Dementia is a general term used to define ageing-associated cognitive disorders and includes diseases that affect mental abilities severely enough that the ability to perform daily tasks is impaired. Among these disorders, Alzheimer’s disease (AD) and vascular dementia are included [1]. Some authors suggest that there is a transition phase referred to as mild cognitive impairment (MCI) in which individuals do not yet meet the criteria for dementia, but are at a greater risk of developing it [2, 3]. The risk of MCI increases with age, and men appear to be at higher risk than women [3]. Moreover, several lines of evidence show that obesity in mid-life is a risk factor not only of MCI at old age, but additionally, late-life dementia and AD [4–7]. Moreover, a growing body of research points towards diet playing a possible role in modulating the risk of cognitive decline and dementia [8, 9]. A macronutrient manifested as an important factor is dietary fat [9, 10], since most studies show that there is increased risk of cognitive dysfunction or late-life dementia with higher intakes of dietary fat, particularly saturated fatty acids [10–12].

The intake of fat-rich food is steadily increasing in Western societies [13]. It is well known that dietary fat intake is positively correlated with the development of obesity and the incidence of its comorbidities (such as insulin resistance, cardiovascular disease (CVD) and/or cancer) [14, 15]. Furthermore, associations between obesity and impaired cognitive function have more recently been recognised [16]. However, the negative effects of an unbalanced diet on the disruption of metabolism are not necessarily correlated with obesity. In fact, there is a group of subjects who present metabolic features related to obesity, but in the absence of increased body weight. These individuals are known as “metabolically-obese, normal-weight” (MONW), and have greater visceral adiposity and increased fat deposition in tissues such as liver, heart, and muscle [17–19], which can lead to hyperinsulinemia, insulin resistance, type 2 diabetes mellitus, and CVD [20, 21]. MONW subjects have recently been reported to represent around 20% of the world population, with greater prevalence among men (25% for men vs 19% for women) [22]; this could be problematic, as it would mean that a significant part of the population could be at high metabolic risk, but most of them are not being diagnosed because they are not overweight or obese. The intake of diets with an unbalanced macronutrient proportion (rich in fats or simple carbohydrates) is one of the main causes of the increasing emergence of MONW subjects [23, 24]. As stated earlier, a high-fat diet appears to play a significant role in cognitive dysfunction and, thus, MONW individuals could be at a higher risk of developing cognitive impairment later in life.

How obesity or high-fat diet intake affect brain physiology and function remains poorly understood, but these conditions are known to have a great impact on the hippocampus [25–27]. The hippocampus is a vital structure for cognition; it has a significant role in memory and learning, and is particularly vulnerable to ageing [28, 29]. Since the diagnosis of MCI or dementia would require a number of established clinical features that are present when the disorder is almost or fully developed, and given the complication of obtaining samples of brain tissue, the need for readily available early biomarkers of poorly invasive samples has proven to be urgent. In this sense, peripheral blood mononuclear cells (PBMC) could be a very interesting biological material. The gene expression pattern of these cells is increasingly used for diagnosis, to predict the clinical outcome of various diseases [30, 31], and in nutrigenomic studies [32–34]. Moreover, PBMC are easily obtained, offering the chance to perform studies at different time points. In this way, it would be possible to find early biomarkers of disease and, thereby, contribute to setting and addressing strategies to prevent or delay the onset and/or progression of dementia.

Here, a moderately high-fat (45% kcal from fat, HF45) or a very high-fat (60% kcal from fat, HF60) diet was administered in isocaloric conditions to a control diet in order to avoid overweight and to mimic the situation of “normal-weight obesity” in Wistar rats. We aimed to evaluate whether MONW-like animals present gene expression alterations in the hippocampus associated with an increased risk of cognitive impairment. Moreover, we aimed to determine whether these changes are reflected in PBMC in order to identify early biomarkers of cognitive dysfunction in blood samples.

It is well known that obesity and/or high-fat diet intake are associated with an increased risk of a wide range of metabolic disorders, such as type 2 diabetes, hypertension, stroke, or some types of cancer. More recently, the relationship between obesity and/or high-fat diet intake and cognitive decline has received some attention. In this study the aim was to evaluate whether the intake of unbalanced diets rich in fat administered in isocaloric amounts compared to a control diet and, thus, not promoting obesity, might be deleterious in terms of cognitive function when offered chronically. Moreover, we decided to identify early biomarkers of cognitive decline in PBMC, which have emerged as valuable for the detection of nutritional imbalance and markers of obesity-associated disorders [33, 34, 41].

