Only mosquitoes identified as members of A. gambiae s.s. S form and homozygous for the L1014F kdr allele were included in the microarray study. Total RNA was extracted from pools of 10 mosquitoes which were either selected against 0.75% permethrin for the LT50 or not exposed to the insecticide. The quality and quantity of all RNA pools was measured by a spectrophotometer (Nanodrop Technologies) and a random subset was also assessed using a 2100 Bioanalyzer (Agilent Technologies). RNA extraction, amplification and labelling protocols followed those described in Müller et al.[21] (link). Labelled targets were hybridised to an updated version of the A. gambiae detox chip[11] (link),[21] (link) which was printed with a physical rearrangement of the probes (ArrayExpress accession A-MEXP-863). The probes on the microarray include 103 cytochrome P450s, 31 esterases, 35 glutathione S-transferases and 85 additional genes such as peroxidases, reductases, superoxide dismutases, ATP-binding cassette transporters, tissue specific genes and housekeeping genes.
The microarray experiment compared RNA pools from selected vs. unselected mosquitoes, comprising six independent replicates with dye-swaps (12 arrays in total). As each probe was spotted in replicates of four and measurements were obtained for both red and green wavelengths in each array, a total of 96 measurements per probe were obtained. After visual inspection of each array, spot and background intensities were calculated from the scanned array images using GenePix Pro 5.1 software (Axon Instruments). Raw intensities were then analysed with Limma 2.4 software package [27] running in R. Any spot that showed a median intensity in one or both channels at saturation was excluded from the analysis. For each spot background intensities were subtracted (i.e. method = “subtract”) from the total spot intensities and adjusted intensities were transformed into intensity log-ratios and normalised. For the comparison between the two groups, selected vs. unselected, estimates for technical replicates (dye-swaps) were first averaged and then compared between the two groups. A detailed description of the methods used for normalisation and statistical analysis is given in Müller et al.[12] . All microarray data has been deposited in ArrayExpress (accession E-MTAB-52).
In terms of absolute fold change our values are likely to underestimate true fold differences between mosquitoes that would survive an LT50 and those that would not. This is a result of the study design whereby the LT50 survivors were compared with a control group that would be expected to be a mixture of 50% mosquitoes surviving and 50% mosquitoes dying after exposure to 0.75% permethrin. It was not possible to select a fully susceptible control group due to the expected RNA degradation postmortem. The underestimation of fold changes may occur wherever resistant mosquitoes are compared with their parental line. Details of how this study design limits maximum fold change are given inFigure S1 . As a consequence we have chosen to rank our genes by statistical significance (i.e., −log10P-value) rather than setting an arbitrary fold change cut-off to filter for candidates.
The microarray experiment compared RNA pools from selected vs. unselected mosquitoes, comprising six independent replicates with dye-swaps (12 arrays in total). As each probe was spotted in replicates of four and measurements were obtained for both red and green wavelengths in each array, a total of 96 measurements per probe were obtained. After visual inspection of each array, spot and background intensities were calculated from the scanned array images using GenePix Pro 5.1 software (Axon Instruments). Raw intensities were then analysed with Limma 2.4 software package [27] running in R. Any spot that showed a median intensity in one or both channels at saturation was excluded from the analysis. For each spot background intensities were subtracted (i.e. method = “subtract”) from the total spot intensities and adjusted intensities were transformed into intensity log-ratios and normalised. For the comparison between the two groups, selected vs. unselected, estimates for technical replicates (dye-swaps) were first averaged and then compared between the two groups. A detailed description of the methods used for normalisation and statistical analysis is given in Müller et al.[12] . All microarray data has been deposited in ArrayExpress (accession E-MTAB-52).
In terms of absolute fold change our values are likely to underestimate true fold differences between mosquitoes that would survive an LT50 and those that would not. This is a result of the study design whereby the LT50 survivors were compared with a control group that would be expected to be a mixture of 50% mosquitoes surviving and 50% mosquitoes dying after exposure to 0.75% permethrin. It was not possible to select a fully susceptible control group due to the expected RNA degradation postmortem. The underestimation of fold changes may occur wherever resistant mosquitoes are compared with their parental line. Details of how this study design limits maximum fold change are given in
Free full text: Click here
