PMC:1409818 JSONTXT

Multiplex degenerate PCR coupled with an oligo sorbent array for human endogenous retrovirus expression profiling Abstract Human endogenous retroviruses (HERVs) can be divided into distinct families of tens to thousands of paralogous loci. The expression of HERV elements has been detected in all tissues tested to date, particularly germ cells, embryonic tissues and neoplastic tissues. Hence, the study of HERV expression could represent added value in cancer diagnosis. We developed a quantitative assay combining a multiplex degenerate PCR (MD-PCR) amplification, based on the relative conservation of the pol genes, and a colorimetric Oligo Sorbent Array (OLISA®). Nine HERV families were selected and amplification primers and capture probes were designed for each family. The features required to achieve efficient amplification of most of the elements of each HERV family and balanced co-amplification of all HERV families were analyzed. We found that MD-PCR reliability, i.e. equivalence of amplification and dose-effect relationship, relied on the adjustment of three critical parameters: the primer degeneracy, the relative concentration of each primer and the total amount of primers in the amplification mixture. The analysis of tumoral versus normal tissues suggests that this assay could prove useful in tumor phenotyping. INTRODUCTION Human endogenous retroviruses (HERVs) constitute about 8% of the human genome (1). HERVs are considered as remnants of past germ-cell line infections by retroviruses that occurred during mammal evolution. Phylogenetic studies have identified at least 31 different HERV families each encompassing tens to thousands of loci (2) which are the result of intracellular retrotransposition of transcriptionally active copies. All are defective for viral replication due to genetic drift and are, as a result, transmitted in a Mendelian fashion. Research over the past two decades has demonstrated that HERVs may have biological functions. It has been suggested that HERVs could be involved in genome shaping as potential chromosomal recombination sites (3). HERV long terminal repeats (LTR), whose original function is retroviral expression control, could provide transcriptional regulatory elements for cellular genes consisting in promoter (4), enhancer (5) or polyadenylation signal functions (6). HERV proteins could also have biological functions as illustrated by the ERVWE1 Env protein, also known as syncytin (7–10). HERV expression was particularly investigated in three specific contexts, i.e. placentation (11,12), auto-immunity (13,14) and cancer (15,16) that are associated with cellular differentiation or immunity modulation. However, it is still generally unclear whether HERVs are triggers or markers in pathological contexts, although it has been suggested that Env HERV-W and Rec HERV-K are involved in multiple sclerosis (17,18) and testicular tumorigenesis (19,20), respectively. Quantitative approaches concerning both HERV transcripts could thus represent added value in current cancer and chronic disease diagnosis. On the basis of functional hypotheses associated with candidate loci, the majority of HERV quantitative expression studies are based on RT–PCR (11,21–23). Nevertheless, global approaches representing the activity of HERV families may elucidate the mechanisms involved in transcriptional regulation and address physio-pathological functions. These approaches would consist in analyzing either the expression of each locus within each family or, alternatively, the overall transcription of each family. The first approach is challenging in terms of specificity due to the features of these repeated elements. The second approach could be tackled in two ways, (i) by developing different HERV-taxon specific real-time RT–PCR assays using consensus (24) or degenerate primers (25) or (ii) by developing a quantitative assay combining multiplex degenerate PCR (MD-PCR) with an oligonucleotide microarray. This could be achieved subject to minimized bias, i.e. efficient amplification of most of the elements of each family and balanced co-amplification of each family. In this way, the retrovirus pol genes reverse-transcriptase region encompasses two conserved sequences, corresponding to the VLPQG and Yv/mDDi/v/lL amino-acid motifs, which enabled the development of pan-retrovirus amplifications using degenerate primers. The first pan-retrovirus PCR assay developed by Shih in 1989 (26) was based on a single pair of degenerate primers. This system was improved by adding a semi-nested PCR step (27) which enabled the detection of HERV-W expression in the serum of multiple sclerosis patients (28). Nevertheless, it was established that this system failed to detect the expression of another HERV family (HERV-H) due to a biased degenerate primer composition (29). More sophisticated systems were developed in order to identify all known retroviral reverse-transcriptase sequences. Independent amplifications of beta and gamma retroviruses were