe sets of interest. Sequence “matches” are determined by comparing the number of base pair matches of the top BLAST result divided by the length of the current primer to a user-adjustable threshold (Table 1). If the sequence matches any existing primer/probe sequences above the threshold, the sequence is discarded and the process restarted. Second, the genome of the organism targeted for insertion is scanned for similar sequences by BLAST. Similarly, a set of additional genome sequences of organisms that may be likely to be present in a sample, such as common environmental background species or human DNA, are scanned. Finally, the entire NCBI database is optionally scanned for similar sequences. The threshold for discarding a candidate sequence based on these BLAST results can be customized by the user, allowing more or less stringent criteria depending on project demands and acceptable CPU time in the case of very strict thresholds.

A sequence that meets all PCR and uniqueness requirements is accepted for use in the barcode. The algorithm cycles through this process to create each primer and the probe sequence, each with its own set of requirement parameters. Optionally, the forward primer can be set as constant for all barcodes in a given project. Once all three primer/probe sequences for a barcode have been generated, two spacer sequences are randomly generated such that the user-defined distance between the forward and reverse primers is met and the G+C content of the full barcode matches the G+C content of the target organism. The full barcode is then assembled by concatenating the forward primer sequence, first spacer sequence, probe sequence, second spacer sequence, and the reverse complement of the reverse primer sequence. The final check scans for potential stem-loop structures in the barcode to limit challenges during genome insertion and during amplification of the sequence. Failing this check triggers regeneration of the spacer sequences (Table 2).

Table 2 Algorithm parameters and default values other than for primer/probe design
Full size table

Experimental validation

The barCoder algorithm was used to generate an initial set of 21 barcodes and corresponding qPCR detection primer/probe sets (sequences listed in Tables S1 and S2 in Additional file 1). Due to code errors that were discovered and corrected after experimental validation was completed, these 21 assays were generated with less stringent checks than intended. Nonetheless, we report that all 21 assays were functional and anticipate that assays generated with the intended parameters would perform as well as or better than those reported here. Tables S3–S6 in Additional file 1 shows the actual properties of the 21 assays compared to what is implemented in the published version of the algorithm. Twelve of these barcodes were designed for B. thuringiensis serovar kurstaki (Btk), a surrogate for the biothreat agent B. anthracis with low-G+C content (35%, ref. [34]). To demonstrate the utility of the barCoder algorithm to create barcodes for other organisms, including those with different G+C compositions, three barcodes each were designed for potential use in Burkholderia pseudomallei 1026b (Bp; 68% G+C content), Yersinia pestis CO92 (Yp; 48%), and Clostridium botulinum Hall A (Cbot; 28%) [35,36,37].

Assay conditions for barcode Btk1 in the pIDTSMART-AMP plasmid backbone were optimized and subsequently standard curves were generated for all 21 TaqMan® qPCR assays using the same conditions (Fig. 2 and Fig. S1 in Additional file 1; Ct values and calculations provided in Additional file 2). All of the assays of the barcodes in plasmids performed well with qPCR efficiencies ranging from 81.1% to 100.0%, strong linear relationships (R2 > 0.99), and no false positive results (Table 3). Limits of detection (LODs) were all below 50 copies (the lowest plasmid concentration tested), except for barcode Btk6, where the LOD was below 500 copies (Table 3). Select barcodes were also markerlessly incorporated into the chromosomes of potential target organisms: barcode Btk1 was integrated into both Btk and B. anthracis Sterne, and barcode Yp1 was inserted into a pgm− derivative of Y. pestis CO92. Again, standard curves were generated for the TaqMan® assays under the same conditions (Fig. 2; Ct values and calculations provided in Additional file 2). Assays using chromosomally-barcoded strains had efficiencies within the range observed for barcodes residing in plasmids (86.5% to 96.5%), R2 values above 0.99, and no false positives (Table 3). LODs were calculated as less than 15 copies and less than 2 copies for barcode Btk1 in the chromosomes of Btk and B. anthracis Sterne, respectively, and less than 25 copies for barcode Yp1 in the chromosome of Y. pestis CO92 pgm− (Table 3). LODs are approximate as lower concentrations and Poisson distribution effects at low copy numbers were not thoroughly interrogated.

