Plaque assay (PA).
PA studies were performed within a seven-day time period. The CVS ranged from 0.11 to 0.24, 0.06 to 0.26 and 0.06 to 0.22 for ECML-117, List-36 and YpP-G, respectively. The mean CVS were tightly grouped: 0.15 for ECML-117, 0.14 for List-36 and 0.13 for YpP-G. The overall range of the CVS was 0.06 to 0.26, with a mean value of 0.15 (). The CVS may be used to estimate the highest and lowest values in a 90% CI for any given sample (using the Z value for 90% confidence), by the following formula (derived from ref. 14
): VH = V × (1 + CVS × Z90)
and VL = V × (1 - CVS × Z90),
where V = sample value, VH
= highest value for 90% confidence, VL
= lowest value for 90% confidence and Z90
= Z value for 90% confidence (Z = 1.6445).15
For example, a phage preparation with a PA-determined titer of ca. 1 × 108
PFU/mL and a CVS of ca. 0.15, would have a low interval of ca. 7.5 × 107
PFU/mL and a high interval of ca. 1.3 × 108
PFU/mL, with 90% confidence. To put this into further perspective, phage specimens whose titers (determined by the same investigator in the same laboratory) are within that range may be considered to be identical to a titer of ca. 1 × 108
CVS for the PA, NS-based and QPC R-based methods
The CVS was reduced as the phage concentration increased: they averaged 0.18, 0.14 and 0.12 for phage preparations containing 108, 109 and 1010 PFU/mL, respectively. The data presented in provide insight into the expected variations and the possible range of titration errors when evaluating PA-determined phage titers obtained by different investigators in the same laboratory.
The reproducibility/precision of the PA-based method among various laboratories was poor. For example, only one dataset (for ECML-117 at an expected concentration of 1010 PFU/mL) was not significantly different (p > 0.05) when the phages were assayed in two different participating laboratories (Intralytix in Maryland and EPI UF in Florida). However, all other matched means were significantly different from one another (p < 0.05). Thus, in order to determine the range of possible variance in phage titers determined by PA, 10 Listeria phages (including List-36) were analyzed by the PA assay. The purpose of the analyses was to determine the variance between paired mean PA values obtained on separate days by different investigators during a time period when the phage preparations should be completely stable; i.e., when changes in their titers should not occur. We chose to compare the mean values because they represent the working values upon which decisions are made, rather than the individual triplicate values obtained by each titration. Twenty-nine paired titrations were performed, within seven days of one another, by at least two different investigators quantitating the 10 different Listeria phages. Each titration value was subjected to a log10 transformation and the calculated difference values between the titrations yielded 29 different values, one for each pair of titrations. Paired two tailed t test was used to calculate the means, standard errors of the means and 95% CI (). We found that the (1) mean difference between titrations was −0.1238 log, (2) standard error of the mean difference was 0.1008 log, (3) upper 95% CI of the mean difference between titrations was 0.3303 log and (4) lower 95% CI of the mean difference between titrations was 0.0827 log. Our data suggest that titrations whose log10 transformation values are within 0.33 log of one another may be considered equivalent since the differences are within the 95% CI of the mean differences between the titrations. Furthermore, if the absolute values of the differences between the titrations were used to calculate the mean differences, the range was even larger at ca. 0.5 log. This analysis took both inter-operator and day-to-day variations into account. To put the above numbers into practical perspective: if investigators in one laboratory determine the mean titer is 1 × 108 PFU/mL, investigators in other laboratories may expect the same phage preparation's titer will range from 4.7 × 107 to 2.1 × 108 PFU/mL (based on mean difference of 0.33 log), assuming that all participants use the same host strain and standardized PA for their analyses.
Range of possible variance in phage titers determined by the PA. The diagonal line represents perfect agreement. The numeric values before and after the semi-colons were obtained by two different investigators.
The possible effect of the agar overlay's volume on phage titer enumeration was examined during additional titration studies performed with each phage preparation (ca. 1010 PFU/mL), using agar overlay volumes of 2, 3 and 4 mL. The values were not significantly different from one another and from the results obtained with the initial 1 mL agar overlay (data not shown). Thus, within those parameters, the volume of the agar overlay does not significantly influence the results.
