• Open Access

Genome-scale expression profiling of Hutchinson–Gilford progeria syndrome reveals widespread transcriptional misregulation leading to mesodermal/mesenchymal defects and accelerated atherosclerosis


  • Antonei B. Csoka,

    Corresponding author
    1. Department of Molecular Biology, Cell Biology and Biochemistry, Brown University, Providence, RI 02912, USA
      Antonei B. Csoka, Pittsburgh Development Center, Magee-Women's Research Institute, Department of Obstetrics, Gynecology & Reproductive Sciences, University of Pittsburgh School of Medicine, Pittsburgh, PA 15213, USA. E-mail: acsoka@pdc.magee.edu
    Search for more papers by this author
    • *

      A.B.C. and S.B.E. contributed equally to this work.

  • Sangeeta B. English,

    1. Children's Hospital Informatics Program, Children's Hospital, Boston, MA 02115, USA
    Search for more papers by this author
    • *

      A.B.C. and S.B.E. contributed equally to this work.

  • Carl P. Simkevich,

    1. Department of Molecular Biology, Cell Biology and Biochemistry, Brown University, Providence, RI 02912, USA
    Search for more papers by this author
  • David G. Ginzinger,

    1. Genome Analysis Core Facility, Comprehensive Cancer Center, University of California, San Francisco, CA 94143, USA
    Search for more papers by this author
  • Atul J. Butte,

    1. Children's Hospital Informatics Program, Children's Hospital, Boston, MA 02115, USA
    Search for more papers by this author
  • Gerald P. Schatten,

    1. Pittsburgh Development Center, Magee-Women's Research Institute, Department of Obstetrics, Gynecology & Reproductive Sciences, University of Pittsburgh School of Medicine, Pittsburgh, PA 15213, USA
    Search for more papers by this author
  • Frank G. Rothman,

    1. Department of Molecular Biology, Cell Biology and Biochemistry, Brown University, Providence, RI 02912, USA
    Search for more papers by this author
  • John M. Sedivy

    1. Department of Molecular Biology, Cell Biology and Biochemistry, Brown University, Providence, RI 02912, USA
    Search for more papers by this author

Antonei B. Csoka, Pittsburgh Development Center, Magee-Women's Research Institute, Department of Obstetrics, Gynecology & Reproductive Sciences, University of Pittsburgh School of Medicine, Pittsburgh, PA 15213, USA. E-mail: acsoka@pdc.magee.edu


Hutchinson–Gilford progeria syndrome (HGPS) is a rare genetic disease with widespread phenotypic features resembling premature aging. HGPS was recently shown to be caused by dominant mutations in the LMNA gene, resulting in the in-frame deletion of 50 amino acids near the carboxyl terminus of the encoded lamin A protein. Children with this disease typically succumb to myocardial infarction or stroke caused by severe atherosclerosis at an average age of 13 years. To elucidate further the molecular pathogenesis of this disease, we compared the gene expression patterns of three HGPS fibroblast cell strains heterozygous for the LMNA mutation with three normal, age-matched cell strains. We defined a set of 361 genes (1.1% of the approximately 33 000 genes analysed) that showed at least a 2-fold, statistically significant change. The most prominent categories encode transcription factors and extracellular matrix proteins, many of which are known to function in the tissues severely affected in HGPS. The most affected gene, MEOX2/GAX, is a homeobox transcription factor implicated as a negative regulator of mesodermal tissue proliferation. Thus, at the gene expression level, HGPS shows the hallmarks of a developmental disorder affecting mesodermal and mesenchymal cell lineages. The identification of a large number of genes implicated in atherosclerosis is especially valuable, because it provides clues to pathological processes that can now be investigated in HGPS patients or animal models.


Hutchinson–Gilford progeria syndrome (HGPS; Online Mendelian Inheritance in Man entry no. 176670) is a rare syndrome affecting approximately one in 8 million births (DeBusk, 1972). The affected individuals appear normal at birth, but during the first few years of their lives recapitulate many of the pathologies of normal human aging, with some exceptions such as neurosensory decline, dementia and cancer. They exhibit normal intelligence. At the age of 5 years and beyond, widespread atherosclerosis, most evident in the coronary arteries and the aorta, develops and rapidly progresses, usually resulting in death by myocardial infarction or stroke in the early teens. The genetic basis for HGPS was recently shown to be a sporadic and recurrent single base change in codon 608 of the LMNA gene encoding the structural nuclear protein lamin A/C. (De Sandre-Giovannoli et al., 2003; Erikkson et al., 2003). This single base substitution activates a cryptic splice site, leading to an altered lamin A protein in which 50 amino acids near the carboxyl terminus are deleted.

Lamin A is a structural component of the nuclear lamina, a polymeric proteinaceous network that covers the inner surface of the nuclear membrane, but also extends as a fibrous network into the interior of the nucleus. The LMNA gene is developmentally regulated, and is expressed in differentiated cells of many, but not all, tissues, whereas the B-type lamins are expressed throughout development (Goldman et al., 2002; Shumaker et al., 2003). The nuclear lamina was originally thought to play mostly a structural role, but it is increasingly being shown to participate in a variety of cellular processes, including DNA replication and transcription. Disorganization of the lamina with dominant negative lamin mutants inhibits DNA replication and transcription by RNA polymerase II (Spann et al., 2002).

