(This is a sample cover image for this issue. The actual cover is not yet available at this time.)
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Journal of Microbiological Methods 91 (2012) 38–44
Contents lists available at SciVerse ScienceDirect
Journal of Microbiological Methods
journal homepage: www.elsevier.com/locate/jmicmeth
Improved real-time PCR estimation of gene copy number in soil extracts using an
artificial reference
T.J. Daniell a,⁎, J. Davidson a, C.J. Alexander b, S. Caul a, D.M. Roberts a
a
b
Ecological Sciences, The James Hutton Institute, Invergowrie, Dundee, DD2 5DA, UK
Biomathematics & Statistics Scotland, Invergowrie, Dundee, DD2 5DA, UK
a r t i c l e
i n f o
Article history:
Received 1 May 2012
Received in revised form 4 July 2012
Accepted 10 July 2012
Available online 17 July 2012
Keywords:
Gene copy quantification
Internal standard
Quantitative PCR
Relative estimation
Soil ecology
a b s t r a c t
Application of polymerase chain reaction (PCR) techniques has developed significantly from a qualitative technology to include powerful quantitative technologies, including real-time PCR, which are regularly used for detection and quantification of nucleic acids in many settings, including community analysis where culture-based
techniques are not suitable. Many applications of real-time PCR involve absolute quantification which is susceptible to inaccuracies caused by losses during DNA extraction or inhibition caused by co-extracted compounds.
We present here an improvement to this approach involving the addition of an artificial internal standard,
prior to nucleic acid extraction. The standard was generated by in-situ mutagenesis from an E. coli template to
ensure it both did not amplify with bacterial primers used for quantification and was short enough to minimise
possible interference with other analyses. By estimating gene target copies by relative abundance, this approach
accounts for both loss during extraction and inhibition effects. We present a novel application of relative real
time PCR, using the internal standard as a reference, allowing accurate estimation of total bacterial populations
both within and across a wide range of soils and demonstrate its improvement over absolute quantification by
comparison of both approaches to ester linked fatty acid analysis of the same soils.
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
Analysis of soil communities is affected by the huge observed levels
of microbial diversity (Curtis et al., 2002) combined with low level of
culturability (Torsvik and Øvreås, 2002) leading to the wide application
of molecular techniques for the analysis of community structure. This
often involves the application of techniques such as denaturing gradient
gel electrophoresis (DGGE) or terminal restriction fragment length polymorphism (T-RFLP) which rely on polymerase chain reaction (PCR)
amplification of diagnostic genomic regions to target groups either phylogenetically or functionally and produce data at a presence/absence or
relative abundance level (for a review see Hirsch et al., 2011). These
techniques are limited to the analysis of community structure and are
incapable of estimating community size which has led to the application
of real-time PCR methods using gene copy number as a proxy of community size (Hermansson and Lindgren, 2001; Hirsch et al., 2011;
Macdonald et al., 2011). Alternatively a wide range of techniques including lipid based methods such as phospholipid-derived fatty acid (PLFA)
and ester linked fatty acid (ELFA) (Frostegård and Bååth, 1996), direct
counting (Bogosian et al., 1996) or chloroform fumigation (Anderson
⁎ Corresponding author. Tel.: +44 1382 568821; fax: +44 1382 568502.
E-mail addresses: [email protected] (T.J. Daniell),
[email protected] (J. Davidson), [email protected] (C.J. Alexander),
[email protected] (S. Caul), [email protected] (D.M. Roberts).
0167-7012/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.mimet.2012.07.010
and Domsch, 1978) have been used to estimate population size or biomass. Typically, application of real-time PCR methods in soil systems targets DNA in order to estimate target-gene copy as a proxy for population
size, for example bacterial quantification using fragments of the 16S
gene (Stubner, 2002), or quantifying genes to target functional groups
such as the ammonium monooxygenase A subunit to estimate nitrifier
population size (Okano et al., 2004). The application of real-time PCR
and other molecular methods for the quantification of functional groups
is reviewed in Sharma et al. (2007). The majority of applications have
applied absolute methods where a standard curve is generated from a
serial dilution of a clone of the target amplicon allowing estimation of
copy number in reactions using the same conditions. Whilst population
estimation using molecular approaches have many advantages over
culture-based or microscopy-based methods, they do have limitations.
They can, for example, suffer from inhibition of any PCR amplification
by co-extracted compounds such as humic acids that inhibit enzymatic
reactions, including PCR amplification (Jackson et al., 1997; Okubara et
al, 2005; Zhou et al., 1996), or inadequacies in the DNA extraction that
can lead to losses that are difficult to predict or quantify. Losses can be
expected at every step of extraction including cell lysis and losses during
purification, for instance where phenol:chloroform extraction and ethanol precipitation steps occur in the DNA extraction process (Mumy and
Findlay, 2004; Tsai et al., 1991). For example, losses during ethanol precipitation of between 70% and 20% dependent on a number of factors including initial DNA levels, temperature and length of centrifugation
(Zeugin and Hartley, 1985) and phenol:chloroform extraction efficiencies
Author's personal copy
T.J. Daniell et al. / Journal of Microbiological Methods 91 (2012) 38–44
of 49–59% have been estimated for community extractions (Weinbauer
et al., 2002). These potential issues make accurate estimation of copy
number difficult with absolute real-time PCR quantification, especially if
experiments include soil from different origins or with treatments likely
to alter the level of inhibitors.
Many applications of real-time PCR, predominantly in relatively
clean systems such as culture based experiments or large organism
studies where samples are taken from single individuals, estimate
expression levels of genes of interest using housekeeping genes to
normalise results (Pfaffl et al., 2002). This option is not available to
estimate gene copy number in soil systems since sample size, in
terms of organism number or biomass, is hard if not impossible to
standardise.