The CVS for the QPCR-based analyses ranged from 0.08 to 0.09, 0.11 to 0.20 and 0.09 to 0.15 for ECML-117, List-36 and YpP-G, respectively (). The mean CVS were 0.09, 0.17 and 0.13 for ECML-117, List-36 and YpP-G, respectively (the differences were not statistically significant [p > 0.05]). The overall range of the CVS for the phage titers determined by the QPCR-based method was 0.08 to 0.20, with a mean value of 0.13 (). The CVS and the above-described formula may be used to calculate the range of titration values obtained by the QPCR-based method. For example, a phage preparation with a QPCR method-determined titer of ca. 1 × 108 PFU/mL and a CVS of ca. 0.13 (), would have a low interval of ca. 7.9 × 107 PFU/mL and a high interval of ca. 1.2 × 108 PFU/mL, with 90% confidence.
The QPCR-based method, which had the lowest CVS, was the most precise of the three assays we examined. However, since that assay involves amplifying phage DNA, we were concerned that it may overestimate the concentration of viable phage particles by detecting free phage DNA and/or DNA inside damaged phage capsids rather than the DNA in viable phage particles. Although, we did observe that the phage concentrations estimated by the QPCR method were higher than those obtained by PA, we found that was not the case with all phages. Thus, the correlation between the results obtained by QPCR (which measures the DNA of both viable and nonviable phages) and PA (which only measure viable phage particles) must be established for each phage before using the QPCR-based assay to quantitate the number of its viable particles in a specific preparation. Moreover, the correlation may need to be updated every time the production, purification and general handling practices that might affect phage viability are changed.
Some shortcomings of the QPCR-based assay include its: (1) requirement for expensive equipment (although real-time PCR machines are becoming increasingly common in many laboratories), (2) requirement for optimizing PCR amplification conditions and ongoing calibration with known standards analyzed concurrently with the test samples and (3) need for phage-specific oligonucleotide primers. The latter necessitates obtaining at least partial phage sequences and poorly designed primers may form self-dimers and decrease PCR efficiency.11
The advantages of the QPCR-based assay include its (1) good precision/reproducibility which correlates well with that of the PA, (2) its requirement for only a few reagents, (3) high throughput (many phage preparations may be analyzed concurrently when the assay is performed in 96-well microtiter plates), (4) fairly fast turn-around time of 3 to 4 h and (5) potential for greater precision than PA performed by investigators in different laboratories ().
Key differences and pros and cons of the PA, QPC R-based and NS-based methods
The CVS of the NS-based assays ranged from 0.22 to 0.30, 0.27 to 0.86 and 0.28 to 0.35 for ECML-117, List-36 and YpP-G, respectively (). The mean CVS values were 0.27 for ECML-117 and List-36, and 0.31 for YpP-G. One specimen of List-36 had an abnormally high CVS of 0.86 at the expected concentration of 109 PFU/mL. After that one sample was rejected using Dixon's Q test with 99% confidence (Q = 0.80), the overall range of the CVS values were calculated to be 0.22 to 0.35, with a mean CVS of 0.28 for the NS method (). Thus, the CVS values and the above-described formula may be used to calculate the range of titration values obtained by the NS method. For example, a NS-obtained value of ca. 1 × 108 PFU/mL and an overall CVS of 0.28 () would have a low interval of ca. 5.4 × 107 PFU/mL and a high interval of ca. 1.5 × 108 PFU/mL, with 90% confidence. In other words, NS-obtained values in the range of 5.4 × 107 to 1.5 × 108 PFU/mL may be considered equivalent to 1 × 108 PFU/mL.
The NS method was found to be affected by background particle distributions, and many commercially available media for propagating bacteria (including BHI broth and LB broth) elicited high levels of “background noise,” which made it very difficult (if not impossible) to interpret the results. Diluting phage specimens to a concentration of ca. 1 × 107 PFU/mL, using 0.9% saline or deionized water, did not entirely alleviate the problem. Diluting further to reduce the background noise was not feasible because the optimal reading of the NS system is in the range of 107 to 109 PFU/mL.
The NS- and QPCR-based assays measure both viable and nonviable phage particles simultaneously. Thus, NS and PA data must be obtained for each phage and its production and storage protocols before using the NS-based method to determine viable concentrations of that phage. Some drawbacks of the NS method include: (1) it requires fairly expensive equipment (although prices are likely to come down in the future), (2) it only reliably detects phages with a size larger than 40 to 50 nm (which is a minor concern since most phages used for food safety, environmental decontamination and clinical applications are larger than that), (3) the phages need to be suspended in a clear medium (which is a significant limiting factor because it negates one of the most attractive potential applications of the technology: using it to optimize phage production) and (4) in order for the results to be reliable, the phage concentrations must be within a range of 107 to 109 PFU/mL. However, NS provides results within an impressive ≤5 min timeframe, which is significantly faster than does PA and the QPCR method (18 to 24 h and 3 to 4 h, respectively), and its performance does not require any additional reagents. Once optimized, it is likely that the NS-based method will be reproducible among various laboratories, with accuracy comparable to PA performed by various investigators (but significantly faster).
