The cognitive and psychological health of older adults is now a major preoccupation for healthcare services and researchers alike. Alzheimer's disease (AD) is the main form of cognitive decline in older persons in Western countries( Reference Blennow, de Leon and Zetterberg 1 ). Age is the main risk factor associated with AD( Reference Blennow, de Leon and Zetterberg 1 ), but other factors also have an effect such as a predisposing genetic polymorphism, i.e. ε4 allele of apoE4 ( Reference Corder, Saunders and Strittmatter 2 ), vascular risk factors including hypertension, obesity and type 2 diabetes( Reference Blennow, de Leon and Zetterberg 1 ), and lifestyle including physical activity and dietary habits( Reference Gomez-Pinilla 3 , Reference Alles, Samieri and Feart 4 ). Among the nutrients closely associated with brain function, the n-3 fatty acids, especially DHA, have attracted special attention. Fatty fish and seafood are the most important dietary sources of both DHA and EPA. DHA is by far the predominant n-3 fatty acid in the brain and is present mostly in various membrane phospholipids (PL) of neurons, especially in synapses( Reference Crawford, Bloom and Broadhurst 5 ). In contrast to other common dietary long-chain fatty acids, DHA is highly conserved and poorly β-oxidised( Reference Cunnane, Ryan and Nadeau 6 – Reference Leyton, Drury and Crawford 8 ). In human subjects, DHA synthesis is relatively inefficient, especially in comparison to rodents( Reference Plourde and Cunnane 9 ).
Low intake of n-3 fatty acids has long been associated with higher risk of CVD( Reference Skeaff and Miller 10 ), and also of suboptimal brain development( Reference Innis 11 ). Much effort has been focused over the past decade on whether a higher DHA intake could decrease the risk of cognitive decline in older adults, or reduce the progression from mild cognitive impairment towards AD. In general, these studies polarise in two directions: randomised clinical trials that are largely negative and epidemiological studies that are more positive( Reference Cunnane, Plourde and Pifferi 12 ) about the DHA's role in maintaining cognition during ageing. Thus, in general, DHA supplementation trials in AD (with or without EPA) have not so far produced any truly positive results( Reference Cunnane, Plourde and Pifferi 12 – Reference Sydenham, Dangour and Lim 15 ). Methodological issues such as dose of n-3 fatty acid, duration of treatment or selection criteria may well have affected the outcomes of these trials. DHA supplementation may have a greater positive effect on memory and learning in healthy adults( Reference Stonehouse, Conlon and Podd 16 ), elderly with subjective cognitive complaints( Reference Yurko-Mauro, McCarthy and Rom 17 ) or with mild cognitive impairment( Reference Sinn, Milte and Street 18 ) than in those with AD( Reference Cunnane, Plourde and Pifferi 12 , Reference Mazereeuw, Lanctot and Chau 14 , Reference Quinn, Raman and Thomas 19 ). However, prospective epidemiological studies have been more positive; they broadly show that habitually low intake of fish and/or DHA is associated with higher risk of developing AD( Reference Cunnane, Plourde and Pifferi 12 , Reference Fotuhi, Mohassel and Yaffe 20 , Reference Barberger-Gateau, Feart, Letenneur and Yaffe 21 ). These results are supported by the neuroprotective role of DHA reported for non-human models of neurodegenerative disease( Reference Kim, Akbar and Kim 22 – Reference Boudrault, Bazinet and Ma 26 ).
Studies with biological samples (human blood and brain) may be able to provide useful leads to explain the divergent results between randomised clinical trials and epidemiological studies. We have previously reviewed at some length the methodological limits on observational or intervention studies on DHA supplementation in older adults or in AD( Reference Cunnane, Plourde and Pifferi 12 – Reference Sydenham, Dangour and Lim 15 , Reference Fotuhi, Mohassel and Yaffe 20 , Reference Barberger-Gateau, Feart, Letenneur and Yaffe 21 , Reference Barberger-Gateau, Samieri and Feart 27 ); so we will review them here only briefly. We will also present an emerging framework showing that DHA homoeostasis changes in older adults and differs in carriers from non-carriers of apoE4, probably before the onset of cognitive decline.
