Effect of BDNF Val66Met on hippocampal subfields volumes and compensatory interaction with APOE-ε4 in middle-age cognitively unimpaired individuals from the ALFA study

Background Current evidence supports the involvement of brain-derived neurotrophic factor (BDNF) Val66Met polymorphism, and the ε4 allele of APOE gene in hippocampal-dependent functions. Previous studies on the association of Val66Met with whole hippocampal volume included patients of a variety of disorders. However, it remains to be elucidated whether there is an impact of BDNF Val66Met polymorphism on the volumes of the hippocampal subfield volumes (HSv) in cognitively unimpaired (CU) individuals, and the interactive effect with the APOE-ε4 status. Methods BDNF Val66Met and APOE genotypes were determined in a sample of 430 CU late/middle-aged participants from the ALFA study (ALzheimer and FAmilies). Participants underwent a brain 3D-T1-weighted MRI scan, and volumes of the HSv were determined using Freesurfer (v6.0). The effects of the BDNF Val66Met genotype on the HSv were assessed using general linear models corrected by age, gender, education, number of APOE-ε4 alleles and total intracranial volume. We also investigated whether the association between APOE-ε4 allele and HSv were modified by BDNF Val66Met genotypes. Results BDNF Val66Met carriers showed larger bilateral volumes of the subiculum subfield. In addition, HSv reductions associated with APOE-ε4 allele were significantly moderated by BDNF Val66Met status. BDNF Met carriers who were also APOE-ε4 homozygous showed patterns of higher HSv than BDNF Val carriers. Conclusion To our knowledge, the present study is the first to show that carrying the BDNF Val66Met polymorphisms partially compensates the decreased on HSv associated with APOE-ε4 in middle-age cognitively unimpaired individuals. Electronic supplementary material The online version of this article (10.1007/s00429-020-02125-3) contains supplementary material, which is available to authorized users.


Introduction
Brain-derived neurotrophic factor (BDNF) is a neurotrophin involved in neurogenesis and synaptic plasticity in the central nervous system, especially in the hippocampus, and has been implicated in the pathophysiology of several neuropsychiatric disorders (Bathina and Das 2015;Autry and Monteggia 2012;Numakawa et al. 2018). The single nucleotide polymorphism (SNP) rs6265 (also known as Val66Met), causes a valine (Val) to methionine (Met) substitution at codon 66 of BDNF protein. Particularly, the study of Val-66Met polymorphism within the BDNF gene is of special interest because of its documented impact on hippocampaldependent functions (Notaras and van den Buuse 2018;Toh The complete list of collaborators of the ALFA Study can be found in the Acknowledgements.

