The Effect of Depression and Neurotrophic Factors On

The Effect of Depression and Neurotrophic Factors on

Hippocampal Subfield Volumes

A study in gene-environment interactions

Hallvard Heiberg

Thesis for the Master’s degree (MSc) in Physiology 60 ECTS credits

Institute of Biosciences The Faculty of Mathematics and Natural Sciences

UNIVERSITY OF OSLO

September 2017

Acknowledgements

The study in this thesis was conducted at the Section for Physiology and Cell Biology, Department of Biosciences, Faculty of Mathematics and Natural Sciences, University of Oslo.

First, I would like to thank all the study participants, without which none of us would be able to do any of this work.

I would like to thank Göran Nilsson for the opportunity to do the work for my master’s thesis in his group. I would also like to thank my supervisors Øyvind Øverli, Cathrine Fagernes and Ida Beitnes Johansen for their contributions to the various aspects of this work. Special thanks are also extended to Rune Jonassen for his guidance outside of what was formally required of him.

To everyone in the Clinical Neuroscience Research Group at the Department of Psychology, University of Oslo, thank you for allowing me insight into your data and for providing a foundation for this project. Thank you Rune Jonassen and Nils Inge Landrø for drafting a project in which I was able to take part. Dani Beck, Luigi Maglanoc and Eva Hilland all deserve thanks for bearing the main weight of the data collection and production for the project, as well as helping me learn how to work with the FreeSurfer software. Thanks should also be directed at the people running the imaging equipment at the Intervention Center at Oslo University Hospital.

I would like to thank my co-student Kristine Rønning for helping me in the lab and for providing the cortisol data which sadly were never used in this thesis. I would also like to thank Kristine for being a therapeutic influence during times of project-related distress.

Finally, I would like to thank my family at home, namely Line and Tuva, who have had patience with me during a period of hours spent writing and great deal of frustration.

Bærum, September 2017

Hallvard Heiberg

iii

Abstract

It has long been recognized that in order to understand the nature of mental processes, one must look at the brain which presumably lies at their origin. Mapping out relationship between the structure of the brain, the genetics of the individual, and their life history, is now thought to be essential to understand the development and state of the psyche.

The systems of BDNF and serotonin have been shown both separately and in interactions to affect the development and individual predisposition to depression. In addition, researchers have demonstrated that altered signaling in pathways involving these molecules may affect the growth of the hippocampus or its subfields. This coincides with the observation that patients suffering from depression will also many times display lower hippocampal volumes than would be expected in healthy people. In this study, a novel segmentation protocol from the FreeSurfer software package has been used to assess the effect of a history of depression as well as selected candidate genes on the volumes of hippocampal subfields. Subjects with a history of clinically significant depression, as well as healthy controls, were brain-imaged by MRI and genotyped for polymorphic regions of the serotonin transporter- and BDNF- genes. Using FreeSurfer v6.0, the brain MRI-images were reconstructed and the hippocampi were segmented by the automatic segmentation protocol based on new ex-vivo data. Decreased volumes in depressed subjects were found for the granule cell layer of the dentate gyrus (left), the parasubiculum (left), and the molecular layer of the subiculum (left). The fimbrial volume was found to have increased bilaterally. There was no effect of the candidate genes on the hippocampal subfields or the whole hippocampi, and no evidence of the candidate genes predicting a past history of depressive illness was found. The BDNF genotypes displayed non- Hardy-Weinberg characteristics leading to speculation about whether certain variants of the gene confer risk or advantage with respect to reproductive success.

These results indicate that there might be characteristic changes the hippocampus that underlie, or result from, depressive illness. Mapping these volumetric differences out in individuals who suffer from the disease may eventually guide clinicians in choosing the correct therapy for this debilitating category of illness.

