The most useful reference genes for standards in gene expression studies should be stably expressed over a range of experimental conditions and should be of wide phylogenetic distribution. Taking these requirements into account, a combined computational and experimental approach was devised to identify reference genes in the Acidithiobacilli (Figure 1). The strategy involves the following steps:

The expression of the 9 selected reference genes was analyzed in the five different experimental conditions described above by the methods of Vandesompele and Andersen as implemented in the Visual Basic Application geNorm 11 and Visual Basic Application NormFinder 7, respectively (Table 1). According to geNorm (which determines a gene expression stability value M to produce a rank where the best genes are those with the lowest M value) map, rpoC, era and gmk showed the least variability in expression in all conditions evaluated (range of expression of 1.029, 1.040, 1.046 and 1.067 fold respectively). The NormFinder approach (which enables identification of the single best genes in a ranking) showed era to be the most stable gene and rpoC and gmk to rank within the four most stably expressed genes. Slight differences observed between the two techniques are to be expected and can be explained by the way in which both methods analyze the data. geNorm selects the gene with the most stable expression independent of the expression of the other genes under analysis. NormFinder instead focuses on the genes with least intra- and inter-group expression variation, thus the selection of the best reference gene is affected by the other genes being analyzed. Similar differences between the results of geNorm and NormFinder have been reported in other studies 4243. Taking into account the results from both geNorm and NormFinder, we can conclude that rpoC and era are the most suitable reference genes for studies with the Acidithiobacilli.

Three genes, recA, alaS and rrs (16S rRNA) have been used in prior studies as internal controls for experimental investigations in A. ferrooxidans 28304445, but there has been no formal report showing that they are reliable references. Conversely, there is evidence showing differential expression of recA and rrs under cellular stress 46 and starvation 4547. In addition, use of rrs as a reference gene has been challenged because it is a very abundant species of RNA present at concentrations outside most calibration ranges 16.

The expression of recA, alaS and rrs was analyzed under the same experimental conditions and their expression values were ranked with respect to the new reference genes derived above using geNorm 11 and NormFinder 7 (Table 2). The variability in expression of rrs and alaS in five different experimental setups shows them to perform well although slightly poorer than rpoC, era and map (Figure 2). Both geNorm and NormFinder identified rrs and alaS among the six more stable genes. On the contrary, recA ranks further down, indicating less stable expression, independently of the ranking method used (Table 2). In addition, the expression of recA varies more than two fold, and together with coaE and trpS exhibits the least suitable expression profile of the genes assessed in this study (Figure 2).

It can be concluded that, among the previously used reference genes in A. ferrooxidans, only alaS is suitable and can be used with confidence as a normalizer. Given the variability observed in the present study, use of recA as a normalizer is not recommended as it would introduce noise to the analysis and eventually produce misleading results. In addition, use of rrs as a reference gene is not recommended despite its stable expression in the five conditions analyzed because its abundance prejudices the analysis of lowly abundant transcripts.

Normalizing gene expression based upon the expression levels of a carefully selected set of reference genes, usually referred to as a normalization factor, performs better than normalizing against any single gene alone 11. This raises the question as to whether a compromise could be discovered in which a pool of genes, fewer than nine but more than one, would provide greater confidence for use as a reference group in A. ferrooxidans. For this purpose, several normalization factors (NF) were calculated following the criteria defined by Vandesompele et al. 11. A NF represents the geometric mean of n genes, and the pairwise variation between sequential normalization factors (NF_n and NF_n+1) gives an idea of how well each of these perform. Geometric means for seven of most stable genes, showing less than 1.5 fold variation in all experimental conditions (map, rpoC, alaS, era, gyrA and nth) were obtained and pairwise variations between any two subsequent values was calculated. As shown in Figure 3, addition of a fourth gene leads to a non-significant change in the average of the gene variance estimates. According to geNorm ranking, it is concluded that the NF derived from the pool of the three genes map, rpoC and alaS (NF₁) is suitable for reliable normalization of gene expression of target genes. In spite of this we posit that era, which ranked first according to the Normfinder method, could also be included in the selected reference gene set.

To assess the value of our study, the relative expression levels of selected target genes was analyzed using the following normalization strategies: a) the three best reference genes selected by geNorm and Normfinder rpoC, era and alaS were used individually, b) a NF derived from the combination of the three genes selected by geNorm, rpoC, map and alaS (NF₁), c) a NF derived from the combination of the top ranking genes selected by geNorm and Normfinder method rpoC, map and era (NF₂) or d) the frequently cited reference genes rrs and recA. For this purpose, four target genes were selected that are known to be differentially expressed in A. ferrooxidans cultures: a) sdrAI (AFE0007) is 95 fold induced in iron 34, b) cyoB (AFE2407) is 12 fold induced in sulfur 22, c) cbbOIa (AFE1408) is 3 fold induced in iron 23 and d) mntH (AFE2920) is 24 fold induced in sulfur (unpublished results).

