Signalling with retinoids in the human lung: validation of new tools for the expression study of retinoid receptors

Background Retinoid Receptors are involved in development and cell homeostasis. Alterations of their expressions have been observed in lung cancer. However, retinoid chemoprevention trials in populations at risk to develop such tumors have failed. Therefore, the pertinence of new clinical trials using second generation retinoid requires prior better understanding of retinoid signalling. This is our aim when validating extensively research tools, focused on Retinoic Acid Receptor beta, whose major role in lung cancer is documented. Methods Biocomputing was used to assess the genomic organization of RAR beta. Its putative RAR-beta1' promoter features were investigated experimentally. Specific measures realized, with qRT-PCR Syber Green assays and a triplex of Taqman probes, were extensively validated to establish Retinoid Receptors mRNAs reference values for in vivo normal human bronchial cells, lung tumors and cell lines. Finally, a pan-RAR-beta antibody was generated and extensively validated by western-blot and immunoprecipitation. Results No promoter-like activity was found for RAR-beta1'. RAR-beta2 mRNAs increase signs the normal differentiation of the human bronchial epithelium while a decrease is observed in most lung cancer cell lines. Accordingly, it is also, along with RXR beta, down-regulated in lung tumors. When using nuclear extracts of BEAS-2B and normal lung cells, only the RAR-beta2 long protein isoform was recognized by our antibody. Conclusion Rigorous samples processing and extensive biocomputing, were the key factors for this study. mRNA reference values and validated tools can now be used to advance researches on retinoid signalling in the lung.


Background
Vitamin A and its active derivatives referred to as retinoids are non-steroid hormones which play a critical role in the development and homeostasis of vertebrate tissues [1][2][3]. They exert their actions by regulating the expression of target genes, influencing likewise: cell proliferation, differentiation and apoptosis through nuclear Retinoid Receptors (RRs) binding [4]. Six different RRs are encoded by separate genes (Retinoic Acid Receptors (RAR) α, β, γ and Retinoid X Receptors (RXR) α, β, γ) with at least two isoforms for each, depending upon promoter usage and alternative splicing [5]. RAR/RXR heterodimers are the functional units transducing the retinoid signal when binding Retinoic Acid (RA), co-activators and DNA response elements (RAREs) of target genes [6,7]. RXRs form homodimers and/or heterodimers with RARs and most other nuclear receptors for extended effects [8,9].
Alterations in RRs expressions are observed in several cancers. The RARα gene is involved in acute promyelocytic leukemia [10] and is also silenced through promoter hypermethylation in human breast carcinoma cell lines [11]. In lung, breast and prostate tumors, the expression of RARβ is found lost or down-regulated through unconstant promoter hypermethylation, but without any deletion or mutation of this gene [12][13][14][15]. The involvement of RARγ in cancer is less known [16]. Down-expressions of RXRs have been reported in prostate and thyroid carcinoma [17,18]. RARβ has been mostly studied in lung cancer [19][20][21]. For this serious disease, primary prevention, that is tobacco eviction, remains the best goal. However, considering the overall poor results obtained by the current treatments, it seems important to offer to "at risk" subjects a secondary chemoprevention. Retinoids have been used with such a goal but have given deceiving results for still unexplained reasons [22,23]. This underlines the need for more knowledge about retinoid signalling in the normal human lung.
In this context, we recapitulated the informations available in silico for the human RARβ in order to update its genomic organization. The activity of a putative promoter RAR-beta1' was assessed experimentally. Reference values for RRs mRNAs were measured in Normal Human Bronchial Epithelial cells (NHBE) with qRT-PCR Syber Green assays and a triplex of Taqman probes. Conceived to save invaluable bronchial mucosa samples, the reliability of this last qRT-PCR tool was extensively validated. Hence, it was used to measure the relative RRs mRNAs levels in lung tumors and cell lines. Finally, contrary to commercial antibodies, our pan-RARβ antibody immunoprecipitated a single band of protein in nuclear extracts of BEAS-2B and NHBE cells. Its molecular weight and further amino acid sequencing demonstrated that it corresponds to the RARβ2 protein isoform. Rigorous samples processing, like extensive biocomputing analysis were the key factors for the realization of this study. Validated tools are now available for future investigations on Vitamin A/retinoid signalling in the lung.