List of abbreviations

5-HT 5-Hydroxytryptamine (serotonin)

5-HTT 5-Hydroxytryptamine Transporter

5-HTTLPR Polymorphic region of 5-HTT gene

BDNF Brain-Derived Neurotrophic Factor

CA1-4 Cornu Ammonis 1-4

CNS Central Nervous System

DG Dentate Gyrus

DG GC Granule Cell layer of Dentate Gyrus

HATA Hippocampus Amygdala Transition Area

HPA axis Hypothalamus-Pituitary-Adrenal-axis

ICV Intra-Cranial Volume

MAO Monoamine Oxidase

MAOI Monoamine Oxidase Inhibitor

MDD Major Depressive Disorder

ML Molecular Layer

MRI nuclear Magnetic Resonance Imaging fMRI functional nuclear Magnetic Resonance Imaging

BOLD-fMRI Blood-Oxygen-Level-Dependent fMRI

NGF Nerve Growth Factor

PCR Polymerase Chain Reaction

PET Positron Emission Tomography

RN Raphe Nuclei rs6265 Val66(G) -> Met66(A) SNP mutation in BDNF prodomain

SERT Serotonin reuptake transporter

SSRI Selective Serotonin Reuptake Inhibitor

TAE Tris base, Acetic acid, EDTA

TrkB Tyrosine receptor kinase B (BDNF receptor)

Contents 1. Introduction ...... 1 Context of study ...... 2 1.1 Structural imaging of the living brain ...... 3 1.1.1 Briefly on current methods ...... 3 1.1.2 Magnetic Resonance Imaging ...... 3 1.1.3 FreeSurfer ...... 4 1.2 The hippocampus ...... 5 1.2.1 Hippocampal volumetrics in depression ...... 6 1.2.2 Imaging and segmenting the hippocampus ...... 6 1.3 Serotonin ...... 8 1.3.1 SSRIs and the serotonin transporter ...... 8 1.3.2 The 5-HTT Long Polymorphic Region ...... 9 1.3.3 Serotonin and hippocampal neurogenesis ...... 9 1.4 Brain-Derived Neurotrophic Factor ...... 10 1.4.1 A functional polymorphism in the BDNF gene correlates with depression ...... 10 1.4.2 The BDNF and 5-HT systems interact ...... 11 2. Aims of study ...... 12 2.1 Hypotheses ...... 12 3. Materials and methods ...... 12 3.1 Recruitment of subjects ...... 12 3.2 DNA acquisition and isolation ...... 14 3.3 PCR of BDNF ...... 14 3.4 Genotyping of BDNF ...... 15 3.5 5-HTTLPR analysis ...... 17 3.6 MRI acquisition ...... 18 3.7 FreeSurfer reconstruction and hippocampal subfield segmentation ...... 18 3.8 Statistical calculations ...... 20 4. Results ...... 21 4.1 Preliminary statistics ...... 21

4.2 Population genetics: 5-HTTLPR/BDNF(Val66Met) and depression ...... 23 4.3 Effects of depression history on subfield volumes ...... 25 4.3.1 Group and gender effects on total hippocampal volume ...... 25 4.4 Biological metrics effects on subfield volumes ...... 31 4.4.1 Sub-hypotheses 1 and 2 ...... 31 5. Discussion ...... 31 5.1 Hippocampal asymmetry...... 32 5.2 Distribution of genotypes and HW disequilibrium ...... 32 5.3 Power of BDNF and SERT genotypes in predicting depression history ...... 33 5.4 Predictive power of depression history on hippocampal subfield volumes ...... 34 5.5 Predictive power of biological metrics on subfield volumes...... 35 6. Conclusions and future perspectives ...... 35 References ...... 36

vii

viii

1. Introduction

“Any man could, if he were so inclined, be the sculptor of his own brain.” ― Santiago Ramón y Cajal

A well-known principle in biology, and indeed in other disciplines such as physics and chemistry, is that structure drives function. From the folding of the intestines and lungs, to the intense branching of the circulatory system, to the most recent acquisition of the heavily convoluted and interconnected brain, structure is intrinsically bound to function. This primacy of structure becomes all the more evident when its integrity is lost, and the consequences present themselves in the form of disease. And so, a natural implication of this important principle is that assessing form will often provide one with clues to make an educated guess at function which can drive experimentation. The opposite must also be true; someone with knowledge of biology should by observing a function (or the loss thereof) be able to make predictions about what structural underpinnings may be its foundation (or have been lost).

In the field of neuroscience, many examples of the importance of structure can be found, both when studying function and dysfunction. On the functional side, many of the properties of different neuronal populations can be attributed to such structural properties; they can vary in their degree of arborization, their tendency to form dendritic spines, or their reach into remote tissues. The isolating properties of myelin are largely structural, and they are fundamental to creating a nervous system that is both capable of rapid action and compaction. On the other hand, loss of structure such as the tissue destruction apparent in Alzheimer’s disease is characteristic and surely integral to the symptoms that we observe. Nervous system damage from trauma also shows us that when a structure loses components, there are often times a decline in the function that it is supposed to serve.