DHA in plasma and post-mortem human brain
In primate, pig and rodent models, when n-3 intake is severely deficient for extended periods, brain DHA also decreases across all cell types and regions, in association with lower scores on cognitive and behavioural tests( Reference Calon and Cole 25 , Reference Brenna and Diau 28 – Reference Novak, Dyer and Innis 31 ). AD is now widely associated with lower fish and DHA intake, so it would be logical that post-mortem brain samples of patients with a definitive diagnosis of AD also contained lower DHA. Indeed, in the hippocampus, which is central to memory processing and learning, AD patients reportedly do have lower DHA( Reference Soderberg, Edlund and Kristensson 32 , Reference Lukiw, Cui and Marcheselli 33 ). However, in the temporal and frontal cortices which are also affected in AD, DHA is almost always the same as in the controls (Fig. 1). Studies reporting lower DHA in the AD brain show that other fatty acids are also lower, particularly n-6 PUFA( Reference Astarita, Jung and Berchtold 35 , Reference Corrigan, Horrobin and Skinner 41 , Reference Guan, Wang and Cairns 44 , Reference Prasad, Lovell and Yatin 46 ). Thus, the effect of AD is not specific to DHA which is contrary to what would be expected if only n-3 fatty acid intake were deficient.
There are many potential methodological reasons for the observed lack of agreement between the apparently low DHA intakes in AD yet frequently normal DHA levels in the brain( Reference Cunnane, Plourde and Pifferi 12 ). Crucial among these are the ‘healthy’ controls against which the AD cases are compared as well as the very marked extent of regional brain atrophy associated with ageing regardless of the presence of neurological disease( Reference Guan, Wang and Cairns 44 , Reference Svennerholm, Boström and Jungbjer 50 ). Furthermore, the basis for classifying a patient as having AD, i.e. whether on clinical cognitive criteria or on neuropathological score, may not give consistent results since senile plaques are increasingly recognised as being present in a significant proportion of cognitively normal elderly persons( Reference Aizenstein, Nebes and Saxton 51 – Reference Duyckaerts and Hauw 53 ). Hence, there is a risk that post-mortem samples from brain banks for which cognitive status is not known at the time of death could be misclassified if based solely on neuropathological scores. It also appears that membrane PL in the cortex can tenaciously retain DHA and that a more discrete and specific subcellular pool or membrane pool of DHA may have to be measured( Reference Fraser, Tayler and Love 43 ). Brain membrane DHA cycles rapidly between PL and free DHA via DHA-CoA( Reference Umhau, Zhou and Carson 54 , Reference Chen, Green and Orr 55 ), and the deteriorating efficacy of this process could theoretically contribute to the neurodegenerative processes. Thus, the key issue in relation to post-mortem tissue analysis is that the time to lipid extraction is rarely less than 4–5 h yet DHA turnover is on the order of minutes, if not seconds. Since the turnover of DHA towards resolvins and neuroprotectins are orders of magnitude lower than the amount of DHA in the brain NEFA pool, truly ‘physiological’ amounts of these products are extremely difficult to measure, especially in human subjects( Reference Bazan, Molina and Gordon 23 , Reference Serhan and Petasis 56 ). As also noted elsewhere, these and other issues severely constrain the validity and hence the utility of DHA measurements on human post-mortem brain samples( Reference Cunnane, Chouinard-Watkins and Castellano 57 ).