Electronic supplementary material
The online version of this article (https ://doi.org/10.1007/s0042 9-020-02125 -3) contains supplementary material, which is available to authorized users. Egan et al. 2003;Hariri et al. 2003). Hence, extensive research focuses on the discovery of associations between BDNF Val66Met polymorphism and several hippocampal phenotypes. However, recent meta-analyses addressing hippocampal volumes for BDNF Val66Met have reported inconsistent statistically significant associations, as well as inconsistencies regarding the direction of the genotype effects across individual studies (Harrisberger et al. 2014(Harrisberger et al. , 2015. Two recent large meta-analyses suggest that the analysis of hippocampal subfield volumes may allow for more accurate detection of genetic effects in genetic association analyses, compared with whole hippocampal volume (van der Meer et al. 2018;Hibar et al. 2017). Moreover, previous studies have shown that different pathological conditions affect subfields differently (West et al. 1994;Jin et al. 2004;Ezzati et al. 2014;Mueller et al. 2010;Hett et al. 2018). In fact, the proven differential expression of BDNF and its receptors in different regions of the hippocampus (Kowiański et al. 2018;Vilar and Mira 2016;Franzmeier et al. 2019), reinforces distinct biological functions of BDNF Val66Met polymorphism on the different subfields. However, to our knowledge no previous studies have addressed the effects of the BDNF Val66Met polymorphism on hippocampal subfields in cognitively unimpaired (CU) individuals. Most of the studies addressing the association of BDNF Val66Met polymorphism and hippocampal volumes (subfields and/or whole hippocampus) included patients of a variety of neuropsychiatric disorders, such as major depressive disorder, schizophrenia and bipolar disorder (Zeni et al. 2016;Cao et al. 2016;Reinhart et al. 2015;Aas et al. 2014;Frodl et al. 2014), showing also inconsistencies concerning the impact of the BDNF Val66Met polymorphisms (Tsai 2018).
The ε4 allele of apolipoprotein E (APOE) gene, the major genetic risk factor for Alzheimer's disease (AD) (Mueller and Weiner 2009), has also an impact on hippocampal subfields. APOE ε4-carriers have reduced volume of the subicular/CA1 region in AD patients (Pievani et al. 2011), as well as in a pool of older adults that included healthy controls and patients with amnestic mild cognitive impairment (aMCI) and AD dementia, after controlling for the diagnostic group (Kerchner et al. 2014). In a recent report in CU participants, we also showed that APOE-ε4 relates to significantly reduced hippocampal tail in a gene dose-dependent manner (Cacciaglia et al. 2018a).
Moreover, recent evidence suggests that APOE genotypes differentially affects the expression of BDNF through the regulation of its maturation in human astrocytes and its secretion (Sen, Nelson, and Alkon 2015). Astrocytes are known to synthesise BDNF, and as brain APOE is primarily produced by astrocytes, studying APOE and BDNF modulation becomes important. Specifically, interactions between APOE-ε4 and BDNF have been suggested to influence their secondary effects on AD pathology (Álvarez et al. 2014), and their influence on hippocampal volume (Li et al. 2016;Shi et al. 2014;Liu et al. 2015a, b). In addition, a significant combined effect of APOE-ε4 and BDNF Val66Met polymorphisms has been reported to moderate β-amyloid-related cognitive decline in preclinical AD (Lim et al. 2015). Episodic memory performance was also found to be impaired in MCI/AD individuals who were also carriers of both the APOE-ε4 and BDNF Met polymorphisms (Gomar et al. 2016), as well as in healthy individuals (Ward et al. 2014). Overall, evidence suggests biological interactions between APOE and BDNF for memory and other brain-related processes that may help to explain the increased AD risk in APOE-ε4 carriers during the period that precedes the development of symptoms.
Therefore, the aim of the present study is to evaluate the impact of Val66Met polymorphism on hippocampal subfields in a large sample of in middle-age cognitively unimpaired individuals CU participants and to assess whether an interactive effect with the APOE-ε4 genotype exists.

Study population and setting
Participants were drawn from the ALFA study (Alzheimer and FAmilies) established at the Barcelonaβeta Brain Research Center (Molinuevo et al. 2016), which aims at identifying the neuroimaging and cognitive signatures in preclinical AD. The ALFA study (Clinicaltrials.gov Identifier: NCT01835717) entangles a cohort of 2,743 cognitively unimpaired participants, mostly adult children of patients with AD, and aged between 45 and 75 years. Cognitive status was assessed at baseline as follows: Mini-Mental State Examination (Folstein et al. 1975;Blesa et al. 2001) > 26, Memory Impairment Screen (Buschke et al. 1999;Böhm et al. 2005) > 6, Time-Orientation subtest of the Barcelona Test II (Quinones-Ubeda 2009) > 68, semantic fluency (Ramier and Hecaen 1970;Peña-Casanova et al. 2009) (animals) > 12 and Clinical Dementia Rating scale (Morris 1993) = 0. A subset of 430 participants from the ALFA study with available information on BDNF Val66Met polymorphisms and APOE genotypes, as well as neuroimaging data (HSv) were included in this study (Fig. 1). The cognitive status of these participants was reviewed if cognitive testing had not been conducted in the last 6 months. For this, mild cognitive impairment (MCI) was ruled out by clinical judgment after interview and accounting for psychometric scores in the main variables of the Free and Cued Selective Reminding Test [FCSRT] (Buschke et al. 2017). The study was conducted in accordance with the directives of the Spanish Law 14/2007, of 3rd of July, on Biomedical Research (Ley 14/2007 de Investigación Biomédica). All participants accepted the study procedures by signing an informed consent form. A subset of 430 participants from the ALFA study with available information on BDNF Val-66Met polymorphisms and APOE genotypes, as well as neuroimaging data (HSv) were included in this study (Fig. 1).