A similar approach has been suggested for molecular approaches including PCR from soil (using DGGE), for example Park and Crowley
(2005) who used Escherichia coli JM 109 cells containing plasmid
which were added to soil and detected using primers specific for the
plasmid in real-time PCR analyses to normalise results. Additionally
the presence of the E. coli band in DGGE was used to normalise intensity
across this analysis. The issue here is that the presence of E. coli genomic
DNA potentially skews estimates of any target contained within this genome, complicating analysis. The ideal situation would be to utilise a
normalisation target and methodology that can both assess extraction
and PCR efficiency whilst not interfering in any way with normal analysis of the samples. To achieve this any spike has to be independent of
any intended analysis methodology which precludes the use of a natural DNA template.
Here we aimed to improve the accuracy of real-time PCR estimation
of gene-copy number in DNA extracted from soil samples through the
application of an artificial reference spike and relative real-time PCR.
To achieve this aim we generated the spike through in-situ mutation
of an E. coli DNA template for amplification of a fragment of the 16S ribosomal gene altering the three 3′ terminal bases of the PCR primer target sequences to their complementary bases. This spike would thus not
be amplified by the normal primer set used to target the 16S gene or,
due to its short length, primers normally used for community structure
analysis allowing extracts to be used for normal analysis with no interference from the added reference spike. The mutated spike was then
added to a wide range of soil types and DNA extracted and used to
test the efficiency of relative real-time PCR targeting the bacterial 16S
gene as a proxy for cell number or biomass using the spike as reference
for relative quantification. We have further contrasted results against
estimates generated using the same amplification but applying absolute
quantification and comparing both with an established method to estimate soil biomass. The application of this methodology should allow far
more accurate control of technical variability recently recognised by
Bustin (2010) as a primary cause for inconsistency and irregularity in
quantitative PCR results.
2. Methods
2.1. Soil collection/storage
The location of each sample site is given in Supplementary Table 1.
These sites were selected to cover a wide range of soil types. Six cores
were randomly taken at each site, to a depth of ten centimetres using
a grass plot sampler (Van Walt Ltd, Haslemere, UK), and combined to
obtain a homogeneous sample from each site. Samples were sieved to
4 mm to remove large debris, mixed and eight replicate sub-samples
taken. Four were used to allow the assessment of the accuracy of estimation of each soil type by providing measurements from each
sub-sample using both real-time PCR techniques. The other set of four
sub-samples were used to measure soil bacteria using the ELFA technique as a ‘gold standard’. Samples were snap-frozen in liquid nitrogen
and aliquots stored at −80 °C prior to DNA or lipid extraction.
39
2.2. ELFA
Ester linked fatty acids (ELFA) were extracted as described by
Schutter and Dick (2000). Briefly, frozen aliquots of soil were extracted
with KOH in methanol, neutralised with acetic acid and fatty acid methyl esters extracted with hexane. Following drying, fatty acid methyl
esters were re-dissolved in 100 μl isohexane and analysed by gas chromatography on a Agilent 6890N Network system fitted with a flame
ionization detector, Agilent HP-5 column (Agilent Technologies UK
Ltd, Stockport, UK). Hydrogen was used as carrier gas and helium as
the make-up gas, with the following programme: 160 °C for 2 min, an
increase by 4 °C per min to 270 °C, hold for 10 min. Fatty acid 23:0
was added as an internal standard. Double bonds of fatty acids were referred to the methyl end (ω) of the molecule and the nomenclature for
ELFA was used as described previously (Zogg et al., 1997). Bacterial biomass was calculated as the concentration of fatty acids that could be related to bacterial groups (Zelles, 1999).
2.3. Preparation of calibration controls and of reference spike
The 16S wild type calibrator control was generated by PCR amplification of a 16S gene fragment from E. coli genomic DNA generated from a
lysate generated by boiling a colony suspended in 100 μl water for
5 min. 1 μl of this preparation was then used as template for PCR using
Primer-1 and Primer-2 (all primer sequences in Table 1) in a final reaction volume of 25 μl containing: 2.5 U DNA polymerase (Expand High Fidelity enzyme mix. Roche Diagnostic Systems, Burgess Hill, UK), 5 pmol
of each primer, 0.5 mM of each nucleotide, 1× Expand High Fidelity buffer (Roche Diagnostic Systems, Burgess Hill, UK) and 10 μg of bovine
serum albumin (BSA) (Roche Diagnostic Systems, Burgess Hill, UK).
PCR was performed with a DNA Engine DYAD thermocycler (MJ Research, UK) with an initial denaturation step of 95 °C for 5 min followed
by 30 cycles of 95 °C for 1 min, 54 °C for 1 min and 72 °C for 30 s.; cycling was completed by a final extension period of 72 °C for 5 min.
The reference spike was generated by PCR in situ mutagenesis
(Vallette et al., 1989) to form a product in which the three 3′ terminal
recognition bases of both PRIMER-1 and PRIMER-2 were altered to the
requisite complementary bases providing a template suitable for amplification with the Mut-F and Mut-R primers.
This was performed in a two stage PCR process using 342FMut and
Primer 2 in the first reaction and Mut-F and 534RMut in the second
using the PCR conditions given above. 1 μl E. coli genomic DNA was
used as template in round one and 1 μl of 100 fold diluted round one
product in the second reaction. PCR success was assessed by 1.5% agarose gel electrophoresis.
Products were cloned into pGEM-T Easy (Promega, Southampton,
UK), following the manufacturer's instructions and transformed into
E. coli DH10B electrocompetent cells prepared following the method
of Tung and Chow (1995). Colonies were screened for mutation success by PCR with relevant PCR primers and conditions outlined
above. Sequence identity of selected clones for both the mutated
and wild-type products was confirmed by sequencing in a total volume of 10 μl using 1:8 dilution of BigDye® Terminator v3.1 Cycle Sequencing kit (Applied Biosystems, Warrington, UK) with vector
primers directed against the SP6 or T7 promoter regions and following manufacturer's instructions. Sequencing reactions were purified
Table 1
Primers used in this study, all primers supplied by Thermo Fisher (Loughborough, UK).
