Five normal and six DS skin fibroblasts were obtained from the American Type Culture Collection (Rockville, MD), from Coriell Laboratories (Camden, NJ), or as de-identified samples from the cytogenetics laboratory, stored by the Columbia University Pathology Department Tissue Bank. All of the cell were maintained and passaged at a ratio of 1:4 in DMEM (Invitrogen, Carlsbad, CA) containing 10% FBS (Sigma-Aldrich, St. Louis, MO). Cell senescence was scored by activity of acid β-galactosidase every 3 to 5 generations. Tissue from fetal hearts at 15 to 23 weeks gestation, from cases of +21 and controls, was obtained as de-identified cryopreserved samples from the Columbia University Pathology Department Tissue Bank. Total RNA was prepared by a two-step procedure using Trizol™ reagent (Invitrogen), followed by RNeasy (Qiagen, Valencia, CA) purification. This work was carried out in compliance with the Helsinki Declaration: since the human fibroblast and heart tissues were de-identified samples, this study was not considered human subjects research by the Columbia University Internal Review Board.

Total RNA was resolved on formaldehyde-containing agarose gels and transferred to Nytran membranes (Schleicher and Schull, Keene, NH). The probes for northern blotting were partial cDNAs prepared from RT-PCR using gene-specific primers. Hybridization was in ULTRAhyb buffer (Ambion) at 42°C degrees. Probes were stripped by boiling the membranes in 0.1 % SDS/0.1 × SSC for 2 minutes, or were allowed to decay without stripping.

Total RNA, 2 μg, was reverse transcribed with oligo-dT primers using Superscript version II reverse transcriptase (Invitrogen). Quantitative real-time RT-PCR was performed with the Master SYBR Green I system (Roche Molecular Biochemicals, Mannheim, Germany) in a LightCycler apparatus (Roche) with an initial denaturation at 95°C for 30 seconds followed by multiple cycles of denaturation at 95°C for 5 seconds, primer annealing at 62°C for 5 seconds, and extension at 72°C for 12 seconds. LightCycler Software 4.0 (Roche) was used for data analysis, with mRNA content determined from inflection points of SYBR Green fluorescence plotted as a function of cycle number. The following oligonucleotide primers were used for the PCR reactions: GART, sense primer 5'-CCATAGCTTTCCTCCAGCAG-3' and anti-sense primer 5'-AGTTCCAACGCCATCTGTTC-3'; HPRT (housekeeping gene control), sense primer 5'-TTGCTCGAGATGTGATGAAGGA-3' and anti-sense primer 5'-TTCCAGTTAAAGTGGAGAGATCA-3'.

After boiling for 10 minutes in a denaturing solution containing 12 mM Tri pH6.8, 5% glycerol, 0.4% SDS, 3 mM 2-mercaptomethanol, and 0.02% bromophenol blue, 30 μg of total protein lysates from +21 and control fibroblasts were subjected to electrophoresis on SDS-denaturating 4–20 % polyacylamide gels (Invitrogen). After transferring to the nitrocellulose membrane and blocking by 5% milk, the membrane was hybridized with mouse antibodies against p78^MX1 or α-actin (Santa Cruz Biotechnology, Santa Cruz, CA) in a solution containing 0.1% BSA, 0.1% Tween-20, and 3% dry milk overnight at 4°C. After washing, the signal was amplified using a peroxidase-conjugated goat anti-mouse IgG (Amersham Pharmarcia Biotech, Piscataway, NJ) and was visualized using a commercial ECL plus detection system (Amersham Pharmarcia Biotech).

