Plasmodium falciparum isolates
Between January 2000 and December 2010, 800 P. falciparum isolates were obtained from the different health centres of French Guiana and collected
by the CNRCP (Centre National de Référence pour la Chimiosensibilité du Paludisme)
hosted by the parasitology laboratory of the Institut Pasteur de la Guyane. Fifty
per cent inhibitory concentration (IC
50
) to doxycycline was determined using the ex vivo isotopic method described by Le
Bras et al. 19]. DNA was extracted from blood samples by QIAamp
®
DNA Blood (Qiagen) according to the manufacturer’s protocol.
Distribution and range of IC
50
The statistical analysis was designed to answer the specific question of whether P. falciparum has a different profile of susceptibility to doxycycline. Parasite susceptibility
is expressed as the IC
50
. As a heterogeneous population is observed, the data are assumed to come from a univariate
Gaussian mixture with k components. Each observation is assumed to come from one of
the k components, and the label of the group from which each observation comes is
unknown. The number of components, the means, variances and weights of the different
components in the model are unknown, as well as the vector of allocations of the observations.
The analysis was performed in two steps. First, reversible jump Monte Carlo Markov
Chains (RJMCMC) samplers were used to choose a suitable number of k components. The
RJMCMC sampler is described by Richardson et al. 20]. The only difference with the algorithm is that we implemented only birth and death
moves, following Cappé et al.’s recommendations 21]. Once a relevant number of components had been chosen, standard Gibbs samplers were
run to obtain estimates of the model parameters and classify the observations 22]. It is well known that these classical Markov Chain Monte Carlo techniques are not
sufficient to cover all the parameter space; it can stay within a neighbourhood of
a local mode and may fail to visit other important modes. In order to improve the
exploration of the parameter space and thereby improve convergence, the RJMCMC and
Gibbs samplers were embedded in a population-based algorithm. Because of the ‘label-switching’
problem caused by symmetry in the likelihood of the model parameters 23], the mixture components should be labelled before making inferences about the parameters.
A classical ordering constraint was used, which is biologically relevant here. The
algorithms were run for 50,000 burn-in iterations and 20,000 post-burn-in iterations.
These numbers are classically sufficient to obtain reliable results. Moreover, each
algorithm was run three times to check that results between two different runs were
similar and that there was no convergence problem. This analysis was performed with
R
®
software (version 2.10.1).
Quantification of the pfmdt and pftetQ gene copy numbers by TaqMan
®
real-time PCR
TaqMan
®
real-time polymerase chain reaction (PCR) was performed using ABI7300 (Applied Biosystems)
to estimate the pfmdt (PFE08254w) and pftetQ (PFL1710c) gene copy numbers relative to the single-copy of pf?tubulin gene (PF10_0084) in 129 isolates from the different phenotypic subpopulations.
Primers and probes were used as describe by Briolant et al. 17]. Individual PCR were performed using 1X TaqMan Universal PCR Master Mix (Applied
Biosystems), 6 µM forward and reverse primers, 4 µM probe, and 1 µL of template DNA
in a final volume of 25 µL. The thermal cycling conditions were 50°C for 2 min, 95°C
for 10 min, then 50 cycles of 95°C for 15 s and 60°C for 1 min for pfmdt, and 55°C for 10 s followed by 60°C for 1 min for pftetQ. Each sample was assayed in triplicate and analysed with SDS software (version 1.3;
Applied Biosystems). The PCR efficiencies of all primers pairs were evaluated using
a dilution series of P. falciparum 3D7 genomic DNA. The method of relative quantification, where C
T
indicates the threshold cycle, was used to estimate the copy numbers of pfmdt and pftetQ genes with the formula 24]. Genomic DNA extracted from P. falciparum 3D7, which has a single copy of each gene, was used for calibration, whereas pf?tubulin served as the control housekeeping gene in all experiments. Data were analysed using
Epi info
®
software (version 3.5.1). Differences in the pfmdt and pftetQ gene copy numbers between the phenotypic groups were tested using the Kruskal–Wallis
test. Genotype proportions were compared using the Fisher exact test. Statistical
significance was denoted by p ? 0.05.
Gene sequence polymorphism analysis
Four genes, pftetQ, pfssurRNA, pflsurRNA and pfrps7, were amplified by PCR with oligonucleotide primer pairs designed using Primer 3
®
software (version 0.4.0) (Table 1). As there was only a few isolates with an IC
50
exceeding the threshold, it was initially chosen to sequence genes from only ten isolates
with the highest IC
50
[median 40.15 µM (min 29.82 µM; max 77.84 µM)] and ten isolates with low IC
50
to doxycycline [median 4.83 µM (min 3.28; max 6.05)]. The reaction mixture consisted
of 10× mix PCR Gold buffer
®
, 200 µM of dNTPs, variable concentration of MgCl
2
(Table 1), 200 nM of forward and reverse primers, 4 UI/100 µL of Ampli Taq Gold DNA polymerase
(Applied Biosystems) and 2 µL of DNA in a final volume of 50 µL. Thermal Mastercycler
®
(Eppendorf) were programmed as follows: 94°C for 10 min then 35 cycles alternating
94°C for 30 s, hybridization temperature (Table 1) for 30 s and 72°C for extension at 1 min per 1,000 bp, followed by a 15-min final
extension step at 72°C. The reaction products were sequenced by Millegen
®
biotechnologies (Labège, France). Sequences were analysed using MEGA
®
software (version 5.05). Epi Info
®
software (version 3.5.1) was used to perform data analysis. Differences in DNA sequences
(pfssurRNA, pflsurRNA, pfrps7, pftetQ) and in amino acid sequences [PfTetQ, PfRps7—itals—as before ? (No because is official
nomenclature)] between the two groups were tested using the Kruskal–Wallis test.
Table 1. Forward and reverse primers, hybridization temperature and MgCL
2
concentration used for polymerase chain reaction and sequence analysis