Genotyping
DNA samples were obtained from whole blood samples by applying salting out protocol. DNA was eluted in 800 µl of H2O (milliQ) and quantified using Quant-iTT PicoGreen® dsDNA Assay Kit (Life Technologies). Integrity of DNA was checked in a subset of samples by running a 1% agarose gel. All the samples were within specification. Genomewide genotyping was performed using the NeuroChip backbone (Blauwendraat et al. 2017), based on a genomewide genotyping array (Infinium HumanCore-24 v1.0) containing 306,670 tagging variants and a custom content that has been updated and extended with 179,467 neurodegenerative disease-related variants at the Cancer Epigenetics and Biology Program (PEBC; IDIBELL). Previous step was to normalize the quantity of DNA from each sample. The analysis was performed by the GenomeStudio (Illumina) software using the genotyping module (standard analysis). PLINK was used for genetic data quality control (Purcell et al. 2007). We applied the following sample quality control thresholds: sample call rate > 97% (N = 6 exclusion) and heterozygosity 5 SD (N = 8 exclusions). Then, we checked sex discordances (N = 4 exclusions). In total, we excluded 18 subjects (less than 2%). None of the individuals presented autosomal dominant mutations in APP, PSEN1, and PSEN2. The final genetic data set consisted of volunteers of European ethnic origin with available information regarding BDNF Val66Met polymorphism and the APOE rs429358 and rs7412 polymorphisms. Genotype and allele frequencies of Val66Met, rs429358 and rs7412 polymorphisms were determined. Moreover, allele frequencies were inspected for potential covariate-related differences. Departures from Hardy-Weinberg equilibrium were also examined (Ryckman and Williams 2008). The APOE allelic variants were obtained from allelic combinations of the rs429358 and rs7412 polymorphism (Radmanesh et al. 2014). According to the genotypes of these polymorphisms, subjects were classified depending on the number of ε4 alleles (non-carriers, one ε4 allele or two ε4 alleles).

Image acquisition and extraction of hippocampal subfield volumes
Scans were obtained with a 3 T scanner (Philips Ingenia CX). The MRI protocol was identical for all participants and included high-resolution three-dimensional structural images weighted in T1 with an isotropic voxel of 0.75 × 075 × 0.75 mm 3 . The acquisition parameters were TR/TE/TI = 9.9/4.6/900 ms, flip angle = 8° and a matrix size of 320 × 320 × 240. Hippocampal subfields were segmented using FreeSurfer version 6.0 (Iglesias et al. 2015). We extracted raw volumes for 12 different HSv per hemisphere: the cornu ammonis region 1 (CA1), cornu ammonis region 2/3 (CA3), cornu ammonis region 4 (CA4), dentate gyrus (DG), fimbria, hippocampal-amygdaloid transition area (hata), tail, parasubiculum, presubiculum, subiculum, fissure and molecular layer. The value of the subfields used as the outcomes of the study were calculated as the sum of the regional value of each hemisphere (mm 3 ). We visually inspected the segmentation of the individuals included in the study (Fig. 2), and we removed outliers and/or abnormal hippocampal subfields volume values. The whole hippocampal volume, as well as, total intracranial volume were also calculated using Freesurfer (v. 6.0).

Statistical analysis
Differences in demographic variables were tested using χ 2 test and F test for gender, age, education, number of APOE-ε4 carriers and total intracranial volume (TIV). The additive, dominant, recessive, and codominant effects of the BDNF Val66Met genotype on the hippocampal subfields volume were assessed using general linear models corrected by age, sex, years of education, number of APOE-ε4 alleles ALFA Study sample N = 2,743 No genetic data or data excluded after applying quality control criteria (N = 1,805).

Genetic data N = 938
No neuroimaging data acquired/pre-processed or data excluded after quality control criteria (N= 508). and TIV. These covariates were selected based on previous associations reported using the ALFA study sample (Cacciaglia et al. 2018b). In brief, the genetic additive model predicts a linear increase of the phenotypic variable depending on the number of Met alleles, whereas the codominant genetic model infers that the heterozygote mean differs from both the homozygote means. The dominant genetic model assumes a common response to 1 or 2 copies of the Met allele. Finally, a recessive genetic model predicts a common response to 0 or 1 copies of the Met allele.