Figure 1 & 2. Left: Macrostructural differences between a healthy brain and a brain severely damaged by Alzheimer’s Disease. The structural difference is conspicuous. (Wikimedia commons license). Right: A parcellated reconstruction of the cerebral cortex of a study participant, produced in FreeSurfer (http://freesurfer.net/.).

Neuroplasticity is another essential principle in neuroscience, and it is intimately linked to structure. Much of what we believe accounts for neuroplasticity is found in structural changes, such as the growth or decay of functionally important processes of the cell body, or indeed the genesis and death of whole groups of cells. These changes and variations in the brain become visible to us through imaging methods, and one such imaging technique is showcased in the current thesis. Presented in this work is also a review of literature pointing to structural plasticity as being important for affective disorders, particularly depression.

A major challenge in human neuroscience is to acquire data about brain structure and function on the level we are interested in, while at the same time minimizing the invasiveness and potential for damage in the individual we study. In animals, although managed by ethical rigor, we are presently able to perform methods which are too invasive for human studies. Animal work has led to landmark discoveries such as LTP1 in the hippocampus of rabbits, and the discoveries of Hubel and Wiesel2 in the visual cortex of cats. Both discoveries have made huge impacts in the field of neuroscience, and to some extent even affected clinical practice. This is a very important aspect of neuroscience, and our knowledge would not be what it is without these bodies of work. However, the animal literature can for the most part provide hints and clues towards human functioning, but the only way to ensure human relevance is to study the phenomenon of interest in humans. The present thesis outlines one such approach.

Context of study

The data in this thesis are based on two clinical projects being conducted by the Clinical Neuroscience Research Group at the Department of Psychology, University of Oslo (Clin.gov ID: NCT0265862, Clin.gov ID: NCT02931487). Both projects study depression but are split into a morphometric project and an intervention-based project. The purpose of the intervention project is to test an emerging method of attention training. This cognitive method is based on an attention bias towards negative stimuli observed in many people suffering from depression and other emotional illnesses3,4. The two related projects seek to unify many different types of data into a complex and multi-faceted model, including the biological data produced and discussed in this thesis. The following section seeks to describe the backdrop of the dataset as this is important for interpretation of biological and demographical data.

Major Depressive Disorder is a heterogeneous clinical syndrome which typically affects mood and motivation in its sufferers in a negative manner. It may, or may not, include a number of other less specific symptoms such as loss of appetite, trouble sleeping, feelings of guilt and hopelessness, or indeed the very opposites5,6. A robust observation in MDD is the presence of a cognitive bias towards negative information and stimuli, shown clinically in a dot probe test which gauges response times

when facing stimuli of different emotional salience3,4,7, see also review by Disner et al.8. Unifying depressive states with biological models of mind is a great challenge, but many discoveries have been made that present possible paths forward.

Several lines of evidence suggest brain structural changes associated with depression, but the evidence is also conflicting9,10 (see 1.2.1). Central to this thesis is the 2015 update of the Hippocampal Subfield Segmentation protocol of the popular FreeSurfer software package developed by Laboratory for Computational Neuroimaging at the Athinoula A. Martinos Center for Biomedical Imaging at Massachusetts General Hospital http://freesurfer.net/. The 2015 update is based on new ex-vivo data, and allows for more precise segmentation of the hippocampus11. Due to the nature of the clinical projects upon which this thesis is built, namely the study of depression, I attempt to contextualize the biology against this backdrop of depression research. Questions about depression are secondary to this thesis, but they are asked in order to connect an affective illness to its putative biological foundations.

