A genetic association analysis of cognitive ability and cognitive ageing using 325 markers for 109 genes associated with oxidative stress or cognition

Background Non-pathological cognitive ageing is a distressing condition affecting an increasing number of people in our 'ageing society'. Oxidative stress is hypothesised to have a major role in cellular ageing, including brain ageing. Results Associations between cognitive ageing and 325 single nucleotide polymorphisms (SNPs), located in 109 genes implicated in oxidative stress and/or cognition, were examined in a unique cohort of relatively healthy older people, on whom we have cognitive ability scores at ages 11 and 79 years (LBC1921). SNPs showing a significant positive association were then genotyped in a second cohort for whom we have cognitive ability scores at the ages of 11 and 64 years (ABC1936). An intronic SNP in the APP gene (rs2830102) was significantly associated with cognitive ageing in both LBC1921 and a combined LBC1921/ABC1936 analysis (p < 0.01), but not in ABC1936 alone. Conclusion This study suggests a possible role for APP in normal cognitive ageing, in addition to its role in Alzheimer's disease.


Background
Individuals differ in their cognitive skills, and in how much these cognitive skills change as people grow older. That is, there are individual differences in the trait (or level) of intelligence, and in the age-related change (or trajectory). We have previously shown that about 50% of the variance in trait intelligence is stable from the age of 11 to the age of 79 [1]. In both the trait and the age-related change, the majority of the between-individual variation is accounted for by a common factor of general cognitive ability (or g) [2,3]. Both mild intellectual impairment (low trait intelligence) and accelerated age-related cognitive decline (increased downward trajectory in intelligence) have a major impact on society, because of the large number of individuals involved who have limited independence. In our increasingly 'ageing society', disabilities linked to cognitive ageing are a growing medical and social problem.
There are environmental and genetic contributions to individual differences in trait intelligence and cognitive ageing [4,5]. Genetic influences account for more than 50% of the variability in adult cognitive abilities [6]. We have shown that genetic variation in some specific genes, e.g. APOE is associated with change in cognitive ability with age, but not with the stable trait of intelligence [7]. Therefore, it is likely that some genetic variants are associated with life-long cognitive abilities and others specifically with variance in age-related cognitive decline. The search for genetic contributions to cognitive ageing can be guided by focussing on mechanisms that affect brain ageing [5].
Oxidative stress is hypothesised to be a significant contributor to cellular ageing. The free radical theory of ageing predicts that, with increasing age, free radicals, reactive by-products of oxidative metabolism, damage macromolecules such as DNA, protein and lipids [8,9]. Support for the free radical hypothesis of ageing comes from a wide variety of sources, including analyses of mutations and transgenic animals (for recent reviews see [10,11]). The brain is particularly vulnerable to oxidative damage as a result of its high aerobic metabolism and high concentrations of polyunsaturated fatty acids that are susceptible to lipid peroxidation [12][13][14][15].
Oxidative damage to mitochondrial DNA accumulates at a ten-fold higher rate than nuclear DNA, although its precise significance to ageing remains controversial [16,17]. The constant leak of reactive oxygen species from mitochondria increases with age, and deficiency of both mitochondrial and cytoplasmic superoxide dismutase are associated with neurodegeneration due to oxidative damage [18][19][20]. A role for oxidative stress has been proposed both in Alzheimer's disease (AD), associated with amyloid plaques [21,22], and in Parkinson's disease, with the presence of iron and auto-oxidised monoamines [23]. A role for oxidative stress has also been proposed in mild cognitive impairment [24,25]. Non-pathological cognitive ageing was found to be related to differences in oxidative stress (measured, for example, by thiobarbituric acid reactive substances) in a large community study of older people [26]. It is also implicated in the "common cause hypothesis of ageing": the recent finding that physical and cognitive capabilities are highly correlated in old age [27].
Expression profiling of large gene arrays in adult and aged mouse brain also supports a role for oxidative damage in cognitive ageing [28,29]. Lee et al [28] examined the expression profiles in neocortex and cerebellum of 6,347 genes in adult (5 months) and aged (30 months) mice. In both brain regions, gene expression profiles showed increased inflammatory response and oxidative stress gene expression in the older mice. These authors concluded that oxidative stress is an important and perhaps underlying cause of the ageing process in post-mitotic (neural) tissues. In a similar study, Jiang et al [29] probed over 11,000 genes in cortex and hypothalamus in 2 month and 22 month old mice and found altered expression for 98 genes (0.9%) in cortex, about 20% of which were also altered in hypothalamus. Significant changes (at least two-fold) were found in a variety of proteins, including eight concerned with oxidative stress response.
We previously identified associations between common functional polymorphisms in genes involved in AD or oxidative stress and cognitive ageing [7,30,31]. However, these studies all involved genotyping small numbers of polymorphisms in a small sample of genes. Technology is now available to genotype easily much larger numbers of polymorphisms. The aim of the present study was to investigate the influence of genetic variation in genes primarily related to oxidative stress and antioxidant defences in two cohorts of relatively healthy older individuals. These are the Lothian Birth Cohort of 1921 (LBC1921) and the Aberdeen Birth Cohort of 1936 (ABC1936), on whom cognitive ability test scores are available at age 11 and in later life; that is, they have data on the lifetime trait of intelligence, and lifetime cognitive change [32]. These cohorts form a unique resource to test for genes associated with cognitive ageing. Both cohorts took an identical mental ability test at age 11 and a different but overlapping series of cognitive ability tests at either age 79 (LBC1921) or age 64 (ABC1936) [32]. To utilise this resource a candidate gene genetic association study was performed by genotyping 387 SNPs in 444 members of LBC1921. We have ~80% power to detect an effect size of 3% at a type-1 error rate of 0.01. Replication of possible associations is important. Therefore, SNPs that showed a positive association with either cognitive ability at age 11 or cognitive ageing were then genotyped in 485 members of ABC1936.

