The serum metabolome mediates the concert of diet, exercise, and neurogenesis, determining the risk for cognitive decline and dementia

2022). The serum metabolome mediates the concert of diet, exercise


K E Y W O R D S
adult hippocampal neurogenesis, cognitive decline, dementia, diet, exercise

INTRODUCTION
Cognitive decline (CD) is an increasing concern, not only due to the resulting deterioration in overall quality of life and independence worldwide, 1 but also because of the well-known elevated risk to develop dementia. 2Given that both CD and dementia are agingrelated conditions 3 and life expectancy continues to increase, CD and dementia are expected to triple in prevalence by 2040. 4 While our understanding of some of the potential mechanisms of cognitive aging has improved, 5 effective strategies for preventing/slowing CD have so far remained limited.However, several lifestyle factors have emerged as potentially modifiable risk factors for CD and dementia, 6 which may offer promising targets for early preventive strategies.
Two key lifestyle modifications that offer considerable promise pertain to exercise and diet.Indeed, several clinical studies have documented the protective effects of increased exercise against CD/ dementia, 7 and revealed how higher adherence to a Mediterranean diet slowed CD and reduced dementia risk. 8,9There is also evidence of a protective association between certain candidate nutrients and foods (e.g., folate and flavonoids in dark leafy vegetables, omega-3 fatty acids, and vitamin D in seafood, caffeine) and cognitive outcomes in aging human populations. 10In this respect, we recently identified a signature of metabolites predictive of CD and found a protective role of coffee, cocoa, fish, and red wine. 11However, before we can effectively exploit the effects of exercise and incorporate protective bioactive compounds into targeted dietary interventions to attenuate/prevent CD and potentially dementia, the role of exercise, and of these dietary modulators, needs to be better understood at the neurobiological level.
Importantly, animal research has shown how diet and exercise can directly influence hippocampal neurogenesis (HN, i.e., the birth of new neurons in the adult brain 12 ), which specifically increases in response to exercise 13 and to diets rich in omega-3 fatty acids, polyphenols, and vitamins D and E. 14 Moreover, HN plays a vital role in memory and cognition 15 and, like diet and exercise, has been associated with CD/dementia in human postmortem research, 16 suggesting that some of the protective effects of lifestyle factors on cognitive health could be mediated by a modulation of HN.However, currently it is impossible to directly test the effects of diet and/or exercise on human adult HN, which is why we turned to a recently developed, well-established in vitro proxy readout of human neurogenesis (Box 1). 17,18erefore, using serum samples collected at inclusion from a longitudinal dementia cohort, we investigated the effect of the systemic environment (i.e., serum) of participants with, and without, subsequent CD and dementia on the neurogenic process using the in vitro assay.
More specifically, we: (1) determined whether markers of HN were associated with CD and/or dementia 12 years after the samples were taken, and (2) whether exercise and key nutritional factors, including our previously identified metabolites 11 and lipids, 19 could modulate the neurogenic process.

Population and study design
Serum samples were from participants of the Three-City (3C) cohort, 20 specifically, from a case-control study on CD (n = 418) nested within the 3C-Bordeaux centre study as described before. 11Briefly, as highlighted in Figure 1A, the 3C cohort (n = 9,294) is a French populationbased cohort on dementia, started in 1999-2000 and consisting of community dwellers aged >65 years.Participants were recruited from three French cities: Bordeaux (n = 2,104), Dijon (n = 4,931), and Montpellier (n = 2,259), and specifically a case-control study nested within the 3C-Bordeaux cohort (n = 418) was constructed for the purposes of this research as previously described 11 and detailed in the Supplementary materials.
As depicted in Figure 1A, at baseline, fasting blood samples, and sociodemographic, lifestyle, and clinical measures were collected from all participants.Follow-up visits were performed every 2 to 3 years over a 12-year period during which cognition status and dementia incidence were assessed.The 3C study protocol was approved by the Consultative Committee for the Protection of Persons participating in Biomedical Research at Kremlin-Bicêtre University Hospital (Paris, France).For further details on all baseline and outcome measures collected on the 3C study, see Table 1 and the Supplementary materials.