obtained initially using two pairs of degenerate primers (30) and at the present time using two complex primer mixtures (31). Interestingly, the report of a broad consistency between the latter qualitative assay and quantitative RT–PCR amplifications (24) suggests that a pan-retrovirus-based procedure could be suitable as a quantitative assay. In this paper, we describe the development of a pan-retrovirus PCR amplification dedicated to the analysis of differential HERV expression profiles. This aim essentially requires the definition of a quantitative method to evaluate HERV expression. Nine HERV families were selected according to their transcriptional activity in various physio-pathological contexts (HML-2, HML-4, HERV-H, ERV-9, HERV-W, HERV-E4.1 and HERV-R) (32), their sequence homology with infectious retroviruses (HML-5) (33) or their genomic complexity (HERV-L) (34). These families belong to three distinct classes of retroviruses: betaretroviruses (HML-2, HML-4 and HML-5), gammaretroviruses (HERV-H, ERV-9, HERV-W, HERV-E4.1 and HERV-R) and spumaviruses (HERV-L). We systematically investigated the key factors involved in the efficient co-amplification of the nine HERV families. Biotinylated PCR products were quantified using an Oligo Sorbent Array (OLISA®) low-density plastic microarray derived from our Enzyme-linked Oligo Sorbent Assay (ELOSA) developed for the detection and quantitation of HIV1 provirus (35,36). Three parameters were found to be critical in order to obtain an almost equivalent amplification efficiency of the different targets and dose-effect relationship: the primer degeneracy level, the relative concentration of each primer and the total amount of primers in the amplification mixture. Finally, the MD-PCR/OLISA assay was used to compare HERV transcript content in differentiation models and in tumoral versus normal RNA samples. MATERIALS AND METHODS Collection of HERV elements HERV-related sequences were collected from GenBank with RepeatMasker (Smit, AFA, Hubley, R & Green, P. RepeatMasker Open-3.0. 1996–2004 ) using a 20% divergence threshold. The query library contained pol reference sequences of the nine HERV families, defined as HML-2 (accession no. Y17832, position 4557–7175), HML-4 (AF020092, 3884–6998), HML-5 (AC004536, 36 079–38 642), HERV-H (AJ289709, 2458–6015), ERV-9 (Z84475, 28 749–32 286), HERV-W (AC000064, 32 379–35 878), HERV-E4.1 (M10976, 2650–6210), HERV-R (AC073210, 57 299–60 388), HERV-L (X89211, 2249–5807). Alignments were performed with ClustalW (37) and adjusted manually with the Seaview multiple alignment editors (38). A matrix indicating the occurrence of each nucleotide for each pol position was generated in order to design amplification primers and detection probes. Cloning of HERV-pol archetypes HERV proviruses were amplified from genomic DNA by means of long-PCR as described previously (10) and pol archetypes were then cloned and sequenced after a nested PCR step. Final HERV archetypes were: HML-2 chromosome 7 from base 103 983 434 to 103 984 261, HML-4 chromosome 10 from base 5 074 724 to 5 075 458, HML-5 chromosome 7 from base 117 350 949 to 117 351 816, HERV-H chromosome 2 from base 166 394 420 to 166 395 285, ERV-9 chromosome 6 from base 115 664 573 to 115 665 345, HERV-W chromosome 7 from base 91 746 541 to 91 747 359, HERV-E4.1 chromosome 19 from base 20 726 359 to 20 727 215, HERV-R chromosome 7 from base 63 899 936 to 63 900 787 and HERV-L chromosome 16 from base 59 440 954 to 59 441 771. Probe design and microarray setup Fifteen 20–25mer probes corresponding to the nine HERV families were designed within the pol amplified region (probe sequences available upon request). For each HERV family, consensus probes were determined using the pol sequence matrix. In addition, for the HML-2 capture probe, positions 12, 6, 19, 22 and 3 were cumulatively changed into cytosine to generate from one to five mismatch-containing probes. This type of probe was also designed for HERV-H by mutating the 9th position. The OLISA® microarrays (39) were purchased from bioMérieux (Grenoble, France). They consisted of 8-well plastic modules in which the probes were spotted in a circular format and tethered onto the support by a 5′ aminated end (Figure 1A). Probe specificity was assessed by hybridization with biotinylated transcripts produced from cloned archetypes. HML-2 and HERV-H variant probes were used to evaluate the mismatch impact on target hybridization. The final version of the microarray used to analyze biological samples contained HERV specific validated probes and HIV1- and HTLV1-negative controls. Primer design Degenerate primer pairs were designed in a highly conserved region of the pol gene RT sub-region. For each of the nine HERV families investigated in the study, the primer sequences were determined according to the corresponding pol sequence matrix. The levels of degeneracy were defined in line with the general parsimony principle highlighted in this study. The amplification efficiency and specificity of each