Sexual-dimorphism in human immune system aging

Differences in immune function and responses contribute to health- and life-span disparities between sexes. However, the role of sex in immune system aging is not well understood. Here, we characterize peripheral blood mononuclear cells from 172 healthy adults 22–93 years of age using ATAC-seq, RNA-seq, and flow cytometry. These data reveal a shared epigenomic signature of aging including declining naïve T cell and increasing monocyte and cytotoxic cell functions. These changes are greater in magnitude in men and accompanied by a male-specific decline in B-cell specific loci. Age-related epigenomic changes first spike around late-thirties with similar timing and magnitude between sexes, whereas the second spike is earlier and stronger in men. Unexpectedly, genomic differences between sexes increase after age 65, with men having higher innate and pro-inflammatory activity and lower adaptive activity. Impact of age and sex on immune phenotypes can be visualized at https://immune-aging.jax.org to provide insights into future studies.


4 8
Only B cells showed the opposite epigenomic remodeling pattern with aging between sexes, where B cell-1 4 9 3 9 8 carefully age-matched healthy adults representing the entire human lifespan. Using novel systems 3 9 9 immunology pipelines, we discovered a genomic signature of aging that is shared between sexes, which is 4 0 0 composed of 1) declines in T cell functions, 2) increases in cytotoxic (NK, memory CD8+ T) and 4 0 1 monocyte cell functions with age. This common signature corresponds to the two hallmarks of immune 4 0 2 system aging: immune-senescence and inflamm-aging 14 . We noted that, despite these similarities, male 4 0 3 immune system goes through greater changes that are not attributable to clinical or cohort differences.

0 4
Most notably, we detected 15 times more monocyte-specific loci to be activated in men compared to 4 0 5 women. Previously, it has been reported from whole blood samples that the DNA methylome of men ages 4 0 6 faster compared to women 27 . Here, we uncovered an accelerated aging phenotype for men in chromatin 4 0 7 accessibility maps and uncovered which cell types, cellular functions, and molecules differentially age 4 0 8 between men and women. In addition to this common signature, our data revealed for the first time that 4 0 9 aging has opposing effects on the B cells of men and women, where B cell-specific loci/genes were 4 1 0 13 modestly activated in women but significantly inactivated in men, potentially linked to sex-differences in 4 1 1 auto-immunity and humoral responses.

1 2 1 3
Our study is unique in its inclusion of data from middle-aged individuals to capture chronological 4 1 4 changes and the timing of these changes, an approach that is essential in distinguishing changes 4 1 5 attributable to aging from those resulting from maturation 28 . We captured correlated chronological 4 1 6 patterns in chromatin accessibility and gene expression maps, including upward trends (i.e., increasing 4 1 7 accessibility/expression with age) associated with cytotoxic cells and monocytes, and downward trends 4 1 8 associated with T cells in both sexes. Combined analyses of male/female temporal patterns further 4 1 9 highlighted the differential aging of the B cell-related loci. Breakpoint analyses uncovered that although 4 2 0 aging-related changes accumulate gradually throughout adult life, there are two periods in the human 4 2 1 lifespan during which the immune system undergoes abrupt epigenomic changes. The first breakpoint was 4 2 2 around the late thirties in both sexes, whereas the second breakpoint was detected earlier in men (mid-4 2 3 sixties) compared to women (early seventies). Interestingly, this difference between sexes (5-6 years) is 4 2 4 comparable to the difference in their life expectancy: 76.9 for men and 81.6 for women according to 4 2 5 World Health Organization 2015 records for USA 29 . In both sexes, the second breakpoint was associated 4 2 6 with accelerated epigenomic changes and occurred ~12-15 years prior to the end of the average lifespan.

2 7
Our findings regarding the timing of age-related changes are novel and can be essential while making 4 2 8 decisions regarding when to start clinical interventions/therapies. 4 2 9 4 3 0 Detailed comparison between sexes at different age groups uncovered that PBMCs of men and women 4 3 1 significantly differed after the age of 65, contrary to expectations due to the decline in sex hormones.