Cognitive outcomes
Participants were retrospectively classified as either cognitively stable (control) or with accelerated CD (case) based on their trajectory of cognitive evolution over 12 years as previously detailed. 11Cases were defined as participants with the worst slopes of CD across the 12-year follow-up and controls as those with CD below median value (i.e., >median slope; Figure 1A).For further details, see the Supplementary materials.

Dementia outcomes
Participants were dementia-free (and cognitively stable) at baseline and were followed-up for incident dementia across years.Clinical diagnosis was established and validated by an independent committee of neurologists, using the Diagnostic and Statistical Manual of Mental Disorders IV as described elsewhere. 21Dementia subtypes included Alzheimer's disease (AD, including cases of probable or possible AD and mixed dementia) or vascular dementia and other (including cases of vascular dementia, Parkinson dementia, Lewy body dementia, and frontotemporal dementia).Due to small numbers the seven subtypes were retrospectively pooled into two main categories for analyses.For further details, see the Supplementary materials.research on the association between hippocampal neurogenesis (HN) and these aging-associated outcomes is in its infancy.Moreover, the relationship between diet, exercise, neurogenesis, CD, and dementia remains elusive.

Interpretation:
We demonstrate that HN is both modulated by diet and exercise and predicts future CD and dementia rates across a 12-year period.These data support the literature revealing that neurogenesis may be a key pathway in the relationship between diet, exercise, and cognitive aging.
3. Future directions: Future efforts are needed to mechanistically understand how diet could modulate HN and subsequent CD/dementia, given that it is modifiable and may represent an effective early preventative strategy against future CD and dementia.

Dietary variables
2][23] The metabolite and lipid data were extracted from serum using untargeted liquidchromatography (LC) mass spectrometry (MS) metabolomics and shotgun MS lipidomics, respectively, as described previously. 11,19For a full list of all dietary variables, see Table 2.

Cell line and culture conditions
We used the immortalized human fetal hippocampal multipotent progenitor cell line HPC0A07/03 (HPC; ReNeuron Ltd, UK) as described before. 17,18HPCs were cultured in medium (constitution as previously described 18 ) and grown on tissue culture flasks (Nunclon, Denmark), incubated at 37 • C, with 5% CO2 and saturated humidity.Cells were routinely passaged at 80% confluency before being plated for experiments.For details on the cell line and culture conditions, see the Supplementary materials and Figure S1.

2.6
In vitro neurogenesis assay HPC0A07/03 cells (of passage number 15 to 21) were treated with participant serum as described before. 17

Image analysis
All immunostainings were quantified using the semi-automated  (j) Assessed using the Center for Epidemiological Studies-Depression scale (CES-D). 1,2High depressive symptomology score ≥ 17 men, ≥23 for women or too depressed to answer.
(k) Includes all antihypertensive drugs, calcium channel blockers, diuretics, beta-blockers, and drugs acting on the renin-angiotensin system.(l) Includes all antidiabetic drugs except insulin.
(o) Practice and intensity of physical exercise was assessed using a physical activity questionnaire for the elderly. 3Regular exercise was classified as doing sport regularly or having at least one hour of leisure or household activity per day.Described in detail in. 4 Dementia incidence: n = 113 participants developed dementia over the 12-year follow-up period (Median of follow-up = 9.1 years, range = 0.6-13.5)with an incidence of 22. as previously described. 18Protocols were set by an experimenter blinded to caseness and parameters were kept constant across all experiments.For details on the protocols and parameters used, see the Supplementary materials.