Lamin A/C has been shown to interact with specific proteins that affect transcription, such as the retinoblastoma (pRb) tumour suppressor (Markiewicz et al., 2002). The lamina becomes depolymerized during the disassembly of the nuclear membrane in mitosis, and improper assembly at the end of mitosis leads to cell death (Steen & Collas, 2001). Intranuclear lamin foci also co-localize with RNA splicing factors, suggesting that lamins might contribute to organizing the RNA processing machinery (Jagatheesan et al., 1999). However, none of these effects has been shown to possess the specificity of action suggested by the phenotypes of human laminopathies. In this communication we extend earlier studies of gene expression in HGPS fibroblasts (Ly et al., 2000; Park et al., 2001) to include the currently measurable human genome.


The spectrum of genes with altered expression in HGPS

We defined a set of 361 genes that were differentially expressed in HGPS fibroblasts (see Supplementary material for complete gene list). Of these, 193 were up-regulated and 168 were down-regulated. The total set of 361 genes corresponds to approximately 1.1% of the 33 000 annotated genes represented on the Affymetrix U133A/B chips. Using the Gene Ontology (GO) classifications we found that the affected genes participate in a wide variety of biological processes (Fig. 1). Approximately 20% of the gene products are located in the nucleus, and over 30% in the various membrane categories.

Figure 1.

Gene Ontology classification of gene expression changes. Top panel: sorted by Biological Process. Bottom panel: sorted by Cellular Component. Analysis was performed on the well-annotated genes found on chip A. The cutoff for inclusion was five or more genes per category. The percentage of affected genes (genes on list) and the percentage of genes in that category represented on the chip (genes on chip) was calculated taking the total number of well-annotated genes as 100%. The categories used in this analysis correspond to the terminal categories adopted by the Gene Ontology Consortium.

We also compared our data with two previously published HGPS microarray studies. Using a single progeria cell line, Park et al. (2001) examined 384 known genes and reported four > 2-fold changes, which were not confirmed by us. Ly et al. (2000) monitored the expression of approximately 6000 genes and found 76 altered in progeria. Our results agree in that approximately 1% of genes appear to be differentially expressed in HGPS, and that many of them encode ECM components. Fifty-eight genes in the Ly et al. list were identified by GenBank accession numbers that are present on our arrays and we found that 17 of these have the same direction of change. We consider the agreement between the two studies to be reasonable, considering that the previous study lacked a statistical filter or replicates, and that only two of the three cell lines used have been shown to contain the HGPS LMNA point mutation (Erikkson et al., 2003).

The largest functional category we found is transcription factors (39 genes), with representation of several major families, such as homeobox, Zn-finger, bHLH, HMG-box and forkhead proteins (Table 1). Interestingly, a significant fraction (29 of 39, 74%) are involved in the regulation of embryonic development and tissue differentiation, and most of these (20/29, 69%) are repressed in HGPS cells. Prominent physiological functions (10/39) are bone and skeletal involvement, limb patterning and myogenesis. The most highly affected gene of the entire 361 gene set, MEOX2/GAX (up-regulated in HGPS; +29.1-fold), is a homeobox factor that functions as a negative regulator of proliferation in several mesodermal tissues (LePage et al., 1994).

Table 1.  Genes implicated in transcription and extracellular matrix functions
SymbolLocusLinkChange*Gene designationAnnotated functionsAB
  • *

    Genes within a category are listed according to fold-change, from the most up-regulated (top) to the most down-regulated (bottom); negative sign indicates down-regulated genes. This Table was compiled by rigorous manual annotation of biological function using all available reference sources; the categories shown thus comprise several terminal GO categories listed in Fig. 1.

  • A = Bone and skeletal involvement.

  • B = Vascular and arteriosclerosis involvement.

