Early detection of viable Francisella tularensis in environmental matrices by culture-based PCR

Background Francisella tularensis is a fastidious, Gram-negative coccobacillus and is the causative agent of tularemia. To assess viability yet overcome lengthy incubation periods, a culture-based PCR method was used to detect early growth of the lowest possible number of F. tularensis cells. This method utilized a previously developed enhanced F. tularensis growth medium and is based on the change in PCR cycle threshold at the start and end of each incubation. Results To test method robustness, a virulent Type A1 (Schu4) and B (IN99) strain and the avirulent Live Vaccine Strain (LVS) were incubated with inactivated target cells, humic acid, drinking and well water, and test dust at targeted starting concentrations of 1, 10, and 100 CFU mL− 1 (low, mid, and high, respectively). After 48 h, LVS growth was detected at all targeted concentrations in the presence of 106 inactivated LVS cells; while Schu4 and IN99 growth was detected in the presence of 104 Schu4 or IN99 inactivated cells at the mid and high targets. Early detection of F. tularensis growth was strain and concentration dependent in the presence of fast-growing well water and test dust organisms. In contrast, growth was detected at each targeted concentration by 24 h in humic acid and drinking water for all strains. Conclusions Results indicated that the culture-based PCR assay is quick, sensitive, and specific while still utilizing growth as a measure of pathogen viability. This method can circumvent lengthy incubations required for Francisella identification, especially when swift answers are needed during epidemiological investigations, remediation efforts, and decontamination verification.

F. tularensis is divided into three subspecies which vary in their pathogenicity and geographic distribution: tularensis (Type A), holarctica (Type B), and mediasiatica with ongoing debate whether to classify F. novicida, as either a fourth subspecies or separate species [7,8]. Type A strains are further divided into two subpopulations, A1 (A.I. or A-east) and A2 (A.II. or A-west) with four distinct genotypes (A1a, A1b, A2a, and A2b) [9]. In the U.S., tularemia can be caused by both Type A and B strains with higher rates of mortality resulting from infections with A1b strains compared to A1a, A2, and Type B strains [9,10].
Due to its low infectious dose, high virulence, and transmissibility by air, misuse of F. tularensis has the capability to cause a high-consequence event with potentially large casualties, negative effects on the economy and infrastructure, and threats to public health and safety. Therefore, F. tularensis is classified as a Tier 1 select agent by the Federal Select Agent Program, which is managed jointly by the U.S. Department of Health and Human Services and U.S. Department of Agriculture [11]. The U.S. Environmental Protection Agency (USEPA) Homeland Security Research Program provides science and technology needed for effective response and recovery from natural, intentional, or accidental environmental catastrophes, including public health threats from microbial pathogens. During these events, numerous and diverse sample types (e.g. aerosol, surface, environmental) will need to be processed and analyzed, underscoring the need for sensitive, specific, quick, and straightforward methods to determine the extent of contamination and efficacy of remediation efforts. Although culturability is a traditional indicator of viability, F. tularensis isolation and identification by culture is challenging because the organism is highly infectious, fastidious, slow-growing (requiring incubations of up to 10 days), difficult to identify, and can be outcompeted in culture medium by other microorganisms present in environmental or clinical samples [10]. Thus, traditional culture methods for F. tularensis would not meet USEPA's response and recovery needs during homeland security incidents.
Molecular detection methods such as polymerase chain reaction (PCR) cannot distinguish between culturable (potentially infectious) and inactivated pathogens since DNA from both types are present in the sample. However, features of PCR, such as rapidity, sensitivity, and specificity, can be combined with the requirement for growth in culture medium to quickly detect low concentrations of viable Francisella bacteria. Specifically, this culture-based PCR method is based on the change in PCR cycle threshold (ΔC T ) which is calculated by subtracting the cycle threshold (C T ) at time 0 (C T0 , i.e. starting DNA levels) from the C T value after incubation of samples in culture medium (C Ti , i.e. starting DNA levels in addition to those that have accumulated during the incubation): ΔC T = C T0 -C Ti [12]. To help achieve the short incubation period required for early detection, an enhanced culture medium, composed of brain heart infusion broth supplemented with Vitox, Fildes, and histidine (BVFH), was previously developed for use in this method to promote early growth and enhanced cultivation of F. tularensis [13]. The method used in this study is similar to the rapid viability PCR method for detection of F. tularensis with the main differences being sample volumes, time points analyzed, and method of preparing the BVFH medium [14]. Other differences included the evaluation of (1) real environmental matrices (e.g. drinking water and well water), which are ecologically relevant to F. tularensis; (2) method sensitivity amidst high concentrations of inactivated target cells; and (3) more F. tularensis strains and PCR primer/probe sets to verify specificity and sensitivity for both Type A and B strains. These method parameters, not tested previously, would be relevant and important features during cleanup scenarios after an actual contamination event. Specifically, the bacterial Type would be unknown, high concentrations of inactivated and viable target organisms would be present, in addition to nontarget microorganisms and inhibitors found in the surrounding environment.
In this study, three Type A and B strains (Schu4, IN99, and Live Vaccine Strain [LVS]) were used to evaluate performance of this culture-based PCR method in various water matrices containing chemical and microbial challenges at low, mid, and high starting concentrations (1, 10, and 100 CFU mL − 1 targets, respectively). Chemical challenges included humic acid and metal oxides which could negatively affect F. tularensis growth, recovery from water samples, and/or interfere with downstream analyses. Various concentrations of inactivated F. tularensis cells were used as a microbial challenge to simulate post-decontamination/remediation scenarios and to determine the minimum concentrations and length of incubations required for viable cells to overcome the PCR signals from the inactivated or unculturable cells. Other microbial challenges included indigenous organisms present in well water, drinking water, and test dust to assess potential overgrowth and negative impact on the culture-based PCR detection of F. tularensis.