Results
384 SNPs were selected for genotyping by the Golden-Gate™ assay. A multiplex assay was successfully designed for 322 SNPs (83.9%). 437 (261 women, 176 men) of the 444 LBC1921 subjects (98.4%) were successfully genotyped for at least 316 SNPs. Genotyping data were obtained, from both samples, for 15 of the 16 subjects who were genotyped in duplicate and no discrepancies were identified. Three further SNPs were genotyped by TaqMan ® technology in 424-434 of the subjects. In summary 325 SNPs were genotyped in 420-437 subjects. 86 SNPs (26.5%) were monomorphic in LBC1921.

Childhood cognitive ability in LBC1921
There was a nominally significant association between three SNPs and age 11 Moray House Test (MHT) score: CTSZ, rs9760 (F = 5.625, p = 0.004, η 2 = 0.025); GSTZ1, rs3177429 (F = 4.820, p = 0.009, η 2 = 0.022); NDUFS4 rs31304 (F = 9.757, p = 0.002, η 2 = 0.022). The genotype frequencies for each of these SNPs did not differ significantly from Hardy-Weinberg equilibrium. Table 1 indicates the effect of each polymorphic SNP (pvalue) on each of the age 79 cognitive outcomes controlling for age 11 MHT score (i.e. the effect on cognitive ageing). Sex was included as a between subjects variable, except in the case of PRDX4 SNP rs552105 which is on the X chromosome. For this SNP men and women were analysed separately. Nine SNPs located in eight genes (APP, GLRX, HSPA9B, MSRB2, NDUFS1, NDUFV2, NDUFV3 and NOS1) showed a nominally significant association (p < 0.01) with one of the cognitive variables (table 2). The two SNPs in NDUFV3 were in almost complete linkage disequilibrium. Therefore, only rs8128440 was taken forward to the next stage. The minor allele frequency of SNP rs9658446 in NOS1 was only 4.58 × 10 -3 , and therefore this SNP was not carried forward to the next stage. The genotype frequencies for each of these SNPs did not differ significantly from Hardy-Weinberg equilibrium.