1.1 Structural imaging of the living brain

1.1.1 Briefly on current methods

Imaging the living brain in any animal is a challenge, and the difficulties are much larger in humans due to the risk and invasiveness of the most powerful methods. Still, the history of neuroimaging in humans dates back some 130 years to the earliest manuscripts of Angelo Mosso12 who became noted for observing pulsations on the brain surface of skull-breach patients. Although this is not a viable technique in routine imaging of humans today, it set a precedent for methods that are in use. Observations of metabolic indicators, such as blood flow, has been carried over to methods such as BOLD-fMRI and PET scans which are currently in use. Intrinsic imaging is an animal parallel which utilizes red light reflection to gauge activity of a brain region. Animal research allows for such permanent surgical interventions as craniotomies and mounting of viewing windows in the skull. Along with techniques of molecular biology, such as transfection and vector injection, many techniques become available in animals. For instance, two-photon calcium imaging can be used to visualize calcium fluxes in cells in deeper layers than what can be seen with conventional imaging techniques13. Imaging with classical fluorophores such as Green Fluorescent Protein (GFP) allows for great structural resolution in shallow regions of the living brain (such as the very outer layers of a rat cortex). Since these methods depend on craniotomies, transgenics, transfection, local injections, or a combination of invasive methods, they are not suited for living humans. For obvious reasons, research-based imaging of the human brain must leave the subject without trauma forcing us to turn to other methods. 1.1.2 Magnetic Resonance Imaging

In structural imaging of living human brains today, Nuclear Magnetic Resonance Imaging (MRI) is by far the most common. MRI utilizes the fact that protons align with magnetic fields based on the quantum property called spin. This aligning happens in one of two possible directions, depending on the spin of the proton. Each proton will align in the most energetically favorable way, and the protons in the field tend to be unevenly distributed between the two alignments. By using electromagnetic

stimulation at a specific frequency which depends on the strength of the magnetic field at a given position, all protons are given enough energy to be forced to align the same way for as long as they are excited by the energy. When the radio pulse ceases, the protons which are in an unfavorable alignment with the magnetic field will fall back to the favorable alignment, and in the process emit photons which form the basic signal for the MRI measurement. The number of protons within a given area of inspection determines the strength of the signal. Complex mathematical transformation and repeated measurements gives rise to sectional images composed of three-dimensional pixels called voxels. Hundreds of such sections together form a complete image of the brain, at spatial resolutions which are very high 14. The strength of the signal in any one voxel is determined by, among other things, the amount of water and fat within the area due to their characteristic proton densities. From this map of varying signal intensities emerge the manual and automatic methods of assessing brain structural integrity or the lack thereof. By manipulating the weighting of signal in the MRI scanner, different contrasts can be produced based on the purpose of the scan. MRI is a widely used clinical tool, not only for brain imaging but also for imaging of other soft tissue types. In clinical brain imaging, MRI can be used to show loss of grey or white matter, ventricular enlargement, signs of ischemia, abnormal cell clusters and tumors, and other structural properties of clinical significance. Due to its low degree of invasiveness, low potential for damage and high spatial resolution, it is among the most powerful methods for measuring the human brain both in the clinic and in research.

1.1.3 FreeSurfer

Figure 3. The FreeSurfer pipeline leading eventually to qualitative and quantitative output at the final stage. Adapted from http://freesurfer.net/

FreeSurfer (http://freesurfer.net/) is an open source software suite developed by the Laboratory for Computational Neuroimaging at the Athinoula A. Martinos Center for Biomedical imaging. The latest stable release was on January 27 2017. FreeSurfer contains a set of tools which performs a series of

complex mathematical and statistical operations as it runs an MRI image through a processing pipeline. The output from this reconstruction is a file containing information about the imaged brain which can be viewed in several interesting ways. The included software FreeView allows for visual inspection embellished with color maps of structures such as the amygdala, the ventricles, the subventricular zone, the hippocampus, the striatum and many other areas. In addition, it provides clear boundaries between gray and white matter, and allows for the calculation of such metrics as the thickness of the cerebral cortex (inner white matter to pial surface). Furthermore, the software can make three-dimensional reconstructions of the cortex which clearly show individual sulci and gyri and provides both a qualitative map of the brain surface as well as being an excellent educational tool. With the latest release of FreeSurfer (v6.0), a new atlas for segmenting the hippocampus has been developed11. This is discussed further in section 1.2.3. 2017-results for the keyword “Freesurfer” on https://www.ncbi.nlm.nih.gov/pubmed yields 191 results, showing that there is a body of literature published which utilizes this tool, yet it has important capabilities outside of hippocampus segmentation.

1.2 The hippocampus

Figure 4. Semitransparent model of human brain, the hippocampi are colored in red. a) Saggital view b) Rostral/Coronal view. Adapted under Wikimedia Commons License.