Evaluation of PCR assays
Twelve strains of Francisella spp. representing Type A1 and A2, Type B, F. novicida, and F. philomiragia were used to determine specificity of three PCR assays targeting the conserved Francisella fopA gene, Type A-specific pdpD gene, and Type B-specific ISFtu2 gene. Total genomic DNA was isolated from each of the twelve Francisella spp. strains and quantitative PCR using cellular standards were performed for each gene assay to confirm amplicon size, specificity, and sensitivity (Table 1). In agreement with previous studies, Type A strains Schu4, F2246, H3563, and KC1482 were positive for the fopA and pdpD, but not the ISFtu2 PCR assays; while Type B strains LVS, IN99, KY99, OR96, and NY98 were positive for fopA and ISFtu2 but not pdpD PCR assays [15][16][17]. Additionally, F. novicida and F. philomiragia were negative for all three PCR assays except for detection of the fopA gene in F. novicida as expected from a previous report [17]. For the Type A (Schu4) and Type B (IN99 and LVS) experimental samples, no statistical differences were observed between C T values from the fopA and pdpD or ISFtu2 gene assays (data not shown); thus, C T values from the fopA gene assay were used and ΔC T calculated as described in the Methods.

Detection of F. tularensis in the presence of inactivated target cells
Inactivated target cells were generated by incubation in isopropanol for use in determining the minimum concentrations and length of incubations required for viable cells to overcome the PCR signals from their inactivated counterparts. Control wells containing 0, 10 2 , 10 4 , or 10 6 of the inactivated IN99 and Schu4 cells, analyzed at 0, 8, 24, and 48 h, confirmed that these cells were not viable and that no change in PCR signal was observed over time (SFigure). In Fig. 1, data from the low, mid, and high targeted starting concentrations of 1, 10, and 100 colony forming units (CFU) mL − 1 are shown in blue, red, and black, respectively; and within each of those groups, open circles, squares, triangles, and diamonds represent samples that contained 0, 10 2 , 10 4 , and 10 6 of inactivated cells, respectively. The actual starting concentrations (mean CFU mL − 1 ± standard deviation [SD]) for strain IN99 were 3 ± 1, 26 ± 9, and187 ± 77; and for strain Schu4 were ≤ 1, 8 ± 0, and 45 ± 7 for the low, mid, and high targets, respectively ( Table 2). Actual starting concentrations for strain LVS was not performed. However, for each of the three targets, strains LVS, IN99, and Schu4 displayed expected CFU increases after 8, 24, and 48 h of incubation indicating growth was not inhibited in the presence of their respective inactivated cells ( Fig. 1 a, e, and i). Because the actual titer of Schu4 was ≤1 for the low target, viable Schu4 cells may not have been inoculated into the wells of the 48-h plate containing 10 2 inactivated Schu4 cells (Fig. 1 i, blue square, < limit of detection (LOD) of 1 log 10 CFU mL − 1 for all three replicates).
At the 100 CFU mL − 1 target, a ΔC T ≥ 6 was achieved by 24 h in LVS and Schu4 samples containing ≤10 4 inactivated cells ( Fig. 1 b and j, respectively), except for IN99 where the mean ΔC T(T0-T24) ± SD in the presence of 10 4 inactivated cells was 3.6 ± 0.6 ( Fig.  1 f, black open triangle). By 48 h, only strain LVS displayed the ΔC T ≥ 6 for all samples (Fig. 1b), while The following isolate was obtained through the NIH Biodefense and Emerging Infections Research Resources Repository, NIAID, NIH: Francisella novicida, Strain Utah 112, NR-13 PCR amplicon/signal "+" detected or "-" not detected. Data are representative of three replicates Sensitivity for each assay based on standard curves was 10 cell equivalents per mL strain IN99 and Schu4 had a mean ΔC T(T0-T48) ± SD of 4.4 ± 1.2 and 3.7 ± 0.6 in the presence of 10 6 inactivated target cells but ΔC T ≥ 6 for all other samples ( Fig. 1 f and j). The same trend was observed at the 10 CFU mL − 1 target where the ΔC T ≥ 6 was achieved by 24 h in the LVS and Schu4 samples ( Fig. 1 c and k, respectively) containing ≤10 4 inactivated cells and not for IN99, where the mean ΔC T(T0-T24) ± SD in the presence of 10 4 inactivated cells was 1.0 ± 0.3 (Fig. 1g, red open triangle). By 48 h, only strain LVS displayed the ΔC T ≥ 6 for all samples (Fig. 1c), while strain IN99 and Schu4 had a mean ΔC T(T0-T48) ± SD of 4.7 ± 0.3 and 2.4 ± 1.5 in the presence of 10 6 inactivated target cells but ΔC T ≥ 6 for all other samples ( Fig. 1 g and k).