Comparison of phage titers determined by the PA and by the QPCR- and NS-based assays.
At the present time, the PA is considered to be the gold standard for determining phage titers and it is the only method among the three approaches we examined that directly enumerates viable phage particles (QPCR and NS methods quantitate both viable and nonviable phages). Therefore, as previously mentioned, correlations must be established between the data obtained by the PA and the other two assays before using only the QPCR and/or NS methods to determine viable phage concentrations. Furthermore, even after the correlations are established, it may be prudent to continue to use PA to verify the outcomes of the QPCR- and NS-based analyses, at least for a period of time until the “correction factor” and reproducibility of testing is established with absolute certainty for the bacteriophage under investigation. The data presented below and in provide preliminary insight into the correlation among the phage titers determined by the three assays.
Figure 2 Correlation between the results obtained by the PA and by the NS- and QPCR-based assays for determining phage titers. The symbols represent matched-pair, ten-point data set means plotted on a log scale. The diagonal lines represent perfect agreement. (more ...)
In all instances, R2 values were calculated with n = 3. In general, the QPCR-based assay under-reported the number of viable phage particles determined by the PA for List-36, over-reported the number for ECML-117 and yielded mixed results for YpP-G (). Among the three phages examined, the strongest correlation observed between the QPCR- and PA-determined means of the titers was R2 = 1 for YpP-G with an average difference of 50%, followed by R2 = 0.99 for List-36 with an average difference of 76% and R2 = 0.98 for ECML-117 with an average difference of 38%. Thus, the values determined by the QPCR method were ca. 0.5-fold different (higher or lower), 0.76-fold lower and 3.9-fold higher for YpP-G, List-36 and ECML-117, respectively, than those determined by the PA. Although the percent- and/or fold-differences (i.e., the “correction factor”) seem high, the R2 values determined for all three phage preparations were excellent (ranging from a very high value of 0.98 for ECML-117, to a perfect match of 1 for YpP-G), which suggests that the results obtained by the QPCR and PA methods correlate very well with one another. In other words the viable phage titers obtained by the QPCR-based assay are essentially identical to those obtained by the PA, for the same sample, when an appropriate adjustment is made with the above-determined “correction factor.” We expected (as mentioned above) that the QPCR-based assay would over-report the data obtained by the PA, which turned out to be the case for many, but not for all, of the specimens examined. Also, half of the readings obtained for YpP-G were over-reported and half-were under-reported. At the present time, it is unclear why some readings obtained by the QPCR method were under-reported, but possible contamination of the phage specimens or PCR reaction mixture, which interfered with PCR-mediated amplification, may be one explanation.
The NS-based assay over-reported the number of viable phage particles determined by the PA, except when List-36 was enumerated at a concentration of 1010 PFU/mL, which was slightly under-reported (). Among the three phages examined, the strongest correlation between the NS- and PA-determined means of the titers was R2 = 1 for YpP-G with an average difference of 544%, followed by R2 = 0.98 for List-36 with an average difference of 52% and R2 = 0.98 for ECML-117 with an average difference of 461% (). In other words, the phage titers determined by the NS method were ca. 0.5-fold higher, 4.6-fold higher and 5.4-fold higher for List-36, ECML-117 and YpP-G, respectively, than they were when the phage preparations were quantitated by the PA. A possible explanation for these differences is that NS enumerated viable phage particles, non-viable phage particles and background noise where as PA only enumerated viable phage particles. The correlation between the phage titers determined by the NS method and PA was very good, with an R2 ranging from a very high value of 0.98 for List-36 and ECML-117, to a perfect R2 of 1 for YpP-G. Thus, as with the QPCR-determined method, we expect that the titers obtained by the NS-based assay adjusted by the above-determined average “correction factor” will very accurately predict PA-determined titers.
The NS method over-reported the number of viable phage particles determined by the QPCR-based assay, except when ECML-117 was enumerated at a concentration of ca. 108 PFU/mL. Among the three phages examined, the strongest correlation between the results obtained with the NS and QPCR methods was R2 = 1 for ECML-117 with an average difference of 27%, followed by R2 = 0.99 for List-36 (with an average difference of 452%) and YpP-G (with an average difference of 568%) ().