The hippocampus is an allocortical brain structure located underneath the medial temporal lobe of the brain. It is shaped somewhat like a seahorse (genus hippocampus) and has been named accordingly. Mammals have two hippocampi, one located on each side of the brain, and they are considered part of an anatomical division of the brain called the limbic system. The limbic system is traditionally considered to deal with emotion, behavior, cognition and other higher functions present in humans but its exact functional boundaries are not obvious15. In the field of functional and anatomical neurobiology, the hippocampus has received more focus than most other regions of the brain. Findings range from the discovery of place cells by John O’Keefe16,17 to the discovery LTP by Tim Bliss and Terje Lømo1 and eventually the finding of grid cells18,19. One recurring theme throughout hippocampus research is the association with memory and learning, as well as its connection to mood and behavior. Perhaps the most cited clinical event in all of neuroscience is the story of patient Henry Molaison’s bilateral hippocampus removal (about 2/3 of each hippocampus) due to epilepsy and his subsequent inability to form new memories. The story serves as a testament to the memory forming importance

of the hippocampus20, but also provides interesting clues to the specific role of hippocampus in memory processes. Patient HM supposedly could recall events from before his surgery, and was able to learn physical skills such as playing tennis. He was, however, completely incapable of learning new names and faces for time spans longer than minutes or hours20. This hints at complex differences between short, and long-term memory, and the hippocampus’ role in these as well as storage and recall20. Many studies have shown specific increases in hippocampal activity during memory tasks using fMRI21,22. In addition to such cases as patient HM, a large body of evidence that points to the hippocampus as integral to memory formation is found in the animal literature. Classic experiments by John O’Keefe pointed out the hippocampus as a structure of spatial coding and mapmaking16, and particularly interesting were the place cells which displayed firing patterns that directly related to the position of the animal17. Replay of spatial memories during both sleep and wakefulness have been shown, and Karlsson & Frank elegantly showed that there is great spatial coherence in the order of place cell firing during replay23. Using a powerful method called optogenetics to silence cells in the CA1 region in a contextual manner, Goshen et al.24 showed that fear memory acquisition and recall were both disrupted during inhibition of CA1 cells in mice. Despite these indications at the hippocampus’ importance in many forms of memories, the details of its contribution to memory formation and recall are still unclear, see Lee et al.25 for review. 1.2.1 Imaging and segmenting the hippocampus

The current method for measuring in-vivo hippocampal volumes is by using structural MRI images and either a manual or an automatic segmentation protocol. These methods are used to determine which pixels (voxels) in an image belong to the hippocampus, and both volumes and subfield segmentation can be performed according to anatomical atlases of the hippocampus. An article by Tae et al.26 from 2008 compares manual segmentation and morphometry to that of the previous version of FreeSurfer, a popular tool for volumetrics. They found that a good overall agreement between manual and automated volumetric methods, yet the software estimated the total volume to be significantly larger. Fortunately, FreeSurfer and other similar softwares have been updated since 2008, and in 2015 Iglesias et al.11 (developers of FreeSurfer) developed a new atlas from ex-vivo data, showing more consistent results and more precise subfield segmentation. Two very important strengths of the automated methods is the lack of bias, and consistency across research projects; even if the automated methods are not entirely in agreement with manual ones, this helps eliminate inter-researcher inconsistencies. Subfield segmentation is currently subject to some controversy with regards to accuracy and generalizability, in particular for automated methods (see review by DeFlores27). One must also bear in mind that manual morphometrics of the hippocampus is very demanding on two important resources in research, namely time and skilled labor. Despite limitations, this is the level of detail about morphology and volumetrics that we can currently acquire by automatic methods in living humans.

1.2.2 Hippocampal volumetrics in depression

At the current level of knowledge about the biology of depression, there is little reason to believe that the plethora of possible symptoms stems from a single brain region. On the contrary, it seems more probable that several regions are involved and that different patients may or may not have overlapping alterations in brain functions. Still, the limbic system is a natural place to direct focus, and the hippocampus and amygdala have both been studied in this context. The amygdala will not be discussed

in any detail in this thesis, but the hippocampus is a focal point as well as the primary dependent variable for the statistical investigations.

