9-14-06 TABASHNIK’S ANSWERS TO MATTEN’S QUESTIONS RE: PCR

FROM SHARLENE 9-8-06

Bruce -- 

Here are some questions of clarifications (I send most of these to Tish

yesterday, who by the way, has been very helpful.)

1.  Is the Molecular analysis method in the submission to EPA, the one

published in Tabashnik et al. 2006?  If so, I need an e-mail back to me

identifying that this is indeed the method used.

Yes.  The molecular analysis method in the submission to EPA is the one
published in Tabashnik et al. 2006.  Key portions of the paper
describing the method are provided below.  Please note that initial
tests use only a small portion of each field-sampled individual
(Tabashnik et al. 2005b), so that re-testing of individuals is possible
if desired.

INTRODUCTION

“In laboratory-selected strains of pink bollworm and at least two
other major lepidopteran pests of cotton, mutations in a cadherin gene
are tightly linked with recessive resistance to Cry1Ac (Gahan et al.
2001, Morin et al. 2003, Xu et al. 2005).  In several
laboratory-selected strains of pink bollworm, three mutant alleles (r1,
r2, and r3) of a cadherin gene (BtR) are associated with resistance to
Cry1Ac and survival on Bt cotton (Morin et al. 2003, 2004; Tabashnik et
al. 2004, 2005b).  Each r allele has a deletion predicted to eliminate
at least eight amino acids upstream of the putative Cry1Ac-binding
region of cadherin protein (Morin et al. 2003).  We previously developed
a PCR-based method for detecting the r1, r2 and r3 alleles in pink
bollworm (Morin et al. 2004).  We isolated, cloned and sequenced the
genomic region spanning the mutation in each r allele and designed
allele-specific PCR primers for each region.  The method can detect any
of the three r alleles in a single heterozygote (r1s, r2s, or r3s)
pooled with DNA from the equivalent of 19 susceptible (ss) individuals
(Morin et al. 2004).”

METHODS:  “DNA Preparation and PCR.  Insects collected from bolls and
traps were stored in ethanol at –20oC.  DNA was extracted using DNAzol
(Tabashnik et al. 2005b) and PCR was done as described by Morin et al.
(2004).  The maximum number of individuals tested per pool was 5 for
samples from 2001-2003 and 11 for samples from 2004-2005.”

2.  The late season sampling method developed by Dennehy and Tabashnik

-- I don't have the specific protocol.  Please send it or perhaps it is

in the Tabashnik et al. 2006 manuscript. I see a couple of sentences

describing the plan, is this it?  Similar to Tabashnik et al. (2006),
details coming soon.

3.  What is the method for estimating false negatives? false positives?

What is the likelihood of non-detection?  Please clarify.

The methods for estimating false negatives, false positives, and the
likelihood of non-detection are detailed in Morin et al. (2004) and
Tabashnik et al. (2006), as well as below:

A. False negatives.  

False negatives are possible from three causes: i) The PCR reaction is
not working properly, ii) The cadherin DNA of field-sampled insects is
not amplified, iii) The PCR is working and cadherin DNA is amplified
from field samples, yet r alleles are present and are not detected.

i) To determine if the PCR reaction is working properly, we use known
positive controls in every set of samples tested.  This is a standard
method.  Known positive controls are samples of DNA from our
laboratory-reared strains that contain r alleles, which are known to the
person running the PCR reaction.  For example, every test of
field-sampled insects for the presence of an r1 allele includes a gel
lane in which DNA from one or more laboratory-reared individuals with
the r1 allele is run simultaneously with the field samples.  

If the known sample of r1 DNA does not yield a positive result for r1,
the test of the field sample is not valid and must be repeated.  In this
case, PCR reaction conditions are corrected until the known controls
yield positive results with the simultaneously tested field samples. 
Such corrections usually involve systematic replacement of reagents
(primers, Taq, etc.) to ensure all are working properly.  Because only
tests yielding positive results for known positive controls are included
in our analysis of the data, this source of false positives has an
effective rate of 0% in the data analysis.

ii) To determine if the cadherin DNA of field-sampled insects is
amplified, we test for amplification of a conserved region of the
cadherin gene that occurs in all known susceptible and resistant alleles
(Morin et al. 2004).  As described in Tabashnik et al. (2006), “We
checked all pools using this approach and >99% tested positive.  Because
as few as one amplifiable allele from a pool of insects could yield a
positive result for this control reaction, we also tested a subset of
insects individually from each of the 59 field samples.  Of the 835
individuals tested, 98.6% were positive.”  