 MEOX2  4223 29.1Mesenchyme homeo box 2Divergent homeobox gene, regulator of limb myogenesisXX
 FOXE1  2304 13.2Forkhead box E1Thyroid morphogenesis, mutations cause hypothyroidism  
 OAZ 23090  5.2OLF-1/EBF associated zinc fingerKruppel-like C2H2 Zn-finger, developmental function unknown  
 BTEB1   687  3.9Basic transc. element binding 1Basal factor, binds GC box elements  
 GATA6  2627  3.9GATA binding protein 6Zn finger, early development, visceral endoderm differentiation X
 TOX  9760  3.6Thymus high mobility group boxArchitectural HMG transcription factor, induces DNA bending  
 STAT1  6772  3.3Signal transducer activator transc. 1Cell viability in response to interferons and cytokines  
 MSC  9242  3.1MusculinE-box binding bHLH factor, B-cell receptor signalling pathway  
 AHR   196  2.8Aryl hydrocarbon receptorXenobiotics metabolism, planar aromatic hydrocarbons  
 TCF7L1 83439  2.5Transcription factor 7-like 1T-cell specific, HMG-box factor, targets uncertain  
 TRIM22 10346  2.3Tripartite motif-containing 22Interferon induced, may mediate antiviral effects  
 TRIM5 85363  2.3Tripartite motif-containing 5Zn-binding, RING and coiled-coil domains, function unknown  
 HES1  3280  2.2Hairy and enhancer of split 1Antagonizes bHLH factors, regulates housekeeping functions  
 NMI  9111  2.0N-myc (and STAT) interactorModifies activity of transcription factors, function uncertain  
 HIVEP1  3096  2.0HIV type I enhancer binding 1Activator of MHC, immunoglobulin, interferon-regulated genes  
 NCOA3  8202  2.0Nuclear receptor coactivator 3Recruits CBP and PCAF, nucleates co-activation complex  
 KIAA1549 57670 −2.0KIAA1549 proteinPutative transcription factor, function unknown  
 PMX2 51450 −2.1Paired related homeobox proteinExpressed in fetal but not adult fibroblasts, dermal regeneration  
 SOX4  6659 −2.1Sex determining region Y-box 4HMG-box factor, parathyroid hormone signalling, bone developmentX 
 MYBL1  4603 −2.2v-myb oncogene homologue-like 1Expressed in testis and peripheral leucocytes  
 TBX3  6926 −2.3T-box 3 (ulnar mammary syndr.)Repressor, limb and skeletal development, also hair, genitalX 
 HOXA9  3205 −2.9Homeo box A9Similar to Drosophila abdominal-B (Abd-B), myeloid differentiation  
 MEIS1  4211 −2.9Viral integration site 1 homologueHomeobox factor, limb development, leukaemogenesis  
 TWIST2117581 −3.0Twist homologue 2bHLH repressor, osteoblast maturation, NF-kB signallingX 
 MSX1  4487 −3.1Msh homeo box homologue 1Cranio-facial, dental development, cleft palateX 
 SNAI1  6615 −3.1Snail homologue 1Zn finger, repressor of ectodermal genes, mesoderm differentiation  
 NKX3-1  4824 −3.2NK3 transc. factor related-1Homeobox factor, prostate, androgen-regulated, tumour suppressor  
 NBL1  4681 −3.3Neuroblastoma tumour suppressor 1Negative cell cycle regulator, up-regulated in osteoblast differentiationX 
 TCF7  6932 −3.4Transcription factor 7HMG-box repressor, thymocyte differentiation  
 FOXF2  2295 −3.6Forkhead box F2Expressed in lung, placenta, activates lung-specific genes  
 TFAP2C  7022 −4.6Transcription factor AP-2 gammaAP2 family, neural crest, skin, extra-embryonic tissues  
 EMX2  2018 −4.9Empty spiracles homologue 2Central nervous system, urogenital development  
 NR1D1  9572 −4.9Nuclear receptor sub 1, group D, 1Circadian rhythms, triglyceride metabolism, atherosclerosis X
 BCL11B 64919 −5.5B-cell CLL/lymphoma 11BC2H2-type zinc finger factor, functions unknown  
 IRX1 79192 −5.7Iroquois homeobox protein 1Multiple roles in embryonic pattern formation  
 PBX1  5087 −6.6Pre-B-cell leukaemia transc. factor 1Pancreatic development, also has roles in limb patterning  
 EYA2  2139 −7.1Eyes absent homologue 2Late eye development, limb connective tissue, myogenesis  
 IRX5 10265 −8.1Iroquois homeobox protein 5Multiple roles in embryonic pattern formation  
 TBX5  6910−10.4T-box 5Heart and limb development, Holt–Oram syndrome  
Extracellular matrix
 NTN4 59277 15.5Netrin 4Component of basement membranes, has laminin domain  
 MGP  4256  7.4Matrix Gla proteinComponent of bone, cartilage, ECM, inhibitor of calcificationXX
 COL4A5  1287  7.4Collagen, type IV, alpha 5Major structural component of basement membranes  
 NID2 22795  5.3Nidogen 2 (osteonidogen)Ubiquitous basement membrane component, binds collagen  
 OSF-2 10631  4.8Osteoblast specific factor 2Ligand for integrins, cell motility, osteoblast differentiationX 
 TNA  7123  4.5Tetranectin (plasminogen binding)Induced during the mineralization phase of osteogenesisX 
 CRTL1  1404  4.3Cartilage linking protein 1ECM, stabilizes aggregates of aggrecan and hyaluronanX 
 RELN  5649  3.7ReelinECM component, cell–cell interactions, cell migration  
 COL4A1  1282  3.6Collagen, type IV, alpha 1Major structural component of basement membranes  
 LAMA5  3911  3.3Laminin, alpha 5Basement membrane component, cell attachment, migration  
 ELN  2006  3.1ElastinComponent of elastic ECM fibres, Williams–Beuren syndrome X
 AGC1   176  3.1Aggrecan 1Chondroitin sulphate proteoglycan core protein, ECM, cartilageX 
 HAS3  3038  3.1Hyaluronan synthase 3Regulates synthesis of the ECM component hyaluronan  
 COL4A2  1284  2.5Collagen, type IV, alpha 2Major structural component of basement membranes  
 MMP14  4323 −2.0Matrix metalloproteinase 14Membrane-attached, activates MMP2, tumour invasion  
 CA12   771 −2.6Carbonic anhydrase XIITransmembrane enzyme, hydration of CO2, bone resorptionX 
 ITGA2  3673 −2.7Integrin, alpha 2 (VLA-2 receptor)Transmembrane protein, cell adhesion, attachment signalling  
 TFPI2  7980 −3.0Tissue factor pathway inhibitor 2ECM component, protease inhibitor, blood coagulation  
 COL10A1  1300 −3.6Collagen, type X, alpha 1Expressed by hypertrophic chondrocytes during ossificationX 
 NCAM1  4684 −3.8Neural cell adhesion molecule 1Cell adhesion, neuron–neuron, neuron–muscle interactions  
 SGCD  6444 −3.9Sarcoglycan, deltaTransmembrane protein, bridges inner cytoskeleton and ECM  
 TNC  3371 −3.9Tenascin CECM protein in tendons, smooth muscle, embryonic boneX 
 SDC1  6382 −4.0Syndecan 1Integral membarne proteoglycan, ECM receptor  
 NINJ2  4815 −4.2Ninjurin 2Cell adhesion, regeneration of peripheral nervous system  
 COL8A2  1296 −4.3Collagen, type VIII, alpha 2Basement membrane component of corneal endothelial cells  
 HAS2  3037 −4.8Hyaluronan synthase 2Synthesizes high-molecular-weight hyaluronic acid  
 DPT  1805 −5.0DermatopontinECM component, cell–matrix interactions, TGF-β signalling  
 COMP  1311 −5.1Cartilage oligomeric matrix proteinNon-collagen ECM protein, defects in pseudoachondroplasiaX 
 HS3ST3A1  9955 −5.7Glucosamine sulfotransferase 3A1Enzyme in heparan sulphate biosynthesis  
 MMP3  4314 −6.7Matrix metalloproteinase 3Breakdown of ECM, function in wound repair, atherosclerosis X

