The NZFACE is situated on a mixed plant species sheep-grazed pasture near

Bulls (40˚14'S, 173˚16'E) in the North Island of New Zealand. The average annual temperature at the site is 12.9˚C and average total annual rainfall 870 mm. The experiment consists of six rings of 12 m in diameter with three of these left at ambient CO₂ (aCO₂) and three enriched with CO₂ during the photoperiod, with enrichment continuous throughout the year. From October 1997 until December 2011 the enrichment was to 475 ppm. The CO₂ enrichment was suspended in December 2011 until July 2013 to conduct technical maintenance and then resumed with a target concentration of 500 ppm. From 1997 to 2011 the rings were grazed periodically by adult sheep under a protocol described in Newton et al (2014) [28] ; thereafter (from 2013 onwards) the rings were cut with a mower and the clippings removed from the site. In October 2012 an irrigation system was installed and all the rings were irrigated to maintain pasture growth during periods of water deficit (usually December-February).

The pasture is a permanent grassland―uncultivated for at least 50 years― containing a mixture of C₃ and C₄ grasses, forbs and legumes [29] [28] . C₃ grasses are the most abundant group with Agrostis capillaris (browntop), a grass typically found in low-fertility situations [30] , one of the dominant species under both ambient CO₂ (aCO₂) and eCO₂ [31] .

The soil at the site is a black loamy fine-sand (a Mollic Psammaquent). During the course of the experiment fertiliser was added―superphosphate, potassium and magnesium sulphate―to maintain soil concentrations at recommended levels [32] (Cornforth and Sinclair 1984). No nitrogen (N) fertiliser was added, the N inputs coming solely from biological nitrogen fixation by the legume species. Details of soil properties and changes over time can be found in [2] .

In January 2014 soil samples were collected for this experiment. Patches of browntop were identified and soil was sampled from these sites. Soil cores were taken to a depth of 75 mm using a 25 mm diameter steel soil corer. There were five browntop patches in each ring and a core was taken from each patch. Three experimental protocols were then used: first (Figure 1(A)) DNA was extracted

from each of the five cores sampled per ring and qPCR completed on all the samples; second (Figure 1(B)), the five soil cores from each ring were bulked and DNA extracted from this single sample per ring; third (Figure 1(C)) as for the first protocol (A) but the DNA was bulked after extraction giving a single sample per ring for qPCR. Note that in each case n = 3 for the CO₂ treatment but there were five qPCR samples per ring for method A) and one qPCR sample for (B) and (C). Importantly, the three approaches used the same soil cores thus allowing a direct comparison of the protocols and extraction kits. The soil samples were stored at −20˚C until DNA extraction. A sub-sample of soil was used to measure soil moisture by oven-drying at 105˚C for 24 h.

Two DNA extraction kits were evaluated in this study, the FastDNA Spin kit for soil (MP Biomedicals, Solon, OH, USA) (hereafter described as the Fast kit) and the MoBio PowerSoil DNA isolation kit (MO BIO, Carlsbad, CA, USA) (hereafter described as the MoBio kit). All operations were conducted according to the manufactures instructions. DNA was extracted from 0.5 g fresh soil (approximately 0.3 g dry soil) for the Fast kit and 0.25 g fresh soil (approximately 0.15 g dry soil) for the MoBio kit. The quantity and quality of extracted DNA was assessed on the basis of absorbance at 260 and 280 nm using the NanoDrop ND-1000 (NanoDrop, Wilmington, DE, USA). An A260/A280 ratio of 1.8 - 2.0 is indicative of high purity [33] [34] . All extracted DNA was stored at −20˚C until required for further analysis.

The abundance of the 16S rRNA gene of bacteria was quantified in triplicate by real-time PCR using a LightCyclerTM 480П (Roche, Vienna, Austria). The real time PCR reaction mixture contained 10 µl 10 × SYBR Green iCycler iQ mixture (Roche Applied Science, Germany), 0.2 µl each of the primers Bacteria 968F; AACGCGAAGAACCTTAC and Bacteria 1378R; CGGTGTGTACAAGGCCCG GAACG [35] [36] , 5µl diluted soil DNA (1 ng/µl) and 4.6 µl ultra-pure water in a total of 20 µl reaction volume. Standard curves were obtained using serial dilutions of a known amount of plasmid DNA containing a fragment of the respective genes. The thermal cycle conditions were 95˚C 10 s, 58˚C 10 s, 72˚C 10 s, 45 cycles. Copy numbers of all gene targets were normalized based on grams of dry soil.