Lower DHA would normally be expected in the blood of those with habitually low DHA intake (whether diagnosed with AD or not). In some AD studies, lower DHA is indeed reported for plasma total lipids( Reference Cherubini, Andres-Lacueva and Martin 38 ), PL( Reference Conquer, Tierney and Zecevic 39 , Reference Corrigan, Van Rhijn and Ijomah 40 , Reference Cunnane, Schneider and Tangney 42 ), cholesteryl esters( Reference Tully, Roche and Doyle 49 ) and NEFA( Reference Wang, Sun and Liu 58 ). However, many other AD studies show no difference in plasma DHA, whether in PL or total fatty acids( Reference Arsenault, Matthan and Scott 34 , Reference Cherubini, Andres-Lacueva and Martin 38 , Reference Cunnane, Schneider and Tangney 42 , Reference Laurin, Verreault and Lindsay 45 , Reference Ronnemaa, Zethelius and Vessby 59 ). Some even report higher DHA in plasma PL( Reference Laurin, Verreault and Lindsay 45 ) or cholesteryl esters (CE)( Reference Corrigan, Van Rhijn and Ijomah 40 , Reference Cunnane, Schneider and Tangney 42 ). Similar inconsistencies are present across DHA levels reported for the erythrocytes in AD (Fig. 1)( Reference Boston, Bennett and Horrobin 36 , Reference Corrigan, Van Rhijn and Ijomah 40 , Reference Selley 47 ). Prospective studies also show this inconsistency: some found a strong association between lower blood DHA level and slower cognitive decline( Reference Heude, Ducimetiere and Berr 60 ) or lower risk of dementia( Reference Schaefer, Bongard and Beiser 61 ), whereas other did not( Reference Laurin, Verreault and Lindsay 45 , Reference Ronnemaa, Zethelius and Vessby 59 , Reference Kroger, Verreault and Carmichael 62 ). It may be that the cognitive domain studied( Reference Dullemeijer, Durga and Brouwer 63 , Reference Beydoun, Kaufman and Satia 64 ) and apoE4 genotype( Reference Whalley, Deary and Starr 65 , Reference Samieri, Feart and Proust-Lima 66 ) contribute to this scatter in the data.
DHA homeostasis during ageing and apoE
We propose that even when collected under hypothetically ideal conditions (zero delay; perfectly matched, cognitively healthy controls, etc.), data obtained from single blood samples are too limited to fully understand possible changes in DHA metabolism due to genotype, ageing or neurodegenerative disease. However, isotopically labelled DHA is emerging as a useful tool to assess how the metabolism of DHA changes with age. Indeed, in a relatively simple study design, it was clear that the clearance of a 50 mg oral dose of uniformly carbon-13 labelled DHA (13C-DHA) from the blood over 1 month was much slower in healthy 76-year-old compared with 27-year-old adults( Reference Plourde, Chouinard-Watkins and Vandal 67 ). These results were similar to our earlier report that the increase in plasma DHA during a short-term treatment with fish oil was higher in healthy older persons( Reference Vandal, Freemantle and Tremblay-Mercier 68 ). 13C-DHA enrichment in plasma NEFA and TAG of older adults was most affected (four- to fivefold higher than in the young adults) but its enrichment in PL and CE 13C-DHA was also affected. The doubling of 13C-DHA enrichment in plasma PL and CE emerged only after about 7 d post-dose, suggesting slower DHA clearance through plasma lipid classes, i.e. an altered plasma ‘DHA wave’ in older adults (Fig. 2)( Reference Plourde, Chouinard-Watkins and Vandal 67 ).
Clearly, therefore, healthy ageing seems to change DHA metabolism and, hence, homeostasis in human subjects. Notwithstanding the limited extent to which the kinetic behaviour of a tracer can be compared with a single plasma fatty acid measurement, the difference in 13C-DHA homeostasis in the elderly seems to reflect the results observed in two studies in which lower DHA was reported in plasma PL yet higher DHA was reported in plasma CE of AD patients( Reference Corrigan, Van Rhijn and Ijomah 40 , Reference Cunnane, Schneider and Tangney 42 ). The minimally invasive nature of this type of experiment makes it difficult to invoke a particular mechanism but one could speculate that the changing sensitivity of endothelial lipoprotein lipase could be involved in this age-associated difference in DHA homeostasis( Reference Millar, Lichtenstein and Cuchel 69 ).