Samples
This study was approved by the French Medical Ethical Committee for Bioresearch of Lorraine. All the samples were provided by the "Centre de Ressources Biologiques du Centre Hospitalier Universitaire de Nancy", an ISO 9001-2000 certified biobank. Informed consents were obtained from patients operated upon for lung cancer who donated their samples to the biobank. Samples collections have been extensively described in [24]. All the reagents were purchased from Sigma-Aldrich (L'Isle d'Abeau Chesnes, Saint-Quentin Fallavier, France), unless mentioned. Twelve Squamous Cell epidermoid Lung Cancer (SqCLC) samples were obtained from different patients. Fourty μm frozen sections were cut and further dissected by JMV, our pathologist, after quick staining with RNAse-free azur blue solution. Contaminating normal tissues were removed and RNA was further prepared to give 90% tumoral mRNA [24].
Trachea and the beginning of the two main bronchi were obtained from ten donors. All subjects were kept under artificial respiration until the specimens were collected. Then, samples were bathed in cold RPMI-1640 complete medium supplemented with 10% Foetal Calf Serum, 50 units/mL penicillin, 50 μg/mL streptomycin and 5 μg/mL fungizone for express transport to the laboratory. There, they were cut open and the mucosa was washed with RPMI complete medium before being scratched with the back of a scalpel blade to give "Normal differentiated Human Bronchial Epithelial cells" or NHBE: our reference tissue sample. Pooled NHBE were kept frozen at -20°C until further processing. Histological examination demonstrated that such samples contained mainly differentiated bronchial cells. The remaining scratched mucosa was dissected out from the bronchi and cut into small explants to give rise to "Cultured undifferentiated Human Bronchial Epithelial cells" (CHBE) in LH9 medium (Cambrex Bio Sciences, Emerainville, France). Cells were grown for two weeks with regular medium change as in [25]. A week later, explants were detached from dishes and cultured again twice. At confluence, CHBE were scrapped and processed like NHBE for RNA purification. Histological examination revealed mostly undifferentiated basal bronchial cells. The smooth muscle sheaths from the back of the trachea were obtained by dissection and were cut frozen as 20 μm sections. Skin samples were collected from three different female patients undergoing fat reduction and were cut also as 20 μm frozen sections parallel to the basal membrane. Quick histological examination of every third section verified that collection was restricted to epidermis. All specimens were processed as NHBE to prepare RNA.

RNA preparation
In order to facilitate RNA solubilization, all frozen cut samples were first passed through a 23-Gauge needle. Total RNA was purified using the RNeasy Mini Kit (Qiagen, Courtaboeuf, France) with DNA digestion and integrity verification steps according to the manufacturer's instructions and our internal quality controls. Then, RNAs were diluted to a final concentration of 125 ng/μL and kept frozen at -20°C.

Drawing the human RARβ genomic region
Genome databanks were mined to construct an extended RARβ genomic organization ( Figure 1 . Finally, the orthologous conservation of the mouse RAR-beta1 and the human RAR-beta1' promoters were investigated using the Eldorado comparative genomics module [28].

Quantitative Real-Time PCR
The qRT-PCR primers and Taqman probes used in this study are presented in the Additional File 1. Primers for β-Actin were chosen in the gene 3' untranslated region as a RNA integrity control. The "RARβ" primers (exon 8) target all the RARβ transcripts generated from the RAR-beta2 promoter, while the "RARβ2" primers (exon 1) are specific to the RARβ2 transcript isoform. Primers used in Syber Green or in Taqman assays are identical for RARα and RXRβ, while they were made in different parts of the 5' end of the RARβ transcripts in order to cross confirm the results obtained with either the Taqman assays (primers located in the exons 6/7 junction) or with the Syber Green assays (primers located in the exon 8). No reliable primers were found for RXRγ (multiple single nucleotide polymorphisms). All primers, 5' FAM-RARβ-3' TAMRA and 5' Texas Red-RXRβ-3' Deep Dark Quencher-2 Taqman probes, were purchased from Eurogentec (Angers, France). The 5' VIC-RARα-3' TAMRA probe was synthesised by Applied Biosystems (Courtaboeuf, France).