Cognitive ability and ageing in ABC1936
Nine of the 10 SNPs that showed a positive association in LBC1921 with either age 11 cognitive ability or cognitive ageing were successfully genotyped in ABC1936 by KBiosciences. The APP SNP rs2830102 was genotyped using TaqMan ® technology. None of the SNPs were significantly associated with either age 11 MHT score or cognitive ageing in ABC1936 (p > 0.01). Table 3 shows the effect of SNPs showing a positive association with at least one cognitive trait at age 79 (LBC1921), controlling for sex and age 11 cognitive ability, on cognitive traits at age 64 (ABC1936), controlling for sex and age 11 cognitive ability. The genotype frequencies for each of these SNPs did not differ significantly from Hardy-Weinberg equilibrium.
A combined LBC1921/ABC1936 analysis to detect associations with cognitive ageing Because larger sample sizes have greater power to detect associations, general linear modelling was performed using combined data from LBC1921 and ABC1936 to investigate the effect of the seven SNPs that showed a significant association with cognitive ageing in LBC1921, on a relatively large sample size (n = 858-886). An effect size of just 2% can be detected with > 80% power at a type-1 error rate of 0.01 using 858 subjects. The effect size of any single polymorphism influencing variation in a complex trait like cognitive ageing may well be relatively small, as many polymorphisms are likely to be involved [5]. A combined LBC1921/ABC1936 univariate analysis was performed for the each of these seven SNP genotypes, with later life Raven's Progressive Matrices score (the only later life cognitive test that was measured in both cohorts) as the dependent variable. All the cognitive tests used to assess LBC1921 are significantly positively correlated [33] and, therefore, associations that were previously identified with tests other than Raven score may be detected with this test when using a larger sample size. Other effects included in the model were age 11 MHT score, sex and cohort (table 4). All interactions were non-significant and removed from the models. As previously shown [31] cohort and sex were significant for all SNP models (p < 0.001), with ABC1936 and males scoring higher than LBC1921 and females. Age 11 MHT score contributed significantly to later life Raven score (p < 0.001). This latter finding reflects the highly significant partial correlation between age 11 MHT score and later life Raven score, controlling for cohort (r = 0.52, df = 892, p < 0.001). APP intronic SNP, rs2830102, was significantly associated with later life Raven score, controlling for age 11 MHT score, sex and cohort (F = 5.988, p = 0.003, η 2 = 0.014). Figure 1 shows the Raven score raw data (A), and the estimated marginal means (B), for later life Raven scores by sex and cohort, controlling for age 11 MHT score. G/G (genotype B/B in tables 2, 3 and 4) homozygotes scored significantly lower than both heterozygotes (p = 0.029) and A/A (genotype A/A in tables 2, 3 and 4) homozygotes (p = 0.002). There was a trend for heterozygotes to score lower than A/A homozygotes (p = 0.057). None of the other SNP genotypes were significantly associated with later life Raven score, controlling for age 11 MHT, sex and cohort (p > 0.01).