Statistical analysis
Data analyses were conducted using SPSS Statistics 26 and R software (version 3.6.3).We first studied the association between each of the proliferation and differentiation markers and caseness for CD and dementia using conditional logistic regressions and logistic regression models, respectively.Models were primarily adjusted for age, sex, and education; in case of association, further adjustment was performed, by including potential confounders that were associated with CD and/or dementia in the present population (those in bold font in Table 1).Additionally, to further explore the relationship between the neurogenic process and exercise, interactions were tested between each HN readout and physical exercise in our models, adjusting for age, sex, and education, and potential confounders as aforementioned.
TA B L E 2 Associations between participant characteristics, exercise, nutritional data and altered in vitro neurogenesis readouts Next, for each marker of the hippocampal neurogenic process that we found to have a significant linear association with CD and/or dementia, we explored associations with nutrient biomarkers and previously identified blood metabolites and lipids using linear regression models initially adjusted for age, sex, and education.Further adjustment was performed including potential confounders associated with the HN and nutrient-related data (those in bold font in Table 2).

Measures
To determine the potential mediating role of HN outcomes in the relationship between diet/exercise and CD/dementia, mediation analyses were conducted using the causal step method 24 and based on recommendations by MacKinnon et al. 25 Models were adjusted for age, sex, and education, and for relevant potential confounders (as above).
Finally, we explored whether neurite morphology was associated with CD/dementia and used principal component analysis to combine (l) Practice and intensity of physical exercise was assessed using a physical activity questionnaire for the elderly. 3Regular exercise was classified as doing sport regularly or having at least one hour of leisure or household activity per day.Described in detail in. 4 (m) Derived from a food frequency questionnaire (FFQ) and a 24-hour dietary recall.A Mediterranean diet score was generated by adding the scores for each food group considered to be part of the Mediterranean diet. 5n) Metabolites we previously identified as being associated with cognitive decline in our sample. 6o) Lipids we previously identified (and validated in an independent cohort) as being associated with cognitive decline in our sample.False discovery rate correction 26 was applied to account for multiple testing throughout and all models were bootstrap enhanced to obtain robust estimates of standard errors.For mediation models, bootstrapping was used to estimate the percentage of the association between diet/exercise and CD/dementia mediated by HN outcomes.

RESULTS
Participants were, on average, 76 years-old at baseline and 66% were female (Table 1).The characteristics of our sample by CD and dementia status are detailed in Table 1.Participant cognition was assessed for an average of 8.5 years and the average age of dementia onset was 85 years.
We first determined the association between individual markers of the neurogenic process and subsequent CD and incident dementia.As indicated in Table 1, levels of %CC3-positive cell density (during proliferation and differentiation) and %SOX2 were all significantly associated with CD and/or dementia.

Cell death during differentiation is associated with future CD while the association of hippocampal stem cell integrity with CD is modified by regular exercise
As shown in Figure 2A-B, increased baseline levels of %CC3-positive cells during differentiation were significantly associated with CD across a 12-year period in a fully adjusted model (p = .006;Model 1A).
Moreover, we found a significant interaction between baseline levels of %SOX2-positive cells and physical exercise on CD (p = .049),such that reduced baseline levels of %SOX2 were significantly associated with CD but only in participants who did not regularly exercise (p = .04;Figure 2A: Model 1B; Figure 2C).

The association between HN and incident dementia depends on specific aspects of the neurogenic process and the type of dementia
Interestingly, no significant difference in any baseline neurogenesis readout was found between cases of dementia and controls.However, when we conducted separate models for each dementia subtype (i.e., AD and vascular dementia/other), we found that particular neurogenic markers were associated with specific dementia subtypes (Figure 3).Specifically, we found that decreased levels of baseline %CC3-positive cells during proliferation were associated with an increased risk of AD (p = .04;Figure 3A: Model 4; Figure 3B), whereas reduced baseline %SOX2-positive cell levels were significantly associated with an increased risk of vascular dementia/other (p = .02;Figure 3A: Model 5; Figure 3C).
Thus far we have shown that only specific HN readouts are associated with CD and incident dementia up to 12 years prior to caseness, and that physical exercise can modify the association of %SOX2 with CD.No differences in any other individual neurogenesis readout, that is, densities of %Nestin-, %Ki67-, %DCX-or %MAP2-positive cells, was observed between groups (Table 1).Moreover, we found no difference in the neurite morphology of %DCX-and %MAP2-positive cells between cases and controls (Supplementary materials and Table S1).
Additionally, principal component analyses revealed no differences in the overall neurogenic profiles between groups at baseline (Supple-

mentary materials).
Next, starting with our HN-associated readouts relevant to CD (i.e., %CC3 during differentiation and %SOX2), we subsequently assessed their relationship with exercise and our baseline dietary variables (Table 2).