Primer name
Sequence (5′ 3′)
Reference
Primer-1
Primer-2
Mut-F
Mut-R
342FMut
534RMut
CCTACGGGAGGCAGCAG
ATTACCGCGGCTGCTGG
CCTACGGGAGGCAGGTC
ATTACCGCGGCTGCACC
CCTACGGGAGGCACGTCTGGGGAATAT
ATTACCGCGGCTGGACCCACGGAGTTA
(Muyzer et al., 1993)
(Muyzer et al., 1993)
(this study)
(this study)
(this study)
(this study)
Author's personal copy
40
T.J. Daniell et al. / Journal of Microbiological Methods 91 (2012) 38–44
by ethanol precipitation and run on an ABI Prism 3700 DNA Analyser
(Applied Biosystems, Warrington, UK). Transformants with the correct insert were placed into long term storage at −80 °C in 20% glycerol
and plasmid generated from a single selected colony by large scale plasmid prep using Plasmid Midiprep System (Promega, Southampton, UK),
following manufacturer's instructions. Plasmid was quantified by absorption at 260 nm using a DU 640 spectrophotometer with a 50 μl
cuvette (Beckman Coulter, High Wycombe, UK) to give a mass concentration and converted to copies using the following formula:
ðTotal number of A=T bases in the insert þ plasmid 601Þ
þ ðTotal number of G=C bases in the insert þ plasmid 618Þ
¼ molecular weight ðg=moleÞ of the plasmid þ insert:
The number of copies of plasmid=μl
¼ ðμg plasmid DNA calculated from 260 nm absorption=
weight of plasmid þ insertÞ
Avogadro’s number
After linearization with NotI (Promega, Southampton, UK), dilution
series of both wild-type and mutant plasmid were used to generate
standard curves and/or as reference spike for real-time experiments.
included in each PCR run as a calibrator to allow comparison to the
wild-type calibration curve. For relative quantification, a second PCR
was performed on each soil-DNA extract using the reference spike
primers (Mut-F and Mut-R) under the above amplification conditions
and using mutated plasmid as a calibrator with the reference calibration
curve.
2.7. Calibration curve generation
Mutated and wild-type plasmids were diluted to produce standards
of known copy-number to allow the generation of curves containing
109–10 1 copies per reaction in a tenfold dilution series. Triplicate replicates of these standards (mutated and wild-type) were amplified, in the
presence of DNA free soil extract to give an equivalent amount of
co-extracted material, using the appropriate primers and the supplied
software (Roche Diagnostic Systems, Burgess Hill, UK) to produce and
store calibration curves for mutated and wild-type templates in a soil
extract background.
2.8. Absolute and relative estimation
An additional sample of soil from Bullion field (Location:
56:27.3397N3:4.6401W, freely drained brown forest soil, same soiltype as sample 2) was processed as described in Section 2.4 and the
resulting DNA-elution was digested with RQ1 RNase-Free DNase
(Promega, Southampton, UK), using the manufacturer's protocol, to remove DNA. Briefly, extract was incubated at 37 °C for 30 min followed
by inactivation by the addition of RQ1 stop solution and heating at
65 °C for 10 min. Extract was then phenol:chloroform extracted and
precipitated with isopropanol/NaOAc as performed for DNA extraction.
Efficient removal of DNA was confirmed by PCR. Following the initial
testing on a selection of samples for optimal amplification (see
Section 2.6 below) soil extracts were diluted 1:10 in water and 1 μl
per reaction added to the PCR master mix when producing calibration
curves for real time PCR to provide a soil extract background.
Absolute estimation using real-time PCR is based on direct estimation of copy number by comparison with a standard curve of diluted
template. Relative estimation involves the separate amplification of
the target gene and a reference of known copy number, here in the
form of a ‘spike’ added at the DNA extraction stage. Both were assessed
using the supplied software (Roche Diagnostic Systems, Burgess Hill,
UK) applying the second derivative maximum method to quantify template DNA.
To estimate the number of 16S copies per gram of dry soil using absolute quantification it is necessary to take into account losses expected
during DNA extraction. With the DNA extraction method applied here
the following factors have to be taken into consideration: the actual
dry-weight of soil extracted, the proportion of water in the soil extracted,
losses involved in avoiding carry-over during the phenol:chloroform extraction and any dilution of the DNA during precipitation or post extraction. These factors were calculated to produce a multiplication factor
specific for each sample. This factor was used to convert the absolute estimation of 16S copies into an estimation of 16S copies per gram dry
weight of soil extracted. As it is impossible to measure the exact loss during the precipitation or the phenol:chloroform stage, we were forced to
assume full recovery leading to an inevitable underestimation of copy
number.
Relative quantification of the target DNA was assessed using the supplied software (Roche Diagnostic Systems, Burgess Hill, UK) for advanced relative quantification using default settings comparing it to
stored standard curves for both the PCR reference and target. Relative
Quantification compares the levels of two different target sequences in
a single sample (16S and mutated 16S spike) and expresses the final result as a ratio of these targets. This ratio is used to calculate the number
of gene copies per gram dry weight soil using (Spike copies ×ratio)/dry
weight soil added.