Several genes involved in DNA replication and chromatin remodelling were also noted: a minichromosome maintenance factor (MCM5, −2.4), regulators of histone acetylation (ING1, −3.0; TWIST2, −3.0) and a silencer of heterochromatin (SALL1, −4.1). Other related genes were also affected, but less significantly (MCM2, −1.7; MCM3, −1.7; MCM4, −2.0; MCM7, −1.6; the licensing factor CDT1, −1.8; CFAF1A chromatin assembly factor, −1.6; TWIST1, −1.9). Interestingly, all were down-regulated in HGPS. The only up-regulated gene in the chromatin category was HDAC9, a histone deacetylase (2.5-fold). Viewed together, these changes are indicative of reduced cellular proliferation.

The second largest category is extracellular matrix (ECM) components (30 genes), including enzymes involved in ECM synthesis or modification (Table 1). Production of the extracellular matrix is an important function of mesenchymal cells, and previous studies have indicated that mesoderm-derived lineages, including cardiomyocytes, cardiovascular endothelium, bone, cartilage and adipose, are severely affected in HGPS. All the major components of the basement membrane, with the exception of perlecan, are significantly up-regulated in HGPS: type IV collagens alpha 1 (COL4A1, +3.6), alpha 2 (COL4A2, +2.5), alpha 5 (COL4A5, +7.4), laminin alpha 5 (LAMA5, +3.3), netrin 4 (NTN4, +15.4) and nidogen 2 (NID2, +5.3). The gene expression changes are suggestive of excess ECM deposition (increased expression of ECM components, decreased expression of ECM remodelling enzymes); evidence for these effects can now be sought in autopsy material. Increased ECM deposition would be expected to have structural effects on tissue function, and could also result in signalling imbalances with adverse effects on neighbouring cells of various lineages.

Candidate genes involved in atherosclerosis

In terms of tissue and organ function, the largest category (31 affected genes) show clear vascular and atherosclerosis involvement: three transcription factors, three ECM components (Table 1) and 25 additional genes with a variety of other molecular functions (Table 2). Skeletal, limb and cartilage involvement is the second most prevalent category (22 affected genes): ten transcription factors, nine ECM components (Table 1) and three additional genes with other functions (Table 2). The increased expression of two transcription factors, MEOX2/GAX (+29.1) and GATA6 (+3.9), both of which are known negative regulators of vascular smooth muscle cell proliferation, may antagonize cell cycle entry of vascular smooth muscle cells and thus impede tissue regeneration. Autopsies of HGPS patients have described extensive depletion of smooth muscle in the aortic media (Stehbens et al., 2001; Ackerman & Gilbert-Barness, 2002). MEOX2/GAX is also involved in skeletal and craniofacial development as well as limb myogenesis (Walsh & Takahashi, 2001); given its significant misregulation and the clinical features of HGPS, this gene warrants further scrutiny.