HGU95A GeneChips were purchased from Affymetrix (Santa Clara, CA). Double stranded cDNA was generated from 5 μg of total RNA using a poly-dT oligonucleotide that contained a T7 RNA polymerase initiation site and the Superscript Choice System Kit (Invitrogen). The cDNA was phenol/chloroform extracted, and biotinylated cRNA was generated by in vitro transcription using the Bio Array High Yield RNA Transcript Labeling System (Enzo, Farmingdale, NY). The cRNA was purified using RNeasy. cRNA was fragmented according to the Affymetrix protocol, and 15 μg of biotinylated cRNA were hybridized to U95A microarrays (Affymetrix). After scanning, the expression values for each gene were determined using Affymetrix GeneChip software v. 5.0. For each sample, the signals for each gene (probe set) were normalized to the values for the entire microarray, using GeneSpring software (Silicon Genetics, Redwood City, CA). Statistical analyses and automated clustering of the microarray data were done using the Filter Genes and Gene Tree functions of GeneSpring. For supervised analysis, we used ANOVA to identify all genes that differed at p < .05 and, in a separate analysis, at p < .01, in a given cell type (early-passage fibroblast, late-passage fibroblast, fetal hearts) as a function of the presence or absence of +21. The resulting small sets of genes were then further filtered for data quality by requiring Affymetrix 'presence calls' (high-reliability hybridization data) in at least 3 samples, and were filtered a third time for magnitude of the differences in hybridization signals by requiring at least 3 samples with signals >1.2-fold above the experiment mean. We did not correct statistically for multiple comparisons (i.e. 12,000 probe sets queried in each experiment) since doing so would reduce the statistical power far below the level needed to detect changes in the 1.5-fold range. We used the Gene Tree function to group the differentially expressed genes according to the extent of similarity in their patterns of expression across the samples, thereby separating the genes into over-expressed and under-expressed clades. Fold over-expression of each gene was calculated as the mean of the normalized signals for the +21 samples divided by the mean of the control samples.

Sections of archival formaldhehyde-fixed scalp biopsies were deparaffinized in xylene and hydrated through graded ethanols. Antigen retrieval was carried out in 1 mM EDTA buffer by boiling the slides in a microwave oven for 8 minutes at the maximum power and, sequentially, boiling for 15 minutes at a reduced power. Endogenous biotin was blocked by two steps of incubation for 10–20 minutes with egg white and 5% skim milk in 1× Tris-buffered saline (TBS) containing 0.5% BSA and 0.1 % NaN3, respectively. Between the two blocking step for biotin, slides were treated with 0.3% hydrogen peroxide in 0.1% NaN3 to block endogenous peroxidase activity. Slides were washed three times in 1× TBS buffer containing 0.1% Tween-20 (TBS-T) and then incubated with 5% of the blocking serum for 30 minutes in a humidified chamber. After 3 TBS-T washes, slides were hybridized with a monoclonal antibody against p78^MX1 at room temperature overnight. After washing in TBS-T, slides were incubated with biotinylated secondary anti-mouse IgG antibody (Vector Laboratories, Burlingame, CA) for 30–40 minutes at room temperature. Antigen-antibody complexes were developed using the Vectastain ABC kit (Vector Laboratories) and a chromogenic substrate, diaminobezidine (Dako). Sections were lightly counterstained with hematoxylin. For assessing the effects of interferon in +21 conditioned medium, we used a goat anti-beta-interferon antiserum (Sigma-Aldrich).

For mRNA expression profiling we used Affymetrix U95A GeneChips, which query ~10,000 human genes (12,000 oligonucleotide probe sets), including 108 unique genes (143 probe sets) from chromosome 21. Data obtained using these microarrays in our laboratory have previously shown highly reliable correlations with independent measures of gene expression, including northern blotting 1415. We analyzed mRNA from an accessible cell type, early- and late-passage fibroblasts, with and without +21. The growth characteristics of the fibroblast cultures are listed in Table 1. This table shows the percentages of cells staining for acid beta-galactosidase, a marker that scores positively in senescent cells and negatively in pre-senescent cells. We first asked whether there might be strong and widespread changes in gene expression due to the extra copy of chromosome 21 in the early and/or in late passage fibroblasts. Inspection of the results of non-supervised hierarchical clustering of the expression data argued against this, since this procedure did not detect the presence or absence of +21 (data not shown). We therefore did a supervised analysis of the data, attempting to identify small subsets of genes that might differ significantly in their expression as a function of the presence or absence of +21. As shown in Figure 1 and Tables 2 and 3, ANOVA with a cutoff of p < .01 yielded small sets of genes that were 'high in +21' or 'low in +21'. We repeated the ANOVA using a less stringent cutoff of p < .05, and the resulting larger gene lists are shown as supplemental tables [see Additional file 1 and Additional file 2].