As expected, for the initial target concentration of 1 CFU mL − 1 , a longer incubation period was needed to overcome the PCR signal detected from the inactivated target cells. A ΔC T ≥ 6 at 24 h could not be achieved in samples containing > 10 2 inactivated target cells for LVS and Schu4 ( Fig. 1 d and l, respectively) while the ΔC T ≥ 6 was only observed for IN99 samples in the absence of inactivated cells (Fig. 1h, blue open circle). After 48 h, ΔC T ≥ 6 was achieved for all LVS samples (Fig. 1d) and for IN99 samples containing ≤10 4 inactivated cells (Fig.  1h) in contrast to Schu4 where the ΔC T ≥ 6 was only achieved in the presence of 10 4 , but not 10 2 , inactivated cells ( Fig. 1l) most likely due to the low starting inoculum where certain samples may not have received any viable Schu4 cells (Table 2). , and Schu4 (i-l) were incubated in BVFH media containing 0 (circle), 10 2 (square), 10 4 (triangle), or 10 6 (diamond) cells mL − 1 of their respective isopropanol-inactivated cells. For each inactivated target, culturable cells were added at targeted starting concentrations of 100 (black symbols), 10 (red symbols), and 1 (blue symbols) CFU mL − 1 . Growth was monitored by CFU enumeration (a, e, and i) and results of the culture-based PCR assay are shown in the bottom three rows. ΔC T of 6 is indicated by the bold black line F. tularensis detection in the presence of environmental inhibitors Humic acid Humic acid was used as a surrogate for natural organic matter to test for possible inhibition of F. tularensis growth and downstream PCR analyses. In these experiments, actual starting concentrations (mean CFU mL − 1 ± SD) for strain LVS were 2 ± 1, 18 ± 3, and 171 ± 48 CFU mL − 1 for the low, mid, and high targets, respectively (Table 2). For strain IN99, the starting inoculum was approximately 1 log 10 unit lower for the low, mid, and high targets displaying ranges of < 1, < 1 to 2, and 7-17, respectively (Table 2). CFU could not be enumerated for Schu4 at the high target due to overgrowth; however, the mean CFU mL − 1 ± SD actual starting concentrations for the 1 and 10 CFU mL − 1 target were 2 ± 1 and 26 ± 6 CFU mL − 1 , respectively ( Table 2).
Growth of F. tularensis strains LVS and Schu4 was not inhibited in the presence of 50 μg mL − 1 humic acid since CFU levels were not statistically different between their respective control and humic acid treated samples at each targeted starting concentration (P > 0.05, Fig Using the culture-based PCR method, a ΔC T ≥ 6 was achieved following 24 h of incubation for LVS and Schu4 at each targeted starting concentration ( Fig. 2 b-d, circle and triangle symbols, respectively) and for strain IN99 at the mid and high targets (Fig. 2 b and c, filled squares). Like the growth trends observed for strain IN99 at the 1 CFU mL − 1 target, the potential exclusion of viable IN99 cells into the samples may have resulted in an observed ΔC T < 6, which could be a false positive indicator of humic acid inhibition (Fig. 2d, open and filled squares).

Arizona test dust
Arizona Test Dust (ATD) was used as an additional environmental test matrix to further evaluate F. tularensis detection with the culture-based PCR method. Strains IN99 and Schu4 were unable to be tested due to time and logistical constraints with the BSL-3 facility; thus, only the reference LVS strain was used for the ATD test matrix experiments. For LVS, actual CFU densities  (mean CFU ± SD) were 1 ± 0, 14 ± 1, 71 ± 22, and for the low, mid, and high targeted starting concentrations, respectively, for each replicate ( Table 2). For the samples containing ATD, LVS CFU levels were < limit of detection (LOD) because ATD background organisms overgrew the agar plate and F. tularensis colonies could not be identified (Fig. 2e, filled in symbols). Fig. 3 a and b show images of ATD microorganisms after incubation in BVFH media for 0 and 48 h, respectively. At 0 h, heterogeneous colony morphologies with various types of form, elevation, and margin was observed (Fig. 3a). In contrast, by 48 h, growth was dominated by a homogeneous mixture of microbes exhibiting large spreading colonies at high concentrations (Fig. 3b). ATD microbes exhibited rapid growth in the BVFH medium, with concentrations (mean log 10 CFU mL − 1 ± SD) of 3.6 ± 0.0 at 0 h; 7.1 ± 0.6 at 24 h; and 7.8 ± 0.3 at 48 h. It is unclear from the culture data if ATD microbes inhibited LVS or masked their growth on agar plates. However, PCR results showed significantly lower levels of LVS in the presence of ATD compared to controls indicating that LVS growth was inhibited or outcompeted Although LVS growth was not detectable, the ΔC T ≥ 6 was achieved in this experiment after 24 h for each targeted starting concentration (Fig. 2 f-h). Although the ΔC T ≥ 6 was achieved by 24 h, the ΔC T did not significantly increase by 48 h, for the mid and high targets, indicating growth of the LVS cells was compromised in the presence of ATD. At the 100 CFU mL − 1 target, ΔC T ± SD values for the ATD containing samples at 24 and 48 h were 6.1 ± 3.2 and 6.8 ± 3.6, respectively (Fig. 2f, filled black circles). Similarly, at the 10 CFU mL − 1 target, ΔC T ± SD values for the ATD containing samples at 24 and 48 h were 7.4 ± 1.9 and 13.0 ± 0.8, respectively (Fig. 2g, filled red circles). At the 1 CFU mL − 1 target, the ΔC T ≥ 6 was achieved at 24 h but not at 48 h (mean ΔC T ± SD of 6.3 ± 1.2 and 3.4 ± 2.1, respectively, Fig. 2h) indicating that inhibition or suppression of LVS growth by ATD microorganisms may be concentration dependent.