Table 2.  Additional genes implicated in skeletal and cardiovascular functions
SymbolLocusLinkChange*Gene designationAnnotated functions
  • *

    Genes within a category are listed according to fold-change, from the most up-regulated (top) to the most down-regulated (bottom); negative sign indicates down-regulated genes. This Table was compiled by rigorous manual annotation of biological function using all available reference sources; the categories shown thus comprise several terminal GO categories listed in Fig. 1.

  • Involved in both vascular development and bone formation.

Bone and cartilage involvement
 TNFRSF11B  4982 3.0OsteoprotegerinTNF receptor superfamily, decoy receptor, regulates bone resorption
 STC1  6781−3.1Stanniocalcin 1Calcium-regulated hormone, calcium homeostasis
 ENPP1  5167−6.8Ectonucleotide pyrophosphatase 1Transmembrane enzyme, broad specificity, calcification of joints
Vascular and arteriosclerosis involvement
 CCL2  634712.1Chemokine (C–C motif) ligand 2Chemotactic activity, psoriasis, rheumatoid arthritis, atherosclerosis
 PTX3  5806 7.3Pentaxin-related geneSecreted protein, vascular inflammation, microbial pathogens
 F2R  2149 6.4Coagulation factor II receptorThrombotic response, endothelial cell function, vascular development
 PTGER4  5734 5.3Prostaglandin E receptor 4Transmembrane receptor, vascular development, also bone formation
 F2RL2  2151 4.8Coagulation factor II receptor-like 2Transmembrane receptor, related to F2R (above), signalling
 EDIL3 10085 4.0EGF repeats, discoidin I domains, 3Integrin ligand, angiogenesis, vessel wall remodelling, development
 PLA2G4A  5321 3.9Phospholipase A2, group IVACytosolic, calcium-dependent, inflammation, haemodynamic regulation
 PTGER1  5731 3.7Prostaglandin E receptor 1Transmembrane receptor, pain perception, blood pressure regulation
 KRT7  3855 3.7Keratin 7Cytokeratin, expressed in simple epithelia, glands, blood vessels
 APOL3 80833 3.7Apolipoprotein L, 3HDL family, cytoplasmic lipid carrier, endothelial cells, TNFα-regulated
 CCR1  1230 3.7Chemokine (C–C motif) receptor 1Seven transmembrane protein, inflammatory response
 THBS2  7058 3.6Thrombospondin 2Secreted protein, tumour suppressor, inhibits microvascular development
 OSBPL10114884 3.5Oxysterol binding protein-like 10Intracellular lipid receptor, responses to cytotoxic oxidized sterols
 PTGES  9536 2.7Prostaglandin E synthaseProstaglandin E2 synthesis, microsomal enzyme, IL-1β induced
 FYCO1 79443 2.5FYVE and coiled-coil domain 1Zn finger, GTPase domain, expressed in heart, skeletal muscle
 CALD1   800 2.3Caldesmon 1Calmodulin, actin-binding protein, regulates smooth muscle contraction
 TEK  7010 2.3TEK tyrosine receptor kinaseEndothelial–smooth muscle cell communication, angiopoietin-1 is ligand
 ACTA2    59 2.3Actin, alpha 2Major constituent of contractile apparatus in smooth muscle of aorta
 PROS1  5627 2.2Protein S (alpha)Plasma protein, inhibits clotting, deficiency predisposes to thormbosis
 LRP8  7804−2.0LDL receptor-related protein 8Internalization of apolipoprotein E
 OSBPL6114880−2.1Oxysterol binding protein-like 6Intracellular lipid receptor, responses to cytotoxic oxidized sterols
 COLEC12 81035−2.7Collectin subfamily member 12Cell surface glycoprotein, binds carbodydrates, oxidized phospholipids
 PTGER2  5732−3.1Prostaglandin E receptor 2Transmembrane receptor, female fertility, blood pressure regulation
 PTGS1  5742−4.0Prostaglandin cyclooxygenase 1Prostaglandin biosynthesis, angiogenesis of endothelial cells
 HMOX1  3162−6.0Heme oxygenase (decycling) 1Oxidantive stress protection, deficiency predisposes to arterial disease

Five genes involved in inflammatory processes were up-regulated: monocyte chemoattractant protein 1 (CCL2, +12.1), pentaxin 3 (PTX3, +7.3), the prostaglandin E receptors 4 and 1 (PTGER4, +5.3; PTGER1, +3.7), and prostaglandin E synthase (PTGES, +2.7); two genes were down-regulated: prostaglandin cyclooxigenase (PTGS1, −4.0) and the prostaglandin E receptor 2 (PTGER2, −3.1). CCL2 (also called MCP-1) is involved in attracting monocytes into the intima at sites of arterial lesions; its role in initiating atherogenesis is supported by reduced lipid deposition in the aorta in atherosclerotic mice deficient in CCL2 (Gu et al., 1998). PTX3 plays a role in the regulation of innate resistance to inflammatory reactions, and is strongly expressed in atherosclerotic arteries (Mantovani et al., 2003). Prostaglandin E2 is involved in the regulation of inflammatory processes, blood pressure, fertility and pain perception. PTGER4 plays a role in suppressing macrophage chemokine production, and PTGER1 and PTGER2 are involved in regulating blood pressure.