Global (mal)nutrition status and exercise are associated with cell death during differentiation whereas cholesterol and carotenoid levels are associated with hippocampal progenitor cell integrity among non-exercisers
Both exercise and diet were found to be associated with levels of %CC3 (during differentiation).Specifically, we found that reduced levels of plasma prealbumin (p = .01)and increased levels of an unidentified metabolite (i.e., M129_06580978T0_904033275275335; p = .04)were both associated with increased levels of %CC3 (during differentiation) in a fully adjusted model (Figure 2D-E).Furthermore, we found that a lack of regular physical exercise was also independently associated with increased levels of %CC3 (during differentiation) (p = .001;Figure2D).Since %SOX2 was modified by exercise in the context of CD in our sample, we subsequently investigated the association between this particular HN marker and the nutritional factors, as stratified by exercise.Specifically, we found that increased plasma levels of low density lipoprotein (LDL) cholesterol (p = .004)and reduced plasma levels of ß-cryptoxanthin (p = .002)were both associated with reduced levels of baseline %SOX2 among non-regular exercisers only in a fully adjusted model (Figure 2F-G).
Finally, we focused on our two HN markers associated with AD and vascular dementia/other (i.e., %CC3 during proliferation and %SOX2 [in the sample as a whole]).Here, we found that diet, but not exercise, was associated with these two neurogenesis readouts (Table 2).

Vitamin D levels and global (mal)nutrition, but not exercise, are associated with cell death during proliferation, whereas carotenoid levels and altered lipid metabolism are associated with hippocampal progenitor cell integrity
As shown in Figure 3D-E, reduced plasma levels of vitamin D (p = .04)and prealbumin (p= .009)were both independently associated with reduced baseline levels of %CC3 (during proliferation).Moreover, increased levels of the same unidentified metabolite found to be associated with %CC3 (during differentiation) was also associated with cell death during proliferation (p = .03).
For results on the associations between these nutritional factors and CD/dementia, see the Supplementary materials.

The effect of regular exercise and prealbumin levels on future CD is partially mediated by cell death during differentiation
In order to more fully understand the relationship between our observed associations with exercise, diet, HN, and CD/dementia, medi-ation analyses were performed.As depicted in Figure 2H-I

