Systems genetics analysis of iron regulation in the brain

Abstract

Iron imbalances in the brain, including excess accumulation and deficiency, are associated with neurological disease and dysfunction; yet, their origins are poorly understood. Using systemsgenetics analysis, we have learned that large individual differences exist in brain ironconcentrations, even in the absence of neurological disease. Much of the individual differences can be tied to the genetic makeup of the individual. This genetic-based differential regulation can be modeled in genetic reference populations of rodents. The work in our laboratory centers on iron regulation in the brain and our animal model consists of 25 BXD/Ty recombinant inbred mouse strains. By studying naturally occurring variation in iron phenotypes, such as tissue iron concentration, we can tie that variability to one or more genes by way of quantitative trait loci (QTL) analysis. Moreover, we can conduct genetic correlation analyses between our phenotypes and others previously measured in the BXD/Ty strains. We have observed several suggestive QTL related to ventral midbrain iron content, including one on chromosome 17 that contains btbd9, a gene that in humans has been associated with restless legs syndrome and serum ferritin. We have also observed gene expression correlations with ventral midbrain iron, including btbd9 expression and dopamine receptor expression. In addition, we have observed significant correlations between ventral midbrain iron content and dopamine-related phenotypes. The following is a discussion of iron regulation in the brain and the contributions a systems genetics approach can make toward understanding the genetic underpinnings and relation to neurological disease.

Introduction

Iron imbalances in the brain, including excess accumulation and deficiency, are associated with neurological disease and dysfunction; yet, their origins are poorly understood. In particular, the genetic influences on these imbalances have not been identified. Over the past five years, using the systems genetics approach and a genetic reference animal model, we have learned a great deal about individual differences in iron homeostasis that show the importance of genetic background in determining an individual’s iron profile in the brain and in the periphery. The following review aims to illustrate the utility of a systems genetics approach to complex trait analysis using iron homeostasis as an example. The review will first broadly discuss the role of iron in neurological disease, and then discuss evidence for genetic-based individual differences in brain iron homeostasis that could contribute to iron-related neuropathology.

Iron Regulation in the Brain

Iron homeostasis is essential to the functioning of the central nervous system (CNS). In the brain, sufficient iron supply is critical to a wide variety of biochemical pathways, ranging from basic cellular metabolism to catecholamine synthesis and myelination [1,2]. Iron is also involved in the production of reactive oxygen species (ROS) and can be toxic in excess [3]. Thus, both iron overload and deficiency must be avoided, yet imbalances do occur in some individuals for reasons that are yet unknown. Iron overload in relevant brain regions is linked to Parkinson’s disease [4,5] and Alzheimer’s disease [6], while iron deficiency is associated with restless legs syndrome (RLS) [7] and attention deficit hyperactivity disorder [8]. Iron deficiency also leads to specific cognitive impairment and emotional effects, which vary across developmental stages [9,10].

A complex network of iron regulatory genes and proteins manages iron’s distribution and homeostasis. This network has been well studied but is not yet completely understood, particularly as concerns the brain. A detailed presentation of iron regulation is beyond the scope of this review, but, in short, iron homeostasis involves absorption, distribution, storage, and export. Iron sensors that signal as iron levels fluctuate must coordinate all of these processes. Many genes are required at each level of iron homeostasis, including the genes for transferrin, the transferrin receptors and divalent metal transporter 1, the storage genes for heavy and light chain ferritin, genes in the iron export pathways including those for HFE, the bone morphogenic proteins, hepcidin, and ferroportin, and genes involved in iron signaling, i.e. those with iron responsive elements that interact with iron responsive proteins [reviewed in 5]. While this list is not all-inclusive, it demonstrates that iron homeostasis is a polygenic trait with several levels of regulation, each of which presents a possible avenue for malfunctioning of the system. For instance, in one genetic disorder of iron overload, hemochromatosis, the classic mutation associated with the disease is a mutation of HFE that decreases hepcidin expression. However, penetrance is highly variable, suggesting involvement of other factors [11]. It was recently shown in one severe case that in addition to being homozygous for the HFE mutation, one patient had another polymorphism in the promoter region of hepcidin, in a BMP responsive element [12]. This polymorphism led to further decreased expression of hepcidin and impaired the interaction of hepcidin with the BMP pathway. Thus, three components of the iron regulatory system were malfunctioning at once, all reducing hepcidin functioning and contributing to severe iron overload. This is the nature of a complex phenotype, and as such, iron homeostasis must be studied from a systems perspective.