Matrix GLA protein (MGP, +7.4) is involved in the spontaneous calcification of arteries and cartilage in mice, exerting a negative regulatory effect (Munroe et al., 1999). Matrix metalloproteinase 3 (MMP3, −6.7) modulates ECM accumulation and is involved in the growth of atherosclerotic plaques (de Maat et al., 1999). Nuclear receptor subfamily 1, group D, member 1 (NR1D1, −4.9) is a repressor of the apolipoprotein CIII gene in the liver. The decreased expression of NR1D1 may be relevant to atherosclerosis in HGPS because of the important role of apolipoprotein CIII in the metabolism of triglyceride-rich remnant lipoproteins. Heme oxygenase (decycling) 1 (HMOX1, −6.0) is a microsomal enzyme that catalyses the rate-limiting step in the degradation of heme, and is induced by a variety of injurious stimuli as part of antioxidant defence mechanisms. Protective effects of HMOX1 in atherosclerosis have been indicated in several animal models (Ishikawa & Maruyama, 2001; Juan et al., 2001). Decreased expression of this protective gene may thus be important in the pathogenesis of atherosclerosis in HGPS, and pharmacological interventions that induce HMOX1 are immediately suggested as therapeutics (Duckers et al., 2001).

The thrombin receptor PAR-1 (F2R, +6.4) and its related receptor PAR-3 (F2RL2, +4.8) are both involved in platelet aggregation and endothelial cell proliferation. In a mouse model, the absence of PAR-1 reduced vascular injury (Major et al., 2003). PAR-1 appears to be involved in the proliferation and accumulation of smooth muscle cells during vascular injury, and inhibition of PAR-1 may be an effective strategy for reducing vascular injury in restenosis. The existing many agonists and antagonists of PAR-1 may represent valuable treatment possibilities.

Our initial studies included strain AG10578, which was subsequently found not to carry the codon 608 mutation, but has instead a region of apparent uniparental isodysomy on chromosome 1 that includes the LMNA gene (Erikkson et al., 2003). Remarkably, AG10578 resembles the control lines in its gene expression patterns (data not shown). The occurrence of four independent reversion events in the relatively small sample size of HGPS cell strains available worldwide, and the discovery of this phenomenon in other diseases (Hirschhorn, 2003), indicates that these events are likely to be under strong positive selection (Luengo et al., 2002; Erikkson et al., 2003).


Comparisons of cellular senescence and organismal aging

We took all possible precautions to ensure that our cultures were free of influences that may promote cellular senescence or quiescence, as well as other possibly stressful conditions: we used early passage cultures well before the onset of senescence, cells were grown at subconfluent conditions in enriched medium with generous serum supplementation, and incubation was in a hypoxic atmosphere of 2% O2 to reduce oxidative stress. The absence of expression changes of positive cell-cycle regulators such as E2F1 and c-Myc as well as the normal expression of cyclins confirms that the HGPS cells were actively proliferating.

The HGPS expression profile shows important differences from the profile of fibroblasts passaged into replicative senescence (Shelton et al., 1999; Ly et al., 2000; Semov et al., 2002; Lindvall et al., 2003). For example, expression of the cyclin-dependent kinase inhibitor p21, a key regulator of the senescence response, is unchanged in HGPS cells. Inflammatory responses are evident in both HGPS and senescent fibroblasts, with both showing increased expression of CCL2 (monocyte chemotactic protein, MCP-1). However, other responses differ: interleukin-1β (IL-1β) and IL-15 expression is increased in senescent but not in HGPS fibroblasts, and PTX3 is increased in HGPS but not in senescent fibroblasts. Interestingly, both ECM components and ECM-remodelling enzymes are affected in HGPS and senescent fibroblasts, but in opposite directions: MMP3 (stromelysin-1) is down-regulated in HGPS and up-regulated in senescent fibroblasts, whereas elastin is up-regulated in HGPS and down-regulated in senescent fibroblasts. Of particular note are that significant changes in the expression of transcription factors in HGPS, especially those involved in the development and differentiation of mesenchymal cells, are absent in senescent fibroblasts. It is evident that the HGPS gene expression patterns do not correspond to those observed in replicative cellular senescence.

Microarray studies of aged muscle in both mouse and rhesus monkey have revealed a selective up-regulation of transcripts involved in inflammation and oxidative stress, and down-regulation of genes involved in mitochondrial electron transport and oxidative phosphorylation (Kayo et al., 2001). Because many of these changes can be reversed by caloric restriction, which significantly extends lifespan in both species, it has been suggested that these gene expression patterns are signatures of organismal aging. The HGPS gene expression profile does not show similarities to these patterns; for example, metabolic and mitochondrial enzymes are minimally affected. Conversely, as is the case in senescent fibroblasts, the changes in the expression of transcription factors in HGPS are not evident in aged muscle cells. These comparisons, however, need to be interpreted with caution. First, many of the expression changes found in aged muscle are not unique to that condition; for example, genes involved in mitochondrial electron transport and oxidative phosphorylation are also affected in diabetic and prediabetic muscle (Patti et al., 2003). Secondly, muscle cells are post-mitotic whereas the fibroblasts we profiled were actively proliferating.