DISCUSSION
Previous in vivo and in vitro parabiosis research supports that the systemic environment plays a significant role in brain aging and may determine the fate of hippocampal progenitor cells (Box 1).Here, we further demonstrate the ability of the human systemic environment, in the context of CD and dementia, to differentially regulate HN in an in vitro model.Using baseline serum samples from a prospective dementia cohort, we show that not only are individual markers of HN associated with CD and dementia 12 years later, but that there is also some degree of specificity with respect to dementia subtype diagnosis.Specifically, we find that increased baseline levels of cell death during differentiation (i.e., %CC3) significantly predicts CD over the following 12 years, while decreased baseline levels of cell death during proliferation and reduced hippocampal progenitor cell integrity (i.e., %SOX2) predict AD and vascular dementia/other, respectively.Moreover, we find that exercise, malnutrition, vitamin D, carotenoid, and lipid levels are associated with these changes in cell death and hippocampal progenitor cell integrity, and that specifically reduced physical activity and increased levels of malnutrition increase cell death during differentiation, which in turn increases the risk for future CD.
One of the brain structures that shows the earliest changes in relation to CD is the hippocampus. 27It is also one of the niches within the human adult brain where neurogenesis occurs, 12 and has consistently been implicated in learning and memory. 15Therefore, it is highly encouraging that we observe such early changes in HN in participants with future CD and dementia.Indeed, while our HN measures are only proxy measures of in vivo neurogenesis, our finding that those with future CD have increased apoptosis during differentiation is in line with human postmortem 28 and human in vitro work. 17Importantly, animal research supports a role for apoptosis in both CD and dementia through promoting neuronal loss, 29 and it has become commonly acknowledged that increased neuronal death due to apoptosis is a key characteristic of CD and AD 30 -outcomes of which could be driven by inflammatory mediators released by microglia and astrocytes, which subsequently compromise the function and structure of neurons. 31r most interesting finding is that in participants who subsequently develop AD, we observe a decrease in apoptosis during proliferation, which was not observed in those with future CD.The likely impact of this change will be a disruption in HN (i.e., an increase), given that less apoptosis occurs in the early trajectory of the neurogenic process. 32Indeed, previous clinical work supports an increase in proliferation and HN in early AD, 33  with previous research.Interestingly, increased platelet levels of antiapoptosis protein, Bcl2, in amnestic mild cognitive impairment (MCI) has recently been reported, implying that abnormal apoptosis may appear in the early stage of AD and increase with the development of the disease. 34Moreover, previous work using the in vitro HN assay found that increased neurogenesis characterized progression of MCI one year before conversion to AD, 17 so it remains possible that these even earlier changes in apoptosis could precede the later neurogenesis outcomes.It is noteworthy that the majority of HN and AD human research evaluates neurogenesis during the later stages of AD, making it difficult to extrapolate these findings to prodromal stages of the disease.In the context of our work this is even more challenging when our measures represent the state of HN up to 12 years prior to AD diagnosis.
It is also exciting that we can distinguish between dementia subtypes (i.e., AD vs vascular dementia/other).Unlike AD, we find that a reduction in baseline levels of %SOX2, a marker of hippocampal progenitor integrity (i.e., the maintenance of the stem cell pool and its proliferative/differentiative capacity), is predictive of vascular dementia/other-an outcome we also find in CD in individuals not regularly exercising.Given the strong implication of SOX2 in regulating the fate of stems cells, it is unsurprising that this transcription factor could play an important role in the development of neurodegenerative disease. 35Indeed, previous animal work demonstrates how SOX2 deficiency not only impairs HN, but can also promote neuronal degeneration. 36Importantly, decreased levels of SOX2 have been found in dementia postmortem, the decrease of which also correlates with disease severity, albeit in the context of AD. 37 Moreover, SOX2-positive neural stems cells may play a key role in cognitive-reserve capacity, shown previously to correlate with human cognition, 38 and increase upon the restoration of cognitive impairment in an AD mouse model. 