PCR has been suggested for diagnosing bacteremia, particularly in the presence of antibiotic treatment (
2,
10,
12,
14,
22). A broad spectrum of pathogens can be detected by using universal primers derived from the conserved DNA sequences of the eubacterial 16S rRNA gene (
3,
9,
20,
28,
29). However, the validity of universal PCRs may be compromised by a high rate of false-positive results (
8,
15), since, in contrast to specific PCRs, the DNA of any bacterial species may cause contamination (
6,
8). Besides UV irradiation (
7,
8,
21,
25,
27) or restriction endonuclease digestion (
6,
8), reagent pretreatment with DNase I is a promising method in overcoming this problem (
8,
11,
15,
24). However, the dosages of DNase I required to eliminate false-positive results reliably are still unclear. Furthermore, the effect of various DNase I dosages on the sensitivity of universal PCRs has not been addressed systematically.
The primers and the Taq polymerase determine the reaction conditions and the efficacy of a PCR. Thus, we hypothesized that they also influence the DNase I dosages that are required to avoid false-positive results in universal PCR assays. We also hypothesized that the adverse effects of DNase I on amplification might be dose dependent. Therefore, we investigated two different Taq polymerases and two different primer pairs as determinants for the necessary DNase I dosages in universal PCRs.
First, the rate of false-positive PCR results was established by using untreated reagents. Then, the dosage of DNase I required to completely eliminate false-positive results (threshold dosage) was assessed. These assessments were performed by using a basic PCR assay (I) that was subsequently varied in its primers (II) and Taq polymerase (III), respectively.
The master mix of PCR (I) contained 10 mM Tris-HCl, 50 mM KCl, 2.5 mM MgCl
2, 200 μM concentrations of each deoxynucleoside triphosphate (Amersham Pharmacia Biotech, Freiburg, Germany), 0.4 μM concentrations each of primers UP1 (5′-CCA GCA GCC GCG GTA ATA CG-3′, corresponding to nucleotides 517 through 537 of the
Escherichia coli 16S rRNA gene) and UP2 (5′-ATC GG(C/T) TAC CTT GTT ACG ACT TC-3′, corresponding to nucleotides 1513 through 1491 of the same gene) (
17), and 1 U of
Taq DNA polymerase (Amersham Pharmacia Biotech). The PCR included an initial denaturation step at 94°C for 10 min; 35 cycles at 94°C for 1 min 30 s, 55°C for 1 min 30 s, and 72°C for 2 min; and a final elongation phase at 72°C for 5 min. After gel electrophoresis and ethidium bromide staining, the amplification product was visualized under UV light. To rule out exogenous DNA contamination (
18,
19), each PCR step was carried out in a separate room by a separate operator wearing protective clothes. The procedures were performed under laminar flow hoods by using dedicated pipetting devices and filter-sealed tips. Four dilutions from 10
0 to 10
−3 pg of
E. coli DNA served as template. For positive controls, 500 pg of
E. coli DNA was used. For negative controls, 5 μl of sterile water was added instead of template. When a negative control delivered a visible amplification product on the agarose gel, the result was classified as a false positive. When no band was obtained from the positive control, the result was classified as a false negative. Only those PCR runs delivering correct results for both controls were considered valid.
The DNase I pretreatment included the incubation of Taq DNA polymerase, MgCl2, 10× buffer, and deoxynucleoside triphosphates with the DNase I (Hoffmann-La Roche Diagnostics, Mannheim, Germany) for 30 min at 37°C, followed by heating at 95°C for 50 min. Primers and template or water for the negative control were added subsequently. Twenty PCR runs each were carried out using 0, 3, and 25 IU of DNase I per master mix.
When the experiments were repeated with other primers (II) (L1 [5′-CAG CAG CCG CGG TAA TAC-3′, corresponding to nucleotides 518 through 536 of the
E. coli 16S rRNA gene] and L2 [5′-CCG TCA ATT CCT TTG AGT TT-3′, corresponding to nucleotides 928 through 908 of the same gene]) (
5), 0, 3, 25, and 70 IU of DNase were tested in 20 PCR runs each. When AmpliTaq Gold polymerase (Applied Biosystems, Weiterstadt, Germany) (III) was used with primers L1 and L2, the tested DNase I dosages were 0, 0.1, 0.3, 0.5, and 3 IU per master mix. Again, with each of these DNase dosages, 20 PCR runs were carried out. Fisher's exact test was used to establish the statistical significance of the increase in valid PCR results by DNase pretreatment; the level of significance was set to 0.05 (
1).