Effect of indigenous microorganisms on F. tularensis detection in various water matrices
Since contaminated water sources have been implicated in tularemia outbreaks, well water and drinking water were used as additional environmental test matrices to evaluate the culture-based PCR method for F. tularensis detection. Microbial and chemical water quality data are presented for each well and drinking water experiment using strains LVS, IN99, and Schu4 (Table 3). Heterotrophic plate count (HPC) bacteria in well and drinking water ranged between 1.6-3.2 and < LOD to 3.2 log 10 CFU mL − 1 , respectively. Total coliforms and Escherichia coli levels in the well water ranged between 24 to 3650 and < 1 to 34 most probable number (MPN) per 100 mL, respectively (Table 3).
Among the chemical parameters measured, pH, temperature, conductivity, hardness, and silicon concentrations were statistically different between the two test water matrices (Table 3, P < 0.01) and most likely alkalinity as well, but this parameter was not measured for the Schu4 experiments, thus statistical analysis could not be performed. Well and drinking water HPC bacteria displayed heterogeneous colony morphologies at 0 h ( Fig. 3 c and e, respectively); however, by 48 h, HPC growth was comprised of a less diverse subset of microbes with well water HPC bacteria growing to high concentrations (Fig. 3d, 10 − 7 dilution of well water versus Fig. 3f, 10 0 dilution of drinking water). Similar to the ATD microbes, those in well water exhibited rapid growth in the BVFH medium reaching high CFU concentrations by 48 h. Specifically, well water HPC levels (mean log 10 CFU mL − 1 ± SD) after 0, 24, and 48 h of growth were 1.7 ± 0.3, 8.2 ± 0.2, and 9.0 ± 0.2, respectively, compared to drinking water HPC levels of 3.3 ± 0.4, 1.8 ± 0.2, and 2.8 ± 0.3 at the same time points.

Well water
Actual starting densities (mean CFU ± SD) for strain LVS were < 1, 4 ± 1, and 25 ± 5; for IN99 were 1 ± 1, 12 ± 3, and Images are representative of three replicates 72 ± 11; and for Schu4 were 1 ± 0, 12 ± 1, and 111 ± 11 for the low, mid, and high targets, respectively ( Table 2). For samples containing well water, CFU levels for all three F. tularensis strains were < LOD because well water microbes overgrew the agar plate and F. tularensis colonies could not be identified (Fig. 4a, filled in symbols). In contrast, control samples containing sterilized well water (sWW), growth of strains IN99 and Schu4 was observed at 8, 24, and 48 h in a starting concentration dependent manner as expected (Fig. 4, open squares for IN99 and triangles for Schu4, with the low, mid, and high targets shown in blue, red, and black, respectively). Except for the 24 h time point at the 1 and 10 CFU mL − 1 target, LVS cells displayed the same growth pattern throughout the time course (Fig. 4a, missing red and blue open circles at 24 h, but present at all other time points). Those 24 h CFU values could not be determined due to the dehydration of agar plates for a subset of the serially diluted samples while the remaining subset were contaminated with morphologically different colonies than those observed with the LVS in well water samples.
For control, sWW samples, a ΔC T ≥ 6 was observed at 24 and 48 h for each strain at all targeted starting concentrations (Fig. 4 b-d open symbols). However, for strains LVS and IN99 in the well water samples, the mean ΔC T was < 6 and ranged from − 3.8 to 3.1 for LVS and − 1.6 to 5.8 for IN99 for all targeted concentrations and time points (Fig. 4  b-d, filled circles and squares). Notably, for strain Schu4, a ΔC T > 6 was observed at 24 h for the high (mean ΔC T ± SD, 12.5 ± 4.0) and mid (12.6 ± 0.7) targets and by 48 h for the low target (17.5 ± 6.7) (Fig. 4 b-d, filled triangles). Moreover, although actual starting concentrations between strain IN99 and Schu4 were not statistically different at the low and mid targets (P > 0.3, Table 2), ΔC T(T0-T48) values were statistically different between the two strains ( Fig. 4 c-d, * P < 0.05) collectively indicating that early detection of F. tularensis in well water type matrices may be strain dependent.