3 2
Annotation of sex-biased loci/genes revealed that older women have more genomic activity associated to 4 3 3 adaptive cells (B/T cells), whereas older men have more genomic activity associated to innate cells, 4 3 4 particularly monocytes and inflammation. Some of these genomic differences, but not all, could be 4 3 5 attributable to changes in cell compositions, e.g., the decline of B cell frequencies exclusively in older 4 3 6 men ( Fig. 1d), which is confirmed in a second independent cohort 12 (Fig. S7d) and a previous study 4 3 7 conducted in a Japanese cohort 30 . This concordance across cohorts suggests that sex-specific aging 4 3 8 signatures described here might be conserved across ethnicities/populations, which needs to be further 4 3 9 investigated in the future. Increase in monocyte associated loci/genes was associated with inflammatory 4 4 0 pathways/genes and was more significant in men, suggesting an accelerated inflamm-aging signature for 4 4 1 men. This signature was confirmed by an independent cohort 11 that showed that older men have higher 4 4 2 levels of pro-inflammatory cytokines in their serum compared to older women (e.g., IL1RA, IL6, IL18) 4 4 3 ( Fig. 6e). Interestingly, neither in our data nor in data from a bigger cohort 12 , could we detect significant 4 4 4 differences between older men and older women in terms of CD14 + and CD16 + monocyte cell numbers 4 4 5 (absolute counts and frequencies) (Figs. S7e, S7e). Therefore, age-related activation of monocytes and 4 4 6 differences between the sexes in the rate of activation likely stems from cell-intrinsic changes. It is 4 4 7 currently unclear why men experience greater inflammation with age. Future studies are needed to 4 4 8 explore potential reasons, including the role of infections, sex hormones, and differential activation of 4 4 9 transposable elements 31 .

5 0
By following a systems immunology approach in human PBMCs, this study uncovered which cell types 4 5 1 and which immune functions are differentially affected with aging between men and women. Future 4 5 2 studies are needed to more precisely describe these differences at single cell resolution and in sorted cells 4 5 3 and to establish the functional implications of these sex differences. Moreover, future studies are needed 4 5 4 to explore the potential of the molecules identified in this study (i.e., IL7R, LEF1, TCF7, STAT5B, IL8, 4 5 5 IL18) as biomarkers of immune system aging in men and women and their role in contributing to sex 4 5 6 differences in immunosenescence and inflamm-aging. Taken together, these findings indicate that sex 4 5 7 plays a critical role in human immune system aging and should be taken into consideration while 4 5 8 searching for molecular targets and time frames for interventions/therapies to delay aging and age-related publication, it will be shared publicly and will be maintained regularly. (negative). PC1 scores from ATAC-seq and RNA-seq data increased with increasing age. Furthermore, we detected 4 9 5 differences in PC1 scores of sexes in older subjects, where older men had higher scores than older women. (c)

9 6
Functional enrichment of PC1-related genes based on immune modules 20 . Note that myeloid/inflammation related 4 9 7 genes were associated with high and positive scores, whereas adaptive immunity/lymphoid related genes were 4 9 8 associated with high negative PC1 scores. These enrichments align well with an age-related increase in the myeloid 4 9 9 lineage and inflammation, and an age-related decline in naïve T cell activity. Hypergeometric test was used to Older: 65+ years old. Wilcoxon signed rank test was used to compare values from female and male subjects, and 5 0 3 older and young individuals. Note that T cell proportions decline with age in both sexes, whereas the decline in B 5 0 4 cell proportions is specific to older men. fold change (older-young) versus average read count for ATAC-seq peaks in women (left) and men (right). Peaks 5 0 8 differentially opening (closing) with age are represented in orange (purple). Note that age has a stronger effect on 5 0 9 male epigenomes, evinced both in the magnitude of fold change values and the number of differential peaks.

1 0
Differential accessibility of ATAC-seq peaks was tested using a generalized linear model based on read counts, with 5 1 1 significance assessed at a 5% FDR threshold, after using Benjamini-Hochberg P-value adjustment. (b) Functional 5 1 2 state annotations for differentially opening/closing peaks in men and women using chromHMM states in PBMCs 5 1 3 and immune cell subsets. Closing peaks tend to be enhancers/promoters across all cell types, whereas opening peaks opening peaks, whereas loci specific to T cells are enriched in closing peaks. Interestingly, B cell specific loci are 5 1 7 more likely to be a opening peak in women and closing peak in men. Note that, a much larger number of monocyte- profiles (grouped by age and sex) around the IL7R locus in women (top) and men (bottom). In both sexes, this locus 5 2 0 is associated with chromatin closing with age. Right: Normalized chromatin accessibility (average of three peaks 5 2 1 shown in the figure) and gene expression levels for IL7R in young and older women (top) and men (bottom). In both 5 2 2 sexes, chromatin accessibility and gene expression levels decline for this molecule with age. The decline is much 5 2 3 more significant in men. (e) Enrichment of differentially expressed (DE) genes using cell-specific gene sets from 5 2 4 scRNA-seq data. Enrichment p values are calculated using hypergeometric test. Note the activation of gene 5 2 5 expression programs for innate cells (NK, monocytes) and inactivation for T cell programs in both sexes. In 5 2 6 agreement with ATAC-seq, B-cell specific genes were downregulated with age specifically in men. (f) Average 5 2 7 normalized expression levels of T cell-specific genes grouped by age group and sex. Note the decline in both sexes.