Discussion
To our knowledge, this is the first large-scale investigation into the possible genetic contributions to the normal variability in cognitive ageing experienced by individuals. We examined genes previously implicated in oxidative stress, dementia and cognitive function. Of 325 gene variants analysed, nine were positively associated with variation in performance on one of four tests of cognitive ability at age 79 (LBC1921), controlling for sex and childhood cognitive ability. Two of these SNPs were in strong linkage disequilibrium and one SNP had a very low minor allele frequency. None of these associations was replicated in a second cohort of 64 year olds (ABC1936) who took a different but overlapping series of cognitive tests. Therefore, the present study should be considered as an informative, null study concerning a coherent set of genes that might have, but do not, affect normal cognitive ageing, beyond the effect size which it was powered to detect. p values are given. p values < 0.01 are in bold. SNPs previously associated with intelligence or AD are indicated. *PRDX4 is located on the X chromosome and therefore men and women were analyzed separately. However, APP intronic SNP rs2830102 genotype was associated with non-verbal reasoning, as measured by Raven's Progressive Matrices, in a joint analysis of LBC1921 and ABC1936 data. It is emphasised, though, that it did not have significant effects in both cohorts separately, and as such it needs replicating in other cohorts. APP genotype accounted for 1.4% of the variance in the Raven scores, after adjustment for sex, cohort and childhood ability differences. Although this is a relatively small effect size, it is what is expected with a complex trait like cognitive ageing, where many variants are likely to be involved. APOE, which is one of the few genes that has been associated with cognitive ageing in several cohorts, including LBC1921 [7], has an effect size of just 1% to 2% and is considered to be important. 12.4% of the variance was accounted for by the cohort of the participants. ABC1936 participants who, at age 64, were 15 years younger, scored significantly better than the 79-year-olds in the LBC1921 (p < 0.001). No significant interaction between year of cohort and APP genotype was identified.
APP encodes the amyloid β (Aβ) precursor protein. Extracellular Aβ plaques, which form in the meningeal vessels of AD patient brains, are a defining feature of the disease. Mutations in both the coding region [34] and the promoter region [35] of this gene have been associated with AD. Aberrant expression of APP has also been implicated in AD [36,37]. AD is characterised by an impairment of multiple cognitive domains. Amyloidogenic peptide derivatives of mutant APP have been implicated in the generation of free radicals and with mitochondrial oxidative damage (reviewed in [38]). It is possible that common variation in APP DNA sequence is associated with variation in oxidative stress in the general population, leading to variation in normal cognitive ageing. This may reflect the possibility that the neurobiology of both cognitive ageing and AD is, to some extent, a continuum. It is also possible that the association between APP SNP rs2830102 and normal cognitive ageing, as measured using Raven's Progressive Matrices, is related to incipient AD in some members of LBC1921 and ABC1936.  SNP rs2830102 is located in intron 1 of APP and may affect regulation of gene expression. Although the SNP does not lie in a predicted promoter region [39] and is not predicted to alter splicing it does occur within a region of sequence conservation [40]. Alternatively it may be in linkage disequilibrium with another functional SNP, possibly in the promoter of the gene. It is important that the APP gene is investigated further for its role in both nonpathological cognitive ageing and AD.
Several of the SNPs investigated in this study had previously been associated with intelligence or AD (table 1). We failed to find any significant association between these SNPs and either cognitive ability at age 11 or cognitive ageing in LBC1921. Such attempted replications are important, because initial reports of genotype-phenotype associations often do not replicate.
To investigate genetic influences on non-pathological cognitive ageing we chose to perform a relatively large scale genetic association study using candidate genes, for which there was strong a priori evidence for their involvement in brain ageing. We focussed on a specific ageing-related mechanism, that of oxidative stress. We were in the invaluable position of being able to test directly for cognitive ageing across a long period of time, as we had cognitive ability scores at both age 11 and in later life. There has been much discussion in the literature regarding larger scale association study designs. We chose a candidate genes approach that allowed the use of smaller numbers of SNPs compared to a whole genome association study. However, it is likely that important regions of the genome were missed by this approach. We followed recent guidelines from a genomewide association scan workshop [41] that concluded that multistage designs, whereby a sub-set of subjects are initially genotyped and additional subjects are then genotyped for SNPs that show a positive association, enhanced the efficiency of such studies. We chose to genotype a limited number of potentially functional SNPs in a larger number of genes rather than to attempt to fully cover a smaller number of genes using, for example, tagging SNPs and may therefore have missed important SNPs whose functionality was not predicted. We considered this a more efficient use of limited genotyping funds. It allowed us to cover more of our candidate genes and increased the likelihood that we would identify a causative SNP, particularly as concern exists over the portability of tagging SNPs across populations. A few recent preliminary studies indicate that it may be possible to use tagging SNPs designed in one population to investigate associations in a second population, but this should only be done with caution [42][43][44][45]. With regard to the analysis, we decided to initially concentrate on the identification of individual SNPs that have a detectable main effect on variation in cognitive ageing. However, in the future we may include newly developed statistical techniques that allow the identification of interlocus interactions [46].
Like all large scale genetic association studies, this study suffers from the problem of multiple testing; we initially investigated 325 SNPs and four cognitive tests in 437 subjects. Because many of the SNPs are in linkage disequilibrium and, moreover, scores on the cognitive tests are positively correlated, it was deemed inappropriate to perform a Bonferroni-type correction. However, we were able to genotype SNPs showing a nominally significant association in the first cohort, with a second equally large and valuable cohort and, we used a relatively stringent p valueof < 0.01.
It is also important, given the relatively small size and younger age of the replication cohort (n = 485), that SNPs that showed a positive association in LBC1921, but not in ABC1936 or the combined cohorts, are investigated in future association studies to identify genetic determinants of cognitive ageing. A further caveat of the study is that ABC1936 did not take exactly the same cognitive tests as LBC1921. Therefore, associations identified in LBC1921

Conclusion
This study has identified a number of genes, for which there was strong a priori evidence for their involvement in cognitive ageing, which have an association with cognitive ageing in a cohort of relatively healthy 79 year old subjects (LBC1921). A significant association with a SNP in the gene encoding APP was also identified in a combined analysis of LBC1921 and a second younger cohort (ABC1936), suggesting its importance in cognitive ageing as well as AD. It is important that the role of this gene in cognitive ageing is investigated further.