Drinking water
In contrast to well water, drinking water microorganisms did not significantly amplify in the BVFH medium (mean log 10 CFU mL − 1 ± SD: 3.3 ± 0.4 at 0 h vs 2.8 ± 0.3at 48 h, P > 0.05 and Fig. 3d vs 3f). In the absence of fast growing HPC organisms, growth of strains LVS, IN99, and Schu4 was observed in the presence of drinking water in a starting concentration dependent manner as expected (Fig. 4e). Moreover, there were no statistical differences between CFU levels in the sterilized drinking water (sDW) controls and the levels of each strain in their respective drinking water samples at each targeted starting concentration and time point (P > 0.05, Fig. 4e). Actual starting densities (mean CFU ± SD) for strain LVS were 1 ± 0, 14 ± 1, and 58 ± 28; for IN99 were 4 ± 0, 40 ± 7, and 329 ± 62; and for Schu4 were < 1 to 1, 8 ± 1, and 43 ± 3 for the low, mid, and high targeted concentrations, respectively ( Table 2). The mean ΔC T(T0-T24) ± SD at the 100, 10, and 1 CFU mL − 1 targeted dose was, respectively, 13.6 ± 0.5, 11.3 ± 0.4, and 11.3 ± 1.1 for strain LVS; 16.2 ± 0.4, 15.7 ± In the absence of fast growing HPC0.9, and 11.4 ± 2.2 for Fig. 4 Detection of F. tularensis in well and drinking water matrices. F. tularensis strain LVS (circles), IN99 (squares), and Schu4 (triangles) were incubated in BVFH containing either well water (a-d) or drinking water (e-h) at targeted starting concentrations of 100 (black symbols), 10 (red symbols), and 1 (blue symbols) CFU mL − 1 . Open symbols represent control wells with sterilized well or drinking water. Filled symbols represent wells with well or drinking water. Growth was monitored by CFU enumeration (a and e) and PCR results are shown in b-d and f-h. ΔC T of 6 is indicated by the bold black line. *, P < 0.05 Schu4 v IN99 well water samples strain Schu4; and 11.3 ± 0.6, 10.2 ± 0.3, and 3.8 ± 3.3 for strain IN99 (Fig. 4 f-h). Thus, the ΔC T ≥ 6 was achieved following 24 h of incubation in the presence of drinking water for each strain at each targeted dose, except for strain IN99 at the 1 CFU mL − 1 target (Fig. 4h, open and filled blue squares, control and drinking water samples, respectively). Because the sDW controls were also below the ΔC T of 6 and lower IN99 CFU levels were observed at the 24 h time point compared to LVS and Schu4 (Fig.  4e, blue squares vs blue circles and triangles, respectively), the ΔC T < 6 was most likely not due to inhibition of IN99 growth in the drinking water matrix. By 48 h, a ΔC T(T0-T48) ± SD of 19.0 ± 4.3 at the low target was observed for strain IN99 (Fig. 4h, filled blue square).