For Syber Green assays, a β-Actin standard curve was first performed using the full cDNA cascade dilutions of each cDNA sample in triplicate and a RotorGene 6000 (Corbett Research, Labgene, Archamps, France). RRs were measured for the sole dilutions: 1/10, 1/50 and 1/100, in triplicates. Ten μL final volume contained: 4 μL of cDNA, 0.14 μL of 10 μM primers (140 nM each), 5 μL of Absolute QPCR SYBR Green Mix (Thermo Fisher Scientific, Courtaboeuf, France) and 0.72 μL of water. The PCR program was: 95°C for 15 min, followed by 50 cycles of 95°C: 10 s, 60°C: 20 s and 72°C: 20 s. Further melt curve analysis was conducted from 70°C to 95°C (0.5°C every 15 s). All amplicons were sequenced for confirmation. Taqman assays were performed for the same cDNA dilutions as above in duplicates. Twenty five μL final volume assays contained: 5 μL of cDNA, 0.5 μL of primers (200 nM each), 0.25 μL of probes (200 nM each), 12.5 μL of QuantiTect Multiplex PCR Mastermix (Qiagen, Courtaboeuf, France) and 3.75 μL of water. The PCR program began with a Taq Polymerase activation step at 95°C for 15 min, followed by 50 cycles of 94°C: 60 s and 60°C: 90 s. Assays with one probe at a time vs the triplex of probes were first compared in order to optimize equal final qPCR efficiencies.
Raw qRT-PCR data were collected with the RotorGene 6000 Series Software version 1.7.87. The comparative quantification approach [30] implemented within the built-in software was used to compute the second derivative of each amplification curve whose peak corresponds to the point at which the most efficient amplicon synthesis occurs. For each measure, the software determines a "Take-Off Point (TOP)" value, 80% before the peak and an "Amplification Efficiency (AE)" from the slope of the section of the curve between the TOP and the peak. For each gene and sample analyzed, all the TOP and AE values from the replicates were averaged to calculate the mean Take-Off Point (mTOP) and the mean Amplification Efficiency (mAE) values with their Standard Deviations (SD) (Additional File 2). The quality of each standard curve was assessed by its associated R 2 value, computed with automatic fluorescence threshold settings by the RotorGene software.
For statistical analysis, the qPCR data were first exported into the Bestkeeper © version 1 software [31] to determine the most stable and common reference genes across all the samples. Validated reference genes were then combined into a Bestkeeper © index for the normalization of RRs expression data. Then, the Rest-RG © software [32] was used to transform the raw qPCR data into normalized RR x-fold expression ratios (R). It uses the mathematical model described by Pfaffl et al. [33], that combines gene Schematic representation of the RARβ genomic context Ratios were computed using a Pair Wise Fixed Reallocation Randomisation Test © and plotted with Standard Error (SE) estimation via a Taylor algorithm. Computations were based upon the sample-specific mAE (mAE×E) and the corresponding mTOP deviations (ΔmTOP×ΔCt) for each sample tested vs the control sample: NHBE. In each sample tested, RRs genes expressions were also normalized by the optimal combination of reference genes previously defined by Bestkeeper © .

Western-Blot and Immunoprecipitation
To produce our pan-RARβ antibody, the following synthetic peptide: GHEPLTPSSSGNTAEHSPSI, corresponding to the F2 region, common to all the human RARβ proteins, was chosen to immunise rabbits and was purified by affinity chromatography (Proteogenix, Strasbourg, France). For western-blot analysis, nuclear extracts were prepared from BEAS-2B and NHBE cells as described in [34]. They were fractionated by SDS (1%) poly-acrylamide gel electrophoresis (12%) and electrotransferred onto nitrocellulose filters. After blocking with PBS containing 3% non-fat powdered milk, the filters were immunoprobed with our polyclonal RARβ antibody (dilution 1:500) for 2 h, extensively washed in PBS containing 0.05% Tween 20, and then incubated for 1 h at room temperature with an anti-rabbit peroxidase-conjugated antibody (dilution 1:40,000) (Jackson ImmunoResearch, West Grove, PE, USA). Specific complexes were revealed by chemiluminescence detection (Super Signal West Dura, Pierce Biotechnology, Rockford, IL, USA) according to the manufacturer's protocol. For immunoprecipitation analysis, NHBE nuclear extracts (1 mg) were incubated with immobilized Protein A cross-linked to the antibody, using the Seize X Protein A Immunoprecipitation Kit (Pierce Biotechnology, Rockford, IL, USA) according to the manufacturer's protocol. The immunoprecipitated proteins were detected as above by immunoblotting and chemiluminescence. Further sequencing of the immunoprecipitates were realized by the "core protein platform" of the University of Nancy.