2 8
Box plots on the top summarizes the data from all genes represented in the heatmap. of data points. (c) Correlation between sexes for age-related ATAC-seq remodeling stratified by cell-specific loci 5 3 5 from chromHMM annotations. Note that the highest correlation is observed in naïve T cells, which is associated 5 3 6 with negative fold changes (i.e., loss in chromatin accessibility with age) in both sexes. Chromatin remodeling 5 3 7 correlates the least between sexes for B and monocyte-specific loci. (d) Correlation between sexes for age-related 5 3 8 RNA-seq remodeling stratified by cell-specific genes from scRNA-seq data. Note that the highest correlation is 5 3 9 observed in naïve T cells, which is associated with negative fold changes (i.e., downregulation with age) in both   mean. K-means clustering was used to group these peaks into three clusters in men and women (M1/F1, M2/F2, 5 5 0 M3/F3). Color bar on the top represents discrete age groupings as defined in this study (young, middle-aged, older).

1
Rows are annotated according to their position relative to the nearest TSS: proximal if within <1kbp distance, distal  Heatmap of ATAC-seq peaks (fitted values from ARIMA models) with a chronological trend in both women and 5 6 5 men (N=3,197). Values represent z-score normalized accessibility values relative to the row mean. K-means 5 6 6 clustering was used to group these peaks into three clusters (C1-C3) using concatenated data from men and women.

6 7
Color bar on the top represents discrete age groupings as defined in this study (young, middle-aged, older). Rows are associates cluster C1 to T cells, suggesting a delayed loss of accessibility in women relative to men; C2 to CD19 + 5 7 2 cells, suggesting the presence of CD19 + specific loci with opposing temporal behavior in men and women; and C3 sequence similarity, and most significant p-value for each family is represented here. (e-f) Expression levels of TFs 5 7 7 associated to cluster 1 (C1) and cluster 3 (C3) grouped by age group and sex. Cluster 2 (C2) is omitted since it lacks 5 7 8 sufficient number of peaks to allow an enrichment test. average read count for RNA-seq data at three age groups using only genes on autosomal chromosomes.

8 5
Transcriptomic differences between sexes increase with age. Genes more expressed in women (men) are represented 5 8 6 in red (blue). (c) Enrichment of sex-biased peaks/genes in older individuals using cell-specific gene sets obtained 5 8 7 from scRNA-seq data. Note the male bias toward increased accessibility/expression for monocytes and DCs and the (GLM) to establish the association between PBMC cell composition data with age group, sex, and their interaction 6 1 3 (age*sex). Darker colors indicate stronger association. Age is most strongly associated with the variation in naïve T 6 1 4 cell proportions, whereas sex, age*sex is most strongly associated with the variation in B cell proportions. (g) Top: 6 1 5 Distribution of subjects by age and sex, showing a comparable spread of ages between sexes and among decades.

1 6
Bottom: Distribution of Rockwood frailty index values 33 stratified by age groups and sexes. No significant 6 1 7 difference detected between men and women at all three age groups. Note that the frailty index values increase with 6 1 8 age, more so in women with none of the individuals having a Frailty Index value greater than 0.25. Most subjects 6 1 9 were non-frail (Frailty index (FI): 0-0.1), a few were prefrail (FI >0.1-0.21) but regardless the frailty did not 6 2 0 influence the observed genomic changes.  test. Note the activation of innate cells (NK and monocytes) and inactivation of T cells with age. The enrichment p-6 3 5 values were more significant in men compared to women. B cell-specific regions were enriched for closing peaks in 6 3 6 men and for opening peaks in women. Grayed out circles are non-significant at a 5% FDR; numbers on circles 6 3 7 represent the number of peaks in each category.   test, ranging from 10 to 20 years. Stronger evidence of a breakpoint emerges as the same age bracket is detected as 6 7 6 significant at multiple window spans. C1, C2, C3 stands for cluster1-3. clustering is used to group these peaks into three clusters (C1-C3) using concatenated data from men and women.