Mini-Mental State Examination (MMSE)
MMSE [52] was used to screen both cohorts for possible dementia. Maximum score is 30. A score of less than 24 was used here as an exclusion criterion because it is often adopted as an indicator of possible dementia.
Both cohorts underwent a series of mental tests designed to examine different cognitive functions: non-verbal reasoning, executive function, and memory and learning. We have previously described this testing in detail [32,53]. The individual cognitive functions of the two independent cohorts (LBC1921 and ABC1936) were examined using a different series of tests as indicated below: Non-verbal reasoning Raven's Progressive Matrices [54] Non-verbal reasoning was examined in all subjects using Raven's Standard Progressive Matrices. The time limit was 20 minutes.

Executive Function
Verbal fluency LBC1921 took the verbal fluency test, which is described as a test of prefrontal executive function [55,56].

Uses of Common Objects
ABC1936 took the use of common objects test, which is described as a test of executive function or purposive action [55].

Verbal Memory and Learning
Logical Memory LBC1921 took the Logical Memory test, which is a verbal declarative memory sub-test from Wechsler Memory Scale-Revised [57].

Rey Auditory Verbal Learning Test
ABC1936 took the auditory verbal learning test which assesses short and longer term memory and learning [55].

Illumina SNP selection
A list of 141 brain-expressed genes was selected and provided to Illumina (table 5). They were selected if they were: a) implicated in antioxidant defence; b) vitagenes (longevity assurance processes); c) associated with cognitive function; d) associated with AD; e) "stress response" genes showing an increased expression in the aged mouse [28]; and/or f) nuclear genes encoding mitochondrial complex 1 proteins. From an initial list of 14,033 potential SNPs, 384 were selected for genotyping using the following criteria: a) all designable (including designability score 0.5) SNPs previously associated with AD and cognitive function; b) all designable (including designability score 0.5) functional SNPs; c) all non-synonymous validated and designable (including designability score 0.5) SNPs; d) all validated and designable (including designability score 0.5) SNPs at exon/intron boundaries that potentially alter splicing; e) all validated and designable (including designability score 0.5) SNPs with percentage identity in mouse >= 80%; f) all validated and designable (excluding designability score 0.5) SNPs with percentage identity in mouse between 60% and 80%; g) remaining SNPs were Illumina validated synonymous SNPs in previously unrepresented genes (see additional file 1). Designability is ranked as 0, 0.5 or 1. A "0" is assigned to SNPs for which an assay cannot be designed, "0.5" indicates the SNP has a designability score low enough to suggest that there might be challenges to the design, and "1" is reserved for those that do not appear to have any challenges in their designability. Validation class is ranked as 1, 2, or 3. "1" means that a SNP is nonvalidated, "2" is a two-hit SNP (non Illumina validated, i.e. it has been validated on some other platform on more than one chromosome), and "3" means two-hit Illumina validated. The percentage identity with mouse is based on a 120 base pair window surrounding the SNP.

Genotyping of LBC1921
Genomic DNA was extracted from blood using standard methods. Genotyping of 384 SNPs was performed using the GoldenGate™ assay by the Illumina BeadLab service facility in San Diego. 444 LBC1921 subjects were genotyped, 16 of them in duplicate. A further three SNPs (MPO, rs7208693; TF, rs3811656 and NDUFAF1 rs3204853) were genotyped at the Welcome Trust Clinical Research Facility Genetics Core, Western General Hospital, Edinburgh [58] using TaqMan ® technology (Applied Biosystems).

Genotyping of ABC1936
Genomic DNA was extracted from blood using standard methods. Genotyping in LBC1921 found seven independent SNPs significantly associated with cognitive ageing (p < 0.01), and three SNPs significantly associated with age 11 MHT score. Genotyping for these SNPs was attempted in ABC1936, using KASPar, by Kbiosciences (Herts, UK). In cases where a KBiosciences assay could not be designed, genotyping was performed at the Welcome Trust Clinical Research Facility Genetics Core, Western General Hospital, Edinburgh [58] using TaqMan ® technology.