Fig. 2
figure 2

Representative qPCR assay standard curves. Curves were generated using a barcode Btk1 in the pIDTSMART-AMP plasmid backbone, b barcode Btk1 inserted into the B. thuringiensis kurstaki chromosome, c barcode Btk1 inserted into the B. anthracis Sterne chromosome, and d barcode Yp1 inserted into the Y. pestis CO92 pgm− chromosome. For each standard curve, data from three replicates and a trendline are shown. Standard curves for the remaining barcodes in the plasmid backbone are shown in Figure S1 in Additional file 1

Full size image
Table 3 Evaluation of qPCR assays from generated standard curves
Full size table

To test the specificity of the TaqMan® qPCR assays for the corresponding barcode, each of the 12 Btk assays were tested against all 12 Btk barcodes in plasmids. This cross-reactivity panel showed unique amplification of each Btk barcode with its cognate primer/probe set (Fig. 3; raw data provided in Additional file 3). The TaqMan® qPCR assay for barcode Btk1 was also tested against a panel of potential pathogens and environmental organisms (Table 4; raw data provided in Additional file 4). Reactions containing the Btk strain with barcode Btk1 inserted in the chromosome, either alone or in the presence of an environmental matrix (DNA from a mock microbial community or DNA extracted from soil) showed robust positive results, while the Btk1 qPCR assay did not cross-react with any of the potential contaminants.

Fig. 3
figure 3

Cross-reactivity of the 12 Btk qPCR assays against the 12 Btk barcodes. For each qPCR reaction, 10–12 g (~ 450,000 copies) of the pIDTSMART-AMP plasmid backbone containing the Btk barcode indicated was used as DNA template. Threshold cycle (Ct) values shown are the median of three replicates; ND, Not determinable

Full size image
Table 4 Cross-reactivity of the barcode Btk1 qPCR assay against a panel of potential pathogens and environmental contaminants
Full size table

Discussion

qPCR has become a standard technique for detection of microorganisms in the environment and for diagnosis of infection [38], and, as such, is an attractive detection technology that also allows a rapid evaluation of the relative abundance of a known microorganism within a sample. However, when conducting environmental fate studies, for example, these assays must discriminate from the endogenous or native microflora, which may be uncharacterized and present signatures similar to or cross-reactive with the signature selected for detection of the experimental strain. We sought in this report to utilize a bioinformatics strategy to generate specific amplicons that require minimal assay optimization and could be introduced into organisms with minimal to no cross-reactivity with environmental and microbial signatures.

Our approach to developing unique qPCR-compatible barcodes expanded upon our previous work, in which we appropriated synthetic signatures from published microarrays and developed PCR assays based on the unique sequences generated both by the tags themselves and by their insertion into the genome [24]. Because those sequences were not designed de novo for use in PCR detection assays, we relied on the presence of a chromosomal primer binding site and the single synthetic sequence to generate suitable amplicons. As a result, considerable optimization of the assay conditions and primer sequences was necessary during the development of those strains, and the assay conditions for each tag required slightly different optimal conditions for detection. This situation was judged as suboptimal for the development of a more diverse panel of barcodes, as new assays would need to be developed for each new sequence.

We therefore sought to develop an algorithm that would enable the high-throughput generation of amplicon sequences that could use a single PCR assay condition, and in which relative proportion of each strain could be compared in a single test, e.g. across a single microplate. The assays would need to be specific to each barcode, and would need to be comparably sensitive with equivalent limits of detection. Using TaqMan® qPCR chemistry and a bioinformatic screening algorithm, we generated a panel of unique primer/probe combinations that exhibited selectivity, specificity, and sensitivity. Using conventional plasmids containing the barcodes as templates for the development of the assays, we demonstrate strong performance in linearity of response, sensitivity, and efficiency across 21 assays using conditions optimized for a single assay. No cross-reactivity was observed across a panel of 12 of these assays. We note that the odds of randomly generating a barcode that would react with a natural sequence is vanishingly small, as three 20 + base primer/probe sequences would need to be closely matched in the correct orientation (> 1036 possible sequences) with spacing appropriate for PCR amplification; nonetheless, sequences are screened for uniqueness to further minimize this possibility. Inserting two of the barcodes into three different genomes, we observed conserved performance compared to plasmid assays and LODs below 25 copies, which we believe to be conservative due to Poisson distribution effects at low copy numbers. We note that the code used to generate the barcodes contained errors that resulted in application of less stringent thresholds for primer and probe design, yet still resulted in highly sensitive and specific assays. The remediated code should therefore reliably produce sequences that function equally well if not better than the barcodes validated here while reducing the possibility of generating suboptimal primer sequences.