Several noteworthy features were seen in the expression data from the early passage cells. First, genes on chromosome 21 were strikingly over-represented, relative to those on other chromosomes, in the 'high in +21' set. The proportion of chromosome 21 genes was 6/14 (43%) by ANOVA at p < .01, Table 2, and 17/58 (29%) by ANOVA at p < .05 [see Additional file 1]. By comparison, the random expectation based on the number of genes on chromosome 21 divided by the number of genes on all human chromosomes represented on the microarray is only ~1%. Second, no genes on chromosome 21 were present in the 'low in +21' sets. Third, in these early passage cells the magnitude of the observed over-expression for most of the genes in the 'high in +21' subsets was small, generally within the range expected from gene dosage (1.5X) (Table 2 and [see Additional file 1]). These results indicate that the cell cannot fully adjust for the presence of the extra chromosome 21, i.e. at least some genes on this chromosome are not subject to compensatory down-modulation in the trisomic early passage fibroblasts.

The expression data in the late passage fibroblasts showed similarities and differences compared to the results in the early passage cells. Only a single chromosome 21 gene passed the 1-way (single parameter, presence or absence of +21) ANOVA test at p < .01 in these senescing cells, and this gene, MX1, was over-expressed far in excess of the 1.5-fold change expected from chromosome dosage in the +21 fibroblasts (mean fold increase compared to control fibroblasts = 16, Table 3). In addition to MX1, a second closely related chromosome 21-linked gene, MX2, passed the less stringent ANOVA at p < .05 [see Additional file 2]. The over-expression of MX1 in senescing +21 fibroblasts, with progressively increasing expression with increasing passage number, was verified by northern blotting (Fig. 2b,c). In normal fibroblasts, we observed a small increase in MX1 expression at late passages, but these cells never achieved the very high levels of MX1 mRNA seen in the +21 lines (Fig. 2a–c). Accordingly, in 2-way ANOVA using both p < .01 and p < .001 criteria, with the 2 parameters of passage (early, late) and genetic characteristic (normal, +21), the MX1 gene did not appear among the genes whose mRNA levels were associated with senescence independently of genetic characteristic, but did appear among the genes whose mRNA levels were associated with the presence or absence of +21. The gene list from the 2-way ANOVA with a cut-off of p < .001 for differential mRNA expression as a function of early vs. late passage, independent of genetic characteristic, is provided as a supplemental table [see Additional file 3]. While we will remain focused primarily on the effects of +21 in this report, this list of candidate general senescence-associated genes may be useful for future studies.

The correlation coefficient for the level of MX1 mRNA (GeneChip data) vs. the percentage of acid beta-galactosidase-positive cells was 0.63 for normal fibroblasts and 0.72 for the +21 fibroblasts. Thus, MX1 expression was correlated to cell senescence in both types of fibroblasts, albeit with a stronger positive correlation in +21 compared to control, and with a substantially greater magnitude of the increased expression in the +21 cells. In addition, western blotting of total cellular protein indicated that the excess of MX1 mRNA in these cells led to the expected excess production of p78^MX1 protein, and we observed a good correlation between MX1 mRNA and p78^MX1 protein (Fig. 3a).

The MX1 gene is a well known target of type-I interferon signaling, as is MX2 1617. While the increase in MX1 mRNA was the most dramatic effect seen during +21 fibroblast senescence, re-probing of the northern blots with an additional interferon-inducible gene, IFI27, revealed a weaker but parallel increase in mRNA expression (Fig. 2a). IFI27 showed a 9-fold mean increase in the late-passage +21 fibroblasts in our microarray data, but this gene was screened out by our data analysis protocol due to an insufficient number of Affymetrix presence calls (see Methods). Although IFI27 is not on chromosome 21, the receptors for interferon alpha and beta are encoded by genes on chromosome 21. While we did not detect increased mRNA for these genes in our survey, fibroblasts and other cells with +21 are known to express an increased density of cell surface interferon receptors at the protein level, and to be hyper-responsive to signaling by exogenous interferon 18192021222324. Gene-gene interaction, namely the presence of an extra gene copy of both an upstream and a downstream component of the interferon-signaling pathway, is a likely explanation for the non-linear over-expression of MX1 mRNA and protein in +21 cells. Indeed, in early work the protein product of the MX1 gene, p78^MX1, was found over-expressed in two lines of +21 fibroblasts, compared to control fibroblasts, after these cells were exposed to exogenous interferon 2526. Our current data add to this information by showing strong constitutive over-expression of MX1 in senescing +21 fibroblasts, and we therefore wished to test whether this increased expression might be maintained by autocrine interferon signaling. Consistent with this hypothesis, conditioned medium from +21 fibroblast cultures induced the expression of MX1 mRNA in control fibroblasts, and this induction was attenuated by the addition of anti-interferon-beta antibodies to the cell cultures (Fig. 3b).