Expression of the four genes of interest in iron versus sulfur grown cells was evaluated using the relative expression analysis software qBase v1.3.5 48. Figure 4 shows a significant increase in the expression of the sdrA1 and cbbOIa genes in cells grown in ferrous iron and of the cyoB and mntH genes in cells grown in sulfur. In all cases, normalization by individual reference genes outlined here (rpoC, era or alaS), by the geNorm derived NF₁ (rpoC plus map plus alaS), by the combined NF₂ (rpoC plus map plus era) and by rrs gave similar results. For example, sdrA1 was up-regulated 60–100 fold depending on whether the normalization strategy was a single stable reference gene or the normalization factors. Conversely, normalization by the recA gene dramatically altered the relative expression ratio of the target gene, revealing an up-regulation of less than 50.

These results demonstrate how the interpretation of bacterial gene expression levels can be affected by the choice of the reference genes in quantitative real-time RT-PCR analysis. If a single gene is to be used, e.g. in studies where only one or a few target genes are being evaluated, the reference gene should be one of the three validated stable reference genes rpoC, era or alaS. In investigations where a larger number of target genes are to be evaluated or a higher degree of confidence is desired, use of the NF derived from the pool of the genes rpoC, map and alaS or era is advisable.

An initial set of candidate reference genes was compiled from a gene list that includes "housekeeping genes" of wide phylogenetic distribution 38. This set was used to textmine the A. ferrooxidans ATCC 23270 genome (GenBank/EMBL/DDBJ accession number CP001219) 39. Additional candidates were identified in the A. ferrooxidans genome by BLASTP and TBLASTX searches. The combined set of candidate genes was then used to formulate bidirectional BLASTP and TBLASTX searches of the genomes of A. thiooxidans and A. caldus. Genomic context for genes present in all three genomes was analyzed using the DNA sequence viewer and annotation tool Artemis v.10 40. Candidate genes belonging to the same predicted operon or gene cluster were excluded from further analysis. Reference genes were classified by function using TIGRfams 41 and one gene per functional category was selected. These last two steps were included to reduce redundancy in the gene set.

Gene expression was evaluated under the following experimental conditions: sulfur at pH 2.5, pH 3.5 and pH 4.5; thiosulfate at pH 4.5 and ferrous iron 200 mM pH 1.6. A. ferrooxidans strain ATCC 23270 was grown in modified 9 K basal salt media (0.7 mM (NH₄)₂SO₄, 0.2 mM K₂HPO₄, 1.6 mM MgSO₄.7H₂O) containing iron (9 K + Fe: 200 mM FeSO₄; adjusted to pH 1.6 with H₂SO₄) or sulfur (9 K + S: 1% ethanol-sterilized powdered sulfur, adjusted to pH 2.5; 3.5 and 4.5 with H₂SO₄). DSMZ71 medium was used for thiosulfate growth (20 mM Na₂S₂O₃.5H₂O, 22 mM KH₂PO₄, 2 mM MgSO₄.7H₂O, 22 mM (NH₄)₂SO₄ and 1.7 mM CaCl₂.2H₂O). All cultures were incubated at 30°C under aerobic conditions on a rotary shaker at 150 r.p.m.

A. ferrooxidans cultures to be used for nucleic acid purification were harvested at 8000 r.p.m. for 10 min. The cell pellet was washed in 9 K basal salt solution (adjusted at the corresponding pH). Washed cells were collected by centrifugation at 12000 r.p.m. for 10 min.

A. ferrooxidans cultures were grown for 72 h until stationary phase. DNA isolation was carried out by phenol-chloroform extraction. Briefly, cells were collected and resuspended in buffer TE (25:10) pH 8.0 with 5 mg/ml lysozyme, and incubated at 37°C for 30 minutes, followed by another hour of incubation at the same temperature with 1% SDS and 0.2 mg/ml proteinase K. Cell lysis was completed by alternate shifting of the suspension from 80°C to -80°C. DNA extraction was performed twice with a mixture of phenol:chloroform:isoamylic alcohol (25:24:1). Removal of the residual phenol was accomplished by one treatment with a mixture of chloroform:isoamylic alcohol (24:1). The DNA contained in the final aqueous phase was precipitated overnight at -20°C with absolute ethanol, washed with 70% ethanol, and finally resuspended in sterilized water. Genomic DNA quality and integrity were assessed by 1% (w/v) agarose gel electrophoresis and standard PCR, and concentration was determined by absorbance at 260 nm.