Results and Discussion
Investigation of the human RARβ genomic region (Figure 1) The different isoforms encoded by RARβ were first described in mice. Two murine promoters have been characterized: RAR-beta1, encoding the RARβ1 and RARβ3 proteins, and RAR-beta2, encoding RARβ2 and RARβ4 [35,36]. Then in humans, the RARβ2 protein [Swiss-Prot:P10826-2], encoded by a CpG island and RARE containing RAR-beta2 promoter region, was reported [37]. The RARβ4 protein isoform [Swiss-Prot:P10826-3], generated from the alternate splicing of the RARβ2 mRNA at the exon 3, was further described [5]. A human foetal protein isoform, RARβ1 [Swiss-Prot:P10826-1], also expressed in small cell lung cancers has been identified [38]. However, neither the corresponding mouse RAR-beta1-like promoter nor the RARβ1 and RARβ3 transcripts have been fully characterized in human cells. Another protein isoform, RARβ5, has been characterized in both normal and malignant human breast cells [39]. It is supposed to contribute to the resistance of the cancerous cells to retinoids. This isoform is encoded by a cryptic distinct RAR-beta3 promoter, located in the first intron of the RARβ2 transcript [39]. Finally, a RARβ1' transcript isoform, with a specific first exon, has been reported [40]. However, the corresponding protein sequence is not described. This transcript is encoded by a distinct promoter RAR-beta1" and is expressed in normal human lung and normal RAsensitive BEAS-2B cells but it is suppressed in RA-resistant BEAS-2B, in some lung cancer cell lines and lung tumors [40].
Based upon these reports, we mined databanks to draw in silico the human RARβ genomic region (Figure 1). Two RARβ Refseq transcripts, [GenBank:NM_000965] and [GenBank:NM_016152], are reported on chromosome 3, strand +. They are respectively associated to the RARβ2 (long) and the RARβ4 (short) isoforms, which are transcribed from the well-characterized RAR-beta2 promoter (represented as P2 on Figure 1). These two transcripts were detected in all the samples tested when using our "RARβ" and "RARβ2" qRT-PCR primers (Additional File 1, Figure 2 (Figure 1). Two of these transcripts have a specific first exon in their respective sequences and share their other exons with RARβ2 ( Figure 1). [Gen-Bank:DC376623] is an Expressed Sequence Tag and [Gen-Bank:DQ083391] is described as a mRNA entry that contains the partial coding sequence of RARβ1'. Interestingly, we found that the qRT-PCR primers used in [40] to amplify the RARβ1' isoform are located in the exons 1 and 2 of this unvalidated trancript. We designed specific PCR primers (Additional File 1, Figure 1) in order to establish the existence of these four upstream transcripts. However, none corresponding amplicons were found in the selected cells: NHBE, CHBE, SMC, A549, H2087, H2342 and BEAS-2B. These results were especially surprising for [GenBank:DQ083391], as the primers we designed are located in the exons 1/2 junction of the transcript and overlap with those previously used in [40] to amplify RARβ1'.
The upstream sequence of each unvalidated transcript (-1500/-50) was further screened using online tools to detect the features of a putative additional RAR-beta promoter region. No canonical promoter features (TATA box, ...) were found, but two degenerate DR5 RARE motifs were identified within the upstream sequences of [Gen-Bank:U52076] and [GenBank:DQ083391], near their respective specific first exons ( Figure 1). Additionally, a CpG island was found in the upstream sequence of [Gen-Bank:U52076]. This last region (indicated as P1" on Figure 1) was previously explored by Petty et al. [40] and it was therefore not reinvestigated experimentally in this study. However, we searched for other possible alternate RAR-beta promoters. In silico comparative genomics indicated an orthologous conservation between the mouse RAR-beta1 promoter region and the upstream sequence of [GenBank:DC376623]. This region (indicated as P1 on Figure 1) corresponds to the region amplified by Houle et al. when cloning the 5' end of the RARβ1 cDNA [38]. Another interesting orthologous promoter conservation was found within the upstream sequence of [Gen-Bank:DQ083391] for chimp RARβ (99%) [Genomatix:GXP_1213141] and rhesus monkey RARβ (95%) [Genomatix CompGen promoter]. This conserved region (indicated as P1' on Figure 1) was further investigated experimentally but no promoter-like activity was demonstrated in reporter assays. This might be a feature of weak nuclear receptors.