Here, evidence is presented not only that excessive caloric intake (i.e. leading to obesity) is related to cognitive dysfunction, but our pair-fed study also shows that fat intake per se is an important harmful factor to be considered. Therefore, individuals with fat-rich unbalanced diets that do not develop obesity, those presenting the MONW phenotype, could potentially be a population at risk of developing not only the better known obesity-related complications, but also cognitive impairment and/or dementia in later life. The underlying mechanisms by which high-fat diet intake is related to cognitive decline are still not fully understood, although systemic and central inflammation has been suggested to be one of the most important pathophysiological mechanisms underlying cognitive impairment and dementia [7, 47]. This is because central inflammation could trigger a number of processes including oxidative stress and neuronal apoptosis [47–49]. Hence, in our study, the focus was on analysing genes associated with a particular cognitive decline-related process, and also on a pool of genes reflecting different, yet entwined, processes, i.e.: amyloid peptide metabolism, neuronal viability/function, synaptic function/signalling, transcription factor-mediated regulation of cognitive-related processes, and inflammation. Therefore, some of the genes have a direct role in modulating cognitive function while others have an indirect impact on such processes, e.g. reflecting central inflammation. All the genes analysed were selected based on data from microarray analysis conducted by our group, or from the literature where their expression changes in response to high fat diets or obesity were related to cognitive dysfunction.

Herein, the intake of a HF diet with 60% kcal from fats induced a clear MONW profile (i.e. greater visceral fat, fat depot in liver, and insulin resistance) in Wistar male rats. In these animals, increased mRNA expression of inflammatory (Tnfα) and pro-apoptotic (Casp3) markers was observed in hippocampus, even in the absence of increased body weight. Moreover, a tendency to astrogliosis was observed in the HF60 pair-fed animals, as seen in other studies using ad libitum HF diets [25, 50]. This is of relevance, as gliosis is a common characteristic of central nervous system damage and neurological diseases that include cognitive impairment [46, 51]. Further, mRNA expression of App, which codes for the precursor of the amyloid protein (APP) whose accumulation is an indicator of AD-like pathology, was also enhanced in the HF60 group, as described for HF diets administered ad libitum [50]. The enhancement of App in itself may not only involve a greater chance of subsequent amyloid deposition but may also exert pro-inflammatory effects [50] and enhance oxidative stress and apoptosis [52], thus promoting a vicious cycle of neuronal dysfunction. The decreased expression seen for Sorl1 in this group, observed at protein level in both the HF45 and HF60 groups, is worthy of highlighting. To our knowledge, this is the first observation showing that the levels of Sorl1 mRNA and protein in the hippocampus are associated with HF diet intake. Sortilin acts as a regulatory gate keeper for the trafficking pathways of App [53], protecting the latter from β secretase activity and thus preventing amyloidogenic peptide production and consequent deposition [54], a typical AD feature. Therefore, in addition to the increase in App mRNA levels in HF60 fed animals, the reduced mRNA and protein expression of sortilin in the hippocampus might worsen cognitive deficits. Isocaloric administration of a more moderate high-fat diet (45% kcal from fats) was not able to elicit these gene expression changes so explicitly as the very high-fat diet, although the same trends were detected. Nevertheless, the HF45 group displayed alterations in the expression of other interesting genes related to cognitive function, even without presenting MONW-related abnormalities as clearly as HF60 animals. For example, the expression of the gene coding for the auxiliary subunit of NatA (Naa16) was reduced in the hippocampus of animals maintained on both unbalanced diets. NatA is a protein complex that is believed to be involved in the regulation of protein turnover [55]. Asaumi et al. [56] demonstrated that, in humans, this complex interacts with APP and inhibits the secretion of amyloidogenic peptides. Here evidence is provided that the expression of the auxiliary subunit of NatA in rat hippocampus is modulated by HF diets and, based on the results from Asaumi et al., its down-regulation may contribute, to some extent, to cognitive decline in these animals. The same down-regulation pattern of expression was seen for Zpr1 and Tmcc2 in animals fed the HF diets. As stated by Nogusa et al [57], Zpr1 may be related to oxidative stress. They reported that Zpr1 is up-regulated in the hippocampus of mice fed a diet rich in saturated fat, which is in contrast with our results. Nevertheless, Zpr1 is required for cell viability and contributes to normal cellular proliferation [58]. Thus, the inhibition of its expression could be associated with neuronal loss and subsequent cognitive deterioration. Little is known about the possible association between HF diets and the expression of Tmcc2, but decreased expression of both TMCC2 mRNA and protein was observed in our MONW-like animals. This gene codes for a protein that interacts with APP and could have a potential contribution to the anomalous beta-amyloid metabolism that occurs in AD; in fact, disruption of the gene Tmcc2 has been proposed as a marker of AD [59, 60].

Obesity and ad libitum intake of HF diets have been related to a decrease in neuronal plasticity [61], which has been attributed to a variety of processes including abnormalities in Bdnf expression [62]. BDNF is a critical mediator of neuronal vitality and function, and has been revealed as a crucial factor in neuronal processes underlying learning and memory [63]. Although isocaloric intake of the high-fat diets did not alter Bdnf expression, the expression of down-stream effectors of BDNF action was reduced in the hippocampus of these animals, including Trkb, coding for BDNF receptor, and Syn1, which codes for an important protein for neurotransmitter release [64] and synaptic function [65]. The reduced mRNA expression of the latter is remarkable as it has been previously related to synaptic dysfunction in animals fed a high-fat, refined-sugar diet [61]; however, it was not accompanied by lower protein levels in our animals.