Our barcodes have a number of potential applications. Marking strains with unique artificial signatures could aid in protecting intellectual property, particularly for production strains whose development has required significant investment in metabolic and/or genetic optimization, perhaps in combination with other techniques such as DNA steganography [39]. While not as information-rich as longer steganographic tags or watermarks [40, 41], qPCR barcodes have the advantage of not requiring further sequencing and informatic analysis to detect and/or verify their presence; they must simply be amplified using appropriate primers and probes. In one scenario, a set of barcodes could be inserted at defined intervals throughout a large DNA molecule used for information storage, and utilized to provide a preliminary indicator of the stability of the archive prior to full sequencing.

These sequences and their associated assays might also find use in forensic applications. In particular, one might imagine their use as molecular taggants that could be spiked into samples by field technicians, and their detection in DNA samples by the reference laboratories would serve to verify the origin of the sample. In a similar vein, these same tools could be used in the future for downstream attribution of accidental or deliberate release of organisms [42]. Select agent strains, in particular, could be tagged, distributed to end-user communities, and then any material from the scene of a biocrime could be rapidly amplified using the library of primers and probes.  Investigators could then focus investigative resources on those potential sources, while excluding the majority of the research laboratories that possess variants containing other barcodes. Any mechanism by which artificial genetic diversity can be introduced into the largely clonal populations of laboratory strains would be useful as all known acts of bioterrorism to date have utilized common laboratory strains (e.g. B. anthracis Ames Ancestor in the 2001 U.S. Mail/Amerithrax case; S. enterica serovar typhimurium 14028s in the Rajneeshi cult attacks of 1984 [43]). In the case of the Amerithrax samples, discriminating between samples present in these laboratories relied on presence of several spontaneous mutants whose discovery and characterization required astute microbiologists and what were at the time Herculean sequencing efforts [44, 45]. The deliberate incorporation of end-user-specific sequences into such commonly available strains could immeasurably speed identification of potential originating laboratories, would help investigators narrow their focus to a subset of potential sources, and would help exclude uninvolved laboratories working on similar research as potential sources. Furthermore, the presence of such signatures (and the knowledge that significant additional effort would be required to disguise the source of a sample) could deter potential malefactors within those laboratories even if the location, sequence, and properties of the sequence were known.

Conclusions

To our knowledge, barCoder represents the first completely in silico method for generating both a synthetic target for qPCR and the primers/probe to amplify the target, and optimal assay conditions for detection of a diverse range of barcodes. We demonstrated that generated barcodes all perform well under a single set of assay conditions and show no cross-reactivity with themselves or environmental contaminants. Insertion of the barcodes into the genomes of three organisms of interest maintained the key properties of the barcodes. We anticipate barCoder finding utility in applications such as environmental fate studies, intellectual property, and microbial forensics.

Methods

Algorithm implementation

All software was written in Perl. G+C and A+T constraints were calculated using Eq. 1 (see above). The algorithm workflow is described in “Algorithm design” (see Results, above) and is shown in Fig. 1b. Most bioinformatics functions were implemented using existing BioPerl modules. EMBOSS software, called by BioPerl, was used to predict stem-loop structures. All BLAST runs used default BioPerl parameters. Software is available on GitHub (https://github.com/ECBCgit/Barcoder).