Relative expressions of RRs mRNAs
Reference values for RRs mRNAs expressions were first determined in normal lung tissues (NHBE), then the relative levels of RRs mRNAs were measured in several cell lines, comparing the results obtained with the two qRT-PCR methodologies: Syber Green and a triplex of Taqman probes. The primers and probes used in this study are listed in Additional File 1. The raw qRT-PCR data are recapitulated in Additional File 2. R 2 values ranging from 0.965 to 1, mTOP SD<15% and mAE SD<10% demonstrates the reproducibility of each assay.
Additional File 3 shows the results obtained from the Bestkeeper © analysis to select the best combination of reference genes used for the normalization of the qRT-PCR data. In addition, when demonstrating the stability of the selected genes, the quality of each sample was also systematically assessed by Bestkeeper © as described in [31]. SD values < 1 demonstrates the overall inter-sample expression stability for β-Actin (0.56), RARα (0.60 with Sybr Green and 0.67 with the Taqman probe) and RXRβ (0.75 with Sybr Green and 0.60 with the Taqman probe). SD value > 1 sustained gene expression variability across samples and therefore exclusion from the computation of the Bestkeeper © index. Using only the three stable genes, their combinations were tested to obtain the most reliable Bestkeeper © index parameters. Although RARα correlated well RARβ and RARβ2 relative expression ratios with the Bestkeeper © index, its inter-sample wide variation did not match with those of β-Actin and RXRβ (r < 0.5, p > 0.05). RARα was therefore not considered as a relevant reference gene in this case. In Syber Green assays, the optimal Bestkeeper © index parameters were found for the combination of β-Actin and RXRβ as reference genes. Indeed, their expressions correlated well together (r = 0.76, p = 0.006) and a good correlation for each β-Actin (r = 0.96, p = 0.001) and RXRβ (r = 0.92, p = 0.001) was observed with the computed BestKeeper © index, which exhibited a moderate SD variation of 0.64. This combination of reference genes was thus further used in each sample for the normalization of RRs expression data obtained from the Syber Green assays.
As a matter of fact, it is not possible to design a multiplex of Taqman including the housekeeping gene, β-Actin, and the RRs genes together because the gap of expression between them is so wide that the amplification of β-Actin will be inhibited when RRs become detectable for the same qPCR cDNA inputs. Since the SD values associated to the RARα and RARβ Taqman probes were > 1 (Additional File 3), they could not be used for the normalization of the data obtained with the triplex of probes.
Therefore, a combination of β-Actin (Syber Green external reference gene) and RXRβ (Taqman internal reference gene) was tested. The related Bestkeeper © index showed a SD variation of 0.59 (Additional File 3). The values of the selected reference genes correlated well together (r = 0.61, p = 0.046) and also with the corresponding Bestkeeper © index (r = 0.94, p = 0.001 for β-Actin and r = 0.84, p = 0.001 for RXRβ). This combination of reference genes was thus used in each sample for the normalization of the data generated from Taqman assays. It also allowed to perform an accurate comparison between the significant relative expression ratios obtained with both qRT-PCR methodologies.
For each sample, the RRs mRNAs expression ratios relative to NHBE were next computed by the Rest-RG © software, with a simultaneous normalization of the data by both reference genes: β-Actin and RXRβ (Additional File 4). The sample-wide relative levels of "RARβ" and "RARβ2" amplicons are represented on Figure 2. Statistical significance was determined using a Pair Wise Fixed Reallocation Randomisation Test © (p-value < 0.05). We chose the method reported by Pfaffl et al. [33] to determine the relative expressions of RRs mRNAs because it allows a reliable normalization of the raw qRT-PCR data with an index of stable reference genes. In this method, the computation of each expression ratio is based upon the crossing point deviation (ΔCP or ΔmTOP) for each sample vs the control sample (NHBE). Furthermore, corrections for exact qRT-PCR efficiencies (E or mAE) are taken into account, which is not the case when computing relative expressions using the standard curve or the ΔΔCP methods [41][42][43]. Such normalizations are indeed very important, especially when the expressions of several genes have to be compared across numerous samples.
