Knowledge of mRNA transcription levels in response to crowding is central to understanding the molecular mechanisms and evolution of density-dependent phase polyphenism in locusts. Now that high-throughput systems are becoming more widely available, gene expression studies and in particular the use of RT-qPCR will become even more important for molecular genetics and evolutionary research in locusts. However, successful RT-qPCR experiments require multiple, carefully validated, reference genes to allow for reliable data normalization and determination of differentially expressed genes.
In this study, we identified eight appropriate reference genes that show invariant expression in neural tissue of the Australian plague locust, whether reared in a solitary state or exposed to different durations of crowding. All of the genes we evaluated were classified as stably expressed as each presented with geNorm stability indexes well below the cut-off of 1.5 . In this study, we provide a comprehensive assessment of these reference genes for use in subsequent RT-qPCR assays.
Amongst the eight candidate housekeeping genes evaluated, Actin and EF1a were selected as the most stable pair by both geNorm and NormFinder. Although Vandesompele et al.  recommend the use of at least three reference genes for reliable normalization, the pairwise variation analysis in geNorm indicated that there is no need to include more than two genes. If we assume that stability is likely to be consistent across locust species, our results may serve as a posteriori validation of the use of Actin as a reference gene in previous studies of rearing density effect on locust gene expression patterns [5, 6, 8]. When comparing our results to the study of reference genes in brains of the desert locust, S. gregaria, we see that EF1a was also validated as stable with age in brains of fifth- instar nymphs and adults of the desert locust, S. gregaria. In contrast, Actin was excluded for normalization of transcriptional studies in brains of ageing adults. Interestingly, despite being amongst the most abundant proteins in all eukaryotic cells, with key roles in cell motility and cytoskeleton maintenance, Actins have been shown to vary in transcript levels with growth and ageing, at least in mammals . The different ranking of the Actin gene across studies underlines the necessity for validation of the candidate reference genes prior to an experiment utilizing different treatments.
The 18S ribosomal subunit was classified as the third most stable gene in our study (Table 3). However, caution should be exercised when using 18SrRNA as a reference control. First, as ribosomal subunits are not polyadenylated they cannot be exploited in purified mRNA samples, i.e. in Reverse-Transcriptase reactions primed with oligo-dT. This was addressed in this experiment by using random hexamer primers only. Second, the high abundance of rRNA molecules compared with target mRNA transcripts, as revealed here by relatively low Cq values for 18SrRNA, makes it difficult to subtract the baseline value in fluorescence data analysis . In the present study, accurate fluorescence treatment was only possible by a large dilution of the RT-qPCR template (i.e. 100-fold, equivalent of 0.25 ng total RNA). Furthermore, because 18SrRNA is so abundant, representing up to 80% of cellular RNA, it is expected that variations from sample-to-sample in initial sample amount, mRNA recovery and integrity, and reverse transcription efficiency, will be more difficult to detect. Furthermore, 18SrRNA is likely to experience RT-qPCR kinetics that are different to those of less abundantly expressed genes . As a consequence of these factors, 18S rRNA can reduce the sensitivity to detect variation in relative levels of expression . Because it is more likely to be expressed at a comparable order of magnitude as the investigated targets, we suggest that Arm will be a best suitable third normalizer in gene expression studies of locust neural tissues that address either the initiation or the maintenance of phase differences.
In our study, the two least stable genes were GAPDH and SDHa that code for the metabolic pathways enzymes, glycolysis and citrate cycle, respectively. Kristensen et al.  showed in Drosophila melanogaster that genes involved in metabolic processes are differentially expressed in response to heat stress. In particular, both glyceraldehyde 3 phosphate and succinate dehydrogenases were down-regulated at 36°C versus 25°C. This indicates that housekeeping metabolism is a dynamic process that can be regulated under different physiological states, thus challenging the usefulness of genes involved in metabolic processes in normalization. Crowded conditions themselves might be predicted to impose physiological stresses that can affect gene transcription. Accordingly, heat shock proteins, which are synthesized in response to stress in insects and act as molecular chaperones to mediate numerous cellular functions, have been shown to be up-regulated in heads of crowded gregarious phase migratory locusts .
RT-qPCR reactions showed a linear concentration-dilution relationship and were highly efficient, as every cycle of reaction amplification increased the amount of DNA template by 1.90-2 fold. The two exceptions were Armadillo and Actin primer pairs which respectively showed a moderate and low 1.81 and 1.66 fold increase per cycle. An exponential amplification of 2 (±10%), as retrieved for most of primer pairs, is needed when using the delta-delta-Cq quantification model as originally described by Livak and Schmittgen . However, the most common efficiency-corrected methods currently in use for relative quantification  do not assume that all targets amplify with the same optimal PCR efficiency. Instead, efficiencies must be calculated prior to quantification, be reproducible and preferably be validated in each target gene RT-qPCR run. Our reaction efficiencies were shown to be reproducible, and therefore should be the same between standards and all of the samples in each of the future qPCR assays. However, we suggest to adopt the more conservative approach of avoiding low efficiency primer pairs (i.e. Actin), which would suggest EF1a and Arm as the best combination of two reference genes. These reference genes do not present any risk of misleading gDNA amplification.
We applied the use of EF1a and Arm for accurate normalization in gene expression studies of locust neural tissue by studying the target gene SPARC shown to be up-regulated in gregarious brains of desert locusts. We showed that 24 hours of crowding did not change the expression of SPARC in brains and thoracic ganglia of the Australian plague locust. Similarly, no expression change was detected with longer crowding. Similar results were found when using Actin as sole normalizer following Rahman et al. . The discrepancy in SPARC expression between this and earlier studies may have arisen from other differences in experimental procedures; in particular, neural tissues were dissected from 4-5 day old adults in Rahman et al. , as opposed to 2-day old 5th-instar nymphs in our study, and SPARC expression is known to be developmentally regulated (reviewed in ). We might also expect S. gregaria and C. terminifera to differ in the density-dependence regulation of the glycoprotein, as their high density gregarious phase express different suites of behavioural [13, 14] and morphological traits . However, since the role of the SPARC protein with respect to density-dependent changes in behaviour or any other trait remain unknown, this alternative cannot be examined further as part of this study.
The reference genes we selected for gene expression normalization will provide a promising starting point for further studies of density-dependent transcriptional changes in the Australian plague locust. Now that appropriate primers had been identified for a range of potential reference genes, validation tests for other gene expression studies involving experimental treatments other than rearing density can also be performed in a relatively straightforward manner. Our findings should also facilitate the selection and use of suitable reference genes in the related migratory locust, L. migratoria, as the primers we used target the cDNA sequences of this species too. The possibility also remains that the primer sequences presented in this study are conserved across other members of the Oedipodinae which includes at least seven other species that show elements of density-dependent phase polyphenism .