SEARCH

SEARCH BY CITATION

Keywords:

  • aging;
  • Ames dwarf;
  • gene expression;
  • Little mice;
  • microarray analysis;
  • mouse models

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Supplementary material
  8. Acknowledgments
  9. References
  10. Supporting Information

Ames dwarf mice (Prop1df/df) and Little mice (Ghrhrlit/lit) are used as models of delayed aging and show significant increases in lifespan (50% and 25%, respectively) when compared with their wild-type siblings. To gain further insight into the molecular basis for the extended longevity of these mice, we used oligonucleotide microarrays to measure levels of expression of over 14 000 RNA transcripts in liver during normal aging at 3, 6, 12 and 24 months. We found that the Prop1df/df and Ghrhrlit/lit genotypes produce dramatic alterations in gene expression, which are predominantly maintained at all ages. We found 1125 genes to be significantly affected by the Prop1df/df genotype and 1152 genes were significantly affected by the Ghrhrlit/lit genotype; 547 genes were present in both gene lists and showed parallel changes in gene expression, suggesting common mechanisms for the extended longevity in these mutants. Some of the functional gene classes most affected in these mutants included: amino acid metabolism, TCA cycle, mitochondrial electron transport, fatty acid, cholesterol and steroid metabolism, xenobiotic metabolism and oxidant metabolism. We found that the Prop1df/df genotype, and to a minor extent the Ghrhrlit/lit genotype, also produced complex alterations in age-dependent changes in gene expression as compared with wild-type mice. In some cases these alterations reflected a partial delay or deceleration of age-related changes in gene expression as seen in wild-type mice but they also introduced age-related changes that are unique for each of these mutants and not present in wild-type mice.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Supplementary material
  8. Acknowledgments
  9. References
  10. Supporting Information

Aging is a highly complex process characterized by an age-dependent increase in the likelihood of death and a decline in the production of offspring. A good experimental approach for studying the mechanisms of aging is provided by alterations that slow down or delay the normal aging process. Ames dwarf mice (Prop1df/df) and Little mice (Ghrhrlit/lit) are established models of delayed or decelerated aging, and both possess single recessive gene mutations that result in a substantial increase in their lifespan. Ames dwarf mice, homozygous for the df allele at the Prop1 locus, show an increase in mean lifespan of 49% in males and 68% in females when compared with their normal siblings (Brown-Borg et al., 1996). Prop1 is a transcription factor required for the proper embryonic development of the pituitary gland. Ames dwarf mice fail to differentiate somatotrophs, lactotrophs and thyrotrophs in the anterior pituitary, and as a result they do not produce growth hormone (Gh), prolactin (Prl) and thyroid-stimulating hormone (Tsh) (Sornson et al., 1996). They also show secondary hormone deficiencies including reduced circulating levels of insulin-like growth factor I (Igf1) (Chandrashekar & Bartke, 1993) and thyroid hormones. Basal insulin and glucose levels are also somewhat reduced (Borg et al., 1995), and this multiple hormone deficiency results in severe growth retardation. As adults they reach only about one-third the weight of their wild-type littermates. Fertility is also affected; females are always infertile and males are mostly infertile. The Ames dwarf mice show a delayed onset of several age-related pathologies; they experience a delayed age-related decline in locomotor activity and cognitive function when compared with wild-type mice (Kinney et al., 2001). They also have a delayed occurrence of fatal neoplastic diseases, especially adenocarcinoma in lung (Ikeno et al., 2003). Ames dwarf are similar in many respects to the Snell dwarf mice. Snell dwarf mice, in which a mutation at the Pit1 locus disrupts the action of Prop1 at the pituitary, show the same collection of hormone deficiencies and also have an extended lifespan compared with wild-type mice (Flurkey et al., 2002).

Little mice show an increase in mean lifespan of 23% in males and 25% in females (Flurkey et al., 2001). They are homozygous for a missense mutation (designated as lit) in the growth hormone-releasing hormone receptor (Ghrhr) gene (Godfrey et al., 1993). Ghrhr is required for the secretion of growth hormone by the pituitary gland in response to growth hormone-releasing hormone (Ghrh) (Frohman & Jansson, 1986; Gelato & Merriam, 1986). As a result, these mice show substantially reduced levels of circulating growth hormone (1% of the normal level) (Flurkey et al., 2001) and have reduced circulating levels of Igf1. These mice show delayed growth and only reach about two-thirds the weight of normal wild-type mice. Although females are generally fertile, males are mostly infertile. Measures in the onset of age-related pathologies, as performed in Ames dwarf mice, have not yet been reported for Little mice.

The biochemical and metabolic pathways that slow aging in these mouse models are still little understood. Gene expression profile analysis provides a useful approach toward delineation of the mechanisms by which the Prop1df/df and the Ghrhrlit/lit genotypes lead to increased lifespan. In this paper we report on studies of gene expression profile analysis based on oligonucleotide arrays to look for alterations in gene expression that discriminate Ames dwarf and Little mice from their respective wild-type strains throughout their lifespan at ages 3, 6, 12 and 24 months. We found that the Prop1df/df and the Ghrhrlit/lit genotypes produced dramatic alterations in gene expression that are largely maintained at all analysed ages. Our data indicate that there is a large overlap of the genes and metabolic processes that are affected by the Prop1df/df and the Ghrhrlit/lit genotypes. Finally, these mutations also produced substantial alterations in age-related changes in gene expression when compared with wild-type mice.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Supplementary material
  8. Acknowledgments
  9. References
  10. Supporting Information

The Ames dwarf and Little mice mutations produce dramatic changes in gene expression

We measured liver expression levels of over 14 000 transcripts in Ames dwarf and Little mice at ages 3, 6, 12 and 24 months using Affymetrix MOE430A gene expression arrays. Three biological replicates were used per time-point for each of the four categories in this experiment: Ames dwarf (Prop1df/df), Ames dwarf control (Prop1+/+), Little mice (Ghrhrlit/lit) and Little mice control (Ghrhr+/lit). Figure 1 shows an unsupervised hierarchical clustering analysis that provides an overview of the general behaviour of the microarray data.

image

Figure 1. Hierarchical Clustering Analysis. The microarray data from Ames dwarf and Little mice were combined to perform unsupervised clustering analysis. Replicates were averaged before clustering. Each row represents a different gene. Bright red and blue indicates high and low expression, respectively. Several gene clusters are of particular interest. Cluster A contains genes that are up-regulated in both Ames and Little mutants. Cluster B contains genes that are down-regulated in both mutants. Cluster C contains genes that are up-regulated preferentially in mutant Little mice. Cluster D contains genes that are up-regulated preferentially in mutant Ames dwarf. Clusters E and F contain genes that discriminate between the two different strains.

Download figure to PowerPoint

We used analysis of variance (anova) to find alterations in gene expression that discriminate between mutant and wild-type mice in each category: Prop1df/df vs. Prop1+/+ and Ghrhrlitlit vs. Ghrhr+/lit. We found 1125 differentially expressed genes in Ames dwarf mice and 1152 differentially expressed genes in Little mice with an anovaP value of less than 0.0001. The complete lists of these genes are presented in Supplementary Tables S1 and S2. At this level of significance, with no multiple testing correction, we would expect to see only one or two genes by chance in an analysis of 14 000 genes. Using more conservative approaches, we found that after Bonferroni adjustment, with an experiment-wise P value of 0.05 (nominal P value < 0.0000035), we obtained 610 genes with a significant genotype effect in Ames dwarf and 628 genes with a significant genotype effect in Little mice. These groups of genes have only a 5% chance of containing a single false positive. Because the Bonferroni adjustment for statistical significance is known to be extremely restrictive and can lead to the generation of a high number of type II errors (false negatives), we selected the anovaP value < 0.0001 as our main criteria for the statistical significance of the genotype effects.

Although arguably inappropriate, in many microarray studies the gene expression fold-change (ratio of expression values between treatment and control) has been used as a criterion for the selection of differentially expressed genes without any measure of statistical significance. Ideally, it is expected that the genes with the highest fold-changes will also possess the lowest P values. In our data, there was a tendency for the genes with the largest fold-changes to have very low P values. The scatter plots in Fig. 2 show that nearly all of the genes with a fold-change in gene expression greater than 2 reached statistical significance. In Ames dwarf, out of the 140 genes that exceeded a two-fold difference, 122 reached the Bonferroni criterion for statistical significance (experiment-wise P < 0.05, nominal P < 0.0000035). In Little mice 104 genes surpassed a two-fold change in gene expression and 94 of these reached the Bonferroni criterion.

image

Figure 2. Relationship between fold-changes and measures of statistical significance. Of the genes with a fold-change greater than 2 in Ames dwarf and Little mice, 87% and 90%, respectively, reach the Bonferroni criterion for statistical significance (experiment-wise P < 0.05, nominal P < 0.0000035). The scatter plots show the expression values for all the genes present in the array. The x-axis shows the values for the mutant Ames or mutant Little mice and the y-axis the values for their respective wild-type controls. The outermost parallel lines correspond to a two-fold-change in gene expression; the intermediate parallel lines correspond to a 1.5-fold-change in gene expression.