Discussion
In this study, the criteria used for early detection of F. tularensis growth is based on the change in PCR cycle threshold (ΔC T ) which is calculated by subtracting the cycle threshold (C T ) at time 0 (C T0 , i.e. starting DNA levels) from the C T value after incubation of samples in culture medium (C Ti , i.e. starting DNA levels in addition to those resulting from growth of viable cells): ΔC T = C T0 -C Ti . For detection of F. tularensis cells, a ΔC T value ≥6 was chosen to indicate growth of viable cells during the incubation period and was similarly used as a criterion for early detection of Bacillus anthracis and Yersinia pestis growth [12,18]. It should be noted that this ΔC T nomenclature is different from the one commonly used in gene expression studies employing the comparative C T , or 2 -ΔΔCT , method where ΔC T is the difference in threshold cycle between target and reference genes within the same sample [19].
Various F. tularensis PCR assays were tested for use in this method. The fopA gene encodes an outer membrane associated protein, pdpD encodes the pathogenicity determinant protein D, and ISFtu2 encodes a multiple copy, insertion element-like sequence [15,16]. The pdpD gene lies within a putative operon consisting of pdpDiglABCD and encodes a 140.7-kDa membraneassociated protein that is required for virulence but not intracellular growth [20,21]. However, in F. tularensis subsp. holarctica (Type B) strains, most of the pdpD gene is missing or absent and may play a role in the difference in virulence between Type A and B strains [22]. The assay for the pdpD gene is absent from Type B strains and contains a 144 bp insert in F. novicida strains [15] which was verified in this study (Table 1). No statistical differences were observed for C T values collected from the fopA and pdpD assays for Schu4 and between the fopA and ISFtu2 assays for IN99 and LVS samples (P > 0.05, data not shown); thus, for consistency, C T values from the fopA PCR assay were reported and analyzed here for strains Schu4, IN99, and LVS.
To evaluate performance under conditions mimicking post-decontamination scenarios, the culture-based PCR method was tested in the presence of low titers of viable target cells and high levels of inactivated target cells which may be present after remediation efforts. Although different results were obtained for each strain, the overall trend indicated the assay can detect growth of 1 viable cell mL − 1 by 48 h in the presence of up to 10 4 inactivated cells mL − 1 for strain IN99 and Schu4 (Fig. 1 h and l, respectively) and up to 10 6 inactivated cells mL − 1 for strain LVS (Fig. 1d). Moreover, at low inactivated background levels of 10 2 -10 4 cells mL − 1 , a ΔC T ≥ 6 was achieved for all strains after 24-48 h incubation in BVFH. However, for low titers of live target cells in the presence of high levels of inactivated target cells, an incubation period > 48 h may be needed to observe a ΔC T ≥ 6 yet would still be less time required for identification using traditional culture methods.
Humic acids are reported to bind DNA in a sequencespecific manner, thus limiting the amount of template DNA available for the PCR reaction [23]. Due to the difficulty in measuring humic acid concentrations directly in water matrices, their concentrations are usually estimated from total organic carbon (TOC) or dissolved organic carbon (DOC) measurements. An Indiana-based study monitored DOC concentrations over a 22-month period in river water before and after conventional drinking water treatment and found DOC concentrations to range between 2 and 12 μg mL − 1 in river water and 1-4 μg mL − 1 in chloraminated, finished water [24]. Thus, even if humic acids constituted a majority of the DOC, the humic acid concentration evaluated in this study, 50 μg mL − 1 , was between 4 and 50 times higher than levels reported for surface and finished drinking water samples. Thus, the evaluation of this high concentration of humic acid in this study represented a worstcase scenario of the possible concentration and carryover of environmental inhibitors into samples for processing and downstream molecular analyses. Without prior removal of the humic acid from DNA extracted samples, PCR reactions were inhibited (data not shown). Thus, these samples required a clean-up step after DNA isolation, after which PCR results were not statistically different from controls with no humic acid (P > 0.05, Fig.  2 b-d). For this chemical challenge, results from the method indicated the ΔC T ≥ 6 was achieved following 24 h of incubation for LVS and Schu4 at each targeted starting concentration and for strain IN99 at the mid and high targets (Fig. 2 b-d). Collectively, despite the issues with the IN99 inoculum, the method was able to detect early amplification of F. tularensis in the presence of humic acid after 24 h of incubation.
Arizona test dust (ATD) is commonly used as a surrogate for atmospheric particles, soil, and household dust due to its consistent particle size distribution and chemical and biological composition [25][26][27]. Thus, ATD was used as a test matrix for its representation of environmental challenges possibly encountered during environmental sampling and effects on the culture-based PCR method. In the 4 mg mL − 1 ATD concentration used in this study and based on the lowest composition percentages of the metal oxides or ions (i.e. 10% Al 2 O 3 , 2% Fe 2 O 3 , and 0.51% TiO 2 ), the ATD samples contained 210 μg mL − 1 of Al, 56 μg mL − 1 of Fe, and 20 μg mL − 1 of TiO 2 , which is approximately 25-400 times above the PCR inhibitory levels previously reported for each inhibitor [28,29]. Moreover, silicon, the main constituent of ATD (68-76% SiO 2 ) was previously reported to inhibit PCR by interfering with Taq polymerase [30]. Thus, similar to humic acid, ATD samples required a clean-up step for extracted DNA prior to PCR analysis in order to remove the multiple PCR inhibitors contained within ATD (data not shown).
In addition to chemical challenges, previous microbial analysis of ATD indicated a composition of yeast, molds, actinomycetes, Bacillus spp., Micrococcus spp., and Streptomyces spp. with HPC concentrations of 4.6 log 10 CFU 10 mg − 1 on Reasoner's 2A agar (R2A) [31], which was approximately 1 log 10 higher than levels observed in this study (3.6 ± 0.06 log 10 CFU 10 mg − 1 on R2A, Methods Section). In this study, CFU enumeration of LVS was not possible in the presence of fast-growing, competing ATD microbes (Fig. 2e, filled  circles v control, open circles). Moreover, although a mean ΔC T ≥ 6 was observed for LVS at each targeted starting dose, method performance was concentration dependent since mean ΔC T ± SD values at 24 h decreased by 48 h at the low target (Fig. 2h, filled circles 6.3 ± 1.2 to 3.4 ± 3.1, respectively) and only increased slightly from 24 to 48 h at the high target (Fig. 2f, filled circles, 6.1 ± 3.2 to 6.8 ± 3.6, respectively). Nonetheless, the culture-based PCR method was able to detect LVS growth after 24 h of incubation demonstrating the specificity and sensitivity of PCR analysis amidst a high interfering background and negative results from traditional culture methods.
Similarly, for the well water test matrix, F. tularensis CFU densities could not be quantified at 8, 24, and 48 h post incubation due to the rapid growth of well water microbes (Fig. 4a). Acid pretreatment of environmental samples has been shown to reduce interference by indigenous microbes and enhance recovery of F. tularensis from natural water samples [32,33]. However, no significant differences in HPC densities were observed between acid treated well water samples and non-treated controls (mean log 10 CFU mL − 1 ± SD: 0.642 ± 0.11 and 0.635 ± 0.37, respectively on heart infusion agar (HIA); P = 0.6; four replicates each). Although the fast-growing, competing microbes negatively affected growth and enumeration of F. tularensis in the well water samples, the culture-based PCR method detected early growth of the Type A1 Schu4 strain, in contrast to the Type B strains LVS and IN99 (Fig. 4 b-d).
Although the persistence and growth in natural waters for both types have been previously reported [reviewed in [6]], it is unknown whether environmental growth fitness and survival mechanisms differ between Type A and B strains. Infections with Type A1 strains result in higher mortality rates compared to Type A2 and Type B strains [9,34]. Thus, virulence mechanisms present in Schu4, but absent in LVS and IN99, could confer a growth/survival advantage as potentially observed in the well water experiments where only Schu4 growth was detected amongst a high background of competing microbes ( Fig. 4 b-d). For future culture-based PCR method studies, more avirulent and virulent F. tularensis Type A and B strains should be further tested in the well water and ATD matrices to confirm this interesting observation of potentially strain dependent fitness to compete with fast-growing microbes.
The occurrence of coliform bacteria in chlorinated systems was reported to be significantly higher at temperatures above 15°C than at 0-15°C [35]. Thus, to assess the performance of the culture-based PCR assay for F. tularensis in the presence of drinking water microorganisms, warm tap water (36.9 ± 0.6°C, Table 3) was concentrated 100-fold to collect a high titer of indigenous drinking water microbes. Their presence in the BVFH medium neither inhibited nor outcompeted F. tularensis growth (Fig. 4e) and did not negatively affect the early detection of F. tularensis using the culture-based PCR assay (Fig. 4 f-h).