Statistical analysis
The power to detect a causative variant at a type-1 error rate of 0.01, for a variant explaining 2-3% of the variance, was estimated by calculating the non-centrality parameter of a non-central χ 2 and the probability that the test statistic under the alternative hypothesis would be larger than the threshold corresponding to the specified type-1 error [59].
The effect of each SNP genotype on LBC1921 age 11 MHT score was analysed using general linear modelling (univariate analysis of variance). The fixed effects (between subjects variables) were: SNP genotype and sex.
The effect of each SNP genotype on each of the four age 79 cognitive outcome variables, for LBC1921, was analysed using general linear modelling (multivariate analysis of variance). The fixed effects were: SNP genotype and sex. Age 11 MHT score was included as a covariate, allowing us to identify associations specifically with cognitive ageing.
General linear modelling, as described above, was used to identify associations between SNPs that showed a positive association in LBC1921 (with either age 11 MHT score or cognitive ageing), and age 11 MHT score and each of the five age 64 cognitive outcome variables (controlling for age 11 MHT score), for ABC1936.
The raw data from LBC1921 and ABC1936 were combined and the effect of each SNP on the Raven's Progressive Matrices Score was analysed using general linear modelling (univariate analysis of variance). In addition to SNP genotype and sex, cohort was added to the model as a fixed effect and age 11 MHT score was included as a covariate.
All general linear modelling was performed using SPSS v12.0. Statistical significance was set at p < 0.01 for all statistical tests.

Authors' contributions
IJD, AFW, LJW and JMS originally designed the study. All authors contributed to the development of the study design and helped draft the manuscript. All authors read and approved the final manuscript. SEH curated the LBC1921 database, performed the statistical analysis and drafted the manuscript. HF curated the ABC1936 database. heat shock 70 kDa protein 6 (HSP70B'). genes associated with cognitive function AR androgen receptor. The steroid hormones and their receptors are involved in the regulation of eukaryotic gene expression and affect cellular proliferation and differentiation in target tissues. CAG repeat polymorphism is associated with cognitive function in older men [71]. CHRM2 cholinergic muscarinic 2 receptor. The muscarinic acetylcholine receptor mediates various cellular responses, including inhibition of adenylate cyclase, breakdown of phosphoinositides and modulation of potassium channels through the action of G proteins. Primary transducing effect is adenylate cyclase inhibition. Polymorphism associated with IQ [60]. CTSD cathepsin D (lysosomal aspartyl protease). Acid protease active in intracellular protein breakdown. Polymorphism associated with AD [61] and general intelligence in a healthy older population [62]. VEGF vascular endothelial growth factor. Growth factor active in angiogenesis, vasculogenesis and endothelial cell growth. VEGF links hippocampal activity with neurogenesis, learning and memory [72].
genes associated with AD AGER advanced glycosylation end product-specific receptor (RAGE). Mediates interactions of advanced glycosylation end products (AGE). Increased expression in AD [73]. APP amyloid beta (A4) precursor protein. Polymorphisms associated with AD (reviewed in [34]). HTR2A 5-hydroxytryptamine (serotonin) receptor 2A. This is one of the several different receptors for 5hydroxytryptamine (serotonin), a biogenic hormone that functions as a neurotransmitter, a hormone, and a mitogen. Polymorphisms associated with episodic memory [63,74] and neuropsychiatric symptoms in AD [64]. IDE insulin degrading enzyme. May play a role in the cellular processing of insulin. May be involved in intercellular peptide signaling. Polymorphism associated with AD [65]. IL1B interleukin 1, beta. Produced by activated macrophages. IL-1 proteins are involved in the inflammatory response, being identified as endogenous pyrogens, and are reported to stimulate the release of prostaglandin and collagenase from synovial cells. Polymorphism associated with AD [66]. PLAU plasminogen activator, urokinase. Polymorphisms associated with AD [68].
stress response genes altered in aged mouse brain [28]. APOD apolipoprotein D. APOD occurs in the macromolecular complex with lecithin-cholesterol acyltransferase. It is probably involved in the transport and binding of bilin. Appears to be able to transport a variety of ligands in a number of different contexts. CRYAB alpha B2 crystallin. May contribute to the transparency and refractive index of the lens. CSNK1D casein-kinase 1 delta. Casein kinases are operationally defined by their preferential utilization of acidic proteins such as caseins as substrates.   Additional material