Download figure to PowerPoint

Overlap between differentially expressed genes in Ames dwarf and Little mice

Because Ames dwarf and Little Mice share some common characteristics, most notably growth hormone deficiency, we anticipated finding genes that are similarly affected in both mutants. When comparing the lists of 1125 and 1152 differentially expressed genes in Ames dwarf and Little mice we found an overlap of 552 genes. Some of the overlapping genes with the lowest P values are presented in Table 1. Notably, as shown in Fig. 3(A), over 99% of these overlapping genes show changes of gene expression in the same direction in both mutants. Furthermore, as shown in Fig. 3(B), we observed that the relative magnitude of the fold-changes of the 547 genes that show parallel changes in gene expression is highly conserved between Ames dwarf and Little mice. (See Supplementary tables)

Table 1.  Sixty-five of the 547 genes with parallel gene expression alterations in Ames dwarf and Little micea
GeneDescriptionAmes DwarfLittle Mice
Fold-changebP valueFold-changebP value
  • a

    The genes shown in this table are ordered according to the P values seen in Ames dwarf. For a complete list of all 547 genes significantly affected by both the Ames and Little mutations see Supplementary Table S3.

  • b

    The fold-change refers to the ratio of the expression values of the mutant mice over the control mice averaged at all ages.

Cyp2b13cytochrome P450, 2b13, phenobarbitol inducible, type c11.7281.17E-2416.7123.86E-29
Fmo3flavin containing monooxygenase 312.6806.31E-23 3.0473.75E-06
Hsd3b5hydroxysteroid dehydrogenase-5, delta<5>−3-beta 0.0494.30E-21 0.0764.49E-12
Elovl3elongation of very long chain fatty acids (FEN1/Elo2, SUR4/Elo3, yeast)-like 3 0.0558.64E-20 0.0717.38E-16
Cyp4a14hypothetical protein LOC26957521.5673.16E-19 1.7002.34E-08
Slc21a1solute carrier family 21 (organic anion transporter), member 1 0.1023.77E-19 0.1409.82E-17
Txn2thioredoxin 2 2.9721.05E-18 2.1772.43E-15
Cyp7b1cytochrome P450, 7b1 0.0901.31E-18 0.2191.87E-15
Stesulfotransferase, oestrogen preferring13.5361.58E-18 5.0422.63E-07
Cyp7b1cytochrome P450, 7b1 0.0638.27E-18 0.1922.80E-15
Slc21a1solute carrier family 21 (organic anion transporter), member 1 0.1721.21E-17 0.2327.49E-14
Hao3hydroxyacid oxidase (glycolate oxidase) 317.8651.44E-1612.1103.70E-19
Serpina3kserine (or cysteine) proteinase inhibitor, clade A, member 3K 0.0841.54E-16 0.4468.69E-11
D1Ertd308enuclear receptor subfamily 5, group A, member 2 0.1591.19E-15 0.2758.45E-13
2610008F03RikRIKEN cDNA 6720460F02 gene 0.1092.40E-15 0.2243.64E-10
Cyp4a10cytochrome P450, 4a1013.6882.61E-15 1.7122.82E-09
Man2b1mannosidase 2, alpha B1 0.3072.95E-15 0.3702.69E-14
Gsto1glutathione S-transferase omega 1 3.3914.54E-15 1.7752.77E-14
Keg1kidney expressed gene 1 0.1479.37E-15 0.1374.67E-16
Ela1elastase 1, pancreatic 0.4121.25E-14 0.6684.17E-08
AI118089expressed sequence AI118089 2.9991.90E-14 2.0875.33E-13
Tubgcp5tubulin, gamma complex associated protein 5 0.2183.20E-14 0.2985.49E-16
Egfrepidermal growth factor receptor 0.1284.21E-14 0.1785.03E-13
0610041O14RikRIKEN cDNA 0610041O14 gene 0.3196.82E-14 0.4701.73E-11
2810428F02RikRIKEN cDNA 2810428F02 gene 0.5057.54E-14 0.5585.31E-11
Rtn4reticulon 4 4.0237.60E-14 2.0026.74E-11
Cyp2c38cytochrome P450, 2c38 2.4221.11E-13 1.4841.48E-13
1700037H04RikRIKEN cDNA 1700037H04 gene 0.3281.21E-13 0.4771.61E-10
Vldlrvery low density lipoprotein receptor 4.2331.48E-13 4.3691.20E-17
Idh2isocitrate dehydrogenase 2 (NADP+), mitochondrial 2.3821.98E-13 1.5939.40E-13
Cyp2a5cytochrome P450, 2a5 1.9302.19E-13 1.7597.04E-07
Abcb1aATP-binding cassette, subfamily B (MDR/TAP), member 1 A 2.4412.52E-13 1.9274.17E-10
HampRIKEN cDNA 1810073K19 gene 3.8042.70E-13 5.9447.39E-09
Slc16a7solute carrier family 16 (monocarboxylic acid transporters), member 7 3.4612.80E-13 2.7411.53E-13
Hsd3b2hydroxysteroid dehydrogenase-2, delta<5>−3-beta 0.2732.91E-13 0.6974.69E-07
Egfrepidermal growth factor receptor 0.4442.93E-13 0.5425.62E-14
hypothetical protein LOC214424 2.3683.65E-13 2.2141.81E-08
2810428F02RikRIKEN cDNA 2810428F02 gene 0.4873.85E-13 0.5712.51E-10
Mtpnmyotrophin 3.2794.73E-13 1.7112.98E-05
Alas2aminolevulinic acid synthase 2, erythroid 0.3685.85E-13 0.4094.24E-16
Cyp2b9cytochrome P450, 2b9, phenobarbitol inducible, type a12.6626.29E-1315.3242.97E-14
Egfrepidermal growth factor receptor 0.3446.40E-13 0.4873.37E-11
4930439B20RikRIKEN cDNA 4930439B20 gene 0.3656.99E-13 0.2082.97E-17
Serpine2serine (or cysteine) proteinase inhibitor, clade E, member 2 0.3017.84E-13 0.3813.69E-09
Cyp2b20cytochrome P450, 2b20 3.7109.03E-13 3.4296.69E-12
Gstp2glutathione S-transferase, pi 2 0.6711.11E-12 0.6005.96E-13
Aldh1b1aldehyde dehydrogenase 1 family, member B1 2.5541.14E-12 1.9822.00E-09
Cyp39a1cytochrome P450, 39a1 (oxysterol 7alpha-hydroxylase) 2.6121.32E-12 2.8296.57E-14
Cyp2b20cytochrome P450, 2b20 5.4061.99E-12 5.2635.73E-11
Scp2sterol carrier protein 2, liver 0.7512.40E-12 0.8151.31E-12
2410041F14RikRIKEN cDNA 2410041F14 gene 0.3642.48E-12 0.4904.39E-09
Mus musculus, clone IMAGE:5136153, mRNA 0.2252.78E-12 0.4067.97E-08
H2-T10histocompatibility 2, T region locus 10 1.3523.54E-12 1.2431.64E-08
BC002292cDNA sequence BC002292 0.4983.61E-12 0.6791.69E-08
Slc21a5solute carrier family 21 (organic anion transporter), member 5 2.8793.96E-12 2.6515.93E-11
Pphnsynaptophysin-like protein 1.7974.87E-12 1.3722.75E-05
CRAD-Lcis-retinol/3alpha hydroxysterol short-chain dehydrogenase-like 2.6355.66E-12 2.5573.22E-12
1200011D03RikRIKEN cDNA 1200011D03 gene 0.3655.71E-12 0.4885.43E-11
Trpm7transient receptor potential cation channel, subfamily M, member 7 0.4985.82E-12 0.5833.46E-09
1300012D20RikRIKEN cDNA 1300012D20 gene 0.4717.37E-12 0.5181.19E-12
Cyp2b10cytochrome P450, 2b10, phenobarbitol inducible, type b 5.2387.76E-12 4.8424.08E-11
Abcd2ATP-binding cassette, subfamily D (ALD), member 2 2.0757.87E-12 1.7419.21E-09
5031400M07RikRIKEN cDNA 5031400M07 gene 1.5957.90E-12 1.6971.43E-09
Sth2sulfotransferase, hydroxysteroid preferring 227.4428.43E-1294.2491.79E-35
Mki67ipMki67 (FHA domain) interacting nucleolar phosphoprotein 0.4688.47E-12 0.4766.70E-13
Nudt7nudix (nucleoside diphosphate linked moiety X)-type motif 7 0.1368.91E-12 0.2654.73E-14
Cthcystathionase (cystathionine gamma-lyase) 1.9631.25E-11 1.9052.44E-12
Rtn4reticulon 4 3.0721.25E-11 1.7563.86E-10
Apobec1apolipoprotein B editing complex 1 0.6861.52E-11 0.8051.63E-06
Cyp2c40cytochrome P450, 2c40 0.4691.55E-11 1.2835.50E-06
Rdh6retinol dehydrogenase 6 1.8902.04E-11 1.8842.51E-08
image

Figure 3. Commonly differentially expressed genes in Ames dwarf and Little mice show parallel changes of gene expression and a strong correlation between their relative fold-changes. The genes analysed in this figure correspond to the set of 547 genes that are affected in both Ames dwarf and Little mice. As shown in A, the vast majority of these genes show the same directional change in Ames dwarf and Little mice; there are only five exceptions. The scatter plot compares the expression values between Ames dwarf (Prop1df/df) and control mice (Prop1+/+). The genes that fall above the diagonal line (x = y) are up-regulated in Ames dwarf, whereas the genes below are down-regulated. The genes are colour-coded according to the directional change in gene expression seen in the Little mice, red for up-regulation and blue for down-regulation. B compares the ranking of these genes according to their fold-changes in Ames dwarf (Prop1df/df vs. Prop1+/+) and Little mice (Ghrhrlit/lit vs. Ghrhr+/lit). The linear trend in this graph shows that the relative magnitude of the fold-changes for these genes in Ames dwarf and Little mice is highly conserved. In other words, the genes that have the highest (or lowest) fold-changes are shared between the two mutants.