Conclusions
In this study, a culture-based PCR method was able to detect early growth of low levels of F. tularensis cells amongst high levels of their respective inactivated target cells and various environmental interferences and inhibitors. Results also indicated that indigenous drinking water organisms were not capable of growth in nutrient rich medium (e.g. BVFH), unlike those in ATD and well water. Thus, this method could also be used to detect early growth of other water-based pathogens, such as Legionella pneumophila, where drinking water is an important source of human exposure [36]. The culturebased PCR method evaluated in this study can be used as an alternative to traditional culture methods for detection and identification of fastidious pathogens by combining the rapidity, sensitivity, and specificity of PCR with culturability as an indicator of microbial viability. Table 1 lists the bacterial strains used in this study. Stock and stationary phase cultures of Francisella spp. were prepared as previously described by Morris et al. [13]. For the test conditions described below, F. tularensis strains LVS, IN99, and Schu4 were grown in a previously developed enhanced growth medium, BVFH, for Francisella spp., which consisted of brain heart infusion broth (Becton Dickinson [BD] Biosciences, USA) supplemented with 2% Vitox™ (Oxoid Hampshire UK), 10% Fildes (Remel, USA), and 1% L -histidine (Fisher Scientific, USA) [13]. Experiments with virulent F. tularensis strains were conducted under biosafety level 3 (BSL-3) conditions at the University of Cincinnati College of Medicine with protocols approved by the university's Institutional Biosafety Committee and the Select Agent Program.

Colony forming units (CFU) enumeration
To determine Francisella spp. densities, an aliquot of the bacterial suspension was serially diluted in Butterfield's buffer (BB; 42.5 mg monopotassium phosphate L − 1 ; Hardy Diagnostics, USA) and plated on chocolate agar plates (BD Biosciences, USA). Plates were incubated for 48 h [for spot plate method [37,38]] or 4-7 d [for spread plate method [39]] at 37°C. The spot plate method was used for evaluation of high CFU titers at later time points. The limit of detection (LOD) was 1 and 1.95 log 10 CFU mL − 1 , for the spread and spot plate method, respectively.
Heterotrophic plate count (HPC) bacteria were enumerated by the spread plate method on heart infusion agar (HIA, Difco Laboratories) following incubation at 35°C for 7 d or Reasoner's 2A agar (R2A, Difco Laboratories) following incubation at 28°C for 7 d. The LOD was 1 log 10 CFU mL − 1 . For untreated well water samples only, total coliforms and Escherichia coli were measured using Colilert® (Idexx Laboratories, USA) following manufacturer's protocols.

Preparation of F. tularensis cell suspensions
Bacterial cultures were grown in triplicate for each strain to stationary phase, washed by centrifugation (3000 relative centrifugal force [RCF] for 10 min at 4°C), and stored in BB for 24 h at 4°C. These resting cells were used as the inoculum to more closely mimic the physiological state of planktonic cells in the natural aquatic environment [40]. Suspensions were enumerated for CFU and the concentrations before and after incubation at 4°C were not statistically different (P > 0.05, STable). Resting cells cultures were diluted in BVFH to yield approximate suspensions of 15, 150, and 1500 CFU mL − 1 and for use in experiments, these suspensions were diluted 15-fold in the various matrices to achieve targeted starting concentrations of 1, 10, and 100 CFU mL − 1 , respectively. Actual starting concentrations are listed in Table 2 and were enumerated by spread plating 200 μL onto five separate chocolate agar plates (BD Biosciences, USA) resulting in a LOD of 1 CFU mL − 1 .

