Descriptive statistics of the derived crossing points (CP), based on BestKeeper program 45, were calculated to investigate the variation level of each candidate gene following Pfaffl et al. 46. According to this analysis (see Table 1), the gene with lowest expression level was actin, for which CP values were obtained around cycles 31–34; while the highest was gapdh, whose CP values were obtained around cycles 21–23. The expression levels of 14-3-3, ccs, gapdh, and rpl7 presented fluctuations of approximately ± 0.6 x-fold (0.52 x-fold < SD < 0.82 x-fold), whereas poly expression showed higher ranges of CP variation (SD = ± 1.39 cycles) as well as up- down-regulation (± 2.09 x-fold). The coefficient of variation (CV) of the assay was 3.57% (total essay variability), which is within the range (from 3.4% to 11.6%) of previously reported values for qPCR 46.

Descriptive statistics for the expression analyses of each reference candidate gene in the five distinct coffee plant tissue/organ types was also obtained (see Additional File 1). According to this analysis, the gene that exhibit the minor gene expression variation among the analyzed tissues was gapdh [SD (± x-fold) = 0.04], while the gene with major variation was ccs [SD (± x-fold) = 0.59]. In addition, ccs also presented the highest expression variation in flowers and fruits tissue samples, and therefore it cannot be used as a reference gene. Numerous studies have shown that the expression of housekeeping genes can vary under given situations 28. This may partly be explained by the fact that housekeeping genes are not only implicated in the basal cell metabolism but also participate in other cellular functions 4748.

The model-based variance estimation approach, a Visual Basic application for Microsoft Excel (termed NormFinder) 849, was used to evaluate expression stability of reference candidate genes. This analysis allowed the ranking of candidate genes since the estimated variation directly indicates the introduced error associated with their use. According to this method, the gene with minimal estimated intra- and intertissue variation was gapdh (expression stability = 0.071) while the gene with the maximal variation was poly (expression stability = 0.592) (see Figure 1), thus corroborating the results obtained in descriptive analysis.

The relationship between the stability value and the intra- and intertissue expression variations is present in Figure 2. This figure clearly demonstrates the distinct specificities of the investigated genes. According to this analysis, the best candidate gene should present the minimal combined inter- and intra-tissue expression variation. Consistent with the descriptive analysis (see Table 1 and Additional File 1) and the model-based variance estimation approach (see Figure 1), gapdh showed not only the highest expression levels but was also the most stable gene studied. In Figure 2, it can be observed that almost all genes presented average of log expression levels near 0 (the thick dashed line). Log difference >0 (as observed for actin) or <0 (as observed for poly) implies that variability in expression levels is significant, so the gene could be incorrectly used as reference gene for normalization. In this context, the poly gene presents the highest intertissue expression variation (SD = ± 0.65; see Figure 2). Thus, among the tested genes, gapdh, followed by rpl7 and 14-3-3, showed the most stable expression over the investigated panel of five distinct coffee tissue/organ types.

The candidate genes were also ranked according to their M values using the geNorm program. The average expression stabilities (M values) of all tested genes were lower than 1.5, with 14-3-3 and actin showing the most stable expression (data not shown). Although actin gene has shown highest stability following geNorm analysis, this gene presented the lowest expression profile according the BestKeeper analysis. Corroborating the previous analysis, poly remained the least stable gene.

As a whole, our analysis indicates that housekeeping genes are differently regulated in different tissues/organs of coffee plants and may exhibit variable expression patterns. The observed differences in gene expression ratios along a comprehensive panel of tissues/organs are consistent with the data presented by Barber et al. 1, Iskandar et al. 50 and Jain et al. 15. Our results also provide evidences that normalizations to the expression level of a single gene in samples from distinct tissue types may induce to errors, thus corroborating previous studies 295152.

In order to validate our potential candidate reference genes (gapdh, rpl7 and 14-3-3), the expression stability of these genes under the influence of a specific biotic stress was investigated. In this case, the obtained results were compared to those dealing with similar coffee gene expression analysis but using ubiquitin as a reference gene for normalization 53545556.

The comparison was conduced by linear regressions analyzes of the CP difference (ΔCP) obtained from the assayed expression levels of the tested genes in leaves of C. arabica inoculated, or not (control plants), with Hemileia vastatrix. The average CP (N = 3) was calculated for each gene and the ΔCP (CP_{inoculated leaf} - CP_{non-inoculated leaf}) was determined for each time-point (8, 12 and 24 h after fungus inoculation).