Download figure to PowerPoint

Our primary interest was to find genes that show parallel changes in gene expression in both Ames dwarf and Little mice. However, because of the unique hormone deficiencies present in Ames dwarf mice, we expected to find genes that were altered exclusively or preferentially in Ames dwarf mice. Using a two-way anova model (P < 0.0001) we found 111 genes that are affected exclusively in Ames dwarf mice and 56 genes that are affected exclusively in Little mice. (See supplementary tables).

Confirmation of selected results by real-time PCR

We used real-time PCR amplification as an independent method to confirm some of the results obtained by microarray analysis. The genes tested were selected over a broad range of fold-changes and P values according to the microarray data. We tested samples from three mutant and three control mice for both Ames dwarf and Little mice at 6 months of age. Of the 23 genes tested in Ames dwarf, 22 genes produced gene expression fold-changes consistent with the microarray data and significant P values of less that 0.05 (Table 2). Of the 20 genes tested in Little mice, 18 produced gene expression fold-changes consistent with the microarray and significant P values of less than 0.05 (Table 3). These results indicate that the data we obtained from the microarray analysis are highly reliable. As shown in Tables 2 and 3, the microarray data have a tendency to underestimate the gene expression fold-changes when compared with the real-time PCR analysis. With the exception of two genes, all of the confirmed genes show more dramatic fold-changes in gene expression according to the real-time PCR data than with the microarray data. This effect was more noticeable for the genes with the most extreme fold-changes. The normalization procedure is probably at least partially responsible for this effect, but the nature of the microarrays probably also plays an important role. Because of signal saturation there is an upper limit for the expression value of highly abundant genes. Additionally, owing to cross-hybridization and non-specific background, there is also a lower limit for the expression of very low-abundance genes.

Table 2.  Ames dwarf: real-time PCR results compared with microarray data
GeneDescriptionReal-time PCRMicroarray data
Fold changeaP valuebFold ChangeaP value
  • a

    The fold-change refers to the ratio of the expression values of the mutant mice over the control mice at 6 months of age.

  • b

    The P values for the real-time PCR refers to a one-tailed t-test for the differences in means between the normalized CT values in mutant vs. control mice (see Experimental procedures).

  • c

    This gene was considered to be expressed exclusively in mutant mice due to its large fold change (17352, P = 1.59E-05).

  • d

    The change in the expression level of this gene was not confirmed by the real-time PCR analysis.

Aco2aconitase 2, mitochondrial  1.5687.73E-04 1.3951.53E-07
Csadcysteine sulfinic acid decarboxylase  0.0621.41E-04 0.2044.45E-09
Cyp2b13cytochrome P450, 2b13, phenobarbitol inducible, type c316.8773.88E-0611.1951.17E-24
Cyp7b1cytochrome P450, 7b1  0.0187.53E-07 0.0691.31E-18
Egfrepidermal growth factor receptor  0.0301.71E-05 0.1124.21E-14
Elovl3elongation of very long chain fatty acids 3  0.0013.99E-05 0.0448.64E-20
Fmo3flavin containing monooxygenase 3829.1889.54E-0614.2286.31E-23
Gpx3glutathione peroxidase 3  2.4273.33E-02 1.6612.81E-05
Gstm3glutathione S-transferase, mu 3  5.9152.03E-05 3.7592.52E-10
Hao3hydroxyacid oxidase (glycolate oxidase) 3380.6241.29E-0618.9581.44E-16
Hsd17b2hydroxysteroid (17-beta) dehydrogenase 2  0.2232.96E-04 0.4158.17E-09
Hsd3b5hydroxysteroid dehydrogenase-5, delta<5>−3-beta  0.0001.56E-06 0.0464.30E-21
Idh2isocitrate dehydrogenase 2 (NADP+), mitochondrial  4.9327.90E-06 2.7941.98E-13
Igfalsinsulin-like growth factor binding protein, acid labile subunit  0.0355.40E-06 0.2725.12E-08
Igfbp1insulin-like growth factor binding protein 1 72.8979.45E-03 5.4376.86E-06
Igfbp2insulin-like growth factor binding protein 2  4.4833.52E-04 2.0231.68E-05
Mor1malate dehydrogenase, mitochondrial  1.4903.21E-03 1.4893.63E-09
Mt1metallothionein 1 29.4911.88E-04 6.1742.28E-07
Nox4NADPH oxidase 4  0.4947.59E-04 0.8281.10E-06
Slc21a1solute carrier family 21 (organic anion transporter), member 1  0.0011.53E-04 0.1023.77E-19
Sth2sulfotransferase, hydroxysteroid preferring 2Only in mutantc35.1738.43E-12
Txn2dthioredoxin 2d  1.0905.06E-02 3.3831.05E-18
Ugt2b5UDP-glucuronosyltransferase 2 family, member 5  0.3654.52E-04 0.4916.63E-11
Table 3.  Little mice: real-time PCR results compared with microarray data
GeneDescriptionReal-time PCRAffymetrix
Fold changeaP valuebFold ChangeaP value
  • a

    The fold-change refers to the ratio of the expression values of the mutant mice over the control mice at 6 months of age.

  • b

    The P values for the real-time PCR refer to a one-tailed t-test for the differences in means between the normalized CT values in mutant vs. control mice (see Experimental procedures).

  • c

    This gene was considered to be expressed exclusively in wild-type due to its dramatic fold change (0.0001, P = 8.90E-07).

  • d

    This gene was considered to be expressed exclusively in mutant mice due to its large fold change (31072, P = 1.14E-07).

  • e

    The change in the expression level of these genes was not confirmed by the real-time PCR analysis.

Aco2aconitase 2, mitochondrial   1.2474.10E-02 1.2464.00E-08
Csadcysteine sulfinic acid decarboxylase   0.0774.24E-04 0.1771.34E-12
Cyp2b13cytochrome P450, 2b13, phenobarbitol inducible, type c 413.9561.55E-0417.8743.86E-29
Cyp7b1cytochrome P450, 7b1   0.0732.79E-05 0.1421.87E-15
Egfrepidermal growth factor receptor   0.0631.48E-05 0.4935.62E-14
Elovl3elongation of very long chain fatty acids 3   0.0252.79E-03 0.0767.38E-16
Fmo3flavin containing monooxygenase 3 138.0351.36E-04 3.0793.75E-06
Gstm3glutathione S-transferase, mu 3   3.9489.13E-03 2.4809.55E-08
Hao3hydroxyacid oxidase (glycolate oxidase) 31033.5085.49E-0613.6393.70E-19
Hsd17b2hydroxysteroid (17-beta) dehydrogenase 2   0.4981.19E-03 0.5993.69E-09
Hsd3b5hydroxysteroid dehydrogenase-5, delta<5>−3-betaOnly in wild-typec 0.0474.49E-12
Idh2isocitrate dehydrogenase 2 (NADP+), mitochondrial   1.8081.38E-03 1.6009.40E-13
Igfalsinsulin-like growth factor binding protein, acid labile subunit   0.1264.00E-04 0.3437.25E-14
Igfbp2einsulin-like growth factor binding protein 2e   1.1412.15E-01 1.6033.06E-05
Mor1emalate dehydrogenase, mitochondriale   1.0395.52E-02 1.6044.71E-07
Mt1metallothionein 1  11.7671.38E-02 6.2852.82E-06
Nox4NADPH oxidase 4   0.4243.15E-03 0.6384.37E-08
Slc21a1solute carrier family 21 (organic anion transporter), member 1   0.0026.48E-06 0.1169.82E-17
Sth2sulfotransferase, hydroxysteroid preferring 2Only in mutantd94.7601.79E-35
Utg2b5UDP-glucuronosyltransferase 2 family, member 5   0.3373.25E-03 0.6832.45E-10

Although the microarray data had a tendency to underestimate fold-changes, there was a strong correlation between the gene expression fold-changes generated by the microarray data vs. the fold-changes obtained in the real-time PCR analysis (Fig. 4). The relative magnitude of the changes in gene expression as determined by microarray data is conserved when analysed by real-time PCR.

image

Figure 4. Relationship between fold-changes obtained by microarray and real-time PCR. There was a strong correlation (R2 = 0.946) between the gene expression fold-changes generated by the microarray data vs. the fold-changes obtained in the real-time PCR analysis. As evident from the smaller range of the y-axis, the absolute fold-change measures are compressed by the microarray data. However, the relative magnitude of these changes is still conserved. The genes with the most dramatic fold-changes according to the microarray data also have the most dramatic fold-changes when analysed by real-time PCR.

Download figure to PowerPoint

The genes altered by the Ames dwarf and Little mice mutations reflect complex alterations in a wide range of metabolic processes

The next step in our analysis was the classification of differentially expressed genes into functional categories. We used the EASE software application (Hosack et al., 2003) to identify over-represented biological themes (Gene Ontology classifications) in the set of 547 genes that show significant gene expression alterations in both Ames dwarf and Little mice. The statistical measure of over-representation used by EASE is a variant of the one-tailed Fischer exact probability. Tables 4 and 5 show, respectively, the over-represented categories corresponding to the Molecular Function and Biological Process Ontology for which the EASE P value is less than 0.05. These analyses reveal that the Prop1df/df and Ghrhrlit/lit genotypes produce significant alterations in many core metabolic processes, such as amino acid metabolism, carbohydrate metabolism, mitochondrial electron transport, fatty acid, cholesterol and steroid metabolism, xenobiotic metabolism and a large number of genes involved in oxido-reductase processes. The genes in some of these processes show a significant tendency to be either mostly up-regulated or mostly down-regulated. The genes in amino acid metabolism, TCA cycle and mitochondrial electron transfer, for example, are mostly up-regulated, whereas the genes classified in humoral defense mechanism are all down-regulated. As shown in Fig. 5 we found that in many cases these trends in gene expression extended to include all of the genes present in the array for a particular process, not only the genes that met our criterion for statistical significance.