Iron regulation in the brain involves the same iron regulatory genes as the periphery, but is unique in two very important ways. First, while the iron regulatory network must be able to supply iron to the circulating plasma and deliver iron to bodily tissues, it must also be able to transport iron across the blood brain barrier (BBB) to enter the brain. In 2003, using a cell culture model of the BBB, Burdo and colleagues showed that iron transport in endothelial cells depends on their iron status; thus, the BBB may play the role of gatekeeper of iron transport into the brain, reducing transport when iron is high and increasing transport when iron is low [13]. Secondly, iron distribution throughout the brain is highly heterogeneous [14]. Also, the distribution of iron follows a regionally specific developmental timeline, wherein iron accumulates with age [14–17]. These together suggest local influences on iron regulation, not only from the classic iron regulatory pathways but also likely from cross talk with unrelated cellular pathways that require iron for their functioning. Thus, in investigating the molecular mechanisms behind individual differences in brain iron homeostasis, it is important to consider brain-specific sources from both the BBB and local systems.

In sum, systemic iron regulation is well described and involves many genes. Local iron homeostatic mechanisms in the brain, on the other hand, are poorly described and could very well involve genes heretofore unnamed or not implicated in iron metabolism. For these reasons, systems genetics is an appropriate approach to identifying genetic influences on iron regulation in the brain.

Systems Genetics and the BXD/TyJ recombinant inbred strains

Systems genetics is a population-based approach to understanding complex heritable phenotypes such as drug addiction or disease [18]. This approach involves quantifying a phenotype across a population, then associating the variance in that phenotype with genetic markers in a quantitative trait loci (QTL) analysis and/or genomic variation from gene expression analyses. Instead of focusing on the effect of one gene, QTL analysis allows multiple genetic loci to be detected, including epistatic interactions. Populations of recombinant inbred strains are well suited for this purpose [19]. Recombinant inbred (RI) strains are generated when two inbred strains are crossed and the F2 progeny are inbred for twenty consecutive generations, after which over 99% of alleles are fixed. In this way, multiple strains are established from the same two parental strains in order to create a genetic reference population. For example, the BXD/TyJ strains were derived from a pairing of C57/BL6 and DBA/2J strains. Recombinant inbred populations have several advantages for quantitative genetic research [19]. First, the progeny strains are each a unique recombination of the parental alleles. Secondly, due to the fixing of alleles, each strain provides a renewable source of biological replicates to be phenotyped for any quantifiable trait. Finally, the RI strains are genotyped and thus QTL analyses with these mice, as opposed to QTL analysis of F2 mice, do not require genotyping. This is the case with the BXD/TyJ strains that we use to study iron homeostasis. We can use a colony of BXD/TyJ mice to measure a multitude of quantifiable phenotypic features, and then use publicly available data to correlate the genotype with the phenotype in a QTL analysis. This is all currently available on Gene Network (http://www.genenetwork.org), a site created for systems genetics analysis using the BXD/TyJ strains as well as other inbred and recombinant inbred strains [20]. This site features an up-to-date database of published phenotypic data and a constantly updated set of microarray data showing gene expression levels for over 40,000 genes in the 70+ BXD strains in whole brain and several brain regions. Because the BXD panel has been profiled for gene expression, QTL analysis of the gene expression as a phenotype can also expose expression QTL, or QTL that influence the expression of a given gene. The QTL, phenotypic data, and genomic data can all be integrated for QTL and correlation analysis to gain more knowledge about a phenotype in minutes than one could gather in years in an isolated laboratory.

Systems Genetics Approach: an example

The ultimate goal of QTL analysis is to identify the genes underlying the QTL that contribute to the phenotype of interest. One obstacle to QTL research is the cumbersome process of candidate gene identification when QTL confidence intervals span hundreds of genes. In the past, this obstacle has been overcome by using breeding strategies to narrow down wide confidence intervals; however, these strategies are not known for their efficiency in time and effort. An alternative candidate gene identification process is to use gene expression experiments in conjunction with the publicly available BXD gene expression datasets to identify differentially expressed genes that underlie behavioral QTL. Palmer and colleagues performed a study of methamphetamine (MA) sensitivity that demonstrated the utility of pairing gene expression experiments with the available gene expression datasets[21–23]. Having identified QTL associated with MA sensitivity, Palmer and colleagues selectively bred two mouse lines differing in sensitivity to methamphetamine for four generations [21]. Gene expression was measured in the nucleus accumbens of fourth generation mice in the high- and low-sensitivity lines, aimed at identifying differentially expressed genes that would account for the differences in MA sensitivity. A number of genes were differentially expressed between lines, and indeed, one of these genes, Casein Kinase 1 Epsilon (Csnk1e), is known to play a functional role in a drug-related biological pathway involving dopamine in the nucleus accumbens. Using the Gene Network gene expression and phenotype databases, the researchers were then able to investigate the source of variation in Csnk1e expression, which turned out to be a cis-regulating expression QTL that co-maps with a behavioral QTL for MA responsivity on chromosome 15. Thus, a strong, testable hypothesis presented itself that an allele at this QTL influences differential MA sensitivity via differential expression of Csnk1e. Recently a role for Csnk1e in the locomotor response to MA was confirmed by administering a selective inhibitor of Csnk1e, which attenuated this response in mice [22]. In addition, a SNP has been identified in CSNK1E that is associated with increased sensitivity to d-amphetamine in humans [23]. This general approach of linking phenotype-associated QTL to differentially expressed, cis-regulated genes within those QTL is the model we will apply to investigating the genetic influences on brain iron homeostasis. The following is a discussion of our findings thus far.