As it can be observed in Figure 3, the regression lines for 14-3-3 and gapdh have slopes close to zero (14-3-3 = 0.085; gapdh = 0.18), indicating similar expression levels in inoculated and non-inoculated leaves, and reinforcing their use as effective normalization genes 15758. In contrast, the slope value for rpl7 was significantly different from zero and higher than the one obtained for ubiquitin (ubiquitin = 0.2467; rpl7 = 0.5389), thus limiting its use as a normalization factor under biotic stress condition.

An additional validation step of the expression levels of gapdh, 14-3-3, rpl7 and ubiquitin was performed using unpooled tissue samples from flower and cherry developmental series. The employed sample set is given in the Additional File 2.

In this assay, the highly expressed gene was gapdh (CP¯ MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGaciGaaiaabeqaaeqabiWaaaGcbaWaa0aaaeaacqqGdbWqcqqGqbauaaaaaa@2E1A@ = 21 cycles) followed by ubiquitin (CP¯ MathType@MTEF@5@5@+=feaagaart1ev2aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGaciGaaiaabeqaaeqabiWaaaGcbaWaa0aaaeaacqqGdbWqcqqGqbauaaaaaa@2E1A@ = 24 cycles), while 14-3-3 and rpl7 presented the same mean CP (25 cycles). The comparison of gene contributions is present in Figure 4. As already observed, gapdh showed the greatest stability in expression among all coffee tissue/organ samples analyzed, while expression of rpl7 and 14-3-3 varied the most, especially in flowers at stage 1 of development (see Figure 4 and Additional File 2). As mentioned earlier, ubiquitin was used as a standard reference gene for comparison.

According to these results, the gene encoding glyceraldeyde-3-phosphate dehydrogenase (GAPDH), an enzyme of glycolysis 59, outperformed all other reference genes tested and should therefore be considered a suitable reference gene for expression studies in arabica coffee plants. This observation corroborates the quantification of gapdh expression in different tissues of sugarcane 50 and Eucalyptus 20. In contrast, several reports in human and animal systems have suggested that this reference gene has limitations for its use as internal control due to its marked variability of expression among tissue types 1165160.

The assayed 14-3-3 gene also showed a stable expression (see Figure 2) that was supported by the descriptive analysis (see Table 1 and also Additional File 1) and confirmed in the biotic stress assay (see Figure 3). However, its expression presented some variation among different stages of coffee flower development (see Figure 4). Papini-Terzi et al. 19 recommended 14-3-3 as a suitable reference gene for expression normalization in a wide range of tissue samples of sugarcane.

The gene encoding the transcription regulator and structural constituent of the 60S subunit of the cytosolic ribosome (rpl7) could also be used as an internal control in gene expression studies in C. arabica, due to its stability (see Figure 2) and acceptable variation among tissue/organ types (see Table 1). Our results are in agreement with previous published data for this gene since small variation among tissue types was detected by descriptive analysis (see Additional File 1). Nevertheless, in leaves of C. arabica inoculated with H. vastatrix, it was observed that the expression ratio of rpl7 (expressed by ΔCP) was not constant and the absolute value of rpl7 linear regression slope was superior to that observed for a commonly used coffee normalization gene (ubiquitin) (see Figure 3). In addition, this gene, like 14-3-3, presented a variable expression level among different stages of coffee flower development (see Figure 4). In sugarcane, the relative expression of rpl35-4, a gene coding for the ribosomal L35-4 60S protein, was also reported to be stable 61. These authors estimated the sugarcane leaf transcriptome using Serial Analysis of Gene Expression (SAGE) and reported that tag associated with the rpl35-4 transcript presented minimum variation among the analyzed SAGE libraries.

The remaining tested genes showed to be unsuitable as internal controls for normalization purposes in C. arabica.

Freshly harvested roots, stems, and leaves were obtained from ten 4 month-old coffee plants (Coffea arabica var. Mundo Novo IAC 388-17-1) grown under greenhouse conditions (28°C, 60% RH) in Campinas, São Paulo, Brazil. Flower and fruit samples, at different developmental stages, were collected from 4–5 year-old plants of var. Mundo Novo grown under field conditions at Botucatu and Campinas, São Paulo, Brazil (see Additional File 2). After harvesting, fresh tissue samples were frozen immediately in liquid nitrogen until RNA extraction.

For the biotic stress assay, equally-aged sets of Coffea arabica var. Mundo Novo plants were kept in growth chamber (16 h/8 h light/dark; 23°C; 70% RH) for at least one week, before being inoculated with the coffee leaf-rust fungus Hemileia vastatrix Berk and Br. race II, that elicits a compatible reaction in coffee. The urediniospores (100 mg) were harvested in a C. arabica field in Campinas, São Paulo, Brazil, and diluted in 10 ml of sterile water under dark conditions.