Neuroimaging data
The assumption of different genetic models was performed to counteract a misspecification of the true underlying genetic model, which could have an adverse effect on the statistical power of an association, and on the effect size  (Gaye and Davis 2017). The goodness-of-fit of each genetic model was evaluated based on the Akaike information criterion (AIC), for which lower numerical values indicate a better fit of the model (Akaike 1998).
We also investigated whether the association between BDNF Val66Met and hippocampal subfield volumes was modified by the number of APOE-ε4 alleles, with a second model that included an interaction term between BDNF Val66Met polymorphism and the number of ε4 alleles, covarying for age, sex, years of education and total intracranial volume potential confounders. In this model, dominant genetic effects were assumed for Val66Met polymorphism and additive genetic effects for APOE-ε4 alleles.
Moreover, in post-hoc analyses, we evaluated the effects of BDNF Val66Met and APOE-ε4 status on cognitive performance.
Statistical significance was set at False Discovery Rate (FDR) corrected p value < 0.05, and all statistical analyses and data visualization were carried out using R version 3.4.4.

Demographic characteristics
Descriptive data of the demographic and BDNF Val66Met polymorphism information are presented in Tables 1 and 2. The mean age of the population was 57.  Table 3 summarizes the hippocampal subfield volumes analyzed in the study by BDNF genotype. All morphometric subfield measures were normally distributed (Kolmogorov Smirnov test, FDR > 0.05) and their variances were homogenous (Levene's test, FDR > 0.05). Figure  S1 shows the pattern of correlation (Pearson correlation statistics) among all subfields included in the study. Hippocampal subfield structures present high correlation among them (r > 0.8) (i.e., structural covariance).

Effect of BDNF Val66Met polymorphism on hippocampal subfields
General linear models revealed that Met carriers showed statistically significant larger bilateral volumes of the subiculum under the dominant model ( dom = 2.53%, p FDRdom = 3 × 10 −3 ) ( Table 4 and Fig. 3). For subiculum subfield, an additive genetic model obtained the lowest AICs score (AIC = 4890.43), indicating that this model is the most parsimonious model for this subfield structure. Moreover, we found statistically significant larger bilateral volumes of the subiculum under the additive genetic model ( add = 2.39%, p FDR add = 0.013) (Table S1). No significant results after FDR-correction were found under recessive and codominant genetic models. In addition, nominal significant results without FDR adjustment (p < 0.05) showed larger bilateral volumes of the molecular layer of the hippocampus (β: 1.55%, p: 0.007), presubiculum (β: 1.74%, p: 0.041), and whole hippocampal volume (β: 1.46%, p: 0.025) for Met carriers under the dominant genetic model. Results of all adjusted genetic models for each HSv can be found in Table S1.

Effect of the interaction between APOE-ε4 and BDNF Val66Met on hippocampal subfields
As expected, APOE ε4 allele was associated with lower bilateral volumes of the hippocampal subfields, even though on a trend-level (Table S2). Interestingly, when this association was studied according to BDNF Val66Met genotypes,  (Fig. 4). Moreover, even though the results for the remainder hippocampal subfields were statistically not significant, changes on hippocampal subfields volumes follow the same general patterns ( Figure S2).

Post-hoc analyses: Effect of BDNF Val66Met and APOE-ε4 on cognitive performance
The post hoc analyses, although not significant, suggested better cognitive performance patterns for Met carriers in most FCSRT domains, and in depression scores of the Hospital Anxiety and Depression Scale (HADS) ( Table 6). In addition, APOE-ε4 status did not significantly influence the effects of BDNF Val66Met genotypes on cognitive performance.