8 2
Color bar on the top represents discrete age groupings as defined in this study (young, middle-aged, older).  Supplementary table legends 7 0 9 Table S1. Detailed metadata on subjects, including types of data generated for each subject.   likelihood that these findings can be translated to the general population 35 . Subjects were carefully screened in order 7 3 8 to exclude potentially confounding diseases and medications, as well as frailty. Individuals who reported chronic or 7 3 9 recent (i.e., within two weeks) infections were also excluded. Subjects were deemed ineligible if they reported a 7 4 0 history of diseases such as congestive heart failure, ischemic heart disease, myocarditis, congenital abnormalities,

4 1
Paget's disease, kidney disease, diabetes requiring insulin, chronic obstructive lung disease, emphysema, and 7 4 2 asthma. Subjects were also excluded if undergoing active cancer treatment, prednisone above 10 mg day, other 7 4 3 immunosuppressive drugs, any medications for rheumatoid arthritis other than NSAIDs or if they had received 7 4 4 antibiotics in the previous 6 months. Beyond these steps to exclude specific chronic conditions we also undertook 7 4 5 further additional efforts to exclude older adults with any significant frailty. Since declines in self-reported physical 7 4 6 performance are highly predictive of frailty, subsequent disability and mortality 36 , all subjects were also questioned 7 4 7 as to their ability to walk ¼ mile (or 2-3 city blocks). For those who self-reported an inability to walk ¼ mile 36 , the 7 4 8 "Timed Up and Go" (TUG) test was performed and measured as the time taken to stand up from the sitting position, 7 4 9 walk 10 feet and return to sitting in the chair 37 . Scoring TUG > 10 sec was considered an indication of increased 7 5 0 frailty and resulted in exclusion from the study 38 . Information on medications in Table S1 illustrates that as 7 5 1 expected medication usage did increase with age. Nevertheless, these medications all reflected their use for common 7 5 2 and controlled chronic conditions unlikely to influence our findings such as hypertension, hyperlipidemia, 7 5 3 hypothyroidism, degenerative joint disease, seasonal allergies, headaches, atrial fibrillation, depression, anxiety, or 7 5 4 ADHD (attention deficit hyperactivity disorder). Finally, smoking history data are not typically collected in these 7 5 5 studies -including ours-since smoking is a rare habit among older adults. 7 5 6 7 5 7 Flow cytometry data generation and analyses: PBMCs were isolated from fresh whole blood using Ficoll-Paque quality-filtered using trimmomatic 40 , and trimmed reads were mapped to the GRCh37 (hg19) human reference 7 7 6 sequence using bwa-mem 41 . After alignment, technical replicates were merged and all further analyses were carried 7 7 7 out on these merged data. For peak calling, MACS2 42 was used with no-model, 100bp shift, 200bp extension, and 7 7 8 bampe option. Only peaks called with a peak score (q-value) of 1% or better were kept from each sample, and the 7 7 9 selected peaks were merged into a consensus peak set using Bedtools multiinter tool 43 . Only peaks called on 7 8 0 autosomal chromosomes were used in this study. We further filtered consensus peaks to avoid likely false positives 7 8 1 by only including those peaks overlapping more than 20 short reads in at least one sample, and peaks for which the 7 8 2 maximum read count did not exceed 500 counts per million (cpm) to account for regions that are potential artifacts. invariant between the two groups of interest (i.e., young and old samples), (2) that it captures a substantial number consistently chosen as poor quality for a range of values chosen to assess the benchmark criteria. After re-applying 7 9 9 the peak selection criteria to the remaining 100 samples, we arrived at a peak count of 86,145 peaks. Prior to 8 0 0 statistical analyses, ATAC-seq read counts were normalized to each sample's effective library size (i.e., the sum of FASTQC tool, which computes read quality using summary of per-base quality defined using the probability of an 8 1 5 incorrect base call 45 . According to our quality criteria, reads with more than 30% of their nucleotides with a Phred 8 1 6 score under 30 are removed, whereas samples with more than 20% of such low-quality reads are dropped from 8 1 7 analyses. Benchmarking is also applied on RNA-seq data using the same benchmark parameters as ATAC-seq, 8 1 8 which resulted in 304 benchmark genes, none of the RNA-seq samples were dropped due to poor quality. Reads 8 1 9 from samples that pass the quality criteria were quality-trimmed and filtered using trimmomatic 40 . High-quality 8 2 0 reads were then used to estimate transcript abundance using RSEM 46 . Finally, to minimize the interference of non-8 2 1 messenger RNA in our data, estimate read counts were re-normalized to include only protein-coding genes. Table   8 2 2 S2 summarizes the quality control measures for our PBMC RNA-seq samples. for the effect of aging between healthy young and healthy old samples by sex, as well as the effect of sex by age 8 2 7 group. In addition to sex and age group (old vs. young), our models included the base-2 log of effective library size 8 2 8 to ensure peakwise normalization. We isolated a batch effect correlated to time period whereby samples were 8 2 9 collected and libraries were prepared, and used ComBat to adjust the data for this effect. Finally, we used Surrogate 8 3 0 Variable Analysis (SVA 47 ) to capture unknown sources of variation (e.g., localized batch effects, subject-level 8 3 1 heterogeneity, variation in library preparation techniques) statistically independent from age group assignments.