Miller et al. (2002) identified 29 genes whose expression was changed both by caloric restriction (CR) and by a mutation that lowers pituitary function. Twenty-five of these mouse genes have an identified human homologue; of these only one also appears on our list, TFPI2, which is down-regulated 3-fold in CR, 8.3-fold in dwarf mice and 3-fold in HGPS fibroblasts. Thus, there is very little overlap between these data sets, although the same cross-species and tissue vs. cell culture caveats discussed above need to be considered.

Insights into HGPS disease processes

Laminopathies display overlapping phenotypes accompanied by significant degeneration of cells that constitute the most affected tissues, be it myocytes in the muscular dystrophies, adipocytes in the lipodystrophies or axonal neurones in Charcot–Marie–Tooth disease (Mounkes et al., 2003a). The clinical features as well as our gene expression data indicate that mesodermal stem cells may be the primary cell type affected in HGPS, and perhaps in mandibuloacral dysplasia. By analogy with the other laminopathies, HGPS might therefore be considered as a more global, systemic ‘mesodermal dystrophy’ with consequent degeneration and accelerated aging of the cardiovascular system and connective tissues. It remains to be established to what extent the phenotype of HGPS is caused by simple dominant loss of function vs. specialized gain of function properties conferred by the HGPS ‘Progerin’ allele. The LMNA knockout mouse develops normally but at 2–3 weeks after birth starts exhibiting a range of deleterious phenotypes, most notably muscular dystrophy (Sullivan et al., 1999). The phenotype of a transgenic mouse carrying a specific LMNA point mutation has recently been reported (Mounkes et al., 2003b); this mutation causes the autosomal dominant form of Emery–Dreifuss muscular dystrophy in humans but displays a progeroid phenotype in the mouse.

Because lamin A has been implicated in interactions with chromatin, widespread effects on gene expression should perhaps not come as a surprise. What we cannot currently explain, however, is the specificity of these changes: why should some genes and some cell types be affected but not others? One intriguing possibility is that transcription factors responsible for co-ordinating developmental processes may be especially sensitive to changes in chromatin contexts. Another possibility is that expressed genes may be hierarchically as well as spatially arranged (Zhang et al., 2003), and that lamin A, by virtue of its architectural influences, may exert more prevalent influences in some subnuclear domains. Finally, because lamin A has already been shown to interact with the pRb protein, additional functional interactions with important global regulators may come to light.

HGPS patients die from myocardial infarction or strokes caused by progressive atherosclerosis of the coronary and cerebrovascular arteries. Identification of gene expression changes that may provide clues to risk factors or disease progression are thus of the highest priority. Our study is the first whole genome analysis of gene expression for any laminopathy, and has identified several genes that would be expected to impact the physiology of the cardiovascular system. Additional studies aimed at other LMNA mutations as well as tissues other than fibroblasts should be very illuminating. The possibility of phenotypic reversion engineered by correcting the LNMA mutation by RNAi or other methods raises obvious implications for gene therapy. In the short term, however, focusing on atherosclerosis may provide significant lifespan extension of HGPS patients, and could be attempted with existing pharmacological agents directed at cardiovascular function.

Materials and methods

Cell lines and culture conditions

Primary fibroblasts were obtained from Coriell Cell Repositories (Coriell Institute for Medical Research, Camden, NJ, USA) and were chosen on the basis of a typical clinical phenotype of the donor individual and the availability of low-passage stocks from the repository. The HGPS cell strains AG10750 (M, 9 years), AG11498 (M, 14 years) and AG11513 (F, 8 years) have been shown to be heterozygous for the codon 608 GGC>GGT mutation (Erikkson et al., 2003). The age indicated in parentheses refers to the age of the donor at the time of biopsy. AG10578 (M, 16.5 years) was obtained from an HGPS patient but has been found to have lost the codon 608 mutation by apparent gene conversion, resulting in a region of uniparental isodysomy on chromosome 1 that includes the LMNA gene (Erikkson et al., 2003). For controls we used primary cell strains derived from age-matched, apparently normal donors: GM00038C (F, 9 years), GM00316B (M, 12 years) and GM08398C (M, 8 years). All cells have been classified as dermal fibroblasts. After receipt from the Coriell Institute, cells were propagated in Dulbecco's modified Eagle's medium (DMEM; high glucose plus glutamine) supplemented with 20% fetal bovine serum (FBS), vitamins, essential and non-essential amino acids, and penicillin/streptomycin. Vitamins and amino acid mixtures were obtained from LifeTechnologies and used at twice the recommended concentration. Incubation was in an atmosphere of 93% N2, 5% CO2 and 2% O2. Cells were passaged 1 : 4 by trypsinization after reaching 80% confluence. RNA was prepared from exponentially growing, subconfluent cultures (40–50% confluency) after a maximum of 2–3 passages after receipt from the repository. To control for biological variability, replicate RNA preparations were extracted from each cell strain on three separate occasions. To avoid variability resulting from freezing/thawing of cells and/or small differences in passage history (in vitro age), each replicate was initiated with a new shipment of cells from the Coriell Institute.