Leaves from the second pair of plagiotropic shoots from the apex of 4 month-old coffee plants were inoculated with an aqueous suspension of fresh urediniospores (10 mg/ml). To allow spore germination, the inoculated leaves were covered with a wet black plastic film for 24 h. Inoculated leaves were not detached from the plants. Mock-inoculated controls as well as non-inoculated controls were performed. The biological samples were obtained from three independent experiments.

Leaves were randomly sampled at different time-points after inoculation: 0, 8, 12, and 24 h, and immediately deep-frozen. To confirm the infection by the leaf-rust fungus, some inoculated leaves were maintained in plants.

Tissue samples of 2.0 to 2.5 mg were weighed and ground to fine powder in liquid nitrogen using a pre-cooled mortar and pestle. Total RNA from the majority of the samples was extracted using TRIzol reagent (Invitrogen) according to manufacturer's instructions. Alternatively, total RNA from seeds was isolated by lithium chloride (LiCl) method, according to Mason and Schmidt 62. Only RNA samples with 260/280 ratio between 1.9 and 2.1 and 260/230 ratio greater than 2.0 were used for subsequent analyses. The integrity of the RNA samples was also assessed on 1.0% agarose/formaldehyde gel electrophoresis.

Five micrograms of total RNA were treated with DNAse I (Promega) and an aliquot of 500 ng of the treated RNA was reverse-transcribed using SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen). Both were used following the manufacturer's instructions. The cDNA sample concentration was determined using the NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies).

Primers were designed for Coffea orthologs of commonly used housekeeping genes representing distinct functional classes, identified by BLAST searches in the Brazilian Coffee EST database 6364 as well as in the public coffee EST database (at the SOL site hosted by Cornell University 65). For primer design, the Primer Express 2.0 software (PE Applied Biosystems, USA) with default parameters was employed. The accession numbers, gene description, primers sequences and amplicon lengths are shown in Table 2. All primer pairs produced a single product and amplified the target transcript with equal efficiency over a 1000-fold range of input material.

The PCR mixture contained 5 μl of a 1:10 dilution of the synthesized cDNA, primers to a final concentration of 700 nM each, 17.5 μl of the SYBR Green PCR Master Mix (Applied Biosystems, USA) and PCR-grade water up to a total volume of 35 μl. The mixes were homogenized and split in three samples of 10 μl, thus each gene reaction was performed in triplicate. PCR reactions in the absence of template were also performed as negative controls for each primer pair. An equimolar pool of cDNA samples of five coffee tissue/organ types (root, stem, leaf, flower, and fruit) was prepared to be used as a common reference for all qPCR. The quantitative PCRs were performed employing the ABI Prism 7300 Sequence Detection System (PE Applied Biosystems, USA). All PCR reactions were performed under the following conditions: 2 min at 50°C, 10 min at 95°C, and 45 cycles of 15 s at 95°C and 1 min at 65°C in 96-well optical reaction plates (Applied Biosystems, USA). Confirmation of amplicon specificity was based on the dissociation curve at the end of each run and by product visualization after electrophoresis on an 8% polyacrylamide gel.

To estimate the expression variation level of the eight candidate genes (actin, adh, ccs, 14-3-3, gapdh, poly, rpl7, or cys) over a five coffee tissue/organ sample set (root, stem, leaf, flower, or fruit), the BestKeeper descriptive statistical method 4566 was applied.

To access the levels of gene expression for each gene in the different coffee tissue types, the method described by Ramakers et al. 67 with modifications was used. Optic data were exported from 7300 Real-Time PCR System (PE Applied Biosystems, USA) into MS Excel. Four cycles at the exponential phase, near and including the crossing point (CP), were used. The fluorescence data were logarithmically transformed, and pasted into statistical software package (SAS version 8e, SAS Institute, Cary, NC, USA) for linear regression analysis, including determination of intercepts, slopes (x), PCR efficiency (E = 10^slope) and their respective standard errors and correlation coefficients (R²). The gene expression values (x-fold) were obtained according to a mathematical model proposed by Pfaffl 68: x-fold = E_{reference gene} ^ΔCP, where ΔCP = CP_{pool of tissues} - CP_{tissue sample}.

Knowing the expression levels, the stability value was assessed utilizing NormFinder 849 and geNorm 1769 programs.