2.6. Real-time cycling conditions
2.9. Statistical analysis
Amplification of 16S DNA and mutated reference plasmid was
performed using Roche Lightcycler 480 SYBR Green I master Kit
(Roche Diagnostic Systems, Burgess Hill, UK) following manufacturer's
instructions. Following initial tests 2 μl of a 1/10 dilution template
DNA was amplified using a Lightcycler 480 (Roche Diagnostic Systems,
Burgess Hill, UK) and the following conditions: an initial denaturing
step at 95 °C for 15 min was followed by 40 cycles of 95 °C for 10 s,
54 °C for 10 s, 72 °C for 10 s and a single acquisition step at 81 °C for
5 s. This was followed by a melt curve: 95 °C for 10 s, 65 °C for 15 s
then heating to 95 °C at a ramp rate of 0.1 °C/s, with continuous acquisition. A reaction containing 107 copies of the wild-type plasmid was
The data to be analysed comprised measurements of bacterial population size on samples of soil taken from locations with differing soil
types. Four sub-samples were taken from each location's material and,
for each of these sub-samples, a measurement of copy number per
gram d.w. calculated using both the absolute and relative real-time
PCR techniques. These eight estimates from fourteen locations gave a
total of 112 real-time PCR measurements. Four further sub-samples
from each location were used to provide measurements of soil bacteria
using ELFA. The mean of these four values at each location provided 14
values used as a gold standard against which comparisons were made.
Both ELFA and real-time PCR data were transformed to log base 10 to
2.4. DNA extraction
DNA was extracted from soil samples by the method of Deng et al.
(2010). Briefly, 1 g of soil was suspended in a solution of 0.12 M
Na2HPO4 1% SDS to form a slurry, 1 ml of which was subjected to bead
beating using 1 mm glass beads (TissueLyser, Qiagen, Hilden, Germany).
The resultant suspension was clarified by centrifugation and the supernatant extracted with phenol:chloroform and chloroform. DNA was precipitated using isopropanol and sodium acetate and further purified with
polyvinylpolypyrrolidone using a spin column. Mutated reference-spike
was added to the lysis buffer prior to addition to soil to give a final spike
of 109 copies per extracted sample.
2.5. Production of DNA-free soil extract
Author's personal copy
T.J. Daniell et al. / Journal of Microbiological Methods 91 (2012) 38–44
stabilise variance prior to analysis. Initially separate tests of the relationship of each real-time PCR measurement versus ELFA were performed
using Analysis of Variance (ANOVA). Measurements from each technique were analysed with a term fitted for ELFA as a covariate and resolving the residual sum of squares into lack of fit and pure error
(Draper and Smith, 1998). Deviations among the four copy number
values for the same soil about their mean represented ‘pure error’ and
were the result of variation within the same experimental material
and a lack of repeatability of the measurement procedure. Deviations
of the observed mean copy number for each soil sample from the line
fitted to the covariate, ELFA, were referred to as ‘observed lack of fit’
and measured the goodness of fit of this line. This lack of fit variation
gave a measure of validity of the technique in its agreement with the assumed ‘gold standard’ of the ELFA values. The estimated variances for
pure error and observed lack of fit were then compared between the
two real-time PCR approaches with an F-test to determine if there
were any significant differences between the two molecular quantification approaches.
Since the measurements of the absolute and relative techniques
were obtained from the same material in each of the four technical replicates for each soil sample, these did not provide independent estimates
of the pure error variance. Similarly, there was no independence in the
estimates of variance of the soil sample means about their respective
fitted lines. This problem was addressed by using a mixed modelling approach on the combined datasets. In this approach pure error had the
same interpretation as with the ANOVAs; however the observed lack
of fit variation was decomposed into contributions due to the pure
error and the further variation of soil sample means not attributable to
pure error ‘underlying lack of fit’. Both pure error and underlying lack
of fit were modelled as random effects but with covariance structures
chosen in order to contain the relevant variance components and to address the lack of independence in the samples. In this way, we could formally test for differences in the variance components between the two
techniques. Additional benefits of this approach were formal statistical
tests for differences in the fixed effects of mean copy number recovered
between the techniques and also differences in the slope of the fitted linear relationship to ELFA.
Three models were fitted. The full model had, within each random
effect, separate variance components estimated for each technique as
well as a covariance term. Two nested models could then be fitted,
each testing one of the two hypotheses: that there is no difference between absolute and relative techniques in (i) the pure error variance
components or (ii) the underlying lack of fit variance components. In
each sub-model this was achieved by constraining the variance components for the two techniques to be equal. The difference in deviance
could then be tested against that of the full model and its statistical significance determined against a chi-squared distribution with 1 degree
of freedom since the number of parameters in the model for random
variation differed by one.
The ANOVAs were fitted using GenStat for Windows, 14th Edition
(VSN International Ltd., Hemel Hempstead, UK) as were the mixed
models using the Residual Maximum Likelihood (REML) facility.
41
Griffiths et al., 2007). ELFA was performed on each soil type to obtain
an estimation of total biomass based on lipid content and this data
was used as a baseline to assess the performance and accuracy of
real-time data. There was excellent reproducibility in lipid data over
four replicates from each soil sample (Table 4), sufficient to justify the
amalgamation of the replicates into a mean-value to investigate the
variability within and between the two real-time quantification approaches and confirming the uniformity of the replicated soil samples.
There was significant variation between soils in biomass estimation by
ELFA (see Table 4), ranging from 10.20 to 623.29 nmol g−1 dry weight
soil (dws). Variation in biomass related to bacterial population sizes has
been reported by others. For example, Hinojosa et al. (2005) in a single
soil type observed similar variation when comparing polluted,
reclaimed and pristine soils with both ELFA and PLFA methods with
extracted ELFA between 22 and 184 nmol g−1 dws with pristine soils
exhibiting concentrations at the higher end of the range. There are
few studies comparing across soil types; Frostegård and Bååth (1996)
analysed a range of soils from forest (Spruce and Beach), grassland, garden and arable fields and using marker PLFA representing the bacterial
populations estimated biomass to be between 164 and 1739 nmol g−1
organic matter in the different soils. Taking account of the measured %
organic matter in each of these soils, these values convert to 34.4–
516.1 nmol g −1 dws. These estimates are within the range determined
by this study but direct comparison is difficult due to their application of
PLFA as opposed to ELFA techniques. Additionally, Frostegård and Bååth
(1996) could find no correlation either between the amounts of bacterial biomass as estimated by PLFA and pH or between the amounts of
bacterial PLFA and the organic matter content of the soils studied similar to the patterns found here.