Table 4.  Molecular Function Ontology of differentially expressed genes in Ames dwarf and Little mice
Functional ontologyaArray hitsbList hitscUpdDowndP valuee
  • a

    Molecular function ontology descriptions are provided for groups that show a significant over-representation in the set of 547 genes whose gene expression is affected in both Ames dwarf and Little mice.

  • b

    Number of genes in each category present in the entire array (Array hits).

  • c

    Number present in the set of 547 differentially expressed genes (List hits).

  • d

    Number of genes whose expression is up-regulated (Up) or down-regulated (Down) in each category.

  • e

    The P value shown refers to the EASE score (see Experimental procedures).

catalytic activity3508223137862.04E-11
 isomerase activity
  intramolecular isomerase activity  30  5  1 44.66E-07
 lyase activity 104 11  8 31.38E-03
 oxidoreductase activity 441 70 50202.13E-03
  acting on paired donors
   with incorporation or reduction of molecular oxygen  63 15 11 42.08E-02
    reduced flavin or flavoprotein as one donor  33 12  8 43.88E-04
  disulphide oxidoreductase activity  23  7  5 26.98E-21
   protein disulphide oxidoreductase activity   9  4  2 24.88E-02
  monooxygenase activity  81 23 15 89.13E-08
  oxidoreductase activity∖, acting on CH-OH group of donors  65 10  8 21.10E-06
   NAD or NADP as acceptor  60  7  6 11.75E-03
  oxidoreductase activity∖, acting on the CH-NH group of donors  14  5  5 9.78E-04
  oxidoreductase activity∖, acting on NADH or NADPH  42  8  7 12.66E-02
   NADH dehydrogenase activity  27  7  7 01.67E-02
   oxidoreductase activity∖, acting on NADH or NADPH∖, other acceptor  30  7  7 02.24E-03
defence/immunity protein activity
 complement activity  20  5  0 53.79E-02
ligase activity
 other carbon-nitrogen ligase activity  14  4  3 11.30E-04
serine protease inhibitor activity  81  9  4 51.05E-02
transporter activity1412 77 45321.79E-02
 electron transporter activity 270 22 15 73.28E-12
transferase activity
 transferase activity∖, transferring alkyl or aryl (other than methyl) groups  45 10  9 15.94E-03
  glutathione transferase activity  20  9  8 18.70E-03
 transferase activity∖, transferring one-carbon groups 102 13  9 42.18E-02
  methyltransferase activity  99 13  9 41.78E-03
   S-adenosylmethionine-dependent methyltransferase activity  63  8  6 22.67E-03
   S-methyltransferase activity   5  3  3 04.19E-02
Table 5.  Biological Process Ontology of differentially expressed genes in Ames dwarf and Little mice
Fuctional ontologyaList hitsbArray hitscUpdDowndP valuee
  • a

    Biological process ontology descriptions are provided for groups that show a significant over-representation in the set of 547 genes whose gene expression is affected in both Ames dwarf and Little mice.

  • b

    Number present in the set of 547 differentially expressed genes (List hits).

  • c

    Number of genes in each category present in the entire array (Array hits).

  • d

    Number of genes whose expression is up-regulated (Up) or down-regulated (Down) in each category.

  • e

    The P value shown refers to the EASE score (see Experimental procedures).

physiological processes36176341991623.57E-03
 metabolism26649871521146.89E-06
  alcohol metabolism 17 166 12  52.81E-03
  amine metabolism 27 178 21  67.01E-08
   amine catabolism 10  40  8  24.87E-05
   amine biosynthesis  8  40  7  11.67E-03
  amino acid and derivative metabolism 24 150 19  51.65E-07
   biogenic amine metabolism
    catecholamine metabolism  4  13  2  21.77E-02
   amino acid metabolism 20 115 17  36.13E-07
    amino acid catabolism  9  35  8  11.14E-04
    amino acid biosynthesis  7  24  7  04.98E-04
    aspartate family amino acid metabolism  4   9  4  05.95E-03
     aspartate family amino acid biosynthesis  3   4  3  01.11E-02
      methionine biosynthesis  3   3  3  05.71E-03
    serine family amino acid metabolism  4  11  3  11.09E-02
    sulphur amino acid metabolism  5  19  4  18.68E-03
     sulphur amino acid biosynthesis  4   6  4  01.56E-03
  aromatic compound metabolism  9  74  6  31.63E-02
  biosynthesis 54 817 25 292.91E-03
  carbohydrate metabolism 21 295 14  73.86E-02
   tricarboxylic acid cycle  6  21  6  01.89E-03
  coenzymes and prosthetic group metabolism 12 114  8  41.21E-02
   coenzyme metabolism 11  96  8  39.86E-03
  electron transport 56 419 39 172.02E-13
   ATP synthesis coupled electron transport  4  15  4  02.64E-02
    mitochondrial electron transport∖, NADH to ubiquinone  4  13  4  01.77E-02
   protein-disulphide reduction  4  17  4  03.70E-02
  lipid metabolism 40 360 23 171.82E-07
   fatty acid metabolism 15  96  9  67.81E-05
    acyl-CoA metabolism  5  12  1  41.42E-03
   lipid biosynthesis 18 144  7 112.06E-04
   steroid metabolism 16  87  8  85.48E-06
    sterol metabolism  6  41  4  23.37E-02
     cholesterol metabolism  6  37  4  22.26E-02
    steroid biosynthesis 10  61  3  71.35E-03
  nitrogen metabolism  6  21  6  01.89E-03
   urea cycle intermediate metabolism  3   7  3  03.55E-02
  organic acid metabolism 42 249 32 101.74E-13
   carboxylic acid metabolism 42 248 32 101.51E-13
    amino acid metabolism 20 115 17  36.13E-07
  protein metabolism
   protein targeting 17 175  9  84.73E-03
 response to external stimulus
  response to biotic stimulus
   defence response
    immune response
     humoral immune response
      humoral defence mechanism (sensu Vertebrata)  7  39  0  76.89E-03
       complement activation  7  37  0  75.29E-03
        complement activation∖, classical pathway  5  22  0  51.48E-02
image

Figure 5. Expression of some functional gene classes with significant trends towards up-regulation or down-regulation. Based on our Gene Ontology analysis of the 547 genes significantly affected in Ames dwarf and Little mice we found several functional gene classes with significant directional changes in gene expression. We used hypergeometric analysis to determine the statistical significance of the over-representation of up-regulated or down-regulated genes in a particular process. The graphs show the expression data of mutant Little vs. wild-type control at 3 months of age of all the genes present in the array for each particular category. The data from Ames dwarf show a similar behavior (not shown).

Download figure to PowerPoint

Age-related changes in gene expression

In addition to investigating the alterations in gene expression that characterize the Ames and Little mutations, we were also interested in studying how these mutations may affect age-related changes in gene expression. Caloric restriction, an intervention that is known to increase lifespan in several model organisms, has been shown to modify and partially prevent the gene expression changes associated with aging (Lee et al., 1999; Cao et al., 2001; Pletcher et al., 2002). We wondered if the mutations present in these mice may have a similar effect. We used regression analysis to look for significant age-related changes in gene expression (genes with a significant age effect) and to locate genes whose age-related expression profiles were affected by these mutations (genes with significant age–genotype interactions).

The age-related changes in gene expression were not as dramatic or as numerous as those associated with genotype. Using regression analyses we found 289, 52 and 81 genes with a significant age effect (P < 0.001) in Ames dwarf (Prop1df/df), Little mice (Ghrhrlit/lit) and wild-type mice (Prop1+/+ and Ghrhr+/lit combined), respectively (see Supplementary tables). The majority of age-related changes in gene expression were increases in all treatments (at least two-thirds of the genes were up-regulated with age) and most of the age-dependent gene profiles were linear, with a fixed tendency to either increase or decrease with age.