degenerate primer pair was checked on cloned HERV-pol archetypes before collecting them to form a global primer mix. HERV-L-specific amplification HERV-L PCR amplifications with a weakly-degenerate (Forward tttactgtcctacytcagggr, Reverse tcarcataatgycatcaatgta) or highly-degenerate (Forward ttyactgtcctacctyarrrr, Reverse tcagyataatrycatyartrya) primer pair were carried out in a total volume of 50 µl containing 10 mM Tris–HCl, 1.5 mM MgCl2, 50 mM KCl, 16.5 pmol of each primer, 0.25 mM of each deoxynucleotide triphosphate, 0.25 U of Taq polymerase (Roche) and 250 ng of genomic DNA. Amplifications were performed in a GeneAmp® PCR System 9700 thermal cycler (Applied) following modified pan-retrovirus cycling (31): initial 5 min denaturation step at 94°C, 3 cycles at 94°C for 30 s, 45°C for 3 min and 72°C for 1 min followed by 35 cycles at 94°C for 30 s, 50°C for 2 min and 72°C for 1 min, followed by a 7 min final extension step at 72°C. The amplification products were separated by electrophoresis on 3% agarose gel. The bands of interest were purified and cloned in pCR® 2.1-TOPO® vector using the TOPO TA Cloning® Kit (Invitrogen). Twenty two ‘weakly-degenerate HERV-L’ clones and 24 ‘highly-degenerate HERV-L’ clones were sequenced. A phylogenetic tree was then constructed from these sequences using the UPGMA (absolute distance) method. Cell culture and differentiation BeWo b30 human choriocarcinoma cells were cultured in F-12K medium (Invitrogen, Carlsbad, CA) supplemented with 10% fetal calf serum (FCS) (Invitrogen, Carlsbad, USA), antibiotics (100 IU of penicillin/ml and 100 µg of streptomycin/ml) (Invitrogen, Carlsbad, CA) and 0.25 µg/ml amphotericin B (Fungizone) (Invitrogen, Carlsbad, USA). At 50% confluence, the medium was then replaced by one containing 50 µM forskolin (Calbiochem, Darmstadt, Germany) or vehicle (DMSO) (Sigma, St Louis, USA), followed by 36 h of incubation at 37°C to induce BeWo differentiation. U937 monocytoid cells were cultured in RPMI 1640 medium containing 10% FCS, antibiotics (100 IU of penicillin/ml and 100 µg of streptomycin/ml) (Invitrogen, Carlsbad, CA) and 0.25 µg/ml amphotericin B (Fungizone) (Invitrogen, Carlsbad, USA). The cells were seeded at a density of 5.105 cells per ml of a culture medium containing 50 ng/ml of phorbol-12-myristate-13-acetate (PMA) or vehicle (DMSO) (Sigma, St Louis, USA), followed by 72 h of incubation at 37°C to induce U937 differentiation. Cultures were performed in triplicate for each set of experiments to assess reproducibility. RNA samples and reverse-transcription BeWo b30 and U937 total RNA were isolated using the RNeasy Midi Kit including an on-column DNAse treatment in accordance with the manufacturer's instructions (Qiagen, Hilden, Germany). Breast, colon, lung, prostate, testis and uterus FirstChoice® Human Tumor/Normal Adjacent Tissue RNA was purchased from Ambion (Austin, Texas). The quantity of RNA was measured at OD260 in a spectrophotometer and the RNA quality was assessed with a Bioanalyzer capillary electrophoresis device (Agilent, Palo Alto, USA). The total RNA (5 µg or 1 µg from cultured cells or commercial samples, respectively) was reverse transcribed in a volume of 20 µl containing 1× first strand buffer, 0.01 M DTT, 40 U RNaseOUT, 0.5 mM dNTP-Mix, 250 ng of random primers (Promega, Madison, USA) and 200 U of SuperScriptII (Invitrogen, Carlsbad, USA) according to the manufacturer's recommendations. Reverse-transcriptase-free reactions were carried out to verify the absence of contaminating genomic DNA. Multiplex PCR Multiplex PCRs were carried out in a total volume of 50 µl containing 10 mM Tris–HCl, 1.5 mM MgCl2, 50 mM KCl, 10 µM (or various concentrations for the optimal concentration range determination) of optimized primer mix, 0.25 mM of dATP, 0.25 mM of dCTP, 0.25 mM of dGTP, 0.188 mM of dTTP, 0.062 mM of dUTP-biotin (Roche), 0.25 U of Taq polymerase (Roche) and 5 µl of the reverse-transcription reaction volume or predefined copy number of cloned archetypes. Amplifications were performed in a GeneAmp® PCR System 9700 thermal cycler as follows: initial 5 min denaturation step at 94°C, 3 cycles at 94°C for 30 s, 45°C for 3 min and 72°C for 1 min followed by 20 to 35 cycles at 94°C for 30 s, 50°C for 2 min and 72°C for 1 min, followed by a final 7 min extension step at 72°C. Amplification products were systematically checked by electrophoresis on 3% agarose gel before the microarray analysis. Microarray analysis The OLISA® microarray handling and protocol were performed according to the manufacturer's recommendations (bioMérieux, Grenoble, France). In sum: 5 µl of biotin-labeled multiplex amplification products were chemically denatured for 5 min at room temperature (RT), and then hybridized on a microarray for 60 min at 37°C in hybridization solution. The microarrays were washed four times with 300 µl of washing buffer at RT. A detection solution, containing a streptavidine–HRP conjugate, was dispensed on each microarray and incubated for 20 min at 25°C. The microarrays were washed