Transcript

A Bayesian Approach to Assess the Tanscriptome of Bone Aging and the Role of the Brd2

Gene in the Regulation of Sex Linked Bone Loss

Amira I. Hussein, Ph.D.1, Joseph Wu, Ph.D.2, Mayetri Gupta, Ph.D.3, Louis Gerstenfeld, Ph.D.1. 1Boston University Medical Center, Boston, MA, USA, 2Boston University, Boston, MA, USA, 3University

of Glasgow, Glasgow, United Kingdom.

Disclosures: A.I. Hussein: None. J. Wu: None. M. Gupta: None. L. Gerstenfeld: None.

Introduction: Aging and sex are among the major risk factors of osteoporosis. Aging involves numerous

biological processes. Longitudinal microarray studies have contributed immensely to the understanding

of aging and the associated pathogenesis. However, one of the main challenges in such studies lies in

the lack of multi-factorial (for example sex and genetic mutation) computation approaches to assess

temporal microarray profiles. An aging study was carried out in male and female mice expressing

reduced expression of the transcriptional regulator bromoDomain2 (Brd2lo). Mice with reduced

expression of BrD2 become obese but do not develop insulin resistance [1] hence eliminating the

confounding effects caused by Type II diabetes. Analyses of the bone phenotype showed that female

mice with Brd2 mutation had an age-dependent bone loss, while no differences were found in male

mice [2]. The goals of this study were to: (1) cluster and then characterize the gene expression profiles

of aging male and female mice with and without the Brd2 mutation using a new Bayesian statistical

algorithm; and (2) assess the biological functions of the genes in each cluster.

Methods: Specimens. Male and female wild type (WT) and Brd2lo (Het) mice on a C57BL/6 background

were used in a longitudinal aging study (3, 6, 9, and 12 months; 4 mice/time point; approved by IACUC).

Microarray Analysis. Total RNAs were isolated separately from the both right and left whole humeri of

mice tissues using procedures developed for murine bones [3]. One microgram of RNA was labeled and

used for hybridization. The GeneChip Mouse Gene 1.0ST Arrays were used for our studies (Affymetrix,

Santa Clara, CA). Out of a total of 21,225 probes, only genes expressing 3-fold change between the

maximum and the minimum values across all samples were used in this study (3,950 genes). Gene

Cluster Discovery Algorithm. A Bayesian modeling approach that captures information on three different

levels (sex, genotype and time) simultaneously was developed in R. The algorithm relies on detecting

groups of genes that behave in a similar way over time, under a particular combination of factors, and

differ from other set(s) of genes in their pattern of behavior. Each group of genes is defined as a cluster.

Biological Functions. The biologic functions of the gene in each cluster were assessed using Ingenuity

Pathway Analysis tool (IPA, QIAGEN Redwood City). All genes expressing 3-fold change as well as genes

from each cluster identified by the gene cluster discovery algorithm were imported into IPA. Gene

functions were grouped into general categories such as skeletal and muscular, immune, and

inflammatory related functions. Only biological functions with p < 0.05 were considered.

Results: Eighteen clusters were identified using the gene cluster discovery algorithm (4 subsets are

shown in Figure 1). Notable differences among the clusters were observed in the patterns of the

differential gene expression. For example, the gene profiles in cluster 4 indicate that the set of genes in

this cluster have similar expression levels within the male mice. While the expression levels for the

female heterozygous mice are different from the wild-type female mice but resemble that of the male

mice. Figure 2 represents a bar plot of all biological functions for each cluster (with 15 genes or more) as

well as the set of all genes used in the gene cluster discovery algorithm (3-fold: 3,950 gene set). Clusters

4, 14, and 18 had the highest percentage of skeletal and muscular related genes. The majority of the

skeletal and muscular related genes in Cluster 4 had skeletal (bone)-related functions, while the genes in

Cluster 18 had muscle-related functions only. Cluster 14 had a small number of genes (15 genes).

Therefore, Cluster 4 was chosen for further assessment of skeletal-related biological functions. Many of

the skeletal related biological processes indicated that over time, the Brd2 female mice resemble the

gene expression levels in male mice. For example, for metabolic bone disease, all genes in this cluster

were down-regulated in male (wild type and Brd2lo) and Brd2lo female mice relative to wild type female

mice (figure 3).

Discussion: Consistent with the known and/or predicted phenotypes ~30% and ~10% respectively of the

genes overall that were differentially expressed showed a relationship to the

hematological/immunological and skeletal and muscular tissue functions (figure 3; 3-fold bar). A

qualitative assessment of the profiles from the gene cluster discovery algorithm facilitated the

identification of gene groups that show commonality by sex or genetic mutation as well as the

interaction between the two factors. The results indicate that over time, the Brd2 female mice resemble

the gene expression levels in male mice. These results are in agreement with quantitative data on bone

loss in the trabecular bone compartment of the tibia of these same groups. Brd2 has no effect on bone

loss in male mice over time, whereas in female mice, the Brd2 mutation resulted in more bone loss over

time compared to wild type in a manner similar to male mice. Taken together, the outcomes indicate

that even though females have decreased bone volume fraction compared to males, their bone

metabolism is increased.

Significance: Clustering of genes based on the differential gene expression among male, female, with

and without Brd2 mutation was achieved through the use of a novel multi-factorial Bayesian statistical

approach. The expression of the skeletal related gene show that female mice with the Brd2 mutation

have similar gene expression profiles to male mice, while no differences were found between wild type

and Brd2 male mice. These data identify a gene (BrD2) that belongs to a gene family that imparts

epigenetic regulation and may be part of the mechanisms that control sex regulated differences in male

and female bone loss with aging.

Figure 1. Factorial time-course differential gene expression patterns relative to female wild type for four

of the eighteen clusters. For each cluster, bottom left plot shows female wild type as reference group;

bottom right plot shows female Brd2lo; top left plot shows male wild type; and top right plot shows

male Brd2lo. For 95% credible intervals, o is lower bound; ^ is mean; and + is upper bound.

Figure 2. Distribution of biological functions for all genes included in the study with significant functional

ontologies (p<0.05); only clusters with at least 15 genes are shown; the number of genes per cluster is

included in parentheses. Biological functions are based on aggregation of multiple functional ontologies

with p<0.05 that were considered with overlapping functions. The color scheme at the blue end of the

scale represents immune-related functional categories, whereas the red shades correspond to skeletal

and developmental categories. The green and other middle categories represent other functions not

known to be directly related to bone aging. The clusters are shown ordered from low to high percentage

of skeletal related functions.

Figure 3. Network Relationship of Expressed Gene Group for the Highest Ranked Disease Associated

Function in Clusters 4. The similarity of the networks in the three groups indicate that similar genetic

mechanisms are turned on in the Brd2lo female mice, as in the males.

ORS 2015 Annual Meeting

Poster No: 1408