The effect of DNase on the PCR sensitivity was examined quantitatively. The detection limit of each assay was derived from the lowest DNA concentration producing a visible PCR product in valid PCR runs without DNase pretreatment. To show the specificity of these amplification products for E. coli DNA, we applied automated sequencing with the ABI Prism 310 Genetic Analyzer (Applied Biosystems) and sequence analysis by using the BLAST program. Subsequently, we assessed the proportion of valid PCR runs (see above) that still became positive at the detection limit, when DNase I was applied.
The rates of false-positive, false-negative, and valid universal PCR results depending on the applied DNase I dosage are given in Fig.
1. Without DNase I, high rates of 85% (assay I), 65% (II), and 70% (III) false-positive findings occurred. Correspondingly, only 15% (assay I) or 30% (II and III) of the runs became valid (Fig.
1). These data substantiate the statement of Corless et al. (
8) that a step to effectively eliminate contaminating bacterial DNA has to be implemented before universal PCRs can be applied for diagnostic purposes. Indeed, by pretreatment of the reagents with dosages of DNase I increasing stepwise, the elimination of false-positive universal PCR results was finally achieved for each assay (Fig.
1). This treatment did not lead to substantial increases in false-negative results. Thus, DNase I application significantly improved the rate of valid PCR results to 70% (Fig.
1A), 85% (Fig.
1B), and 100% (Fig.
1C).
The DNase I dosages needed to improve PCR results varied considerably between the assays (Fig.
1). In assays I, II, and III, the threshold dosages of DNase I necessary to eliminate false-positive results completely were 25, 70, and 0.1 IU per master mix, respectively. The differences in the required DNase I dosages in assays I and II are attributed to the use of different primers, since all other reagents and the handling of the reagents were kept constant. The influence of the primers on the DNase I requirement can be explained by the fact that the change in the primers resulted in different detection limits of the PCR (1 pg of
E. coli DNA in assay I and 10 fg in assay II). Therefore, the measures to reduce the amount of contaminating DNA below this detection limit had to be more powerful in assay II than those in assay I (Fig.
1A and B). Similarly, the difference between DNase I dosages in assays II and III can be attributed to different
Taq polymerases. These enzymes have been shown to contain bacterial DNA (
4,
8,
15,
16,
21,
23,
24,
26). Thus, the influence of the polymerase enzyme on DNase I requirement is most likely due to varying levels of DNA contamination in different
Taq polymerases, as suggested by Hilali et al. (
15).
The 700-fold difference between the required DNase I dosages, which depended on the applied
Taq polymerase, also implies that the contamination found in the
Taq polymerase might be the main reason for the false-positive results of universal PCRs. By comparison, the importance of contamination from exogenous sources might be marginal. This notion is corroborated by the rigorous hygienic barrier precautions applied in this study (
18,
19). Furthermore, the rate of false-positive results in a specific PCR assay for the detection of
E. coli in our laboratory was <1% (
14), indicating high standards for handling PCR technology.
Although the discussion above has revealed the beneficial aspects of the use of DNase I in universal PCRs, a negative aspect also has to be mentioned: increasing amounts of DNase I hamper the sensitivity of the PCRs (Fig.
2). The detection limit of assay I was 1 pg of
E. coli DNA; this amount produced a visible amplification product in each of the valid runs (100%) without the addition of DNase (Fig.
2A). However, when the master mix was pretreated with 25 IU of DNase I, the same amount of template gave a positive result in only 2 of 14 (14%) valid runs. Similarly, sensitivities in assays II and III (Fig.
2B and C, respectively) were reduced with increasing amounts of DNase I. This dose-dependent effect of DNase I on the sensitivity of the PCRs most likely appears due to inhibition of the amplification by the DNase I itself, as mentioned by Corless et al. (
8). However, the persisting or recurrent enzymatic activity of DNase I (
13) might be excluded, since we applied an extended inactivation step at a high temperature (95°C for 50 min).
Our results suggest that the lowest possible amount of DNase I should be used in limiting false-positive universal PCR results. This is further illustrated in comparing assays II and III. Due to the rather small threshold dosage (see above) of DNase I in assay III (0.1 IU) versus assay II (70 IU) (Fig.
1B and C), the sensitivity of assay III was much better preserved: at the detection limit of 10 fg of
E. coli DNA, 80% of the runs in assay III were positive versus 30% of the runs in assay II at the same detection limit (Fig.
2B and C). The importance and necessity of titrating DNase I for universal PCRs is further shown in Fig.
1C: if amounts higher than 0.1 IU are used, the rate of false-negative results sharply increases up to 100%.
In conclusion, false-positive findings for different universal PCR assays are reliably eliminated by DNase I pretreatment. Careful titration of DNase dosages, depending on the primers and the Taq polymerases, is needed to achieve the optimal validity and sensitivity of different assays.
Acknowledgments
We are indebted to C. Görke, P. Krüger, and C. Wolz for discussions; C. Meisner for assistance in the statistical evaluation of the data; and D. Blaurock for language corrections in the manuscript.