The 98.6% amplification rate of the conserved region of the cadherin
gene indicates that DNA was not amplified from 1.4% of field-sampled
insects.  We take this into account in estimating the likelihood of
non-detection by adjusting the sample size accordingly.  For example, if
1000 alleles are screened from 500 individuals and the amplification
rate of the conserved region is 98.6%, the corrected sample size is 986
(see C below).

iii) As described in Tabashnik et al. (2006): “In addition to standard
positive controls for each of the three r alleles in all tests, we
included “blind” positive controls as follows:  Two researchers
analyzed each field sample.  One researcher prepared DNA and added
individuals with one or two r alleles from laboratory-selected resistant
strains in zero to three (usually one) of the pools tested from each
field site.  The other researcher performed PCR and did not know which,
if any, of the pools contained these blind positive controls.  The
detection rate for blind positive controls was 97% (97/100).”  The
rate of false negatives (3%) caused by failure to detect r alleles
present in pools is incorporated in the estimate of the likelihood of
non-detection, as described below (C).

B. False positives.  To detect false positives, we use a standard
technique.  All tests of field samples include blanks, which are gel
lanes containing all of the PCR reagents, but no DNA.  If a blank yields
a positive result, this indicates contamination (i.e., a false
positive).  In this case, PCR reaction conditions are corrected and the
field samples are retested.  Results are included in the data analysis
only if the blanks do not yield positive results.  

Of the 5,571 field-sampled insects tested in Tabashnik et al. (2006),
none yielded positive results.  Thus, the problem of false positives is
minimal to nil.  When a pool of field-sampled insects yields a positive
result for an r allele (e.g., r2), each individual in the pool will be
tested separately to verify the positive result and to more precisely
estimate the frequency of resistance in the pool.

C. Non-detection.  As described in Tabashnik et al. (2006), the
likelihood of non-detection is estimated as follows:

“The probability of detecting no r alleles in a sample of N
individuals was calculated as (1-[F X D])2N X A, where F is the
frequency of resistance alleles, D is the probability of detecting an r
allele present in screened individuals (0.97, based on the data from
blind controls), 2N is the number of alleles screened, and A is the
probability of amplifiable cadherin DNA occurring in field-sample
insects (estimated as 0.986, based on the proportion of positive results
for amplification of a conserved sequence in 835 insects tested
individually).  We assumed that the probability of an r allele occurring
was an independent event at each cadherin allele screened.  For example,
with an r allele frequency of 0.001, the probability of detecting no r
alleles in the sample of 5,571 individuals (11,142 alleles) is 0.000023
= (1- [0.001 X 0.97])11,142 X 0.986.  Analogously, with an r allele
frequency of 0.0003, the probability of detecting no r alleles in the
sample of 5,571 individuals is 0.041 = (1 – [0.0003 X 0.97] )11,142 X
0.986.”

The goal in 2006 is to screen 500 field-sampled individuals with PCR
(i.e, N=500).  Assuming no r alleles are detected and values for D and A
similar to those above, the probability (P) of non-detection is
estimated as:

i) for true r allele frequency of 0.00316 (frequency of rr = 0.00001), 

P = (1- [0.00316 X 0.97])1000 X 0.986 = 0.048

ii) for true r allele frequency of 0.01 (frequency of rr = 0.0001), 

P = (1- [0.01 X 0.97])1000 X 0.986 = 0.000067

iii) for true r allele frequency of 0.001 (frequency of rr = 0.000001), 

P = (1- [0.001 X 0.97])1000 X 0.986 = 0.38

Below please find additional discussion of the potential for
non-detection from Tabashnik et al. (2006):

“It is important to consider potential underestimation of r allele
frequency based on DNA screening.  DNA screening based solely on males
caught in pheromone traps could cause underestimation if the probability
of capture in traps was lower for rr or rs males than for ss males. 
However,  tests conducted in large cages (64 m3) in the field refuted
this hypothesis for pink bollworm (Carrière et al. 2006).  Furthermore,
DNA screening of pink bollworm from bolls, which was independent of
males caught in traps, also detected no r alleles (n = 1,344; Table 1).

If alleles other than cadherin mutants r1, r2, and r3 confer pink
bollworm resistance to Bt cotton, the results of our DNA screening could
underestimate the frequency of resistance.  For example, resistance to
Cry1Ac in some strains of diamondback moth is not linked with cadherin
(Baxter et al. 2005).  However,  in four laboratory-selected
Cry1Ac-resistant strains of pink bollworm tested so far, all resistant
individuals screened have two copies of the known r alleles (i.e., r1r1,
r2r2, r3r3, r1r2, r1r3 or r2r3) and no other resistant alleles have been
detected (Morin et al. 2003, Tabashnik et al. 2004, 2005b).  Although
the presence of additional resistance alleles at the cadherin locus or
other loci cannot be excluded, such alleles appear to be more rare than
the three known resistance alleles.”

 