39To our knowledge, our finding that SOX2 is specifically associated with vascular dementia/other is novel.Therefore, it is difficult to extrapolate our findings to the wider literature on other forms of dementia specifically, which have largely been understudied in the context of HN research. 40terestingly, SOX2 also plays a key role in vascularization, 41 and it is noteworthy that vascular dementia, which is associated with reduced cerebral blood flow (CBF), 42 was most prevalent within our vascular dementia/other category.Animal research shows how SOX2 deficiency can lead to abnormal vasculature development, 43 and has been specifically linked to vascular disease brought on by abnormal endothelialmesenchymal transitions, for which SOX2 is an essential mediator. 44rthermore, given that exercise promotes CBF, 45 it might explain why SOX2 can predict CD in participants that do not regularly exercise.
Indeed, previous clinical research reports memory and executive function improvement in individuals with exercise-associated increases in CBF. 46sed on these findings, we sought to determine whether exercise could be modulating these HN outcomes and find that reduced exercise specifically increases cell death during differentiation which then subsequently increases the risk for future CD within our sample.This is supported by both animal and human research showing how exercise inhibits apoptosis, 47 via suppression of the immune system, 48 and importantly, can attenuate CD in a clinical sample 49 -findings all observed in the context of our work.However, while we show that exercise plays an important role in CD-associated apoptosis, we find no evidence to support an association between exercise and AD or vascular dementia/other in our sample.
With respect to diet, we find that overall nutritional status may be a key factor in determining apoptosis at any point during the neurogenic process.Specifically, we find a positive association between prealbumin-a biomarker for malnutrition 50 -and apoptosis during proliferation but a negative association with apoptosis during differentiation-suggesting a biphasic relationship between these two factors.Importantly, malnutrition has been frequently associated with increased apoptosis and immune system activation both clinically and in animal models. 51,52Moreover, poor nutritional status is known to accelerate CD/dementia in aging human populations, 53 and altered apoptosis could represent one of the mechanisms of action contributing to CD/dementia risk.Indeed, in our sample, we find a significant indirect effect of reduced prealbumin levels on future CD mediated by increased cell death during differentiation, emphasizing that the observed association between apoptosis and CD is influenced by the negative effect of poor overall nutrition.This supports the idea that global nutritional status may be more effective in slowing CD/dementia in clinical populations 54 by promoting brain health.However, it is noteworthy that while we find significant independent associations between prealbumin levels, apoptosis, and AD (see the Supplementary materials), we find no evidence to support that apoptosis during proliferation mediates the observed relationship between AD and poor nutrition, pointing to other key mediators not identified in the context of our work.
In addition to the observed positive association with prealbumin, we find that vitamin D levels are also positively associated with cell death during proliferation, which is in line with human in vitro cellular work showing that increased vitamin D promotes apoptosis in a dosedependent manner. 55Furthermore, we also find that AD cases within our sample have significantly lower levels of vitamin D at baseline (see the Supplementary materials), which is also consistent with previous clinical research that supports vitamin D deficiency as a risk factor for AD. 56However, similar to prealbumin, while we observe independent associations between apoptosis during proliferation, vitamin D levels, and future AD, we find no statistical evidence to support that the relationship between vitamin D and AD is mediated by apoptosis.Therefore, other (as yet unidentified) mediators are likely playing a more substantial role in the trajectory of AD.However, we are mindful that insufficient power may have limited mediation analyses in the context of AD, increasing the likelihood of type II error. 57Future work should aim to replicate these findings in a larger sample while using a longitudinal design to maximize the identification of other key mediators of HN in the trajectory of dementia.