Using the expression profiles of the 357 genes that changed significantly with age in Ames dwarf (Prop1df/df) or wild-type mice, we performed hierarchical clustering analysis with respect to age and genotype (Fig. 6). It is clear that the Prop1df/df genotype produced dramatic alterations in age-dependent changes in gene expression compared with wild-type mice. However, these alterations were more complex than a simple delay of the age-associated changes in gene expression present in wild-type mice. As seen in the gene clusters A and B of Fig. 6, the major effect of the Ames dwarf mutation was actually to introduce age-related changes not present in wild-type mice. Interestingly, the genes contained in these two clusters have a characteristic behavior. Their expression value is initially (at 3 months) considerably different (either higher or lower) between Ames dwarf and wild-type mice. As the mice age, however, the gene expression values observed in Ames dwarf have a tendency to approach the values observed in wild-type mice (which do not change significantly with age). By 24 months, the expression values of these genes are very close or overlapping between Ames dwarf and wild-type mice. We also observed smaller clusters that contain genes for which the age expression profile is negligibly affected by the Ames dwarf mutation and increase with time in both mutant and wild-type mice (cluster D in Fig. 6). The age–genotype interaction P values obtained by linear regression were particularly significant for the genes contained in clusters A, B and E, which reflects the dissimilarity between the age-related profiles of these genes in Ames dwarf vs. wild-type mice. This contrasted with the low significance of the age–genotype interaction P values of the genes contained in cluster D. The number of genes in each cluster with an age–genotype interaction of less than 0.01 is indicated in Fig. 6 (see Supplementary Table S7 for a complete list of the age–genotype interaction P values for the genes in Fig. 6).

image

Figure 6. Age-related expression profiles in Ames dwarf vs. wild-type mice. The expression profiles analysed in this figure correspond to the 357 genes that were found to change significantly (linear regression analysis, P < 0.001) with age in Ames dwarf or control mice. Hierarchical clustering analysis was performed with respect to age and genotype. Each row represents a different gene. Bright red and blue indicate high and low expression, respectively. The averaged expression profiles corresponding to the indicated clusters are plotted as scaled expression values as a function of age. N is the number of expression profiles contributing to the observed pattern. I is the number of genes in each cluster with an age–genotype interaction P value < 0.01. On the averaged expression profiles, the profiles corresponding to Ames dwarf are coloured in blue and the wild-type control mice profiles are coloured in red.

Download figure to PowerPoint

We performed gene ontology analyses on each one of the gene clusters shown in Fig. 6. Table 6 shows some of the gene categories with a significant over-representation in each gene cluster. This analysis reveals that the genes that show more dramatic changes with age in Ames dwarf compared with wild-type mice (clusters A and B) are indicative of increased lipid metabolism, including increased cholesterol biosynthesis and steroid biosynthesis in old Ames dwarf mice. In contrast, the genes that show more similar age-related profiles between Ames dwarf and controls (clusters C and D) are indicative of an increased immune and stress response.

Table 6.  Gene ontology classification of genes with a significant age effect in Ames dwarf or wild-type mice
ClusterGene ontologyGene categoryaList hitsbArray hitscP valued
  • a

    Gene ontology descriptions are provided for groups that show a significant over-representation in the set of 357 genes with a significant age effect in Ames dwarf or wild-type mice. The genes are grouped according to the gene clusters of Fig. 6.

  • b

    Number of genes in each category present in the entire array (Array hits).

  • c

    Number of genes present in this set of 357 genes (List hits).

  • d

    The P value refers to the EASE score (see Experimental procedures).

AGO Biological Processcarboxylic acid metabolism 82481.22E-04
GO Biological Processfatty acid metabolism 5 969.51E-04
GO Biological Processenergy pathways 41312.21E-02
GO Biological Processlipid metabolism 63602.39E-02
GO Cellular Componentmitochondrion106002.29E-03
GO Molecular Functionoxidoreductase activity 84424.49E-03
BGO Biological Processlipid biosynthesis121449.60E-07
GO Biological Processlipid metabolism173604.60E-06
GO Biological Processsterol metabolism 7 417.64E-06
GO Biological Processalcohol metabolism111662.52E-05
GO Biological Processcholesterol metabolism 6 376.79E-05
GO Biological Processsteroid metabolism 8 877.15E-05
GO Biological Processsteroid biosynthesis 6 617.40E-04
GO Biological Processcoenzyme metabolism 7 969.31E-04
GO Biological Processcholesterol biosynthesis 4 179.68E-04
GO Biological Processwater-soluble vitamin metabolism 4 242.71E-03
GO Biological Processisoprenoid biosynthesis 3  94.72E-03
GO Biological Processcarboxylic acid metabolism 92489.00E-03
GO Biological Processcomplement activation 4 379.32E-03
GO Biological Processhumoral immune response 4 704.97E-02
GO Cellular Componentendoplasmic reticulum163593.06E-05
GO Molecular Functionmonooxygenase activity 7 814.31E-04
GO Molecular Functionoxidoreductase activity144422.63E-03
GO Molecular FunctionN-acetyltransferase activity 3 253.64E-02
CGO Biological Processcholesterol metabolism 3 379.38E-03
GO Biological Processcarboxylic acid transport 3 391.04E-02
GO Biological Processorganic acid transport 3 391.04E-02
GO Biological Processsterol metabolism 3 411.14E-02
GO Biological Processcell-cell adhesion 41451.95E-02
GO Biological Processheterophilic cell adhesion 3 642.65E-02
GO Biological Processalcohol metabolism 41662.78E-02
GO Biological Processsteroid metabolism 3 874.65E-02
GO Cellular Componentlysosome 41007.22E-03
GO Molecular Functioncysteine-type peptidase activity 41161.13E-02
DGO Biological Processimmune response124144.96E-08
GO Biological Processresponse to biotic stimulus125641.13E-06
GO Biological Processresponse to pest/pathogen/parasite 82288.08E-06
GO Biological Processresponse to external stimulus138621.10E-05
GO Biological Processhumoral immune response 5 708.10E-05
GO Biological Processantigen processing 3  93.85E-04
GO Biological Processresponse to stress 84707.62E-04
GO Biological Processantigen processing 3 201.98E-03
GO Cellular Componentlysosome 31004.15E-02
GO Molecular Functiondefense/immunity protein activity 51226.80E-04
GO Molecular Functionantigen binding 3 161.25E-03
E No over-represented classes   

We performed a similar analysis using the expression profiles of the 130 genes that changed significantly with age in Little mice (Ghrhrlit/lit) or wild-type mice. As seen in Fig. 7, the Little mutation also produced alterations in age-dependent changes in gene expression; however, the effect was not as dramatic as with the Ames dwarf mice. In contrast to what we observed in Ames dwarf mice, most of the alterations in age-dependent changes in gene expression in Little mice did not represent an alteration on the profile (shape) of gene expression but rather a shift of the profile. This is particularly evident in gene clusters C and D, in which the age-dependent gene expression profiles are shifted toward lower expression levels in Little mice. The end result, however, is a delay in the elevation of the gene expression of these genes, and in this sense this can be considered as a partial delay of the age-related changes seen in wild-type mice. The genes contained in these clusters (C and D) are indicative of an increased immune and stress response in old mice (Table 7).

image

Figure 7. Age-related expression profiles in Little mice vs. wild-type mice. The expression profiles analysed in this figure correspond to the 130 genes that were found to change significantly (linear regression analysis, P < 0.001) with age in Little mice or wild-type mice. Hierarchical clustering analysis was performed with respect to age and genotype. Each row represents a different gene. Bright red and blue indicate high and low expression, respectively. The averaged expression profiles corresponding to the indicated clusters are plotted as scaled expression values as a function of age. N is the number of expression profiles contributing to the observed pattern. I is the number of genes in each cluster with an age–genotype interaction P value < 0.01. On the averaged expression profiles, the profiles corresponding to Little mice are coloured in blue and the wild-type mice profiles are coloured in red.

Download figure to PowerPoint

Table 7.  Gene ontology classification of genes with a significant age effect in Little mice or wild-type mice
ClusterGene ontologyGene categoryaList hitsbArray hitscP valued
  • a

    Gene ontology descriptions are provided for groups that show a significant over-representation in the set of 130 genes with a significant age effect in Little or wild-type mice. The genes are grouped according to the gene clusters of Fig. 7.

  • b

    Number of genes in each category present in the entire array (Array hits).

  • c

    Number of genes present in this set of 130 genes (List hits).

  • d

    The P value refers to the EASE score (see Experimental procedures).

AGO Biological Processsteroid metabolism3 877.95E-03
GO Biological Processlipid metabolism43601.72E-02
BGO Biological Processimmune response54149.65E-03
GO Biological Processdefense response55262.17E-02
GO Biological Processresponse to biotic stimulus55642.72E-02
GO Biological Processiron ion homeostasis2 173.50E-02
GO Biological Processtransition metal ion homeostasis2 234.71E-02
GO Biological Processacute-phase response2 244.91E-02
CGO Biological Processhumoral immune response6 701.45E-07
GO Biological Processimmune response94143.44E-07
GO Biological Processdefense response95262.13E-06
GO Biological Processresponse to pest/pathogen/parasite72282.51E-06
GO Biological Processresponse to biotic stimulus95643.59E-06
GO Biological Processresponse to stress74701.53E-04
GO Biological Processcomplement activation∖, alternative pathway2 112.05E-02
GO Biological Processcytolysis2 173.15E-02
GO Biological Processcomplement activation∖, classical pathway2 224.06E-02
GO Molecular Functiondefense/immunity protein activity61221.97E-06
GO Molecular Functionantigen binding3 163.69E-04
GO Molecular Functionendopeptidase inhibitor activity41151.09E-03
GO Molecular Functionprotease inhibitor activity41161.12E-03
GO Molecular Functioncopper ion binding3 361.89E-03
GO Molecular Functionenzyme inhibitor activity41773.76E-03
GO Molecular Functioncomplement activity2 203.58E-02
GO Molecular Functionenzyme regulator activity44434.40E-02
DGO Biological Processelectron transport24192.01E-01
GO Biological Processmetabolism549872.15E-01
EGO Biological Processresponse to heat4 213.17E-07
GO Biological Processresponse to temperature4 255.48E-07
GO Biological Processresponse to abiotic stimulus43431.44E-03
GO Biological Processresponse to stress44703.56E-03
GO Biological Processresponse to external stimulus48621.94E-02
GO Molecular Functionheat shock protein activity4 342.04E-06
GO Molecular Functionchaperone activity41219.52E-05
FGO Biological Processlipid metabolism63603.13E-04
GO Biological Processcarboxylic acid metabolism42488.95E-03
GO Biological Processorganic acid metabolism42499.05E-03
GO Biological Processfatty acid metabolism3 961.23E-02
GO Cellular Componentmicrosome5 992.34E-05
GO Molecular Functionglucuronosyltransferase activity2  61.03E-02
GO Molecular Functionaminopeptidase activity2 284.70E-02