A previous genetic association study linked intronic polymorphisms in the MX1 gene to the inflammatory hair loss disorder alopecia areata 27. This condition has long been recognized to have an increased (up to 90-fold) prevalence in DS 28293031. We therefore used immunohistochemistry to assess the expression of p78^MX1 in scalp biopsies. In control scalp biopsies, p78^MX1 expression was restricted to the capillary endothelial cells (data not shown). In both of two cases of sporadic alopecia areata that we examined there was high expression both in endothelial cells and in a subset of the infiltrating lymphocytes and in epithelial cells (outer root sheath cells and matrix cells) of the inflamed hair follicles, with no immunoreactivity in the adjacent uninvolved hair follicles (Fig. 4a–c and data not shown). We did not have access to alopecia areata biopsies from DS, but examining such cases will be desirable in future studies.

A similar supervised analysis of microarray data from 5 fetal hearts with +21 compared to 8 control fetal hearts matched for gestational age (mean G.A. in +21 cases = 16.8 weeks; mean G.A. in controls = 18.5 weeks, t-test p = .2) also revealed a systematic over-expression of chromosome 21-linked genes in the +21 cases. Using ANOVA at a cutoff of p < .01, among 37 genes in the 'high in +21' category, 7 (19%) mapped to chromosome 21; while no chromosome 21-linked genes were in the under-expressed category (Fig. 5 and Table 4). Similarly, the statistical analysis of the microarray data for these samples at p < .05 showed 17/110 (15%) of the over-expressed genes mapping to chromosome 21, a proportion much higher than the 1% expected from the overall gene representation on this small chromosome [see Additional file 4]. The chromosome 21 gene with the strongest over-expression in this dataset was GART, encoding a metabolic enzyme with three catalytic activities (phosphoribosylglycinamide formyltransferase, phosphoribosylglycinamide synthetase, phosphoribosylaminoimidazole synthetase), which acts at several steps of purine biosynthesis 32. In addition, several other chromosome 21 genes, including SH3BGR, encoding a coiled-coil and SH3 domain-containing protein, were modestly over-expressed on average in the +21 hearts (Table 4 and [see Additional file 4]). GART mRNA, while not an abundant transcript, is detectable by northern blotting with prolonged exposures, and we were able to verify the strong (non-linear) over-expression of this gene, and the modest over-expression of SH3BGR mRNA, by northern blotting and Phosphorimaging in 3 of the hearts with +21 for which ample RNA remained after the GeneChip experiment, compared to 4 control hearts closely matched for gestational age (Fig. 6a). This result was further confirmed by real-time RT-PCR comparing GART mRNA expression (normalized to mRNA from the HPRT housekeeping gene) in 4 of the +21 fetal hearts compared to 6 of the control fetal hearts from our GeneChip series (Fig. 6b).

As seen by comparing Tables 2, 3, 4 and the supplemental tables [see Additional file 1], [see Additional file 2], [see Additional file 4], the genes differentially expressed as a function of +21 in the fetal hearts did not overlap substantially with the genes that were differentially expressed in the early- or late-passage fibroblasts. The two exceptions were SH3BGR, which was modestly (i.e. linearly) over-expressed in the hearts and early passage fibroblasts with +21, and MX2, which was modestly over-expressed in both the hearts and the late passage fibroblasts. The limited overlaps between the gene lists from the supplemental tables are highlighted by the Venn diagram in Figure 7. Among the few genes that appear in the regions of overlap in this diagram, a strikingly large proportion map to chromosome 21.