four times with 300 µl of washing buffer at RT. An aliquot of 50 µl of staining solution containing the horseradish peroxidase (HRP) substrate was added to each microarray and incubated for 20 min at 25°C. The microarray image acquisition was then performed using an Apimager® densitometry reader (bioMérieux, Grenoble, France) and the data analysis was performed using the ApiAnalyser software (bioMérieux, Grenoble, France). Statistical analysis HERV transcriptome analyses of in-vitro differentiation models and tumor/normal adjacent tissue RNA were performed in triplicate. The statistical analysis was performed using Student's t-test. The mean values of two groups (mock-treated cells versus treated cells or normal tissue versus tumoral tissue) were considered to be significantly different if P < 0.05. RESULTS Probe design and validation The RT moieties of retroviral pol genes encompass two highly conserved and closely positioned regions corresponding to the VLPQG and Yv/mDDi/v/lL reverse-transcriptase functional motifs. This area was selected to design a multiplex amplification system dedicated to the study of nine HERV families. The alignments of the collected sequences led to the determination of the probability of the presence of the 4 nt on each pol consensus position (pol sequence matrix). In this way, probes corresponding to the most conserved part of the VLPQG–Yv/mDDi/v/lL inner region of each HERV family were designed. Probe specificity was confirmed by the absence of cross-reactions after hybridization with synthetic biotinylated targets (Figure 1B–J). However, the sequence variability existing within each HERV family may result in imperfect complementarities between probes and their homologous targets. Therefore, we evaluated the impact of possible mismatches on target detection using mutation-containing probes. Equivalent signals were measured for HML-2 and HML-2 1MM probes and a weaker but significant signal was measured for HML-2 2MM probes. Similarly, HERV-H and HERV-H 1MM probes displayed comparable signals. Conversely, no signal was detected using HML-2 probes containing more than two mutations. Degenerate primer design The primer design was based on the determination of the nucleic acid sequence composition corresponding to the VLPQG and Yv/mDDi/v/lL motifs. This constituted a rationale for determining the primer degeneracy. It should be noted that the linear increase of the primer degeneracy level is detrimental to amplification efficiency (sensitivity) due to the geometric increase in combinatory possibilities. The amplification product diversity was used to evaluate the impact of the degeneracy criteria on PCR behavior. The HERV-L family was chosen as a model because it contains the highest number of copies (990) among the nine HERV families and is sufficiently divergent (22%) to discriminate between cloned PCR products. Briefly, PCRs were performed on genomic DNA either with a weakly-degenerate HERV-L-specific primer pair hereafter referred to as Dg4 (level of degeneracy = 4 for each primer) or with a highly-degenerate HERV-L-specific primer pair referred to as Dg128 (level of degeneracy = 128 for each primer) and the PCR products were cloned and sequenced. No bias was obtained using either the Dg128 or the Dg4 primer sets. Firstly, 87.5% (21/24) and 95.5% (21/22) of the sequenced populations consist of distinct clones for Dg128 and Dg4, respectively. Secondly, Dg128 and Dg4 amplification products were found uniformly distributed throughout the phylogenetic tree (Figure 2). In the Dg4 primer set, the lowest frequency observed for a nucleotide position was 83% of that present in the whole HERV-L population alignment. Consequently, for each family, we designed primers by minimizing their level of degeneracy as follows: a single nucleotide or a combination of nucleotides, representing from 100 to 80% of the base frequency observed in the whole population, was determined for each primer position. This led to a mean degeneracy level of 13 ± 16 for the 18 primers. The 80% rule was empirically bypassed in two cases in order to avoid an alteration of the assay sensitivity. For HERV-W, the 78% frequency tolerated for one position of the reverse primer made it possible to halve the primer pair degeneracy. For HERV-E4.1, a 75% frequency tolerated for one position of the forward primer together with a 67% frequency tolerated for one position of the reverse primer made it possible to keep the primer pair degeneracy comparable with other HERV primer pairs (Table 1). Degenerate primer mix optimization In a second step, we developed a complex mix of degenerate primers in order to perform a multiplex PCR. The heterogeneity of the level of degeneracy within a primer pair and the heterogeneity of intrinsic amplification efficiencies of primer pairs were taken into account. The primer mix composition was established empirically by PCR amplification of a synthetic model consisting of nine cloned HERV-pol archetypes. The amplification performances were compared