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Supplementary material
  8. Acknowledgments
  9. References
  10. Supporting Information

Using gene expression array analysis we found that the Prop1df/df and Ghrhrlit/lit genotypes produce substantial alterations in transcript representation of a large number of genes. The majority of these gene expression changes are maintained throughout all ages examined. As might be predicted, we found that alterations in the gene expression profile produced by the Prop1df/df and Ghrhrlit/lit genotypes are strikingly similar in several ways. There was a large overlap of 552 genes that are present in both lists of 1125 and 1152 differentially expressed genes in Ames dwarf and Little mice, respectively. In all but five genes, they showed changes of gene expression in the same direction. The relative magnitude of the fold-changes for these overlapping genes in Ames dwarf and Little mice was also highly conserved. The number of genes that show parallel gene expression alterations in both mouse models of delayed aging is likely to be larger than this. Our criterion for the selection of genes that are affected by both mutations is highly restrictive (anovaP < 0.0001 in both gene lists). In similar reports a more relaxed significance criterion has been used for the comparison between gene lists generated in different systems (Miller et al., 2002). Indeed, we observed that the similarities between these gene lists extend beyond the group of overlapping genes. A gene ontology analysis run separately on each of these gene lists (Prop1df/df vs. Prop1+/+ and Ghrhrlit/lit vs. Ghrhrlit/+) reveals that both mutations affect the same biological processes in a similar fashion. Our data suggest that growth hormone deficiency is the most important deficiency that contributes to the phenotypic alterations seen in these mice, probably including lifespan extension. The relatively small group of genes that are affected exclusively in Ames dwarf mice probably reflects the unique hormone deficiencies of these mice.

The gene ontology analysis revealed that the Prop1df/df and Ghrhrlit/lit genotypes produce significant alterations in several metabolic pathways and functional gene classes. We observed a statistically significant over-representation of differentially expressed genes in several categories, including: amino acid metabolism, TCA cycle, mitochondrial electron transport, lipid and steroid metabolism, xenobiotic metabolism and oxidant metabolism. In many of these groups there was a significant trend for directional changes.

Major metabolic alterations in Ames dwarf and Little Mice

An increase in resistance to several forms of stress has been correlated with extended longevity in several model organisms (Parsons, 2003; Tatar et al., 2003). Our data reveal a generalized increase in several indices of stress resistance in Ames dwarf and Little mice. We observed up-regulation of several genes involved in reactive oxygen species (ROS) detoxification such as metallothionein (Mt1), thioredoxin (Txn2), glutathione s-tranferases, glutathione peroxidase (Gpx3) (only in Ames) and mitochondrial superoxide dismutase (Sod2) (only in Little). We also observed down-regulation of ROS-generating genes such as NAD(P)H oxidase (Nox4). The accumulation of oxidative damage is considered one of the basic mechanisms of aging. The observed up-regulation of genes involved in oxidant metabolism is consistent with the increase in oxidative stress resistance observed in these mice (Brown-Borg & Rakoczy, 2000, 2003; Hauck & Bartke, 2000; Romanick et al., 2004).

Another major alteration in Ames dwarf mice and Little mice was related to xenobiotic metabolism. Detoxification and elimination of xenobiotics and endobiotics is a major function of the liver and is important in maintaining the metabolic homeostasis of the organism (Willson & Kliewer, 2002). This system is composed of a large number of metabolizing enzymes and transporters classified into four groups: phase 0, phase I, phase II and phase III. Phase 0 genes are involved in the uptake of xenobiotics by the liver. Phase I genes are primarily a large group of monooxygenases that convert these hydrophobic chemicals into hydrophilic molecules, phase II genes further convert these modified products into amphiphilic anionic conjugates and phase III genes export these products out of the liver (Francis et al., 2003). We see a number of significant alterations in gene expression in all four groups. We observe an up-regulation of several cytochrome p450s (Cyp2b9, Cyp2b13, Cyp2b10, Cyp2b20, Cyp2c38, Cyp2a5, Cyp2d22, Cyp4a10 and Cyp4f14), flavin monooxygenases (Fmo1 and Fmo3), glutathione s-transferases (Gstm3, Gsto1, Gsta2, Gsta4, Gstm2 and Gsst1), sulfotransferases (Sth2, Ste and Sult1a1) and transporters (Slc21a1, Slc22a1 L, Abcb1a and Abcd2). Importantly, most of these genes are among those with the greatest fold-changes and lowest P values in the set of commonly differentially expressed genes in Ames dwarf and Little mice. The up-regulation of genes involved in xenobiotic metabolism may represent an increase in the ability of these mice to neutralize both exogenous toxic compounds and endogenous deleterious byproducts of metabolism. Because many of the drugs and xenobiotics detoxified by these groups of genes are oxidants in nature and are able to produce oxidative injury to cells (Jaeschke et al., 2002), it is possible that these changes also contribute to an increased resistance to oxidative stress. For example, the Ames dwarf mice are resistant to oxidative injury produced by paraquat (Bartke, 2000); however, it is not clear if this reflects an increased ability to detoxify ROS, an increased ability to metabolize paraquat before it can produce damage or, probably, a combination of both. Some reports are also indicative of an increased ability to metabolize toxic compounds in Little mice: it has been reported that they have an increased resistance to N,N-diethylnitrosamine (DEN)-induced hepatocarcinogenesis (Bugni et al., 2001). The nuclear receptors Constitutive Androstane Receptor (Car) and Pregnane X Receptor (Pxr) are known to regulate several of the xenobiotic metabolism genes whose expression is affected in Ames dwarf and Little mice. Although there was no change in the expression of Pxr, Car was up-regulated in both mutants. Car activity can be regulated by several steroid hormones; in particular, testosterone has an inhibitory effect. In this regard it is interesting to note that production of testosterone by the testes is significantly lower in Ames dwarf compared with wild-type mice (Chandrashekar & Bartke, 1993).

Mitochondria are the major source of ROS within the cell. The concerted up-regulation of several genes involved in oxidative phosphorylation may have a significant impact on mitochondrial function or ROS production. Interestingly, some of the most significant alterations in gene expression within this group of genes correspond to components of complex I (NADH Co Q reductase) of the respiratory chain (Ndufv1, Ndufv2, Ndufc1, Ndufb4). Complex I may be one of the major physiologically and pathologically relevant ROS-generating sites in mitochondria (Kushnareva et al., 2002; Liu et al., 2002). We also saw an up-regulation of several other components of the electron transport chain, such as succinate dehydrogenase in complex II (Sdhb), cytochrome c reductase in complex III (Uqcrc2) and ATP synthase in complex V (Atp5g1, Atp5h, Atp5f1, Atp5o). Interestingly, in parallel with these changes in oxidative phosphorylation genes, we observed a concerted up-regulation of genes involved in the TCA cycle (Aco2, Idh2, Idh3b, Mdh2, Sdhb, Suclg1). In Little mice there was an additional up-regulation of the TCA cycle genes: Suclg2, Fh1, Idh3g and Sdhc.

The alterations in amino acid metabolism in Ames dwarf and Little mice were suggestive of an altered methionine metabolism and an increased cysteine and glutamate biosynthesis. Several genes involved in the biosynthesis of cysteine from methionine were up-regulated: methionine adenosyltransferase (MAT), S-adenosylhomocysteine hydrolase (Ahcy) and cystathionase (Cth). Cysteine sulphinic acid decarboxylase (Csad), an enzyme involved in biosynthesis of taurine from cysteine, was down-regulated. The first enzyme in the conversion from proline to glutamate, proline dehydrogenase (Prodh), was up-regulated. Glutamate oxaloacetate transaminase 1 (Got1), which catalyses conversion of aspartate and 2-oxoglutarate into oxaloacetate and glutamate, was also up-regulated. Glutamate and cysteine are precursors to the biosynthesis of glutathione, which plays a key role in the cellular defense against oxidative stress. The gene expression alterations in these genes suggest that Ames dwarf and Little mice have potentially a better capacity for the production of glutathione, which can translate into an increased ability to deal with oxidative stress. In fact, glutathione synthetase (Gss), the second step enzyme in the synthesis of glutathione, was up-regulated in Ames dwarf mice. The observations regarding an altered methionine metabolism are in agreement with a previous study in Ames dwarf mice (Uthus & Brown-Borg, 2003).