ApoE4 carriers are at significantly higher risk of AD( Reference Jack, Knopman and Jagust 70 , Reference Jack, Petersen and Xu 71 ). It is now emerging that the apoE4 status also affects DHA metabolism in human subjects( Reference Barberger-Gateau, Samieri and Feart 27 , Reference Chouinard-Watkins, Rioux-Perreault and Fortier 72 , Reference Plourde, Vohl and Vandal 73 ). This interaction may help explain why the protective association of higher dietary intake of fish( Reference Huang, Zandi and Tucker 74 , Reference Barberger-Gateau, Raffaitin and Letenneur 75 ) or higher erythrocyte total n-3 fatty acids( Reference Whalley, Deary and Starr 65 ) is generally limited to non-carriers of apoE4. Measuring expired 13C-CO2 after dosing with 13C-DHA permits the estimation of the whole body half-life of DHA in healthy older adults, which is of the order of 32 d in carriers of apoE4 and 140 d in non-carriers of apoE4 ( Reference Chouinard-Watkins, Rioux-Perreault and Fortier 72 ). Ageing seems not to affect the whole body half-life of DHA although the small sample size makes these results still somewhat preliminary( Reference Plourde, Chouinard-Watkins and Vandal 67 ).
Using positron emission tomography and the tracer, 11C-DHA, human brain turnover of DHA has been estimated to be about 4 mg/d, giving rise to a half-life of brain DHA of about 2·5 years( Reference Umhau, Zhou and Carson 54 ). Hence, the half-life of brain DHA is much longer than its whole body half-life. Perhaps further assessments of the brain or whole body half-life of DHA could provide some insight into the current ineffectiveness of DHA supplements in AD despite the fact that these supplements typically supply several fold the brain's apparent daily turnover of DHA( Reference Cunnane, Plourde and Pifferi 12 ).
ApoE4 seems also to supress the plasma DHA response to a fish oil supplement( Reference Plourde, Vohl and Vandal 73 ) and the metabolism of an oral dose of 13C-DHA( Reference Chouinard-Watkins, Rioux-Perreault and Fortier 72 ), an effect somewhat opposite to that observed with healthy ageing. For up to 28 d after a single oral dose of 13C-DHA, carriers of apoE4 have a slightly lower concentration of plasma 13C-DHA compared with non-carriers( Reference Chouinard-Watkins, Rioux-Perreault and Fortier 72 ). When the tracer is given both before and again after a 5-month period of DHA+EPA supplementation, the apoE4 carriers had a greater accumulation of the tracer in plasma after supplementation compared with the non-carriers, again suggesting slower clearance of DHA to and/or use by tissues (Fig. 3). Hence, two established risk factors for AD (ageing and apoE4) both significantly change DHA homeostasis but possibly in different ways.
Linking dietary and plasma DHA
Habitual DHA intake is commonly estimated to be <250 mg/d( Reference Harris, Mozaffarian and Lefevre 76 – Reference Meyer 78 ) but this is a difficult and laborious measurement and subject to high day-to-day variability depending on the frequency of fish or shellfish consumption. Hence, it would be useful if DHA intake and plasma DHA were highly correlated because plasma DHA measurement is now technically simple and reliable, so it could potentially be a surrogate for dietary DHA measurement. There is indeed a good positive correlation between dietary and plasma intake for DHA consumption, especially at DHA intakes towards 1000 mg/d, during which plasma DHA rises to a maximum of about 4 % in plasma total lipids. However, DHA always seems to be present in plasma, even when DHA intake is negligible; thus, vegans consuming no known dietary sources of DHA still have about 0·5 % DHA in plasma total lipids( Reference Sanders, Hinds and Pereira 79 ). The problem is that there are relatively few reports on which to build the relationship of dietary to plasma DHA. At DHA intakes between 0 and 50 mg/d, DHA is between 0·5 and 1·2 % of plasma total lipids, but the spread in these data is large( Reference Cunnane, Chouinard-Watkins and Castellano 57 ). A value of 0·5 % DHA in plasma total lipids therefore seems to be at or close to the lower limit possible of plasma DHA in healthy adults.