One of the characteristic features of the MONW phenotype is insulin resistance, which can be experimentally assessed using the HOMA index. Insulin resistance, a hallmark of obesity and of the intake of unbalanced diets, seems to play an important role in the modulation of gene expression observed in the hippocampus. Pearson’s correlation analysis shows that the degree of insulin resistance correlates negatively with the expression of Bdnf, Naat16, Syn1, Tmcc2, Trkb, and Zpr1, and positively with App mRNA levels when considering the HF60 group; and negatively with Bdnf, Naat16, Tmcc2, and Zpr1 when considering the HF45 group. As a matter of fact, peripheral insulin resistance has been related to cognitive deficits in both humans [66, 67] and rodents [68]. Insulin receptors are distributed across the brain, particularly in the hippocampus and in other areas related to cognitive processing, where insulin signalling through its receptors is involved in synaptic plasticity and behaviour, proving it is a key component of hippocampal memory processes [16]. Arnold et al. [69] reported that central insulin resistance as a result of HF diet intake is associated with a decreased expression of synaptic markers in the hippocampus, suggesting that insulin acts as a synaptic modulator. Moreover, insulin is known to modulate Aβ processing within the brain, competing with this protein for degradation, and regulating its removal [70]. Thus, our results are in line with the above, suggesting that insulin plays a critical role in regulating cognitive processes; therefore, the impairment of its modulation would have deleterious effects on such processes.

Taken together, these molecular changes in the hippocampus suggest cognitive impairment in MONW-like animals that could eventually lead to neurodegenerative diseases. This was emphasised by our behavioural data, which demonstrates impaired spontaneous alternation in a T-maze after HF60 pair-feeding, thus providing additional support for hippocampal-based memory dysfunction resulting from isocaloric HF diet intake. These results are in agreement with studies on ad libitum administration of HF diets causing obesity, showing that cognitive performance is impaired when assessed using this or other spatial and non-spatial memory tasks [25, 71, 72].

The diagnosis of AD or dementia requires invasive and expensive techniques. Moreover, dementia is diagnosed when the disease is already present and sometimes fully developed. For these reasons, there is an imminent urgency to find early and valid biomarkers to detect cognitive disruption before dementia shows up. Based on our expertise in previous studies searching for the suitability of PBMC as a biological source of material to detect biomarkers of metabolic and/or homeostatic alterations after nutritional imbalances [32, 33, 41], we hypothesised that PBMC could be a suitable source of biomarkers of fat-induced neurodegenerative disease. PBMC are easily obtainable material that expresses 80% of the genome [73]. Moreover, these cells are exposed to different metabolites secreted by the central nervous system [74, 75]. It is known that about 500 ml of cerebrospinal fluid is taken up by blood every day [74]. Furthermore, dysfunction of the blood-brain barrier, an early pathologic feature of AD and vascular dementia, may enhance the exchange of metabolites between brain and blood in both directions [75]. Thus, PBMC could reflect gene expression changes occurring at a central level. In fact, it has previously been shown that PBMC can accurately reflect the gene expression pattern of orexigenic NPY which occurs in the hypothalamus in response to fasting-refeeding [32]. Here, our data show that PBMC express genes involved in cognitive function, and that their expression is regulated at different time points during rat development by chronic exposure to isocaloric HF intake in a similar manner to what is expected and observed in the hippocampus. Some of the gene expression changes were especially reflected at an early age in PBMC from animals maintained on both diets; this is the case of Sorl1, Syn1, and Tmcc2. Other genes, such as Nrf2 and Naa16, displayed changes in their expression at later stages (five and six months of age) in response to both diets. It is worth mentioning that Sorl1 and Syn1 were revealed as potential early biomarkers of cognitive impairment, since their mRNA levels in PBMC remained significantly low with the intake of the HF diets throughout the dietary intervention. These results are in accordance with those observed in the hippocampus, and seem to be indicative of the disruption of cognitive processing that occurs at a central level after chronic exposure to HF diets even in the absence of obesity. Moreover, PBMC seem to be especially sensitive to nutritional imbalance, as although HF45 animals do not yet display changes in Syn1 expression in the hippocampus they do show reduced mRNA levels of this gene in PBMC at different ages comparable to those of HF60 animals. Syn1 could be a particularly interesting biomarker of cognitive impairment associated with the intake of HF diets in PBMC, as its expression in these cells has a strong negative correlation with insulin resistance-related parameters. More research is needed, either in rodents or directly in humans, to confirm the usefulness of Syn1 and to validate it as an early marker of diet-related cognitive impairment.

Continuous intake of unbalanced diets with an excess of fat has a great impact on the expression of cognitive-related genes in the hippocampus, suggesting cognitive impairment in these animals. To our knowledge, this is the first time a link has been established between the MONW phenotype and cognitive dysfunction. Moreover, PBMC are able to reflect impairment in the expression of cognitive-related genes observed in the hippocampus. Although further studies are needed, considering that neither MONW-like individuals nor dementia are easily diagnosed, and the lack of effective treatments against the latter, PBMC open the possibility to perform human studies with minimum invasion, since they could be a promising approach for the identification of accessible early predictors that reflect cognitive impairment before any other pathological signs appear.