There were several alterations in lipid metabolism in these mutants. We observed a concerted up-regulation of genes involved in several steps of the fatty acid beta-oxidation cycle: Acadl, Echs1, Ehhadh and Hadhsc. An increase in fatty acid beta-oxidation can potentially lead to an increased production of the TCA cycle substrate acetyl coenzyme A. This would be consistent with the up-regulation of genes in the TCA cycle and oxidative phosphorylation in these mice. We also observed evidence of a decreased biosynthesis of bile acids: cytochrome 7b1 (Cyp7b1), cytochrome 27a1 (Cyp27a1) and peroxisomal thilase 2 (Scp2), which have central roles in the biosynthesis of bile acids from cholesterol, were down-regulated in both Ames dwarf and Little mice. We also observed a dramatic down-regulation of several hydroxysteroid dehydrogenases involved in the inactivation of steroid hormones, such as oestrogens and androgens, by the liver (Hsd17b12, Hsd17b2, Hsd3b2, Hsd3b5, Hsd3b6). The down-regulation of these enzymes may be an indication of alterations in the levels of sex steroid hormones in these mice, such as the reduced levels of testosterone in Ames dwarf (Chandrashekar & Bartke, 1993). The diverse metabolic alterations in Ames dwarf and Little mice seem to be co-ordinated to promote a generalized increase in stress resistance and an increased capacity for energy production. Our data indicate that these long-lived mutants are characterized by an increased, rather than a decreased, metabolic activity, suggesting that mutations in the insulin-like pathway may increase lifespan while simultaneously increasing metabolic activity.

Our data agree with previous published results examining the gene expression profile of other mouse models of delayed aging. We observe a partial overlap with genes previously reported to be significantly affected in mice under caloric restriction (Miller et al., 2002). In Ames dwarf mice, of the 352 genes reported to be affected in calorie restriction, 27 showed parallel changes in gene expression, with 13 up-regulated and 14 down-regulated genes. In Little mice, 27 genes showed parallel changes in gene expression, with 16 up-regulated genes and 11 down-regulated genes. In a gene expression profile of Snell dwarf, a mouse model of delayed aging that is phenotypically very similar to Ames dwarf mice, 60 genes were found to be affected by the dw mutation (Dozmorov et al., 2002). Our data confirm changes in 13 and 11 of these genes in Ames dwarf and Little mice, respectively. In a recent study in Ames dwarf mice (Tsuchiya et al., 2004), 309 genes were found to be affected. Our data agree with 67 and 51 of these gene expression changes in Ames dwarf and Little mice, respectively.

Effect of the Prop1df/df and Ghrhrlit/lit genotypes on age-related changes in gene expression

Age had a considerably smaller effect on gene expression than the Prop1df/df and Ghrhrlit/lit genotype effects. It has been suggested that alterations or mutations that extend lifespan in model organisms may prevent or delay age-related changes in gene expression, as has been observed in caloric restriction (Lee et al., 1999; Cao et al., 2001; Pletcher et al., 2002). Our data show that the Prop1df/df genotype had dramatic and complex effects on age-dependent changes in gene expression compared with wild-type mice. Surprisingly, the Prop1df/df genotype actually introduced age-related changes in gene expression not present in wild-type mice. This is an interesting phenomenon that has been previously observed in flies under caloric restriction. Although a large number of age-dependent changes in control flies were ameliorated under caloric restriction, caloric restriction introduced an equally large number of changes not seen in control flies (Pletcher et al., 2002). We observed a situation in which Ames dwarf and control mice follow different age-dependent expression profile trajectories, and in some senses Ames dwarf mice can be considered to be aging differently to control mice. Although the alterations in age-related changes in gene expression produced by the Prop1df/df genotype are likely to be a reflection of the dramatic life-long alterations in gene expression that are present in these mice since an early age, these changes may also prove important in understanding the delayed aging seen in these mice. Interestingly, the majority (∼70%) of the genes with a significant age effect in Ames dwarf mice have a characteristic behavior: initially their expression level differs greatly between mutant and control mice, but with time, the expression levels of these genes in Ames dwarf tend to approach and overlap with those seen in wild-type mice. Age-related changes in gene expression are frequently assumed to be deleterious, because they represent a departure from what is considered to be a positive, young pattern of gene expression. Therefore, these gene expression changes can be seen as a departure from a state that favors longevity (Ames dwarf mice) to a state that favors aging (wild-type mice). It is noteworthy that this group contains genes that have been strongly associated with lifespan extension. In particular, Igf1 transcript levels increased with time in Ames dwarf mice, whereas Igf1 binding protein 1 and 2 decrease, suggesting a partial recovery of Igf signaling in Ames dwarf mice at 24 months. The survival curve for the Ames dwarf mice reflects a shift in the age at which the age-dependent increase in mortality risk first becomes appreciable, from around 15–18 months in wild-type to around 25–28 months in Ames dwarf mice (Bartke et al., 2001). The relatively rapid change of the level of expression of these genes in Ames dwarf mice between 12 and 24 months may be causally related to this increase in mortality rate. Although not as dramatic as with Ames dwarf mice, the Ghrhrlit/lit genotype also had an effect on age-dependent changes in gene expression compared with wild-type mice. However, similarly to Ames dwarf mice, the main effect of this mutation was not a generalized delay of the age-related changes seen in wild-type mice.

We found only a very small number of genes whose wild-type age-related change in gene expression was actually delayed by these mutations. Only two genes in Ames dwarf: Enpp3 (ectonucleotide pyrophosphatase/phosphodiesterase 3) and Mucdhl (mucin and cadherin like); and only five genes in Little mice: Slc3a1 (solute carrier family 3, member 1), Lgmn (legumain), Npc2 (Niemann Pick type C2), Evc (Ellis van Creveld) and Lyss (lysozyme). Of these genes, only Enpp3 came close to delay by both mutations. Interestingly, abnormal Enpp3 expression is involved in pathological mineralization, crystal depositis in joints, invasion and metastasis of cancer cells, and type 2 diabetes (Goding et al., 2003). It might be possible that the increase in expression with age of this gene in wild-type mice may be related to some age-related impairment in function. Most of the genes that show a highly consistent age-related expression profile across all genotypes corresponded to immune and stress response genes and all of them were up-regulated with age. These data are in agreement with previous reports that suggest that inflammation and a generalized stress response is characteristic of the aging process in the liver (Cao et al., 2001).

Taken together, the data from Ames dwarf and Little mice suggest that these life-extending mutations fail in general to prevent many of the age-related changes in gene expression present in wild-type mice. This implies that aging is delayed even though the expression of these gene changes similarly during aging in the mutant mice. This would then bring into question whether the changes in gene expression of these genes in wild-type mice are actually causally related to age-related impairments in function or mortality. Although this might be the case with some of these genes, one has to be cautious about reaching such a conclusion because the dramatic changes in gene expression (over 1000 genes, maintained at all ages) induced by these mutations may be able to compensate or surpass any detrimental effects that these age-related changes in gene expression may have, even if they were actually related to impairments in function or mortality.

Finally, it is important to bear in mind that at least some of the alterations in gene expression reported here as a consequence of the Prop1df/df and Ghrhrlit/lit genotypes may be particular to the liver tissue. Analysis of additional tissues, particularly post-mitotic tissues such as brain or muscle, is likely to provide a more complete picture of the alterations that contribute to the extended longevity of these mice. Our data represent a genome-wide analysis of the alterations in liver gene expression produced by the Prop1df/df and Ghrhrlit/lit genotypes. This molecular characterization should be useful as a guide for the development of future hypotheses and experiments for the study of lifespan extension mechanisms in these mice.

Experimental procedures

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Supplementary material
  8. Acknowledgments
  9. References
  10. Supporting Information

Mice

Prop1df/+ female and male heterozygote breeders, purchased from Jackson Laboratories, were crossed to produce the Prop1+/+ and Prop1df/df male mice used in these experiments. Little mice breeders were also purchased from Jackson Laboratories. Ghrhrlit/Ghrhrlit females and Ghrhrlit/+ males were crossed to produce the Ghrhrlit/+ and Ghrhrlit/Ghrhrlit male mice used in these experiments. We used wild-type Prop1+/+ littermates as controls for the Ames dwarf mice. Heterozygous Ghrhr+/lit littermates were used as controls for the Little mice; the phenotype of these mice is indistinguishable from wild-type Ghrhr+/+ mice. All mice were genotyped according to established protocols (Dolle et al., 2001). The animals were weaned at the age of 21–30 days and housed in groups of 3–5 animals of the same gender. Animals were housed in a room with controlled photoperiod of 12 h light−12 h darkness (lights on from 07:00 to 19:00 h) and a temperature of 22 ± 2 °C. Animals were given free access to water and pelleted diet (5010 rodent diet, LabDiet, PMI Nutrition International, Brentwood, MO, USA). At the ages of 3, 6, 12 or 24 months, the animals were anaesthetized with isoflurane and killed by cervical dislocation; livers (and other organs used for other studies) were collected, flash frozen in liquid nitrogen and stored at −70 °C. Three biological replicates were used per time-point for each one of the four experimental groups: Ames dwarf (Prop1df/df), Ames dwarf control (Prop1+/+), Little mice (Ghrhrlit/lit) and Little mice control (Ghrhr+/lit). Each sample was hybridized to an individual oligonucleotide array for a total of 48 arrays.

Microarray analysis and statistical analysis

Total RNA was extracted from frozen livers using the Qiagen RNeasy (Qiagen, Valencia CA, USA) purification kit in accordance with the manufacturer's protocols. RNA quality assessment was performed using the Agilent 2100 Bioanalyser (Agilent Technologies, Palo Alto, CA, USA). Specific mRNA transcript levels were determined using Affymetrix MOE430A high-density oligonucleotide arrays according to standard Affymetrix protocols (Affymetrix, Santa Clara, CA, USA). Briefly, double-stranded cDNA was synthesized from total liver RNA. An in vitro transcription reaction was then performed to produce biotin-labelled cRNA from the cDNA. The cRNA was then fragmented and hybridized to the probe array during 16 h of incubation. The arrays were scanned using an Affymetrix GeneChip Scanner 3000. The image files were analysed for probe intensities and converted to tabular formats (CEL files) using the Microarray Suite Expression Analysis software from Affymetrix. Quality control, normalization and calculation of expression values for each probe set were performed using the DNA-Chip Analyser software (Li & Wong, 2001). Before statistical analysis the data were filtered to include probe data sets that were ‘present’ (as determined by the detection algorithm incorporated in the DNA-Chip Analyser software) in at least 75% of the arrays per experimental category.