3 2
SVA decomposes the variation that is not accounted for by known factors like age group or sex, into orthogonal 8 3 3 vectors that can then be used as additional covariates when fitting a model to test for differential accessibility or 8 3 4 expression. Using the built-in permutation-based procedure in the R package sva, we choose to retain one SV to 8 3 5 include as covariate in the GLM model for PBMC ATAC-seq and none for RNA-seq data analyses 48 . GLM models 8 3 6 where implemented using a negative binomial link function, including both genome-wide and peak-specific 8 3 7 dispersion parameters, estimated using edgeR's "common," "trended," and "tagwise" dispersion components,  the chromHMM-generated states with our set of consensus peaks, and solved conflicting cases where multiple 8 5 0 chromatin states overlap the same ATAC-seq peak so that each peak was assigned a single annotation, according to 8 5 1 the following priority rules: if a peak overlaps both an active TSS and enhancer region, which state takes priority 8 5 2 depends on whether the peak is proximal (i.e., within 1,000 bp of the nearest TSS), in which case it is annotated as a 8 5 3 promoter, or distal (distance to nearest TSS greater than 1,000 bp), when it is annotated as an enhancer instead. For annotate peaks as cell-specific for a given subset obtained from one of the three datasets listed above, we determined 8 6 2 for each peak whether it was annotated as an active promoter or an active enhancer in a single cell population or 8 6 3 lineage, and in such cases labeled the peak accordingly as cell-or lineage specific. For example, if a peak is 8 6 4 annotated as an active enhancer in both naïve and memory CD4 + T cells but as another state (e.g., repressed) in 8 6 5 every other subset, then the peak would be considered CD4 + T-specific. For gene-based analyses, HOMER was used 8 6 6 to assign each ATAC-seq peak to the nearest TSS, as measured from the peak center. To improve confidence on 8 6 7 these annotations, gene-based analyses were further restricted to include only peaks located within 100 kb of their 8 6 8 corresponding TSS. ATAC-seq peaks were also annotated using gene sets provided by curated immune function-8 6 9 related co-expression modules 13 . These gene sets comprise 28 modules defined from multiple compiled 8 7 0 transcriptomic data sets, which were originally annotated based on expert knowledge of representative functions of 8 7 1 the gene ensemble in each module. In this study, we have preserved and used these annotations to test for 8 7 2 enrichment of these modules in gene sets of interest, such as the set of genes associated to chromatin peaks gaining 8 7 3 or losing accessibility with aging. We assessed enrichment using the hypergeometric test followed by Benjamini-