3.2. Real-time PCR
Calibration curves for both the target and spike were produced using
a linearised plasmid containing a PCR fragment of either E. coli 16S or a
mutated version respectively. The PCR efficiency, in a DNA-free soil extract background, of both the wild-type and mutated PCR targets was
high confirming the robustness of both PCRs. The calibration curves
for the mutated template was linear over nine orders of magnitude
(101–109), the calibration curves for the wild-type template was linear
over eight orders of magnitude (102–10 9). This difference is likely due
to absence of template from the PCR kit as the variation did not increase
with decreasing template copy number. The sensitivity of the calibration curves reliably attained 102 copies per reaction (Table 2).
3.3. Efficiency of recovery of DNA from soil as estimated by real-time PCR
A number of groups have used real-time or competitive PCR to investigate the efficiency of recovery of DNA from soil or sediment extractions
(Table 3). These studies used a range of added targets to estimate the efficiency of recovery and estimation. Targets include genomic DNA, bacterial cultures containing a high copy number plasmid and bacterial cells
seeded into sterilised soil samples, and vary in extraction methodologies
3. Results and discussion
3.1. ELFA
To assess the suitability of absolute and relative quantification of 16S
copy number to estimate soil bacterial populations it was necessary to
compare these molecular-based estimations against an established
technique. The use of ELFA to estimate total bacterial populations in
soil has previously been established (Schutter and Dick, 2000). ELFA
analysis was used in preference to the more routinely employed phospholipid fatty analyses (PLFA) because ELFA analysis is more rapid
and both PLFA and ELFA approaches are comparable in their ability to
discriminate between microbial communities (Hinojosa et al., 2005;
Table 2
Showing the mean crossing point (±SE) for the wild-type and the mutated calibration
curves generated from linearised plasmid containing wild type or mutated insert respectively. Means are based on three replicates for each concentration.
Copies
Wild-type mean Cp
Mutated mean Cp
108
107
106
105
104
103
102
101
10.30 ± 0.04
13.74 ± 0.05
16.01 ± 0.01
20.21 ± 0.04
22.46 ± 0.16
25.95 ± 0.03
28.28 ± 0.05
27.87 ± 0.09
9.63 ± 0.15
13.48 ± 0.06
17.30 ± 0.26
20.66 ± 0.17
24.43 ± 0.01
27.65 ± 0.01
30.90 ± 0.29
35.0 ± 0.01
Author's personal copy
42
T.J. Daniell et al. / Journal of Microbiological Methods 91 (2012) 38–44
Table 3
Comparison of DNA extraction/detection efficiencies reported by previous groups and
this study for different soil types. Details of soil types (this study) are given in Supplementary Table 1. Efficiency (this study) is calculated as average % of spike recovered
over 4 replications and standard deviation.
Reference Extraction
method
Quantification Target
method
Lee et al., Chemical
1996
Competitive
quantitative
PCR
Competitive
touchdown
PCR
Rose et
al.,
2003
Chemical
Wizard®
Tsai et al., Chemical
1991
Mumy
and
Findlay,
2004
This
study
Membrane
hybridisation
UltraClean QC-PCR
FastDNA
SoilMaster
Chemical
Sybr green
real-time PCR
100 ng−1
E. coli DNA
Soil type
Sterilised
pasture soil
180 ng g−1
Sterilised
E. coli DNA
silt loam
1.8 pg g−1
E. coli DNA
1.6 × 109 cells Sterile
southern
California
soil
Lake
E. coli +
sediment
pBR322-λ
9
1 × 10 cells
1 × 109 linear Soil 1
DNA
Soil 2
Soil 3
Soil 4
Soil 5
Soil 6
Soil 7
Soil 8
Soil 9
Soil 10
Soil 11
Soil 12
Soil 13
Soil 14
Efficiency of
recovery
(%)
3.68 ± 0.11
34 ± 7
6.1 ± 1.3
60 ± 5
14.9 ± 16.0
28.3 ± 10.5
2.4 ± 0.1
10.93 ± 2.87
9.13 ± 5.29
4.10 ± 1.01
7.08 ± 6.7
2.77 ± 1.9
5.84 ± 2.04
4.09 ± 2.44
11.48 ± 5.40
5.64 ± 4.39
5.67 ± 2.93
13.66 ± 4.44
2.99 ± 1.21
2.61 ± 2.17
8.19 ± 3.51
with one study comparing several methods. The majority of these studies do not speculate on the likely source of variation in the estimation or
cause of losses although the choice of extraction method (Mumy and
Findlay, 2004) and/or the amount of added target (Rose et al., 2003)
can clearly play a role. Soil, when compared to cultures, as a starting material also negatively affects both the recovery and variability of extraction of both DNA and RNA (Mumy and Findlay, 2004; Okubara et al.,
2005; Tsai et al., 1991) highlighting the need to understand issues
with variability or performance of extraction and subsequent PCR inhibition both between and within soil types.
Our studies found recovery efficiencies of the spike of between 2.61%
and 13.66%, dependent on, but with no clear link to, soil-type, which is of
a similar range to recoveries observed in other studies (Table 3). Analysis of our spike-recovery data suggests that limitations in the efficiency
of the soil extraction driven by differential effects of different soils play
a part. Fig. 1 shows a significant variation in recovery of spike between
soil types (pb 0.001) but suggests that variation was not driven by different perceived organic matter content between soil types. For example
the two peat soils (13 and 14) vary significantly. A high degree of variation is found in the recovery of this internal standard for the wide range
of soil-types included in this study as well as the high level of variation in
recovery between replicates. There was also no correlation between efficiency of recovery and the amount of soil extracted (data not shown).
This difference in performance is likely due to variation in the distribution of inhibitors between aliquots or variation in the performance of
the extraction between aliquots. The observed variability in recovery
as estimated by an absolute real time PCR reinforces the improved performance of the relative tool discussed below since, unlike an unspiked
soil, a fixed amount of template for the test PCR reaction was added to
the extraction removing the possibility that variation was driven by natural variation in population size.