Genotype effects

We used analysis of variance (anova, parametric test, assuming equal variance) to find significant alterations in gene expression that distinguish between mutant and wild-type mice in each category: Prop1df/df vs. Prop1+/+ and Ghrhrlit/lit vs. Ghrhr+/lit (for this analysis all the different ages in each category were grouped). We selected a nominal P value of 0.0001 as our criterion for the selection of significant genotype effects. To test for unique alterations in gene expression produced by the Prop1df/df or Ghrhrlit/lit genotypes we performed a two-way analysis of variance. Genes with a significant (P < 0.0001) genotype (mutant or wild-type) – strain (Ames dwarf or Little mice) interaction were considered to be differentially affected by the Prop1df/df or Ghrhrlit/lit genotypes. Out of this group of genes, those with a significant genotype effect in Ames dwarf (P < 0.0001) but not in Little mice were considered to be affected exclusively in Ames dwarf mice. Similarly, those genes with a significant genotype effect in Little mice (P < 0.0001) but not in Ames dwarf mice were considered to be affected exclusively in Little mice. This sort of analysis was required because a simple comparison of the lists of genes affected in Ames dwarf vs. Little mice cannot be used as a reliable method for finding unique alterations in gene expression. Although we can be very confident that the overlapping genes represent common alterations in gene expression, we cannot conclude that all of the non-overlapping genes represent unique alterations in gene expression for each mutant because of the high proportion of type II errors (false negatives) characteristic of most microarray experiments. For the genotype effects, the fold-change for each gene was calculated as the ratio of its average expression values in mutant and wild-type mice: Prop1df/df/Prop1+/+ for Ames dwarf and Ghrhrlit/lit/Ghrhr+/lit for Little mice.

Age effects

A series of linear regression analyses were used to find age-related changes in gene expression in Ames dwarf (Prop1df/df), Little mice (Ghrhrlit/lit) and wild-type mice (Prop1+/+ and Ghrhr+/lit, combined data). These analyses were done using the simple linear regression model of the statistical programming language R (http://www.r-project.org) assuming normality and homoscedasticity. Linear regression analysis was also used to test for age–genotype interactions in Prop1df/df vs. Prop1+/+ and Ghrhrlit/lit vs. Ghrhr+/lit.

The Gene Ontology over-representation analyses were performed with the help of the software application EASE (http://david.niaid.nih.gov/david/ease.htm). The statistical measure of over-representation used was a conservative adjustment of the Fisher exact probability, which strongly penalizes the significance of categories supported by few genes and negligibly penalizes categories supported by many genes, yielding more robust results (Hosack et al., 2003). Clustering analyses were performed on the DNA-Chip Analyser software. For all clustering analyses the expression values for each gene across all samples were standardized to have mean 0 and standard deviation 1. Genes were clustered according to the correlation coefficient between the standardized expression values across the samples used. The standardization and clustering methods used have been described previously (Eisen et al., 1998; Golub et al., 1999).

Real-time PCR

Total RNA was extracted from frozen livers using the Quiagen RNeasy purification kit (Qiagen) in accordance with the manufacturer's protocols. DNase-treated total liver RNA was reverse transcribed using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA) according to the supplier's protocol. Real-time PCR was performed using SYBR Green PCR Master Mix and the ABI Prism 7000 Sequence Detection System (Applied Biosystems, Foster city, CA, USA). B-actin was used as a normalizer. Amplification specificity of all targets was confirmed by melting curve analysis. Thermal cycling conditions consisted of an initial step at 95 °C for 10 min to activate the Taq DNA polymerase and 40 cycles of sequential denaturation at 95 °C for 15 s and annealing/extension at 60 °C for 60 s. Data analysis was performed using the ABI Prism 7000 SDS Software (Applied Biosystems). The primers used are listed in the Supplementary Table S9. In all cases we tested samples from three mutant and three control mice for both the Ames dwarf and Little mice at 6 months of age. The real-time PCR analysis was performed according to the comparative CT method. The fold-changes were calculated according to the expression: 2–ΔΔCT. In our experiments: ΔΔCT = ΔCT, mutant − ΔCT, wild-type, where, ΔCT, mutant = Average CT, target gene − Average CT, β-actin and ΔCT, wild-type= Average CT, target gene − Average CT, β-actin. The average CT values for both mutant and control wild-type mice were based on three biological replicates. The CT value for each biological replicate represents the average of three technical replicates. The P values reported for these changes refer to a one-tailed t-test (assuming homoscedasticity) between the normalized CT values (CT, target gene − CT, β-actin, three biological replicates) in mutant vs. control mice.

Supplementary material

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Supplementary material
  8. Acknowledgments
  9. References
  10. Supporting Information

The following tables and figures are available as supplementary material from: http://www.blackwellpublishing.com/products/journals/suppmat/ACE/ACE125/ACE125sm.htm. The raw microarray data is available online from Array Express (http://www.ebi.ac.uk/arrayexpress) with the accession number E-MEXP-153.

Table S1 Differentially expressed genes: Ames dwarf mice

Table S2 Differentially expressed genes: Little mice

Table S3 Common affected genes in Ames dwarf and Little mice

Table S4 Age-related changes in gene expression: Ames dwarf

Table S5 Age-related changes in gene expression: Little mice

Table S6 Age-related changes in gene expression: wild-type mice

Table S7 Age–genotype interactions: Ames dwarf mice

Table S8 Age–genotype interactions: Little mice

Table S9 Primers used for real-time PCR assays

Table S10 Unique alterations in gene expression in Ames dwarf

Table S11 Unique alterations in gene expression in Little mice

Table S12 Raw data for real-time PCR analysis

Fig. S1 Average CT values and standard deviation values from real-time PCR measurements in Ames dwarf mice.

Fig. S2 Average CT values and standard deviation values from real-time PCR measurements in Little mice.

Fig. S3 Effects of the Prop-1 and Ghrhr mutations on wild-type age-related changes in gene expression. The gene clusters show the diverse types of alterations that the Prop-1 and Ghrhr mutations have on the wild-type age-related changes in gene expression. The expression profiles analyzed in this figure correspond to the 81 genes that change significantly (P < 0.001) with age in wild-type mice. The averaged expression profiles corresponding to the indicated clusters are plotted as scaled expression values as a function of age. The table below indicates the total number of genes that contribute to each profile, the number of genes with an age-genotype interaction (P < 0.05).

Acknowledgments

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Supplementary material
  8. Acknowledgments
  9. References
  10. Supporting Information

We thank Scott Pletcher and Leif Peterson for helpful discussions. This work was supported by The Ellison Medical Foundation and grant A619254.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Supplementary material
  8. Acknowledgments
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Supplementary material
  8. Acknowledgments
  9. References
  10. Supporting Information

Table S1. Differentially expressed genes: Ames dwarf mice. Table S2. Differentially expressed genes: Little mice. Table S3. Common affected genes in Ames dwarf and Little mice. Table S4. Age-related changes in gene expression: Ames dwarf. Table S5. Age-related changes in gene expression: Little mice. Table S6. Age-related changes in gene expression: wild-type mice. Table S7. Age?genotype interactions: Ames dwarf mice. Table S8. Age?genotype interactions: Little mice. Table S9. Primers used for real-time PCR assays. Table S10. Unique alterations in gene expression in Ames dwarf. Table S11. Unique alterations in gene expression in Little mice. Table S12. Raw data for real-time PCR analysis. Fig. S1. Average CT values and standard deviation values from real-time PCR measurements in Ames dwarf mice. Fig. S2. Average CT values and standard deviation values from real-time PCR measurements in Little mice. Fig. S3. Effects of the Prop-1 and Ghrhr mutations on wild-type age-related changes in gene expression. The gene clusters show the diverse types of alterations that the Prop-1 and Ghrhr mutations have on the wild-type age-related changes in gene expression. The expression profiles analyzed in this figure correspond to the 81 genes that change significantly (p<0.001) with age in wild-type mice. The averaged expression profiles corresponding to the indicated clusters are plotted as scaled expression values as a function of age. The table below indicates the total number of genes that contribute to each profile, the number of genes with an age-genotype interaction (p<0.05).

FilenameFormatSizeDescription
ACEL_125_sm_FigS1.pdf69KSupporting info item
ACEL_125_sm_FigS2.pdf63KSupporting info item
ACEL_125_sm_FigS3.pdf66KSupporting info item
ACEL_125_sm_tableS1.pdf158KSupporting info item
ACEL_125_sm_tableS2.pdf430KSupporting info item
ACEL_125_sm_tableS3.pdf278KSupporting info item
ACEL_125_sm_tableS4.pdf199KSupporting info item
ACEL_125_sm_tableS5.pdf24KSupporting info item
ACEL_125_sm_tableS6.pdf39KSupporting info item
ACEL_125_sm_tableS7.pdf170KSupporting info item
ACEL_125_sm_tableS8.pdf70KSupporting info item
ACEL_125_sm_tableS9.pdf17KSupporting info item
ACEL_125_sm_tableS10.pdf51KSupporting info item
ACEL_125_sm_tableS11.pdf34KSupporting info item
ACEL_125_sm_tableS12.pdf236KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.