In Syber Green assays, the levels of "RARβ" amplicons varied as follows: H69(0.  (-8.33). In H2087, the level of "RARβ2" amplicons was so weak that it was not detected. In all the samples, the relative amounts of "RARβ" amplicons were always higher than those observed for "RARβ2" (Figure 2). This is consistent with the fact that the "RARβ" primers target the common part of all the RARβ transcripts generated from RAR-beta2, whereas the "RARβ2" primers are specific for the sole RARβ2 transcript isoform. When compared to "in vivo" differentiated NHBE, the relative levels of both "RARβ" and "RARβ2" amplicons were always found at lower levels in the samples tested. This, especially for the "in vitro" undifferentiated CHBE cells, indicating that RARβ is indeed very important in lung cells and that an increase in RARβ expression signs the normal differentiation of the human bronchial epithelial tissue. In addition, the levels of "RARβ" and "RARβ2" amplicons observed in skin and SMC were very similar to CHBE and their respective amounts were also significantly lower than in NHBE. Concomitant and significant low levels of "RARβ" and "RARβ2" amplicons were observed in H1648, H2087, and H2342 (Figure 2), confirming the importance of RARβ in NSCLCs [21,34,44,45]. That was not the case in both H69 and A549 cell lines whose "RARβ" and "RARβ2" amplicons levels corresponded to those observed in NHBE.
The results obtained with the RARβ Taqman probe varied as follows: A549(0.24) > H69(-0.40) > SMC*(-2.06) > skin*(-2.21) > CHBE*(-2.87) > BEAS-2B*(-3.8). In H1648, H2087, H2342 and H513, RARβ was found highly down-expressed and still detected by the Syber Green assays but not by the Taqman assays ( Figure 2). In order to compare the data obtained with the RARβ Taqman probe vs the corresponding Syber Green data, a tissue specific variation rate (VR) was computed when RARβ amplicons were detected in both kinds of qPCR assays and when the computed expression ratios were both significant. All the resulting VR values are < 25% (Additional File 4). This confirmed that, when detected, the RARβ expression ratios generated by the two qRT-PCR methodologies were consistent. The triple Taqman assay was initially developed to save rare and invaluable mRNA samples by simultaneously performing the detection of three RR genes. Since there is twice as much cDNA in Syber Green assays for a single RR amplification as there is in the Taqman probes assays to amplify three RRs, the sensitivity of the Taqman assay is therefore at least twice that of the Syber Green assay.
As Corresponding RRs protein levels were also previously reported as highly expressed in this tissue [46].
Significant down-expressions of RXRs mRNAs were previously reported in lung tumors by comparison to their respective matching normal tissues [47] and RXRβ was reported as a biomarker of aggressivity for NSCLCs [48]. Nevertheless, in all the lung cancer cell lines tested in this study, RXRα and RXRβ were not found differentially expressed from the control tissue NHBE. Furthermore, RXRβ, was demonstrated as stably expressed across all the samples, and with both qRT-PCR methodologies. In this context, it was therefore considered as a relevant reference gene.
12 dissected SqCLC tumor samples were also analyzed using β-Actin in Syber Green and the triplex of Taqman probes. The resulting raw qRT-PCR data are recapitulated in Additional File 5. β-Actin amplicons were detected in all the samples tested with Syber Green as well as RARα amplicons with the triplex of Taqman probes. Corresponding qRT-PCR data were analyzed with Bestkeeper © to check the expression stability of the genes across samples. The results are indicated in Additional File 6. SD values < 1 demonstrated the overall expression stability for both β-Actin (0.89) and RARα (0.86) across tumoral samples. The related Bestkeeper © index showed a SD variation of 0.85. The values of the selected reference genes correlated well together (r = 0.78, p = 0.003) and with the corresponding Bestkeeper © index (r = 0.96, p = 0.001 for β-Actin and r = 0.92, p = 0.001 for RARα). As previously, the quality of each tumor sample was also confirmed by the software [31]. Interestingly, although RARα amplicons were detected for each tumor sample analyzed with the triplex of Taqman probes, no amplicons could be detected for both RARβ and RXRβ (Additional File 5). This suggests the deep concomitant down-expressions of these two RRs genes in epidermoid lung tumors.

