Initial inhibition tests (using undiluted extracts and no PCR facilitators) showed complete inhibition in all soil, feces, and hair extracts, with only bone extracts permitting any amplification of the IPC (Supplementary Table 2). Both bone extracts (B1, B2) had good amplification efficiency (E = 94%), but with ΔCq values of 2.4 or 6.1 cycles (Table 2); thus, even the least inhibited extract limited our expected recovery to <20% under these conditions.
As indicated in Table 2, the use of additional Taq improved the ΔCq values (increasing the ER), but led to a decrease in amplification efficiency for the bone extracts. The cause of this decrease is unknown; however, adding BSA improved the ΔCq values comparably and with optimal efficiency. Incorporating both facilitators brought the ΔCq values to <1 (62 and 69% ER), such that no dilution would be beneficial in subsequent PCR runs with these extracts.
For the feces (F1, F2) and hair (H1, H2) extracts, both facilitators and dilution were required to reduce inhibition sufficiently, although BSA alone was responsible for the majority of the improvement for extracts F1, F2, and H1. Even at very high dilutions, additional Taq was unable to facilitate amplification in the feces extracts, but was particularly effective against inhibitors in H2. Unfortunately, the dilutions necessary to overcome inhibition in these extracts drop the maximum ER values below 2%, which may render the optimization process trivial for highly degraded and other low-copy templates.
In terms of reaction success versus complete inhibition among the soil extracts, Taq had no significant effect as a facilitator in contrast to BSA (P = 3 × 10-15). Not surprisingly, BSA had a less noticeable effect on reaction success when the extracts were diluted (P = 3 × 10-3 and P = 5 × 10-1 with 1/10 and 1/50 dilutions, respectively). Optimal ER values (from 20 to 77%) were achieved with BSA alone in ~31% of the extracts (Figure 1, solid black triangles), while the other 69% benefited from additional Taq as well (Figure 1, solid black circles). A subset of those more inhibited extracts showed optimal ER values using a 1/2 dilution in PCR (Figure 1, solid blue circles). Although the amplification efficiencies were improved by diluting the soil extracts further (Supplementary Table 2), conditions providing higher ER values were more relevant for subsequent PCR runs.
A comparison of inhibition levels across soil DNA extraction methods is shown in Table 3. In all cases, the method of Willerslev et al. (27) demonstrated the lowest inhibition levels. Although we expect some variation in inhibition levels between these extracts due to the heterogeneity of soil, comparable tests of duplicate extracts (i.e., biological replicates) of sample P15 (Supplementary Table 2) indicate that this variation is minimal, with SDΔCq = 0.3 cycles and SDE = 1.6%.
From the data in Figure 2, there is an apparent correlation between E and ΔCq, which is expected given the threshold-based method of calculating Cq; as the amplification efficiency decreases, the placement of the fluorescence threshold becomes more sensitive, and the resulting ΔCq values more variable. Thus, many of these poor efficiency reactions have deceptively low ΔCq values, even though they are highly inhibited.
Interestingly, many reactions display a high ΔCq value without a corresponding drop in reaction efficiency (Figure 2). This effect of inhibition is apparently overcome by the addition of BSA, as it is only visible in reactions lacking BSA, which required considerable dilution of the extracts to facilitate any amplification. Examples of these atypical efficiency-Cq relationships are shown in Figure 3.
By incorporating both ΔCq and efficiency measurements in our experiments, and by analyzing their results through pairwise comparisons, we have detected inhibition with various responses to PCR facilitators. We have typically observed Cq shifts in response to decreased amplification efficiency, as expected using threshold-based quantification. In some instances, however, a low ΔCq and high ER can be associated with poor efficiency due to placement of the threshold. In these cases, and in any reaction where the efficiency differs from 100% (relative to a standard reaction), the ER value should only be considered an upper limit, but whose accuracy becomes greater as E approaches 100%.