Collection of water samples
Potable municipal drinking water samples were collected from laboratory taps and were derived from river water treated by coagulation/flocculation/sedimentation, rapid sand filtration, granular activated carbon filtration, and chlorination. Ten liters were concentrated by filtration through a 0.45 μm polycarbonate filter (GE Osmonics, USA) and resulted in a 100 mL volume of 100-fold concentrated drinking water. A 1 mL aliquot of a 10% (w/v) sodium thiosulfate (Sigma-Aldrich, USA), prepared using distilled water, was added to the concentrate to neutralize disinfectant residuals.
Well water samples were collected on a private farm from a well supplying nonpotable ground water from a deep aquifer. The well water pump was operated for a few minutes before two 1 L samples were collected in sterile collection bottles. Water temperature was recorded right after sampling with a sterilized thermometer. Samples were kept at 4°C for up to 4 h before transport to the laboratory. A 10 mL portion of the drinking and well water was filtered through a 0.1 μm Supor® (hydrophilic polyethersulfone) membrane Acro-disc® syringe filter (Pall Corporation, USA) resulting in sterilized drinking water (sDW) and well water (sWW) for use as controls.

Preparation of inactivated F. tularensis cell suspensions
For each F. tularensis strain, static cultures were pelleted and washed twice by centrifugation at 3000 rcf for 10 min at 21°C with 25 mL of BB. Aliquots of the suspensions were incubated in either BB (for controls) or 70% isopropanol (for inactivation) at 21°C for 2 h. Suspensions were then washed twice by centrifugation and resuspended with 20 mL of BB. Three control and isopropanol treated replicates were performed for each strain. To confirm inactivation, suspensions were enumerated for CFU by spread plating 200 μL onto five separate chocolate agar plates (BD Biosciences, USA) resulting in a LOD of 1 CFU mL − 1 . For control suspensions, the mean log 10 CFU mL − 1 ± SD concentrations for LVS, IN99, and Schu4 were 6.9 ± 0.1, 7.0 ± 0.0, and 6.7 ± 0.1, respectively. CFU concentrations for all inactivated suspensions were < LOD. For use in experiments, the inactivated cell suspensions were diluted (based on the concentrations of the respective strain's control suspensions) and added to samples to yield a final concentration of 10 2 , 10 4 or 10 6 inactivated cells mL − 1 .

Water quality analysis
Drinking and well water samples were analyzed for free and total chlorine (Cl 2 ), pH, temperature, conductivity, percent dissolved oxygen (DO), hardness, and turbidity (Table 3) [41]. Measurements were made using DPD colorimetric analysis for free and total chlorine; a glass electrode 440 pH meter (Corning® Electrochemistry Products) for pH; a YSI 566 Multi Probe System (YSI Environmental) for temperature, conductivity, and DO; ethylenediaminetetraacetic acid titration for hardness (Hach Company); and a portable 2100 Turbidimeter (Hach, USA) for turbidity. Additionally, water samples were submitted to the National Risk Management Research Laboratory at the US EPA in Cincinnati, OH to assay total organic carbon (TOC) (EPA Method 415.3 rev 1.1); trace metals (EPA Method 200.7); phosphate (PO 4 ) (EPA Method 365.1); and nitrate (NO 3 ) (EPA Method 353.2) ( Table 3).

Isolation and preparation of total DNA
DNA was extracted from bacterial cells using the Mas-terPure™ Complete DNA purification kit (Epicentre Biotechnologies Inc., USA) according to manufacturer's protocol. The DNA pellet was resuspended in 50 μL of molecular grade water. To remove PCR inhibitors from samples containing humic acid and ATD, an additional purification step was performed by using the OneStep™ PCR Inhibitor Removal Kit (Zymo Research, USA) following manufacturer's protocols. Final DNA concentrations were measured with a Nanodrop ND-1000 Spectrophotometer (Thermo Scientific, USA) and analyzed as described below.

Incubation of F. tularensis in the presence of each test matrix
Ninety-six deep well polypropylene plates (Fisher Scientific, USA), with a working volume of 1.8 mL per well, were pre-loaded with a sterile 4 mm glass bead per well (to aid in aeration of the samples). In each well, 0.1 mL of the F. tularensis suspensions and 0.1 mL of suspensions containing either inactivated target cells, drinking water, well water, humic acid, or ATD (except for control wells), was added to 1.3 mL of BVFH in triplicate. For control wells, 0.1 mL of BB, sDW, sWW, or Milli-Q®, for the humic acid and ATD experiments, was used. Duplicate wells without bacteria (0.1 mL of BB) for each test matrix were also included. The final volume in each well was 1.5 mL where both bacterial cells and test matrix or controls were diluted 15-fold.
For each experiment, four 96-well plates were generated for each time point (0, 8, 24, and 48 h) to minimize the loss of volume due to evaporation and sampling for culture and molecular analyses. After samples were added, plates were covered with an air-permeable Aera-Seal™ (Excel Scientific Inc., USA) and incubated at 37°C with shaking (Innova® 42 shaker/incubator, New Brunswick Scientific, USA). After each time point, 1 mL from each well was transferred to 1.7 mL tubes, processed for DNA isolation, and analyzed by real-time PCR. The remaining volume (approximately 400 μL) in the well was used for CFU enumeration.