Although we show that global (mal)nutrition may be important with respect to apoptosis, we find that single dietary nutrients may be more relevant to %SOX2.Specifically, we see a positive association between %SOX2 and ß-cryptoxanthin-a xanthophyll and biomarker for fruit and vegetable intake. 58This is particularly interesting, given that both have been associated with cancers, and cardiovascular and neurodegenerative disease, with increased levels of ß-cryptoxanthin known to have a protective effect against these conditions. 35,59Furthermore, reduced carotenoid levels have been associated with dementia, 22 suggesting that this particular dietary measure could play a key role in modulating SOX2 levels that subsequently promote or aggravate disease pathology.However, similar to AD, we could only demonstrate associations between these factors and did not find a mediatory effect of %SOX2.Moreover, we did not find a difference in ß-cryptoxanthin levels at baseline between controls and individuals diagnosed with vascular dementia/other, although it remains possible that the effect of ßcryptoxanthin on hippocampal stem cell integrity could influence other downstream processes that subsequently promote dementia pathology years later, an effect not captured due to our study design.
Interestingly, carotenoids are also key modulators of lipid membranes and play a vital role in their protection and in the maintenance of the lipid bilayer, 60 which is consistent with our finding that ß-cryptoxanthin levels are negatively correlated with PCO32:0(16:0/16:0) (p = .008,Supplementary materials)-an ether phospholipid that forms a key component of the lipid membrane and lipid rafts. 61Importantly however, we find that %SOX2 is positively associated with unsaturated PCO34:1(16:1/18:0) and negatively associated with saturated PCO32:0(16:0/16:0)-all of which play a critical role in neurotransmission and synaptic plasticity. 62Moreover, we find that the positive association between ß-cryptoxanthin and %SOX2 is mediated through levels of PCO32:0(16:0/16:0), such that reduced ß-cryptoxanthin is associated with increased PCO32:0(16:0/16:0), which in turn is associated with decreased %SOX2 (Supplementary materials).This partial mediation suggests that disruption in lipid homeostasis could be one of the mechanisms contributing to the observed changes in %SOX2.Although the dysregulation of the aforementioned PCO species has been consistently observed in CD/dementia, 63 due to limited sample size we were unable to further explore the complex interactions that exist between %SOX2, ß-cryptoxanthin, the two PCO species, and this dementia category.
The strength of our work lies in the use of a well-characterized prospective cohort to evaluate the impact of exercise and nutrition (including metabolomics and lipidomics) on in vitro neurogenesis measures early (i.e., up to 12 years) in the trajectory of CD and dementia.However, our study also has limitations.First, our HN measures are only proxy measures of in vitro neurogenesis; therefore, they might not mirror HN in vivo.Moreover, we recognize that our assay does not reconstitute the neurogenic niche in its entirety, and future work should expand the model to include other important players (e.g., microglia) and extend its duration to also monitor synaptic formation and plasticity.Finally, we only assessed a single baseline measure of the neurogenic process, deemed necessary to determine temporality.Due to time and logistical constraints, serum samples at later times points were unavailable for analysis.To more fully understand the impact of the neurogenesis-associated outcomes across the trajectory of CD and dementia, it would be highly profitable for future research to adopt a longitudinal approach and compare HN markers and overall neurogenic profiles at multiple timepoints.
In summary, we show that HN, altered by exercise and diet, can predict future CD and dementia, and can potentially distinguish between dementia diagnoses.The observed changes in apoptosis and hippocampal progenitor cell integrity, up to 12 years prior to the development of CD and/or dementia, could signify the start of the pathological process and potentially represent early functional biomarkers for these conditions.However, more work is needed to more fully understand how exercise and diet might be modulating HN and subsequent CD/dementia, given that it is modifiable and may represent an effective early preventative strategy.