with a reference equimolar primer mix. An iterative procedure was applied, by constituting several primer mix generations, to define a primer mix resulting in signals as close to each other as possible (Table 1, Figure 3A). The average intensity for the nine HERV families was used as a reference and all intensities measured were expressed as a percentage of this reference. In this way, the multiplex amplification using the equimolar primer mix resulted in signal intensities widely dispersed around the 100% reference value. The extreme values were 0% for HERV-H and 459% for HML-4. This overall range was dramatically reduced with the final version of the complex primer mix, ranging from 43% for HERV-R to 240% for HML-4. More specifically, six out of nine signals were improved (HML-4, HML-5, HERV-H, HERV-W, HERV-E4.1 and HERV-L), two out of nine signals were unchanged (ERV-9 and HERV-R) and one out of nine was slightly degraded (HML-2: 55% for the equimolar primer mix and 179% for the complex primer mix). Taken together, these results demonstrated that empirical determination of primer quantities in a complex mix compensated for both degeneracy variation in primer pairs and intrinsic amplification efficiency differences. Hence, the complex primer mix developed enabled an equivalent multiplex PCR amplification of nine HERV families. We also studied the impact of the primer concentration on the amplification efficiency of each HERV archetype. We amplified 106 copies of each archetype separately with an optimized primer mix concentration range (Figure 3B). The band intensity measurement on the agarose gel electrophoresis pattern led to the determination of an optimal primer mix concentration range in the final reaction volume between 8 and 13 µM. Primer concentrations above 15 µM led to a decrease in the PCR amplification efficiency. Multiplex amplification characterization In order to check whether the defined amplification system co-amplified the nine targets on a 100-fold range of copy numbers efficiently, we performed the amplification of 10-fold serial dilutions of an equimolar mix of archetypes (Figure 4A). These amplifications led to signal variations correlated with target amounts for all HERV archetypes, i.e. with a dose-effect relationship, although with variable dynamic ranges. The broadest dynamic ranges were found for HML-4 (signal intensities from 1773 to 10 320) and HERV-E4.1 (signal intensities from 781 to 9500) and the narrowest was found for HERV-H (signal intensities from 1043 to 1675). However, HML-4 and HERV-H had linear dynamic ranges (R > 0.99). Conversely, although it had one of the broadest dynamic ranges, HERV-E4.1 was exponential rather than linear (R ∼ 0.73). Nevertheless, all these dynamic ranges were found to be compatible with a further quantitation objective. Furthermore, we evaluated the efficiency of the quantitation process for unbalanced target quantities. To this end, four different mixtures of HERV-pol archetypes were amplified. Mix A and B contained equimolar amounts of each archetype, 105 and 106 copies, respectively. Mix C contained 105 copies of each HERV archetype as for mix A, except for HML-4 and HERV-E4.1 for which the amounts were raised to 106 copies. Mix D contained 106 copies of each HERV archetype as for Mix B, except for HML-4 and HERV-E4.1 for which the amounts were reduced to 105 copies (Figure 4B). The ability of the assay to amplify and to detect the targets independently was evaluated by calculating, for each HERV, the C/A and D/B ratios of the intensities measured. C/A and D/B ratios close to 1 were expected for targets with identical quantities under different experimental conditions. In this way, HML-2, HML-5, HERV-H, ERV-9 HERV-W, HERV-R and HERV-L displayed C/A and D/B ratios ranging from 0.80 to 1.43 and 0.87 to 1.39, respectively. Hence, this demonstrated that the amplification of these seven targets was only affected slightly by the altered amount of the two other targets. Conversely, the increase in the HML-4 and HERV-E4.1 copy numbers under experimental condition C was correlated with a C/A ratio of 2.64 and 2.06, respectively. A reverse alteration of the ratio was observed when the HML-4 and HERV-E4.1 copy numbers were reduced, with the D/B ratio falling to 0.35 and 0.38, respectively. Altogether, these results confirmed the dose-effect relationship previously highlighted and strongly suggested that the amplification system developed could enable the detection of the variation in HERV sequence quantity between different samples. HERV transcriptome analysis of in-vitro differentiation models and tumor/normal adjacent tissue RNAs MD-PCR/OLISA was used for comparative studies of in-vitro cell differentiation models and biological samples. In order to evaluate the reproducibility of the assay, three biological replicates were performed from independent cultures of BeWo, forskolin-induced BeWo, U937 and PMA-stimulated U937 cells. The coefficients of variation (CV = SD/mean) calculated for BeWo cells