7 4
Hochberg FDR adjustment for P-values. Further functional enrichment analyses were carried out using 8 7 5 Wikipathways pathways 51 , immune modules 20 , gene sets from DICE database 10 and single cell RNA-seq data in 8 7 6 PBMCs. Gene sets from PBMC scRNA-seq data is driven from one-vs-all cell cluster comparisons. First, we 8 7 7 identified 17 clusters from PBMC scRNA-seq data (n=26 samples) using the Louvain clustering in the ScanPy 8 7 8 toolkit 52 . These clusters were manually inspected and assigned to different cell types by studying the expression of 8 7 9 known marker genes. For each cluster, differentially expressed genes were identified using one-vs-all approach 8 8 0 based on T-test followed by FDR. Marker genes for a cluster (or cell population) were found by using FDR <=5% 8 8 1 and logFC > log2(1.25) using differential analyses results. Similarly, marker genes from the DICE database are 8 8 2 obtained using their differential analyses results and the same cutoffs for cell-specific genes (FDR <=5% and logFC 8 8 3 > log2(1.25)). Gene sets used in our differential analyses are provided as a supplementary table in Table S14. For 8 8 4 each gene set, we tested for enrichment using the hypergeometric test, against a background defined by the set of 8 8 5 genes that are expressed, as determined by RNA-seq data, or potentially expressed, as given by promoter 8 8 6 accessibility, in PBMCs. We used the Benjamini-Hochberg FDR multiple test correction to assess significance of 8 8 7 hypergeometric P-values. Congruence between chromatin accessibility and transcription data: Gene expression (RNA-seq, see above) 8 9 0 data was generated for a subset of subjects with ATAC-seq profiles (summarized in Table S1). These data were 8 9 1 normalized to protein-coding transcripts, and annotated to ENSEMBL GRCh37 gene symbols. Genes for which at 8 9 2 least three normalized reads per million were obtained in at least two samples were considered as expressed, all 1 0 0 0 for different w values. Finally, we used Gaussian mixture modeling on the distribution of these maxima, as 1 0 0 1 implemented in R-Mclust package, to group loginvp maxima obtained from different window spans into cohesive 1 0 0 2 breakpoint intervals, whose medians and ranges we report herein for each cluster. Since breakpoints are Analyses of public data from 500GG and MI projects: We obtained publicly available ELISA data measuring 1 0 0 7 serum protein levels by the 500 Human Functional Genomics (500FG) consortium 11 . We only retained individuals 1 0 0 8 who are matching our cohort in terms of the age span, which resulted in data from 267 individuals. These 1 0 0 9 individuals are grouped together using the age brackets defined in our study: Young 22-40 years old, middle-aged: 1 0 1 0 41-64 years old, older: 65 years old. We compared data from 1) men and women at all age brackets; 2) young men 1 0 1 1 to old men; and 3) young women to old women using Wilcoxon Rank Sum non-parametric test. Note that flow 1 0 1 2 cytometry data from this same cohort was not publicly available; hence cannot be included into the analyses.

0 3
Similarly, publicly available flow cytometry data from Milieu Intérieur Consortium 12 cohort was obtained. We used 1 0 1 4 data that is already processed by this study and just used individuals whose ages are matching our cohort, which 1 0 1 5 resulted in data from 892 individuals. We built linear models using R (lm function) to quantify the association 1 0 1 6 between each flow cytometry measurement to age group (young, middle-aged, older), sex (F, M) and their 1 0 1 7 interaction (age*sex). Significant associations with age (at FDR 5%) and with sex (at FDR 10%) are plotted in data from this study. Plots were generated using ggplot2 57 , using graph aesthetics used throughout the manuscript 1 0 2 2 figures. Statistics for boxplots were calculated using the wilcox.test function in R and presented without correction 1 0 2 3 for the multiple (n = 5) comparisons. Statistics for scatterplots were calculated by fitting a linear model with the lm 1 0 2 4 function in R using the formula (measurement ~ age:sex + sex). For ATAC-seq data, only the peak closest to the 1 0 2 5 gene TSS was considered.  3  8  R  o  c  k  w  o  o  d  ,  K  .  ,  A  w  a  l  t  ,  E  .  ,  C  a  r  v  e  r  ,  D  .  &  M  a  c  K  n  i  g  h  t  ,  C  .  F  e  a  s  i  b  i  l  i  t  y  a  n  d  m  e  a  s  u  r  e  m  e  n  t  1  1  1  1  p  r  o  p  e  r  t  i  e  s  o  f  t  h  e  f  u  n  c  t  i  o  n  a  l  r  e  a  c  h  a  n  d  t  h  e  t  i  m  e  d  u  p  a  n  d  g  o  t  e  s  t  s  i  n  t  h  e  C  a  n  a  d  i  a  n  s  t  u  d  y  1  1  1  2  o  f  h  e  a  l  t  h  a  n  d  a  g  i  n  g  .   T  h  e  j  o  u  r  n  a  l  s  o  f  g  e  r  o  n  t  o  l  o  g  y  .  S  e  r  i  e  s  A  ,  B  i  o  l  o  g  i  c  a  l  s  c  i  e  n  c  e  s  a  n  d   1  1  1  3   m  e  d  i  c  a  l  s  c  i  e  n  c  e  s   5  5   ,  M  7  0  -7  3  (  2  0  0  0