barCoder-designed elements and sources of DNA

The properties of the barcode modules designed in this study using an earlier version of the barCoder algorithm are listed in Tables S3–S6 in Additional file 1. For the BLAST step against the target organism genome sequence (Table S5 in Additional file 1), the following NCBI RefSeq accession numbers were used: B. thuringiensis serovar kurstaki, NZ_CP010005.1; B. pseudomallei 1026b, NC_017831.1 (chromosome 1) and NC_017832.1 (chromosome 2); C. botulinum Hall A, NC_009495.1; Y. pestis CO92, NC_003143.1. All barcodes, primers, and probes were obtained from Integrated DNA Technologies, Inc. (IDT, Coralville, IA) and their sequences listed in Tables S1 and S2 in Additional file 1. SacI and NheI restriction sites flanking the Btk barcodes were added to facilitate later subcloning. Barcodes were received as “minigenes” inserted in the pIDTSMART-AMP plasmid backbone and were propagated in NEB® 5-alpha E. coli (New England Biolabs, Inc., Ipswich, MA) on Luria–Bertani (LB) agar and in LB broth with 100 µg/ml ampicillin at 37 °C. Plasmid DNA was isolated using the QIAprep Spin Miniprep Kit (Qiagen, Hilden, Germany). DNA probes were ordered as PrimeTime double-quenched qPCR probes containing the 5ʹ FAM fluorophore, 3ʹ Iowa Black FQ quencher, and internal ZEN quencher. The sources of the DNA used for the cross-reactivity panel of pathogenic and environmental organisms are given in Table S7 in Additional file 1.

qPCR

All qPCR experiments were run on an Applied Biosystems 7900HT Real-Time PCR System (Applied Biosystems, Foster City, CA) using Applied Biosystems MicroAmp optical 384-well reaction plates (catalog number 4309849) sealed with Applied Biosystems MicroAmp optical adhesive film (catalog number 4311971). Optimized 20 µL reactions included Applied Biosystems TaqMan® Universal PCR Master Mix (catalog number 4304437), forward and reverse primers each at a final concentration of 900 nM, DNA probe at a final concentration of 250 nM, 1 µL DNA template at the indicated concentration, and nuclease-free water. TaqMan® assays used the following thermocycler protocol: 1 cycle of 50 °C for 2 min, 1 cycle of 95 °C for 10 min, and 40 cycles of 95 °C for 15 s and 55 °C for 1 min. The standard curve properties of each assay were assessed by performing tenfold serial dilutions of the template DNA in nuclease-free water. Efficiency and linearity (R2) values for each qPCR standard curve were calculated using the median Ct of three replicates for each template DNA dilution. Data points corresponding to the highest amount of template DNA tested (10–8 g for genomic DNA, 10–9 g for plasmid DNA) were omitted from these analyses in all cases as the Ct values tended to be non-linear with the other data points of the standard curve. LODs were conservatively estimated using the lowest amount of template DNA tested that produced a Ct value < 40 for all three replicates. Ct values and calculations used to generate and analyze all standard curves for the 21 barcodes in this study are provided in Additional file 2.

Construction of genomically-barcoded strains

B. thuringiensis and B. anthracis strains were routinely cultured on Brain Heart Infusion (BHI) agar and in BHI broth at 30 °C (B. thuringiensis) or 37 °C (B. anthracis). Unless otherwise indicated, Y. pestis strains were grown on BHI agar or Tryptic Soy Agar (TSA), and in BHI broth at 28–30 °C. Genomic DNA was extracted using the UltraClean Microbial DNA Isolation Kit (MOBIO Laboratories, Inc., Carlsbad, CA). Barcode Btk1 was selected to construct a strain in which the barcode was markerlessly incorporated into the chromosome of B. thuringiensis serovar kurstaki HD-1 [24, 46] (obtained from the DoD Unified Culture Collection (https://www.usamriid.army.mil/ucc/)). The insertion was generated at the same locus that was identified and modified in our previous report (within Target 1, [24]). This corresponds to an insertion between positions 4,834,064 and 4,834,065 of NCBI RefSeq accession number NZ_CP010005.1. Plasmid pRP1028-T1-PL (GenBank accession number MW055899), a derivative of pRP1028 [47], was designed specifically for incorporating synthetic elements within this target region of the Btk chromosome and was synthesized by DNA2.0 (Menlo Park, CA). Plasmid pRP1028-T1-PL contains 1,550 bp of DNA homologous to the Btk chromosomal insertion region between the pRP1028 HindIII and BamHI sites, as well as a 36-bp polylinker within the homology region. Following digestion of plasmid pIDTSMART-AMP:Barcode Btk1 with SacI and NheI, the Btk1 barcode was gel extracted (QIAquick Gel Extraction Kit, Qiagen, Hilden, Germany) and ligated with pRP1028-T1-PL that had been digested with the same restriction enzymes. This pRP1028-T1-PL derivative containing barcode Btk1 was introduced into Btk, and the barcode was incorporated into the chromosome using the markerless allelic exchange strategy described previously [47]. Successful barcode integration into the Btk chromosome was verified by PCR amplification of the target locus and SacI/NheI digestion of the resulting amplicon. Construction of a strain of B. anthracis Sterne 34F2 with barcode Btk1 in the chromosome was previously published [48].