ranged from 3.3 to 17%, the forskolin-treated BeWo cell CVs ranged from 0.5 to 22.3%, the U937 cell CVs ranged from 3.3 to 24.3% and the PMA-treated U937 cell CVs ranged from 0.5 to 12.2%. In BeWo cells, forskolin had a significant prominent effect on (i) HML-2, HML-5, and HERV-E4.1 whose expression decreased and on (ii) HERV-W and HERV-R whose expression increased (Figure 5A). In U937 cells, PMA only downregulated HML-5 and upregulated HERV-R expressions demonstrably (Figure 5B). In this way, inducible modulations of HERV expression in BeWo and U937 cells are specific. Surprisingly, no signal was detected for HML-4 in native or differentiated BeWo and U937 cells. All of the tumor types also demonstrated unique differential expression profiles. The HML-4 family remained untraceable in any of the six sample pairs. Reproductive tissue tumors exhibited unique expression profiles with breast-derived tumor demonstrating statistically significant increases in HML-2, HML-5, HERV-W, HERV-E4.1, HERV-R and HERV-L (Figure 5C), uterine tumor demonstrating HERV-R and HERV-L upregulation (Figure 5D), testicular tumor demonstrating HML-2, HERV-W and HERV-E4.1 upregulation and ERV-9 downregulation (Figure 5E), and prostate tumor demonstrating HERV-W, HERV-E4.1, HERV-R and HERV-L upregulation (Figure 5F). Additional tumors, such as those derived from lung and colon also demonstrated unique expression patterns. Lung tumor exhibited an increase in HML-2 transcription and a decrease in HERV-E4.1 expression (Figure 5G) whereas only HERV-H expression increased in colon tumor (Figure 5H). In this way, each cell type or tissue exhibited a unique differential HERV expression profile as opposed to a general increase or decrease in expression. DISCUSSION Diagnostic methods have to compromise between specificity and sensitivity, according to the assumed risk of false negative and false positive results, respectively. The biological question addressed in this paper involved two additional problems due to (i) the number of variable loci within each HERV family and (ii) the need for a balanced amplification of all families. Hence, the exhaustiveness and sensitivity of the overall assay are closely linked. This study relates to the identification of general rules allowing the definition of a quantitative assay based on a MD-PCR, coupled with the OLISA® microarray detection format. In the OLISA® detection format, the exhaustiveness of the capture probe relies upon the empirical selection of the most conserved region, i.e. the consensus, and the experimentally observed ability to tolerate the hybridization of slightly divergent targets as shown using mismatch-containing probes. In this way, the one to two authorized variable sites would make it possible to hybridize not strictly identical loci, e.g. containing polymorphic sites, without generating cross-reactions between different HERV families as depicted. Similarly, no cross-reaction resulted from the use of degenerate amplification primers (see below) as validated using pol archetypes. Additionally, a single band of ∼130 bp detected by pol probes was observed in the agarose gel electrophoresis of MD-PCR products derived from sixteen different RNA samples, strongly suggesting the absence of non-retroviral sequence amplification (data not shown). In the amplification step, high primer degeneracy should lead to exhaustive but poorly sensitive amplification and, conversely, low primer degeneracy should lead to a poorly exhaustive but sensitive amplification of an HERV family, as highlighted by Forsman et al. (25). We defined three critical amplification criteria enabling the quantitation of pol gene transcripts of nine HERV families, (i) the primer degeneracy determining exhaustiveness, sensitivity and specificity, (ii) the concentration of each primer determining sensitivity and balanced amplifications and (iii) the total amount of primers in the amplification mixture determining sensitivity. The impact of primer degeneracy on the diversity of the generated amplicons was assessed using the complex HERV-L family. The two experimental populations derived from DNA amplification using either low or high-degenerate HERV-L primers depicted the whole HERV-L population similarly. This enabled us to propose an acceptability threshold to design each primer position, defined as follows: each single nucleotide or combination of nucleotides may represent at least 80% of the base(s) observed at that position in the whole population. In order to minimize the degeneracy level, a lower frequency could be tolerated for a single primer position. In our assay, this concerned 3 out 18 primers and the lowest frequency was 67%. The degeneracy heterogeneity within each single primer pair may impair the amplification efficiency due to unbalanced availability of reverse versus forward primers toward a given target. This was efficiently counterbalanced by increasing the relative concentration of the more degenerate primer. However, for a primer pair, no linear relationship