RNA was isolated from A. ferrooxidans mid-logarithmic cultures grown in modified 9 K basal salt medium in the presence of iron 200 mM, sulfur (1%, pH 2.5, 3.5 and 4.5) or 0.5% thiosulfate. Briefly, cells were collected and resuspended in ice-cold buffer TE (25:10) pH 8.0 with 1× Extraction Buffer (per liter: 1% SDS, 50 mM Tris-HCl pH 8.0, and 2 mM EDTA). Cell lysis was accomplished by incubation at 100°C for 5 minutes. The suspension was treated with TRIzol (Invitrogen), and the recovered aqueous phase was treated with chloroform followed by two extractions with acid phenol and chloroform. RNA was precipitated with absolute ethanol overnight at -20°C, washed with 70% ethanol, and finally resuspended in sterilized water. Samples were treated with DNase and purified with the Roche High Pure RNA Isolation Kit, following the manufacturer's recommendations and checked for DNA contamination by standard PCR, including a genomic DNA positive control. RNA quality was evaluated by 1.0% agarose gel electrophoresis and its concentration was measured by absorbance at 260 nm.

cDNA was prepared from 1 μg total RNA using random hexamers and Superscript II reverse transcriptase (Invitrogen) according to manufacturer instructions. The resulting cDNA was diluted 1:10 in distilled water and stored in aliquots at -20°C until further use.

Primers for real-time PCR assays and amplification efficiencies (E) are shown in Table 3. The real-time PCR reactions were performed in the Mx3000P QPCR System (Stratagene) using the SYBR GreenER qPCR SuperMix Universal Kit (Invitrogen). The 20 μl PCR reactions contained 2 μl of a 1:100 diluted cDNA sample; 200 nM of each primer and 1× SYBR GreenER qPCR SuperMix Universal (Invitrogen). The reference dye ROX was included at a final concentration of 5 nM. The cycling protocol was as follows: initial denaturation for 10 min at 95°C followed by 40 cycles of 30 s at 95°C, 15 s at 52°C; 30 s at 72°C. Fluorescence was measured after the extension phase at 72°C. The PCR products were subjected to a melting curve analysis, that commenced at 52°C and increased at 0.5°C s-1 up to 95°C, with a continuous fluorescent measurement. Specific amplification was confirmed by a single peak in the melting curve. For each experimental condition total RNA was extracted from two independent A. ferrooxidans cultures. Each RNA sample was retro-transcribed and the expression of all genes was assessed on the same cDNA sample. The reactions for each target gene were performed in triplicate and in the same PCR run. Thus, data sets consist of 6 values per gene per experimental set-up generated under standardized PCR cycling conditions. Stationary phase genomic DNA 10-fold dilutions (ranging from 10 ng to 1 pg) were used to generate a 5-point standard curve for every gene by using the Cycle Threshold (C_t) value versus the logarithm of each dilution factor. Reaction efficiency (E = [10(-1/slope)]-1) for every gene was derived from the slope of the corresponding standard curves. Transcript quantities were calculated from the standard curve by the software accompanying the MxPro3000P QPCR System (Stratagene) set with default parameters. Each experiment included a no template control.

The stability of gene expression was evaluated using the Excel-based applications geNorm 11 and Normfinder 7 and the relative expression was calculated with qBase 1.3.5 48. Briefly, the geNorm method is based on a pairwise comparison approach and depends on the calculation of an M value, defined as the average pairwise variation of a particular gene with all others. The NormFinder method is based on a different mathematic model that considers the intra- and inter-treatment variation in expression for gene ranking. The normalization factors were calculated following the criteria defined by Vandesompele 11. These include: a) to use the geometric mean (n numbers are multiplied and then the nth root of the resulting product is taken) as this controls better for possible outliers and abundance differences between genes and b) to define the minimum number of reference genes needed for a reliable calculation by evaluating the pairwise variation between sequential normalization factors including three, four, five or more stable reference genes (NF_n/NF_n+1).

A. ferrooxidans gene expression was evaluated under three experimental conditions: (1) cells were grown in 9 K medium containing 62 mM FeSO₄ at pH 1.6 versus cells grown in 9 K medium 1% elemental sulfur at pH 3.5 containing; (2) cells were grown in 9 K medium containing 62 mM FeSO₄ plus 1% elemental sulfur at pH 1.6 versus cells grown in 9 K medium containing 1% elemental sulfur at pH 1.6 and (3) cells were grown in 9 K medium containing 200 mM FeSO₄ at pH 1.6 versus cells grown in 9 K medium containing 62 mM FeSO₄ at pH 1.6. Construction, experimental and data analysis protocols for A. ferrooxidans type strain specific oligonucleotide microarrays have been previously described 22 and deposited in the ArrayExpress database under the following accession numbers (A-MEXP-1478, A-MEXP-1479).