Where applying absolute quantification we attempted to account
for all measurable losses in the DNA extraction process and yet this
variation of DNA recovery efficiency demonstrates the limitations of
simple absolute-quantification for estimation of DNA copies present in
samples of different soil-types. For relative quantification this variation
is countered by the internal standard used, as demonstrated by the significantly reduced variation between replicates. This reduced variation
allows a far more precise estimation of population size.
3.4. Estimation of soil-bacterial population size
ELFA and real-time PCR data were converted to log base 10 to ensure
residual variance remained constant prior to comparing the absolute
(Fig. 2a) and relative (Fig. 2b) real-time quantifications against the
ELFA data. The results of the separate ANOVAs for each technology are
shown in Table 5 and demonstrate that there was evidence of significant differences between both lack of fit and pure error variation.
Also shown in Fig. 2a and b are the fitted lines from the REML analysis using the full model with separate variance components estimated
for each of the respective estimation techniques. This analysis demonstrates the improved performance of the relative technique on a number
of criteria. Total copy number per gram d.w. recovered were at least an
order of magnitude higher for all soil types, ranging from a factor of
11.4 for soil type 6 to a factor of 23.0 for soil type 3 (intercept for the relative line was 8.61 compared to 7.70 for absolute). The slope of the fitted
relationship for the relative technique was also significantly steeper
(pb 0.001) — slope parameter was 0.91 for relative technique and 0.75
for absolute. This steeper slope demonstrates the greater ability to discriminate amongst soil types of the relative technique compared to absolute estimates. For example, the mean detected copy-number for
samples of soil type 3 was greater than for soil type 1 when using relative estimation (which agrees with the gold standard) but this ranking
is reversed when using the absolute approach. These differences are
likely due to differences in extracted inhibitors between soil types.
Since REML estimates variance components for (i) pure error, or, the repeatability of each technique and (ii) underlying lack of fit we fitted
models which tested for the statistical significance of any difference in
performance. The pure error variance component was tested for significant differences between the techniques by comparing deviance between models where pure error was identical (Model II in Table 6)
and where different variances were fitted (Full Model, Table 6). This
showed a highly significant difference (pb 0.001) with the best fit
model estimating the absolute technique's variance component at
0.173 compared to 0.011 for the relative; there was also a high correlation of 0.68 between the estimates from the two techniques performed
on the same technical replicates. This evidence shows the relative
Fig. 1. Efficiency of recovery of added template, showing the recovered copy-number
(E+06) of the spike for each soil-type estimated by absolute quantification
(1 × E+09 added prior to extraction). Bars indicate ±1 standard error, letters indicate
significant classes generated by post-hoc testing using a Fisher LSD test following
ANOVA where soil types was significantly different (p b 0.001). Soil types as described
in Supplementary Table 1.
Author's personal copy
T.J. Daniell et al. / Journal of Microbiological Methods 91 (2012) 38–44
43
Fig. 2. Fitted linear terms from REML analysis showing: (a) copy number from 16S absolute quantification and (b) copy number from 16S relative quantification versus the same
ELFA measurements. For the four replicates from each soil type, ELFA values were set to the mean ELFA value. All axes are scaled on log base 10.
technique's ability to achieve more consistent results on repeated analysis of the same soil samples.
In the comparison of sample mean squares for observed lack of fit in
the ANOVAs, although the stratum variances were highly different (as
shown in Table 5, F ratio was 5.33), this included a contribution from
the pure error variance in the lower stratum. REML decomposed this
variance and allowed the testing of the contribution from underlying
lack of fit alone. This was done by testing a model with separate variance components for underlying lack of fit for the two techniques (the
Full Model in Table 6) against one constrained to have the same variance component for both (Model I in Table 6). The Full Model gave variance components of 0.0766 for the absolute and 0.0197 for the relative
techniques but there was no evidence for a difference when compared
with Model I which fitted a single variance component of 0.0228. The
reduction of ‘noise’ coupled with the more significant positive correlation between the ELFA data and the relative quantification estimation
of population numbers also would allow accurate comparison of
populations between soil types which would otherwise have significant
error attached due to differences in efficiency of estimation. This type of
analysis would not be reliable with the absolute quantification approach due to the variance between repeats of the same soil type. The
lack of fit between the ELFA data and the absolute quantification
makes this approach less appropriate when working across soil types.
The relative quantification approach is far more suitable for this type
of analysis.
Table 4
Data summary of bacterial gene copy number for the soils used in this study. Estimations are by absolute and relative real-time PCR and are ranked by the relative estimate. Also shown is the estimate of ELFA. Details of soil types (this study) are given
in Supplementary Table 1.
Land use
Soil
number
ELFA nmol g−1
dws mean,
(%CV)
Absolute copies
Relative copies (108)
(108) per gram mean per gram mean
(%CV)
(%CV)
Non-arable
Non-arable
Arable
Arable
Arable
Non-arable
Non-arable
Arable
Arable
Arable
Non-arable
Arable
Arable
Arable
13
14
10
11
9
12
5
6
7
8
4
2
3
1
Average
532.67 (7.1)
623.29 (3.3)
92.87 (4.7)
107.14 (6.5)
88.59 (11.4)
156.70 (7.6)
58.24 (2.7)
62.65 (6.6)
66.70 (4.2)
67.88 (6.8)
58.65 (3.7)
35.85 (3.5)
38.11 (4.5)
10.20 (15.6)
146.29
47.27 (92.9)
129.32 (47.1)
40.65 (63.4)
100.26 (52.3)
31.64 (99.4)
9.47 (99.3)
8.60 (93.9)
13.64 (53.3)
12.48 (89.4)
27.67 (81.3)
12.56 (103.2)
9.33 (64.6)
3.08 (14.3)
3.43 (41.1)
32.10
1110.47 (33.9)
1030.87 (23.6)
539.17 (15.6)
415.85 (24.4)
313.27 (24.2)
306.28 (17.0)
230.82 (17.0)
205.09 (24.4)
200.55 (7.5)
192.50 (25.5)
119.35 (19.8)
90.39 (28.7)
84.89 (36.0)
26.17 (25.7)
347.55
Published estimates of bacterial population size in soil are variable,
even when targeting the same gene. For example, estimates of gene
copy number in agricultural soil based on 16S-targeted real-time PCR
range from 6 × 108 to 2.2 × 109 copies per gram d.w. soil (Dandie et al.,
2007; Okano et al., 2004). As described in Section 3.3, this range can
be due to different soil-types supporting different population sizes of
bacteria, effect of soil-type on efficiency of DNA extraction and/or PCR
amplification or on limitations inherent in the measurement approach.