The effects of atmospheric CO₂, sampling protocol and DNA extraction kits were tested by a mixed effects model with heterogeneous variance structure using the packages “nlm” and “predictmeans” in R version 3.1 [37] . The effect of DNA extraction kit and the CO₂ on the DNA yield (ng DNA/g dry soil) and quality (A260/A280 ratio) were tested using the data from individual soil cores. The effect of elevated atmospheric CO₂, and sampling protocol on bacterial 16S rRNA gene abundance were conducted separately for the two DNA extraction kits. The mixed effects model used had the CO₂ ring nested within block as a random effect while atmospheric CO₂ concentration, sampling protocol and their interaction were considered as fixed effects. Mean separation was performed using Fisher’s LSD test.

The DNA yield from the MoBio kit was significantly lower than that from the Fast kit (P < 0.001) (Figure 2(A)). The DNA yield from both kits was significantly lower in the eCO₂ soil and the percentage reduction was similar (22% for MoBio and 25% for Fast kit) (Figure 2(A)). The DNA extracted from aCO₂ and eCO₂ soils was of the same quality (CO₂ P = 0.357) but the MoBio kit extracted significantly (P < 0.001) higher quality DNA than the Fast kit irrespective of the CO₂ treatment (Figure 2(B)).

Comparing the values between Figure 3(A) and Figure 3(B) (note the different scales) shows bacterial abundance measured as 16S rRNA copy number was much lower measured from the DNA extracted from the MoBio kit compared to the DNA from the Fast kit. There was no significant effect (P = 0.722) of the experimental protocol on bacterial abundance measured from the DNA of the MoBio kit (Figure 3(A)); by contrast, a significant difference between the sam-

pling method was evident when using the Fast kit (P < 0.001) (Figure 3(B)). The DNA from both kits showed soil bacterial abundance at eCO₂ was marginally lower (P = 0.100 MoBio and 0.084 Fast) compared to aCO₂. However, only the individual soil core method was able to detect this difference at a significant level as calculated using Fisher’s LSD at the P ≤ 0.01 level (Figure 3).

The DNA extraction rates convert to about 7 µg DNA g dry soil-1 for the Fast kit and 2 µg DNA g dry soil-1 for the MoBio kit. These values are higher than those reported byJossi et al. (2006) [16] in another grassland under elevated CO₂ where the average values for two perennial grasses were 6.0 and 8.1 µg DNA g fresh soil-1. The amount of DNA extracted was significantly greater using the Fast kit and this kit extracted a significantly greater amount from the aCO₂ compared to the eCO₂ soil. The greater amount of DNA extracted by the Fast kit is consistent with much higher 16s rRNA gene copy numbers using this kit. Similar to our studyVishnivetskaya et al (2014) [38] extracted more DNA using the Fast kit than the MoBio kit from permafrost soils in the Arctic; although in their study both the Fast and MoBio kits extracted high quality DNA (see Table 1 of Vishnivetskaya et al (2014) for comparison of the kits) [38] . The DNA extracted by the MoBio kit in our study was of a higher quality than that extracted using the Fast kit. The quantity and quality of the soil DNA extracted are important for downstream techniques to analyse microbial community structure [39] . Feinstein et al (2009) found that incomplete DNA extraction from soil resulted in biased estimates of ribosomal copy number and bacterial community composition [40] . The quality of DNA characterised by A260/280 ratio is important as it indicates co-extraction of impurities such as humic substances with DNA that can inhibit the activity of polymerase chain reaction [41] .

We found that bacterial abundance measured as copy number of the 16S rRNA gene and showing whether the result was significant or non-significant was more impacted by the soil sampling approach than the DNA extraction kit (Table 1). The soil sampling approach most widely used in CO₂ studies is the bulked soil method providing a single soil sample per FACE ring or OTC; examples where bacterial abundance has been measured are given in Table 1 and there are further examples where the object was to study bacterial community structure and diversity [42] [16] . The results for bacterial abundance in eCO₂ field experiments shows a consistent reduction in abundance with eCO₂; the only exception in the experiments listed being a slight and non-significant response in one set of measurements in this paper (Figure 3(B)). However, despite the consistent response across experiments there were only three examples where the bacterial abundance was significantly different between aCO₂ and eCO₂ and this was when individual soil cores were studies rather than a bulked sample (Table 1).

Using the individual core method meant that in our case 30 samples were required to be analysed compared to 6 for the bulk sampled method. The approximate total cost for single gene abundance (DNA extraction and qPCR) for 30 samples was 5 fold higher than for 6 bulked samples but similar for the two kits. There was also 3 fold increase in technical time required for the individual soil core approach compared to the bulked soil sample approach. Although the cost of analysis of individual soil core approach can be reduced by 1.5 fold by extracting DNA from individual cores and then bulking the DNA for qPCR (method C) in Figure 1) the statistical analysis indicated that this method resulted in high variability and ultimately a non-significant difference between treatments although the same trend was observed as with individual soil core method.