could be established between the relative degeneracy levels and the optimal primer concentration ratio, as previously mentioned (40). In addition, primer pair concentration was increased for the primer pairs generating weak signals and was decreased for the primer pairs generating strong signals, in order to compensate for their intrinsically heterogeneous amplification efficiencies. This result was found to agree with rules proposed for the development of multiplex PCR (41). The optimal overall primer concentration was found to range between 8 and 13 µM, 10 times higher than in conventional PCR. This MD-PCR primer concentration lies between the 20 µM concentrations described in the original pan-retrovirus degenerate PCR protocol (26) and the 0.5–4 µM concentration range described in various multiplex PCR assays (41–43). The MD-PCR/OLISA assay detected at least 105 copies of each individual archetype in an equimolar mix of all archetypes. This was in the same range as general retroviral PCRs developed by Tuke et al. and Seifarth et al. (about 1–2.106 copies) (25,27,30). This sensitivity is lower than that observed in single-target RT–PCR/ELOSA assays [50 copies, for a review, see (44)], due to primer degeneracy, and lower than that observed in HERV broadly targeted real-time reverse-transcription PCRs (1–10 copies) (25). Nevertheless, this sensitivity is compatible with the detection of moderately expressed genes (10 copies per cell) in view of the nucleic acid quantities used for this assay (∼5.105 copies in 0.25 µg RNA). A dose-effect relationship was observed within the 105–107 target range. In addition, a 10-fold difference in target concentrations did not affect the overall amplification efficiency of each distinct target. The robustness of the MD-PCR/OLISA method was assessed on biological replicates consisting of independent RNA extractions, cDNA syntheses, MD-PCR amplifications and OLISA detections. The calculated coefficients of variation ranged from 0.5 to 24.3%. These CVs were found to be comparable to those described using single-target PCR or RT–PCR amplifications coupled with ELOSA monodetection, from 3.0 to 24% (44,45). Hence, the reliability of the MD-PCR/OLISA method is suitable for use as a clinical assay. Moreover, this indicates that the use of both accurate RNA quantitation and quality assessment of RNAs are sufficient to establish total RNA normalization as a suitable method, as previously suggested (46). We thus investigated the HERV transcriptome modifications occurring in the cell differentiation process and the malignant transformation process. The human choriocarcinoma BeWo cell line undergoes fusion and extensive morphological differentiation, mimicking cytotrophoblast differentiation into syncytiotrophoblasts, following their treatment with cyclic AMP metabolism effectors (47). With the MD-PCR/OLISA assay we confirmed the upregulation of HERV-W and HERV-R in forskolin-treated BeWo Cells (8,48). Similarly, we confirmed HERV-R upregulation in PMA-treated U937 cells (49), an inducible monocyte differentiation model. However, we did not find significant increases in HERV-W and HERV-H, as previously shown with RT–PCR using pol and gag primer sets, respectively (21). This may reflect the use of poorly representative pol and gag primers, inducing increased sensitivity but not reflecting the overall behavior of each family. The MD-PCR/OLISA analysis of normal/tumoral RNA samples led to discriminating profiles between tissues as observed for the two differentiated cell lines. Within these profiles, the observed variations were in agreement with previous independent studies. In this way, the MD-PCR/OLISA assay demonstrated an increase in HML-2 expression in breast tumors (16) and in testicular tumors (50), in HERV-E4.1 expression in prostate tumors (15) and in HERV-H expression in colon tumors (51,52). Surprisingly, no HML-4 expression was detected as opposed to the ubiquitous expression described (24,31), questioning the exhaustiveness/specificity features of both systems. This discrepancy was previously observed in northern blot analyses, showing disparate HML-4 expressions in the lung and placenta (53,54). This illustrates that none of the existing systems are probably exhaustive and that further developments or combinations of assays will be necessary for a comprehensive evaluation of HERV expression. Nevertheless, all profiles were found to be different for the six paired tissues and the two differentiation models. This suggests that the processes involved in the modulation of HERV transcription in the course of (de)differentiation are tissue-specific, as previously suggested (52). Although, it is not known whether identical or different locus subsets of one given HERV family were mobilized, these results demonstrate that the pol-based global approach is valid as a profiling tool, as similarly demonstrated for the brain (55). In conclusion, the reliable MD-PCR/OLISA assay we developed provides general rules for the design of multiplex