The contrasting results between absolute and relative estimations observed in this study suggest that the application of relative methods
may control for a number of these causes of error. Our estimations for
soil-bacterial populations are 3.2× 109 and 3.5× 1010 for absolute and
relative quantification respectively (see Table 4), averaged over all soil
types. Our estimation by absolute quantification is higher than that
published previously, due, we believe, to our having made stringent attempts to account for all known losses during the DNA extraction
(Section 2.8). Our estimation by relative quantification is above the
published range, due to the relative quantification approach accounting
for losses and inhibitory affects and therefore allowing a far more precise estimation of population size.
4. Conclusion
The application of relative quantification methods is standard in expression studies where housekeeping genes are used to control for
differences in extraction efficiency or assay performance. Here we demonstrate that the addition of an artificial spike to soil extractions allows
relative quantification methods to be applied in such a non-defined
starting material by providing a fixed term of reference. This allows
both a significant improvement in the estimation of gene copy number
and a decrease in the variability in the estimation. The difference between relative and absolute methods of estimation suggests that the application of absolute measures, commonly applied in soil studies, can
Table 5
ANOVA table of results for separate analyses on data from absolute and relative techniques with ELFA as covariate.
Absolute
Relative
Source of variation
df
ss
ms
df
Soil sample stratum
log10(ELFA)
Fitted line
Lack of fit
ss
ms
F-ratio
Sig.
13
1
12
11.745
5.998
5.746
0.903
5.998
0.479
13
1
12
9.886
8.808
1.078
0.760
8.808
0.090
5.33
0.003
Technical rep stratum
Pure error
42
Total
55
7.248
18.993
0.173
42
55
0.465
10.351
0.011
15.58
b0.001
Author's personal copy
44
T.J. Daniell et al. / Journal of Microbiological Methods 91 (2012) 38–44
Table 6
Comparison of models for testing equality of variances between absolute and relative techniques. The full model has no constraints on its parameters. Variance components
constrained to be equal (shown in italics) are: soil sample for Model I and the technical replicates for Model II. Variance components are labeled vaa for the variance of the absolute
technique, vrr the variance for the relative and var the covariance between the two. LRS is the likelihood ratio test statistic (equal to the change in deviance relative to the full model),
df = degrees of freedom and Sig. = significance.
Soil sample
Technical rep
Model
vaa
vrr
var
vaa
vrr
var
Deviance
LRS
df
Sig.
Full
I
II
0.0766
0.0228
0.0973
0.0197
0.0228
0.0040
0.0189
0.0098
0.0194
0.173
0.199
0.090
0.011
0.012
0.090
0.029
0.033
0.028
−206.62
−204.05
−125.03
2.57
81.59
1
1
0.109
b0.001
underestimate the gene copy count by factors ranging from 11.4 to 23.0.
Additionally, the generation of an artificial spike negates concerns relating to erroneous detection of added spike as part of the soil community
under analysis.
Supplementary data to this article can be found online at http://
dx.doi.org/10.1016/j.mimet.2012.07.010.
Acknowledgements
This work was financially supported by the Scottish Government
Rural and Environment Science and Analytical Services Division. We
would like to thank both internal and external reviewers for comments
which have improved the manuscript.
References
Anderson, J.P.L., Domsch, K.H., 1978. A physiological method for the quantitative measurement of microbial biomass in soils. Soil Biol. Biochem. 10, 215–221.
Bogosian, G., Sammons, L.E., Morris, P.J.L., O'Neil, J.P., Heitkamp, M.A., Weber, D.B.,
1996. Death of the Escherichia coli K-12 strain W3110 in soil and water. Appl.
Environ. Microbiol. 62, 4114–4120.
Bustin, S.A., 2010. Why the need for qPCR publication guidelines? The case for MIQE.
Methods 50, 217–226.
Curtis, T.P., Sloan, W.T., Scannell, J.W., 2002. Estimating prokaryotic diversity and its
limits. Proc. Natl. Acad. Sci. U. S. A. 99, 10494–10499.
Dandie, C.E., Miller, M.N., Burton, D.L., Zebarth, B.J., Trevors, J.Y., Goyer, C., 2007. Nitric
oxide reductase-targeted real-time PCR quantification of denitrifier populations in
soil. Appl. Environ. Microbiol. 73, 4250–4258.
Deng, H., Zhang, B., Yin, R., Wang, H.-L., Mitchell, S.M., Griffiths, B.S., Daniell, T.J., 2010.
Long-term effect of re-vegetation on the microbial community of a severely eroded
soil in sub-tropical China. Plant Soil 328, 447–458.
Draper, N.R., Smith, H., 1998. Applied Regression Analysis, 3rd edition. John Wiley &
Sons, Inc., New York, U.S.A.
Frostegård, A., Bååth, E., 1996. The use of phospholipid analysis to estimate bacterial
and fungal biomass in soil. Biol. Fertil. Soils 22, 59–65.