Western-Blot and Immunoprecipitation of RARβ
Commercial antibodies against RARβ gave in our hands conflicting results in western-blots, with several protein bands that are either attributed to phosphorylation or degradation. Therefore, we validated extensively a new pan-RARβ antibody directed toward a specific peptidic sequence common to all the RARβ protein isoforms described (F2 region). Using nuclear extracts of BEAS-2B and NHBE cells, the antibody recognized a single band of about 52 kDa in both western-blot (Figure 3, panel A) and immunoprecipitation ( Figure 3, panel B) experiments. The amino acid sequencing gave the following sequence: SSADHRVRLDLG (exon 5 of [NM_000965]) also common to all the RARβ proteins described, attesting for the specificity of the antibody. The results obtained correspond to the molecular weight expected for the RARβ2 protein, which is reported as a nuclear isoform of RARβ in databanks. The RARβ4 protein, whose molecular weight was previously reported at about 37 kDa [5] is reported as a cytoplasmic isoform of RARβ in normal cells [49] and it was accordingly not detected by our antibody in BEAS-2B and NHBE nuclear extracts. The RARβ1' protein isoform was previously characterized using whole cell extracts of BEAS-2B and it was reported with a molecular weight of about 47 KDa [40]. This isoform was also not detected by our antibody in the nuclear extracts analyzed. These results suggest that, if it is actually expressed in BEAS-2B and NHBE cells, the RARβ1' protein is another short cytoplasmic isoform of RARβ. That is particularly interesting as truncated isoforms of RARβ with "tumorogenic activity" were previously reported over-expressed in malignant cells [50]. Indeed, these truncated isoforms are thought to compete with RARβ2 in the nuclei of cancer cells; inhibiting its cellular functions, enhancing the proliferation process and contributing to the resistance of the cells to retinoids.

Conclusion
In this study, we used qRT-PCR analysis to measure RRs mRNAs expressions in several samples and compared the results obtained in Syber Green assays to those given by a triplex of Taqman probes that was conceived to save invaluable mRNA samples. RARβ was the only RR gene whose mRNA levels were found significantly different when compared to NHBE. In these differentiated lung cells, the RARβ transcripts were found at their highest levels of expression, indicating that an increase in RARβ mRNAs signs the normal differentiation of the human bronchial epithelium. In all the samples analyzed, RARβ2 was the major transcript isoform that was detected. In addition, the levels of RARβ2 varied concomitantly with the whole levels of RARβ mRNAs measured in each sample. To evince the existence of additional RARβ isoforms, we recapitulated the informations available in silico for the human RARβ. Then, we investigated for the presence of unvalidated RARβ transcripts and for the activity of a putative RAR-beta1' promoter region. However, such transcripts were not detected in the samples tested and no activity was found for RAR-beta1'. This allowed to demonstrate that the whole level of RARβ mRNAs observed in each sample is represented by the transcripts isoforms generated from the RAR-beta2 promoter. RARβ transcripts were found highly down-expressed in NSCLC cell lines and SqCLC tumors, which is a reported event in the process of lung carcinogenesis. β-Actin and RARα expressions were found stable across samples. As for RXRβ, although its expression was found stable in normal lung cells and cancer lines, a concomitant down-regulation was observed with RARβ in SqCLC samples. This suggests a possible role for this RR as an heterodimerization partner for RARβ in the bronchial mucosa. Finally, RARβ2 was the only protein isoform detected by our pan-RARβ antibody in NHBE and BEAS-2B nuclear extracts, indicating that short protein isoforms of RARβ were not present within the nuclei of these cells. In our view, the research for specific human micro-RNAs targeting each of the RRs, as well as RNA interference experiments, and the over-expressions of full or truncated versions of RRs, would be of great interest in order to complete our understanding of the molecular mechanisms involved in the process of lung carcinogenesis.