RNA isolation

Total RNA was isolated using the TRIzol reagent (Life Technologies) and further purified on RNeasy mini columns (Qiagen). Five micrograms of total RNA was used in each labelling reaction. Reverse transcription was with the SuperScript cDNA Synthesis Kit (InVitrogen). The cDNA was purified from the synthesis reaction using PhaseLock gels (Eppendorf). Conversion of double-stranded cDNA into cRNA was accomplished with the T7 BioArray transcript labelling Kit (Enzo Diagnostics). The cRNA was purified on RNeasy mini columns. Fragmentation of the cRNA for target preparation, hybridization to GeneChip™ HG-U133A and HG-133B probe arrays, washing, staining and scanning were performed according to the manufacturer's protocols (Affymetrix).

Data analysis

Analysis of the scanned GeneChip™ images was performed using the Affymetrix Microarray Suite 5.0 software. All the chips were scaled to the same target intensity of 500. All possible pairwise correlation coefficients above 0.9 were computed. For each chip, probe-set signal intensities for all genes were plotted against the signal intensities for all other chips, and in all cases the calculated slopes were close to unity. In addition to the high linear correlation, the average signal intensities for all chips are very similar, as indicated by the very small coefficient of variance of 0.018. Further normalization was thus deemed unnecessary. Owing to the large number of variables (genes) and small sample sizes, the analysis of gene expression microarray data is problematic with standard statistical techniques. Unlike, for example, a high-quality clinical study which will involve thousands of cases in which tens of variables are measured, in microarray studies tens of thousands of variables are measured using only a small number of samples (Kohane et al., 2003). This leads to low signal/noise ratios. We chose the Significance Analysis of Microarrays (SAM) method (Tusher et al., 2001) because in addition to performing gene-specific t-tests, SAM adjusts for multiple testing by using permutations of the repeated measurements (replicates) to estimate the False Discovery Rate (FDR), defined as the proportion of false positives in a set of potentially significant genes. By adjusting a sliding threshold, one obtains a set of differentially regulated genes with a desired median FDR. Within this set of differentially regulated genes of desired FDR, each gene is assigned a q-value (Storey & Tibshirani, 2003), which represents the expected proportion of false positives when calling that gene significant. FDRs are defined for the set of genes that are deemed significant, and q-values are defined for each individual gene. In order to focus on genes changes that were statistically as well as biologically significant, we further filtered the output of SAM. A 2-fold cutoff was applied to the set of differentially regulated genes defined with median FDR of approximately 5%. In this way, a stringently selected set of 218 genes was identified for the UG133 A and B chips. To this list, we added a subset of 143 genes with a median FDR of up to approximately 10%; to reduce the number of false positives in this subset we imposed a higher fold cutoff (3-fold) for inclusion. Only those genes with LocusLink identifiers were included in the final list. This excluded a large number of the probes on the B chip.

RT-PCR validation

Quantitative reverse transcription PCR (qPCR) was performed in an ABI7900 instrument using the two-step TaqMan system according to the manufacturer's instructions (Applied Biosystems). The housekeeping gene β-glucuronidase was used as the internal standard. All PCR efficiencies were > 90%, and were determined for each primer set using an RNA reference pool obtained from Stratagene. Eleven genes with fold-changes greater than 3 were chosen for qPCR verification (MEOX2, PRG1, KRT18, OSF-2, STAT1, BAALC, IRX5, CLGN, ENO1, CamKIINalpha and PPP1R14A); in all cases the direction of change was reproduced, although the magnitude of the change showed some differences from the fold-changes predicted by Affymetrix GeneChips. The ENO1, CamKIINalpha and PPP1R14A genes do not appear on our final gene list (see Supplementry table) because in spite of a significant fold-change, they subsequently did not pass the adopted FDR q-value cutoff. The qPCR data for the eight genes listed in the Supplementry table are (expressed as fold-change relative to normal cells): MEOX2, 94.0; PRG1, 104.6; KRT18, 6.5; OSF-2, 28.5; STAT1, 13.1; BAALC, −38.5; IRX5, −3.9; CLGN, −27.8.


S.B.E and A.J.B. thank Dr Peter Park for critical comments and discussions. This work was supported by a grant from the Progeria Research Foundation and NIH grant R01 AG16694 to J.M.S. A.J.B. and S.B.E. were supported by NIH grants U01 HL066582, R01 DK060837 and K12 DK063696, and by the Lawson Wilkins Pediatric Endocrinology Society and the Harvard Center for Neurodegeneration and Repair.

Supplementary material

The following material is available from:


Table S1. All genes found to be differentially expressed in HGPS fibroblasts