degenerate PCR applications and an alternative system for HERV transcription analysis which may be useful in tumor phenotyping. The authors are grateful to C. De Putter for technical assistance and N. Barnes for reading the manuscript. The authors thank anonymous referees for their useful comments. J.Ph.P. was supported by doctoral fellowships from bioMérieux and L'Association Nationale de la Recherche Technique and from L'Association pour la Recherche sur le Cancer. Funding to pay the Open Access publication charges for this article was provided by bioMérieux. Conflict of interest statement. None declared. Figures and Tables Figure 1 Microarray design. (A) Microarray map. The OLISA® microarray comprises a spotting control consisting in a biotinylated oligonucleotide and a hybridization control consisting in an oligonucleotide complementary to the spiked target. HERV-pol probe specificity was assessed by hybridization with biotinylated targets: (B) HML-2, (C) HML-4, (D) HML-5, (E) HERV-H, (F) ERV-9, (G) HERV-W, (H) HERV-E4.1, (I) HERV-R and (J) HERV-L. Figure 2 UPGMA (absolute distance) phylogenetic tree of HERV-L amplification products obtained by PCR on genomic DNA. Amplicons generated using weakly-degenerate primers (degeneracy level = 4) are depicted in grey and amplicons generated using strongly-degenerate primers (degeneracy level = 128) are depicted in black. Figure 3 Complex primer mix optimization. (A) Equimolar versus complex primer mix amplification product analysis on microarray. A target mix consisting of 106 copies of each cloned archetype was amplified either with an equimolar mix of degenerate primers (open triangle and dashed line) or a complex primer mix of degenerate primers (open circle and black line). Band intensities were measured using the Photo-Capt™ software (Vilbert-lourmat, Marne-la-Vallée, France). The average intensity for the nine HERV families was used as a reference and all intensities measured were expressed as a percentage of this reference. (B) Optimal complex primer mix concentration range. Each cloned archetype and U937 cDNA were amplified with various final primer mix concentrations (HML-2 large indent, HML-4 open rhombus, HML-5 star, HERV-H open triangle, ERV-9 cross, HERV-W open circle, HERV-E4.1 plus, HERV-R open square, HERV-L small indent and cDNA black rhombus). Figure 4 Complex primer mix characterization. (A) PCR co-amplification of various target quantities with a complex primer mix. PCRs were performed on target mixtures containing 107, 106 and 105 copies of each HERV archetype, respectively. The mean and standard deviation were calculated from experiments conducted in triplicate. (B) Amplification of unbalanced target quantities. Condition A: 105 copies of each target, condition B: 106 copies of each target, condition C: 105 copies of each target, except HML-4 and HERV-E4.1 (106 copies) and condition D: 106 copies of each target, except HML-4 and HERV-E4.1 (105 copies). Black bars represent the C/A ratio of the intensities measured and grey bars represent D/B ratio of the intensities measured. Ratios were calculated from the means of experiments conducted in triplicate. The continuous black line represents a reference value of 1. Figure 5 MD-PCR/microarray HERV transcriptome analysis of in-vitro differentiation models and paired normal/tumoral RNA. (A) Forskolin-treated BeWo b30 cells, (B) PMA-treated U937 cells, (C) breast, (D) uterus, (E) testis, (F) prostate, (G) lung and (H) colon. Each bar represents a percentage of signal intensity variation calculated from experiments conducted in triplicate between the differentiated cells or the tumoral tissue and the mock-treated cells or the normal tissue, respectively. The ∞ symbol indicates that a signal was detected for the differentiated cells or the tumoral tissue, whereas no signal was detected for the corresponding mock-treated cells or normal tissue. The Student's t-test statistical significance (P < 0.05) was indicated by *. Table 1 Degenerate primer sequences and primer mix composition HERV Forward primer Dg levela Final concentration (µM) Reverse primer Dg levela Final concentration (µM) HML-2 TRGAAAGTGTTRCCTCARGGA 8 0.51 CACAYAAAATATYATCAAYAYA 16 1.02 HML-4 TGGAAAGTCCTACCACAAGGC 1 0.03 TGCAGAGGAGATCATCCATGTA 1 0.03 HML-5 TRGAAAGTRCTTCCTSAARGR 32 1.02 CYAGTARRATATCATCCATAAA 8 0.51 HERV-H TGGRCTGTRCTGCYRCAAGGY 32 2.03 AAAGWAGAAGSTCRTCAAWATA 16 1.02 ERV-9 GTCTTGCCHCAAGGGTTT 3 0.13 AGTAAATCATCCACATAYTGAAG 2 0.13 HERV-W TGGACTGTTTTACCCCAAGGG 1 0.13 AAgTAAATYATCCMYRTACYRAc 64 1.02 HERV-E4.1 TGGACCsRGCYTCCCCAARGGb 16 1.02 CCAGCARRAGGTCATCaAYSTAd 16 1.02 HERV-R TGGACTAGTCTCCCACAAGGG 1 0.06 CCAAAAGAAGGTTGTCTATGTA 1 0.06 HERV-L TTTACTGTCCTACYTCAGGGR 4 0.25 TCARCATAATGYCATCAATGTA 4 0.51 aLevel of degeneracy. b‘s’ represents 75% of the base(s) observed at that position in the whole population. c‘g’ represents 78% of the base(s) observed at that position in the whole population. d‘a’ represents 67% of the base(s) observed at that position in the whole population.