Discussion
To the best of our knowledge, this is the first study to show a phenotypic effect of the BDNF Val66Met polymorphism in the hippocampal subfields of cognitively unimpaired (CU) individuals. Moreover, the present study is also the first in CU to find an effect modification by BDNF Val66Met polymorphism of associations between APOE-ε4 status and hippocampal subfield volumes. We first found significantly larger bilateral subiculum volumes in CU middle-aged/late-middle-aged BDNF Val66Met carriers in a dose-dependent manner. The direction of the effects is consistent across different subfields and the entire hippocampal formation, as shown by the nominally significant difference between BDNF Val66Met carriers and noncarriers involving the molecular layer of the hippocampus, as well as the pre-subiculum and the whole hippocampus.
Given that BDNF Val66Met polymorphism has been related to impaired hippocampal long-term potentiation which underlies learning and memory (Spriggs et al. 2018), our results may underline compensatory mechanisms in the Met-carriers to achieve normative episodic recall, which is highly specialized in the subiculum (Eldridge et al. 2005;Suthana et al. 2015). However, although most studies showed that large hippocampal volumes lead to better memory performance and may protect from dementia (Pohlack et al. 2014;Whitwell 2010;Erten-Lyons et al. 2009), the impact of hippocampal volume on  2004). Thus, our results could suggest a moderating role of BDNF in the neurobiology of hippocampal subfields, which may stress the importance to consider the hippocampal formation at the subfield level to disentangle potential opposite effects leading to the aforementioned conflicting results. In addition, we cannot rule out that the BDNF Val66Met polymorphism may differentially influence the morphology of other brain areas. This calls for additional whole-brain voxel-wise studies addressing distinct genetic models of penetrance of BDNF Val66Met. Second, we also observed that Met-carriers compensate for the deleterious impact of the number of APOE-ε4 alleles on hippocampal subfield volumes. As expected, we observed that APOE-ε4 homozygotes showed a tendency towards displaying reduced volumes of the subiculum and hippocampal tails, in accordance with previous reports (Kerchner et al. 2014;Cacciaglia et al. 2018a;Pievani et al. 2011). These individuals are at increased higher risk (× 15) to develop AD as compared to APOE-ε4 non-carriers. The lower hippocampal volumes in APOE-ε4 carriers are often interpreted as brain marker that confers vulnerability towards developing the clinical picture of AD. Strikingly, we found that APOE-ε4 homozygotes who were also Met-carriers countered the effect of the APOE genotype and presented HSv within the ranges expected for APOE-ε4 non-carriers, particularly in the CA4, GC-ML-DG and the hippocampal tails. It could be argued that Met-carriers can counter the deleterious effect of the APOE-ε4 genotype in the age range of the studied sample.
Another possible explanation to this finding could raise from an interaction of the BDNF Val66Met polymorphism with pathological markers of AD, as the APOE-ε4 allele has also been strongly linked to a dose-dependent increase in the prevalence of abnormally elevated cerebral amyloid deposition in CU individuals (Reiman et al. 2009). By the mean age of our APOE-ε4 homozygote group (56.62 ± 5.71y), about half of them are expected to display abnormally high amyloid levels (Jansen et al. 2015). However, previous longitudinal reports on amyloid-positive CU individuals have described that the BDNF Val66Met allele was associated with a steeper decline in cognitive function and hippocampal atrophy (Yen Ying Lim et al. 2013). Moreover, this deleterious effect is more severe in APOE-ε4 carriers (Lim et al. 2015). These studies, however, were performed in significantly older cohorts (average age of 70y) than that in our work. Another recent study performed in subjects with an age range similar to ours (55y) confirmed this longitudinal pattern of decline in cognition, particularly in amyloid-positive CU individuals (Boots et al. 2017). However, in this work, Boots et al. also reported that, at baseline, Met carriers showed a significantly better cognitive performance (verbal learning and memory, speed and flexibility, and working memory), even if amyloid positive. Similar patterns of effect were observed in our sample. Specifically, we found that Met carriers suggested patterns of better cognitive performance on the Free and Cued Selective Reminding Test (FCSRT) domains, and on the Depression score of the Hospital Table 4 Main effects of Val66Met genotype on hippocampal subfields (mm 3 ) All models were adjusted by sex, years of education, number of APOE-ε4 allele and total intracranial volume CA1 cornu ammonis region 1, CA3 cornu ammonis region 23, CA4 cornu ammonis region 4, GC-ML-DG granule cells