The AD cases we have studied had 1·0 % DHA in plasma total lipids, which empirically corroborates very low DHA intake, yet they had ‘normal’ DHA in the PL of brain cortical grey matter( Reference Cunnane, Schneider and Tangney 42 ). Plasma DHA does not rise in human subjects given EPA or α-linolenic acid supplements, even in vegans( Reference James, Ursin and Cleland 80 – Reference Horrobin, Fokkema and Muskiet 82 ). The conundrum therefore is: how are plasma (or brain) DHA levels maintained when DHA intakes are very low to negligible? We speculate that with changes in DHA metabolism and homeostasis during age-related cognitive decline, the diet–plasma relation of DHA may shift, explaining why with lower DHA intake, a population with age-related cognitive decline appears to have the same plasma DHA concentration as healthy elderly even though the availability of DHA to the tissues may be reduced( Reference Lopez, Kritz-Silverstein and Barrett Connor 83 ). The lack of an established reference lipid class in blood (PL, CE, TAG or NEFA, erythrocytes, etc.) for DHA measurements relative to intake still hampers the extent to which plasma DHA data from various reports can be compared in relation to ageing, genotype and risk of cognitive decline. This area clearly needs further research but suffice it to say that it is becoming increasingly important to take into account changing DHA homeostasis in the study of ageing population, especially in a context of age-related cognitive decline or AD.
We have sought to briefly highlight some of the methodological challenges and potential future directions for the study of DHA in ageing and AD. The emerging evidence for changing DHA half-life in older adults and in carriers of apoE4 should encourage more basic research on DHA metabolism in human subjects. Molecular, cellular and animal models have contributed enormously to understanding the complexity of DHA biology, but none of them seem to represent the changes in DHA homeostasis reported in elderly human subjects. The human brain is able to strongly retain DHA in membrane PL despite very low DHA intake and advanced AD( Reference Cunnane, Schneider and Tangney 42 ), so the classical dietary n-3 deficiency model used to probe the function of DHA in the animal brain appears inappropriate for research into AD. In human subjects, DHA in the post-mortem brain is unlikely to correctly reflect what is happening during ageing and AD, due to its fast turnover in neuronal membrane PL. Ageing- and apoE4-associated changes in DHA metabolism strongly suggest altered DHA homeostasis involving a decrease in plasma DHA clearance during age-related cognitive decline and AD. Therefore, a shift in the relationship between plasma and dietary DHA may be occurring during age-related cognitive decline and AD, one that needs to be considered when looking at plasma DHA as a measure of dietary DHA intake. In the future, the availability of innovative tools for studies of DHA half-life and metabolism in human subjects will be needed to understand in a better manner the changes in DHA metabolism occurring during human ageing and AD and the potential protective role of DHA on cognitive decline. As shown in prospective studies, a protective role of DHA in cognitive health of older persons may depend on consuming a healthy diet throughout adult life.
Mélanie Fortier, Conrad Filteau, Christine Rioux-Perreault, Jennifer Tremblay-Mercier and Sébastien Tremblay provided excellent technical support.
CIHR, NSERC, FQRNT (CFQCU), INSERM, CFI and CRC provided financial support for S. C. C.'s research.
Conflicts of Interest
M. H. and S. C. C. conceived and wrote the first draft with all the authors contributing to the revisions and final version of the manuscript.