Griffiths, B.S., Caul, S., Thompson, J., Birch, A.N.E., Cortet, J., Andersen, M.N., Krogh, P.H.,
2007. Microbial and microfaunal community structure in cropping systems with
genetically modified plants. Pedobiologia 51, 195–206.
Hermansson, A., Lindgren, P.-E., 2001. Quantification of ammonia-oxidizing bacteria in
arable soil by real-time PCR. Appl. Environ. Microbiol. 67, 972–976.
Hinojosa, M.B., Carreira, J.A., Garcia-Ruíz, R., Dick, R.P., 2005. Microbial response to
heavy metal-polluted soils: community analysis from phospholipids-linked fatty
acids and ester-linked fatty acids extracts. J. Environ. Qual. 34, 1789–1800.
Hirsch, P.R., Mauchline, T.H., Clark, I.M., 2011. Culture-independent molecular techniques for soil microbial ecology. Soil Biol. Biochem. 42, 878–887.
Jackson, C.R., Harper, J.P., Willoughby, D., Rosen, E.E., Churchill, P.F., 1997. A simple, efficient method for separation of humic substances and DNA from environmental
samples. Appl. Environ. Microbiol. 63, 4993–4995.
Lee, S.Y., Bollinger, J., Bezdicek, D., Ogram, A., 1996. Estimation of the abundance of an
uncultured soil bacterial strain by a competitive quantitative PCR method. Appl.
Environ. Microbiol. 62, 3787–3793.
Macdonald, C.A., Clark, I.M., Hirsch, P.R., Zhao, F.-J., McGrath, S.P., 2011. Development of
a real-time PCR assay for detection and quantification of Rhizobium leguminosarum
bacteria and discrimination between different biovars in zinc-contaminated soil.
Appl. Environ. Microbiol. 77, 4626–4633.
Mumy, K.L., Findlay, R.H., 2004. Convenient determination of DNA extraction efficiency
using an external DNA recovery standard and quantitative–competitive PCR.
J. Microbiol. Methods 57, 259–268.
Muyzer, G., Waal, E.C.D., Uitterlinden, A.G., 1993. Profiling of complex microbial
populations by denaturing gradient gel electrophoresis analysis of polymerase
chain reaction-amplified genes coding for 16SrRNA. Appl. Environ. Microbiol. 59,
695–700.
Okano, Y., Hristova, K.R., Leutenegger, C.M., Jackson, L.E., Denison, R.F., Gebreyesus, B.,
Lebauer, D., Scow, K.M., 2004. Application of real-time PCR to study effects of ammonium on population size of ammonia-oxidizing bacteria in soil. Appl. Environ.
Microbiol. 70, 1008–1016.
Okubara, P.A., Schroeder, K.L., Paulitz, T.C., 2005. Real-time polymerase chain reaction:
applications to studies on soil-borne pathogens. Can. J. Plant. Pathol. 27, 300–313.
Park, J.W., Crowley, D.E., 2005. Normalization of soil DNA extraction for accurate quantification of target genes by real-time PCR and DGGE. Biotechniques 38, 579–585.
Pfaffl, M.W., Horgan, G.W., Dempfle, L., 2002. Relative expression software tool
(REST©) for group-wise comparison and statistical analysis of native expression
results in real-time PCR. Nucleic Acids Res. 30, e36.
Rose, P., Harkin, J.M., Hickey, W.J., 2003. Competitive touchdown PCR for estimation of
Escherichia coli DNA recovery in soil DNA extraction. J. Microbiol. Methods 52,
29–38.
Schutter, M.E., Dick, R.P., 2000. Comparison of fatty acid methyl ester (FAME) methods
for characterizing microbial communities. Soil Sci. Soc. Am. J. 64, 1659–1668.
Sharma, S., Radl, V., Hai, B., Kloos, K., Fuka, M.M., Engel, M., 2007. Quantification of functional genes from procaryotes in soil by PCR. J. Microbiol. Methods 68, 445–452.
Stubner, S., 2002. Enumeration of 16S rDNA of Desulfotomaculum lineage 1 in rice field
soil by real-time PCR with SybrGreen™ detection. J. Microbiol. Methods 50,
155–164.
Torsvik, V., Øvreås, L., 2002. Microbial diversity and function in soil: from genes to ecosystems. Curr. Opin. Microbiol. 5, 240–245.
Tsai, Y.L., Park, M.J., Olson, B.H., 1991. Rapid method for direct extraction of mRNA from
seeded soils. Appl. Environ. Microbiol. 57, 765–768.
Tung, W.L., Chow, K.C., 1995. A modified medium for efficient electrotransformation of
E. coli. Trends Genet. 11, 128–129.
Vallette, F., Mege, E., Reis, A., Adesnik, M., 1989. Construction of mutant and chimeric
genes using the polymerase chain reaction. Nucleic Acids Res. 17, 723–733.
Weinbauer, M.G., Fritz, I., Wenderoth, D.F., Höfle, M.G., 2002. Simultaneous extraction
from bacterioplankton of total RNA and DNA suitable for quantitative structure
and function analyses. Appl. Environ. Microbiol. 68, 1082–1087.
Zelles, L., 1999. Fatty acid patterns of phospholipids and lipopolysaccharides in the
characterisation of microbial communities in soil: a review. Biol. Fertil. Soils 29,
111–129.
Zeugin, J.A., Hartley, J.L., 1985. Ethanol precipitation of DNA. Focus, 7 (4). Bethesda
Research Laboratories/Life Technologies, Inc., Maryland, U.S.A.
Zhou, J., Bruns, M.A., Tiedje, J.M., 1996. DNA recovery from soils of diverse composition.
Appl. Environ. Microbiol. 62, 316–322.
Zogg, G.P., Zak, D.R., Ringelberg, D.B., MacDonald, N.W., Pregitzer, K.S., White, D.C.,
1997. Compositional and functional shifts in microbial communities due to soil
warming. Soil Sci. Soc. Am. J. 61, 475–481.