For construction of a strain of Y. pestis CO92 pgm− with barcode Yp1 markerlessly inserted in the chromosome, the locus between the convergently transcribed genes YPO0388 and YPO0392a (NCBI RefSeq accession number NC_003143.1) was selected using rules adopted from Buckley et al. [24], in combination with the PATRIC database [49] and available transcriptome sequencing (RNA-seq) data (NCBI SRA accession numbers SRR1013703, SRR1013704, SRR1013705, SRR1041589), to identify a potentially neutral insertion region. The barcode was then inserted into the chromosome between positions 406,742 and 406,743 via the method described by Sun et al. [50], which utilizes λ Red recombination and sacB counterselection. Briefly, plasmid pKD46 (CGSC #7739, [50]) containing the genes for λ Red recombination was electroporated into a strain of Y. pestis CO92 pgm− (strain R88, Robert Perry, University of Kentucky). A linear DNA fragment containing a cat-sacB cassette flanked by DNA homologous to the Y. pestis chromosomal insertion region was electroporated into this pKD46-containing strain of Y. pestis, and successful integrants were selected on media containing chloramphenicol. Following electroporation with a linear DNA fragment containing barcode Yp1 flanked by homologous DNA and selection on media containing sucrose, the cat-sacB cassette in the chromosome was replaced with the barcode. The resulting strain was subsequently cured of pKD46, and successful barcode insertion was verified by PCR amplification and sequencing. Whole-genome sequencing (MiSeq, Illumina) was also performed to confirm the absence of off-target modifications. Primers used to construct this barcoded strain of Y. pestis are listed in Table S8 in Additional file 1. To generate the cat-sacB cassette, the cat gene was PCR amplified from plasmid pKD3 (CGSC #7631, [51]) with primers #1 and #2, and the sacB gene was PCR amplified from plasmid p88171 (synthesized plasmid with pJ207 backbone and sacB gene from Bacillus subtilis, DNA 2.0, Menlo Park, CA) with primers #3 and #4. The two PCR amplicons were purified (QIAquick PCR Purification Kit, Qiagen, Hilden, Germany) and joined together by overlap extension PCR [52] using primers #1 and #4. The cat-sacB cassette was gel extracted and cloned between the SacI and BamHI sites of pUC19 to create plasmid pCBV4. To construct the cat-sacB cassette flanked by homologous DNA, the cat-sacB cassette was PCR amplified from pCBV4 with primers #5 and #6. Approximately 500 bp flanking each side of the barcode insertion point were separately PCR amplified from the Y. pestis CO92 pgm− chromosome; primers #7 and #8 were used to amplify upstream DNA, and primers #9 and #10 were used to amplify downstream DNA. The three purified PCR amplicons (up flanking region, cat-sacB cassette, and down flanking region) were joined together by overlap extension PCR [52] using primers #7 and #10, and the resulting amplicon was gel extracted and cloned into the pCR™4Blunt-TOPO® vector (Invitrogen, Carlsbad, CA) to generate plasmid pCBV6. The linear DNA fragment containing the cat-sacB cassette flanked on both sides by Y. pestis CO92 pgm− DNA was PCR amplified from pCBV6 with primers #7 and #10. To create barcode Yp1 flanked by homologous DNA, the barcode was PCR amplified from the synthesized plasmid pIDTSMART-AMP:Barcode Yp1 using primers #11 and #12. Approximately 500 bp flanking each side of the barcode insertion point were separately PCR amplified from the Y. pestis CO92 pgm− chromosome; primers #7 and #13 were used to amplify upstream DNA, and primers #10 and #14 were used to amplify downstream DNA. The three purified PCR amplicons (up flanking region, barcode, and down flanking region) were joined by overlap extension PCR [52] using primers #7 and #10, and the resulting amplicon was gel extracted and cloned into the pCR™4Blunt-TOPO® vector (Invitrogen, Carlsbad, CA) to generate plasmid pCBV9. The linear DNA fragment containing barcode Yp1 flanked on both sides by Y. pestis CO92 pgm− DNA was PCR amplified from pCBV9 with primers #7 and #10.