Systems Genetics of Iron in the BXD/TyJ strains

In 2003, we published the first study to specifically address individual differences in iron regulation [24]. We used fifteen strains of the BXD/TyJ strains to measure iron concentrations in four brain regions: the ventral midbrain, caudate-putamen, nucleus accumbens, and prefrontal cortex. With the advantages of using these strains we were able to have 8–10 biological replicates in almost all of the strains, thus reducing environmental influences. We measured iron concentrations at 120 days of age in mice that had been on a standard laboratory diet containing 240 ppm iron. In QTL analysis there is no control group; rather, the strains are compared interse. In terms of baseline iron concentrations in the various brain regions, we found significant differences in iron concentrations in all of the regions with moderate correlations between regions. Ventral midbrain iron concentrations in this study were correlated with several dopamine-based phenotypes, including cocaine-induced hyperlocomotion (r=−0.60, p<0.0156) [25]. Microarray data on Gene Network also revealed a correlation between ventral midbrain iron and striatal dopamine D1a receptor expression (r=0.59, p<0.0196; Illumina striatum dataset; Nov, 06). QTL analysis revealed non-significant but suggestive QTL on chromosomes 3, 7, 11, 13, and 17 that were associated with ventral midbrain iron concentrations. The QTL on chromosome 17 was also associated with iron in the caudate-putamen and nucleus accumbens and has later been also associated with copper and zinc in various brain regions [26]. In several studies to follow, we confirmed that copper and zinc concentrations are positively correlated with iron concentrations in the brain. We also followed up with a study of iron, copper, and zinc in the hippocampus of over 20 strains and found significant differences and multiple QTL, but no correlation between the hippocampal iron concentrations and those in other brain regions [27]. Finally, we measured systemic iron parameters in 31 of the BXD/TyJ strains, including hemoglobin, hematocrit, plasma Fe, total iron binding capacity, liver iron, and spleen iron [28]. For each of these variables, significant differences among strains were observed in a continuous distribution across the population. Principle components analysis enabled the grouping of QTL into functional categories, such as iron transport, for example. Interestingly, the systemic variables were not correlated with brain iron concentrations. This was compatible with other evidence from clinical studies that brain iron concentrations can be independent of systemic iron status [29,30]. Overall, our work shows that even in a small population of mice generated from the same parental strains, substantial genetic-based variation in iron homeostasis exists, not only in the brain but in the periphery as well. The variation in the brain is linked to several QTL and dopamine-based phenotypes. The challenge now is to identify the genes underlying the QTL identified.

Systems genetics of iron and Restless legs syndrome

Our work shows links between iron and dopamine and may have applications for iron-related neuropathology. With the possible exception of Alzheimer’s disease, iron-related neurological conditions and effects of iron deficiency can all be primarily tied to dopamine malfunction, which is interesting as dopamine and iron have been shown to be related in a number of studies [31–38]. Iron is distributed heterogeneously throughout the brain and is most highly concentrated within the substantia nigra, a region where dopaminergic cell bodies reside [32]. It is now clear that iron deficiency has a negative impact on the nigrostriatal development and maintenance of dopaminergic systems in the basal ganglia [33–38] as well as sensorimotor functioning in the rat [36]. Specifically, the D1 and D2 receptors and DA transporter (DAT) densities are reduced in the caudate-putamen of rats with neonatal iron deficiency, and D2 receptor density is also reduced in the nucleus accumbens [33,34]. Dopamine neurons in the substantia nigra are also the target of iron overload and thus provide the link between iron and Parkinson’s disease [4,5]

RLS is a dopamine-based disorder related to iron deficiency in the midbrain [39]. It is a sensory-motor syndrome involving odd sensations and urges to move the lower limbs, particularly during the transition to bedtime in the evening. RLS can (but does not necessarily) also involve periodic limb movements (PLMs) during sleep. RLS is currently treated with dopamine agonists, which reduce symptoms [40].