in the molecular layer of the dentate gyrus, hata hippocampal-amygdaloid transition region, HP hippocampus, CI95 confidence interval, FDR95 false discovery rate corrected p value < 0.05. AIC akaike information criterion Anxiety and Depression Scale (HADS). Although these differences were not statistically significant, one potential explanation to reconcile our findings with the existing literature would be that the Met genotype might provide a limited beneficial effect during middle age. When no longer capable of compensating for the deleterious downstream effects of amyloid accumulation, then Met carriers would experience faster hippocampal atrophy and a steeper decline in cognition. Neither shall it be excluded that APOE-ε4 and Met carriers could be the most vulnerable to an inflammatory response at the beginning of the amyloid-pathology. Nevertheless, the current unavailability of core AD biomarkers and the cross-sectional nature of this study constitute a limitation for the full interpretation of the interaction between the BDNF Met and APOE-ε4 genotypes. Nevertheless, this will be mitigated in the longitudinal follow-up of the cohort here studied, as a subset of our participants will undergo a lumbar puncture to assess cerebrospinal fluid levels of core AD biomarkers (Aβ42, total Tau, and phosphorylated Tau).
A strong feature of our study that may sustain our ability to detect a significant effect of Val66Met genotype (and its interaction with APOE) on hippocampal subfields is that the studied cohort presents a higher prevalence of BDNF Met and APOE-ε4 homozygotes compared with previous studies. While most of the studies reported allele frequencies between 0% (studies without MetMet carriers) to 6% (Harrisberger et al. 2015), the minor allele frequency in our study achieves 23%, which is even higher to the population frequency (15-19%, Source: Genome Aggregation Database https ://gnoma d.broad insti tute.org/varia nt/11-27679 916-C-T). Similarly, the high number of APOE-ε4 carriers in the ALFA participants compared to the general population (19% vs. 14%, respectively; p < 0.001) has allowed us to *Significant difference between groups at a nominal level (p < 0.05). ***Significant difference between groups after multiple comparison correction (FDR < 0.05). CA1 cornu ammonis region 1, CA3 cornu ammonis region 3, CA4 cornu ammonis region 4, GC-ML-DG granule cells in the molecular layer of the dentate gyrus, hata hippocampal-amygdaloid transition region, HP hippocampus disentangle specific effects in ε4 homozygotes, while most studies pool them with heterozygotes in a single APOE-ε4 carrier group. In our study, the high prevalence of these less frequent genotypes has allowed us to achieve a relatively higher inferential power, allowing for testing additive, recessive and codominant genetic effects, as well as, gene-gene interactions.
Another substantial strength is the use of a high-resolution T1 scan as compared to previous studies, combined with the use of the most recent version (v.6.0) of the hippocampal subfield segmentation toolbox in Freesurfer, which overcomes significant shortcomings of previous versions (Iglesias et al. 2015;Wisse et al. 2014;Mueller et al. 2018). Thus, subfield volumes available for this analysis are of substantially better quality, which, combined with a considerably higher sample size, has allowed us to achieve a significantly superior statistical power than previous reports. Moreover, only a few studies have assessed hippocampal subfield volumes and compared them to whole hippocampal volumetry, even though the independent genetic variation specific to hippocampal subfields (Elman et al. 2019). Thus, our analyses based on hippocampal subfields increased the sensitivity of the results, which make our study more robust and consistent than previous ones.
Finally, in contrast to previous studies in which the diagnostic value of hippocampal subfield volumes related to Val66Met polymorphism was only assessed by comparing patients with psychiatric disorders, our study includes CU middle-aged/late-middle-age participants. This is also a relevant strength because the misrepresentation of the general population could constitute a bias in the assessment of the diagnostic utility of hippocampal subfield volumes, due to potential etiologies associated with neurodegenerative processes (de Flores et al. 2015).
Altogether, our findings suggest that the BDNF Met allele might confer a time-limited resilience, which protects the  hippocampi from the downstream deleterious effects of ageing and/or amyloid accumulation, thus mediating the risk effect of APOE-ε4. Hence, these results prompt us to further explore hippocampal atrophy rates and cognitive trajectories of BDNF Met carriers compared to Val homozygotes, also as a function of their APOE genotype and long-term accumulation of amyloid beta.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.