BOX 1 : 1 .
The concept behind the in vitro human neurogene-Systematic review: The authors reviewed the literature using PubMed, Embase, and PsycINFO.While the impact of diet and exercise on cognitive decline (CD) and dementia has been consistently demonstrated (cited herein), ,18 Briefly, as detailed in Figure 1B, 1% serum was added to the cell culture during both proliferation (48 hours) and differentiation (7 days) before being fixed in 4% paraformaldehyde and stained for proliferation-and differentiationspecific markers, respectively.All experiments were carried out in technical triplicates by an experimenter blinded to caseness.For further detail on the in vitro assays, see the Supplementary materials.

TA B L E 1
CellInsight NXT High Content Screening (HCS) platform (Thermo-Scientific) and Studio Cell Analysis Software (ThermoScientific) F I G U R E 1 Population, study design, and cellular assays.(A) Three-City (3C) cohort and sample: The 3C cohort is a French population-based cohort on dementia that started in 1999-2000 and consists of male and female community dwellers aged >65 years (n = 9,294).Participants from the 3C study were recruited from three French cities, Bordeaux (n = 2,104), Dijon (n = 4,931), and Montpellier (n = 2,259), and a case-control study on cognitive decline (CD) nested within the 3C-Bordeaux cohort (n = 418) was constructed for the purposes of this study.Participants in our sample were all free from dementia at baseline, and cases and controls (for CD) were matched at baseline for age, gender, and education; therefore, our sample consisted of 209 cases and 209 controls.Using a linear mixed model, individual slopes of cognitive change for each participant over the 12-year follow-up period were estimated.The primary outcome represents the change in a composite score of global cognition, defined as the average of Z-scores of five neuropsychological tests (Mini-Mental State Examination [MMSE], Benton Visual Retention Test [BVRT], Isaac's Set Test [IST], Trail-Making Test part A [TMT-A], Trail-Making Test part B [TMT-B]) across the five follow-up visits.Cases were defined as participants with the worst slopes of cognitive decline and controls as those with a more stable trajectory (i.e., >median slope).(B) Longitudinal study design and measures: At baseline, face-to-face interviews were conducted to collect sociodemographic, lifestyle, and medical data, plus cognitive testing, blood pressure, and anthropometric measurements from all participants.Additionally, fasting blood samples were collected for constituting a plasma and serum biobank, from which serum samples were used for the metabolomics, lipidomics, and in vitro cellular assays.Follow-up visits were performed every 2 to 3 years over a 12-year period and included in-person neuropsychological assessments carried out by a trained psychologist.(C) In vitro cellular assay: Proliferation timeline: Twenty-four hours after seeding, cell medium was replaced with fresh medium containing 1% serum and 1:100 penicillin streptomycin (PenStrep; 10 000 U/mL) and subsequently left to incubate for 72 hours before being fixed in 4% paraformaldehyde (PFA) and stained for proliferation specific markers.Specifically, to assess proliferation, sex determining region Y (SRY)-box 2 (SOX2) and Nestin were used to measure hippocampal progenitor cell integrity and maintenance, while Ki67 was used to measure proliferation; cleaved caspase-3 (CC3) was also used to assess cell death during these early stages of the neurogenic process.(D) In vitro cellular assay: Differentiation timeline: After 48 hours of proliferation in the presence of 1% serum and 1:100 PenStrep, cells were washed and treated with another serum supplementation, this time in medium lacking 4-hydroxytamoxifen (4-OHT) and growth factors (epidermal growth factor and basic fibroblast growth factor), to allow cells to differentiate spontaneously.Serum-treated cells were subsequently left to differentiate for a further 7 days before being fixed in 4% PFA and stained for differentiation specific markers.To assess differentiation, cells were stained with doublecortin and microtubule-associated protein 2 to look at early differentiation overall, and CC3 to assess cell death.Abbreviations: M, male; F, female; LCMS, liquid chromatography mass spectroscopy.Image created using BioRender software.Baseline participant characteristics and in vitro neurogenesis readouts according to caseness
This project was part of the EU consortium DCogPlast "Diet Cognition and Plasticity" funded by JPI-HDHL (Medical Research Council UK: MR/N030087/1; French National Research Agency ANR-15-HDHL-0002-05; PCIN-2015-229-MINECO; CiberFES-Cofund by FEDER Program from EU, 2017SGR1546 & ICREA 2018 Academia Award from the Generalitat de Catalunya), the BMWFW under BMWFW-10.420/0009-WF/V/3c/2015.This represents independent research part funded (ADP) by the National Institute for Health Research (NIHR) Maudsley Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London.The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.PJL and AK are supported by Alzheimer Nederland and the NWO Food & Cognition program; PJL is supported by the Urban Mental Health program of the University of Amsterdam.SLA was part of the University Research school (Ecole Universitaire de Recherche, EUR) Digital Public Health PhD program, supported within the framework of the French National Research Agency (ANR) "Programme d'Investissement d'Avenir" (Investment for the Future) PIA3 (17-EURE-0019).The Three-City Study was conducted under a partnership agreement between the Institut National de la Santé et de la Recherche Médicale (INSERM), the Institut de Santé Publique et Développement of the Victor Segalen Bordeaux 2 University and Sanofi-Aventis.The Fondation pour la Recherche Médicale funded the preparation and initiation of the study.The 3C Study was also supported by the Caisse Nationale Maladie des Travailleurs Salariés, Direction Générale de la Santé, Mutuelle Générale de l'Education Nationale, Institut de la Longévité, Regional Governments of Aquitaine and Bourgogne, Fondation de France, Ministry of Research-INSERM Programme "Cohortes et collections de données biologiques," French National Research Agency COGINUT ANR-06-PNRA-005, the Fondation Plan Alzheimer (FCS 2009-2012), and the Caisse Nationale pour la Solidarité et l'Autonomie (CNSA).
(c) Estimated using conditional logistic regression, conditioned on age, gender, and education to account for matched cases and controls.(d) Estimated using logistic regressions controlling for age, gender, and education.False discovery rate correction was applied to control for multiple testing; p values represent adjusted p values.(e) ApoE genotype was considered dichotomously: presence of at least one ε4 allele.(f) Blood pressure ≥ 140/90 mmHg or antihypertensive medication use.(g) Glucose ≥ 7.2 mmol/L or antidiabetic medication use.(h) Fasting plasma total cholesterol ≥ 6.2 mmol/L or lipid-lowering medication use.(i) History of cardiovascular or cerebrovascular disease.