Challenges to peripheral iron homeostasis, such as nutritional iron deficiency, pregnancy and hemodialysis-related iron deficiency anemia can produce RLS [41,42]. Alternatively, many RLS patients present with normal systemic iron status; however magnetic resonance imaging and postmortem analysis reveal low iron concentrations in the cerebrospinal fluid and substantia nigra [29,30]. Very little is known as to what exact role iron plays in the etiology of the disorder. One hypothesis is that iron deficiency in the brain impairs dopamine functioning and thus leads to the symptoms of RLS. Oral iron treatments are not always effective in treating RLS, but recent clinical studies have shown that intravenous iron dextran treatments can alleviate symptoms of RLS in some patients for up to six months [43,44].

RLS can be idiopathic or can run in families. Genetic association studies in humans have recently revealed polymorphisms within several genes that are predictive of RLS. These include Ptprd, Map2k5, Meis1, and Btbd9 [45–48]. Btbd9 was found in two independent studies to have a SNP associated with RLS [46,47]. This finding was recently replicated in a third study of three European populations [49]. In the study by Stefansson and colleagues, this association was shown to be particular to PLMs with or without RLS. Interestingly enough, Stefansson and colleagues also found an association between btbd9 and serum ferritin, a measure of bodily iron stores [46]. The function of BTBD9 is not well characterized, and none of the genes associated with RLS are known iron regulatory genes [46]. Understanding how these polymorphisms may play into altered iron homeostasis and the PLMs of RLS is the current challenge.

In an example of how our work can be applied to neurological disease etiology, the mouse homolog, btbd9, is positioned within one of the QTL we associated with iron concentrations in the ventral midbrain [24]. This QTL envelops a region of mouse chromosome 17 that is homologous to human chromosome 6p, which contains BTBD9. In addition, our data in the BXD strains show that striatal expression of the mouse homolog, btbd9, is positively correlated with iron concentrations in the ventral midbrain. It is also correlated with iron in the caudate-putamen and nucleus accumbens, and with copper and zinc in the brain. Shown in figure 1 are the strain values for iron concentrations in the ventral midbrain plotted against Illumina-based striatal expression of probe ILM101980400 in the promoter region of Btbd9 (http://genenetwork.org). The correlation is r=0.73, p<0.001. While the microarray data must be validated, the correlation is quite interesting in consideration of the association of a variant allele of BTBD9 with RLS and serum ferritin and the QTL in the btbd9 region that is also related to ventral midbrain iron. We are currently following up on validating the correlations observed with this gene and will continue to search for links between our work and genetic analyses in humans that may be helpful in generating animal models of disease.

Figure 1.

Ventral midbrain (VMB) iron in female BXD/TyJ mice is correlated with the expression of Riken gene 0710001D07, which lies in the Btbd9 promoter region on murine chromosome 17 (r=0.731; p<0.001). A one-unit difference on the x-axis corresponds to a twofold difference in expression. Graph adapted from http://genenetwork.org.

Future directions: candidate gene selection with gene expression data

The studies leading up to identification of CSNK1E as a QTgene for methamphetamine sensitivity provide the framework with which we plan to follow up on the QTL analyses and correlations we have observed. Instead of selective breeding, we will take advantage of the existing variation in the BXD/TyJ strains in ventral midbrain iron concentrations and use microarray analysis and real-time pcr to test for underlying gene expression differences that map back to our QTL. We also plan to induce iron deficiency in our panel and observe gene by environment interactions that will be subject to QTL and gene expression analyses. The systems genetics approach, including gene expression analysis, shows promise for identifying genes involved in brain iron regulation and any other phenotype that can be quantified in a genetic reference population.

Conclusions

Iron is an important metal for brain function and its homeostasis is critical. Iron imbalances occur in individuals for reasons that are sometimes not apparent. Genetic analysis may help to uncover genetic differences between individuals that lead to differences in iron homeostasis. Systems genetic analysis is an approach that takes into account the complexity of iron regulatory systems, including the large number of genes known to be involved and those that are yet unidentified. We have applied the principles of systems genetics to describe variation in iron homeostasis among RI strains, identify related QTL, and perform correlation analyses. Future work will follow up on these results by integrating QTL analyses with gene expression correlates and testing candidate genes. The identification of genes involved in altered iron homeostasis may shed light on the role of iron in iron-related neurological disease.