Being a region of high plasticity and the most well-known example of a brain structure with adult neurogenesis28-33 (in the dentate gyrus), it is not unreasonable to expect the hippocampus or some of its sub regions to undergo structural and volumetric changes throughout the life of an individual. There are many studies that point to hippocampal volume differences between individuals which cannot simply be ascribed to random variation34,35. Han et al. 36 found reduced volumes of several hippocampal subfields in medication-naïve, depressed female patients when comparing to healthy controls. A meta- analysis by Cole et al. found a 4.0-4.5% average reduction in hippocampal volumes in patients with first period depression compared to healthy controls. Findings by Chen et al. indicate that groups considered to be predisposed yet without having developed the clinical disorder also have lowered hippocampal volumes bilaterally37. MacMaster et al.38 found reduced hippocampal volumes in both hippocampi in pediatric patients suffering from familial MDD compared to controls, yet no differences in amygdala size were found. Colle et al.39 found that Early Life Adversity (ELA) and MDD correlated with reduced hippocampal volumes in male in-patients. Some studies have investigated hippocampus volume in relation to therapeutic response. For instance, a study by MacQueen et al.40 found a positive correlation between posterior hippocampal volumes and remission from MDD during an 8-week period. Hayasaka et al. found increased hippocampal volume after 2-weeks of lateral repetitive Transcranial Magnetic Stimulation (rTMS) treatment in MDD patients, but only on the side stimulated10. The dentate gyrus shows promise as a region of volumetric plasticity with relation to depression due to a number of studies showing effects on neurogenesis and volume31-33. These examples point to the hippocampus as a region that may undergo volumetric differences and depression appears to be one of factors that can affect this variation.

There are inconsistencies, however, which underline the need for further research. Several studies have failed to find any relationship between hippocampal volumes and depression41-43, and the effects in the previously cited research articles tend not to be very strong (a meta-analysis by Cole et al. reports 4-4.5%44). Take note, that small effect size is to be expected in any system which is influenced by many different factors and the size of effects must be compared to what is normal in the field of study. Still, it warrants caution when interpreting results when studying depression and volumetrics in general as the actual cause of variation in hippocampal volume is very hard to pinpoint. In this thesis, I attempt to use information about some of the most studied genes in depression to provide an additional angle at what causes volumetric differences in the hippocampus.

1.3 Serotonin

Figure 5. The biosynthesis of serotonin. The essential amino acid tryptophan goes through an oxidation step to 5-hydroxy-tryptophan prior to decarboxylation forming serotonin. A number of further syntheses are based on this pathway. Wikimedia Commons License.

Serotonin (5-hydroxytryptamine) is a tryptophan-derived, small-molecule, monoamine neurotransmitter found mainly in the CNS, smooth muscle, blood platelets and the enteric nervous system45,46. The transmitter molecule is involved in a host of mechanisms ranging from aggression and feeding behavior to hemostasis, smooth muscle tone and consciousness (citations!). In the CNS, a cluster of brainstem nuclei called the raphe nuclei serve as the main origins of brain-wide serotonergic projections across animal groups 46-49. It is commonly accepted that fourteen receptors exist in seven 50 different families named 5-HTT1-7 , most of which are metabotropic receptors. This thesis does not discuss the role of receptors in depression or hippocampal morphology, but the reader should know of the complexity and ubiquity of the serotonin system and appreciate the potential for both system- wide and local effects as a result of serotonin perturbation. There is also a literature on the neurotrophic effects of serotonin, and its role in brain structural plasticity in depression. The following sections review the findings that have made serotonin a popular topic in depression literature.

1.3.1 SSRIs and the serotonin transporter

The serotonin transporter (SERT) is a membrane-bound protein found in the presynaptic neuron and surrounding astrocytes. It is a sodium-dependent cotransporter that pumps serotonin from the synaptic cleft and back into the presynaptic neuron. This regulates the strength and duration of serotonin signaling through the management of synaptic concentrations of serotonin. The fate of serotonin molecules is either oxidation in mitochondria by Monoamine Oxidase A, or further pumping back into secretory vesicles51. Selective Serotonin Reuptake Inhibitors block the action of the serotonin transporter52,53, and are typically the first line of defense in drug-based interventions, examples are Prozac and Citalopram. However, the efficacy of these drugs is not clear cut54,55 and controversy remains in part due to severe side effects such as suicidal ideation seen in some, particularly young, patients56,57. An interesting and controversial point in the SSRI debate is the fact that SSRI interventions tend to produce their results after 2-4 weeks58, although their inhibitory action sets in within hours53. This indicates that the acute synaptic levels of serotonin alone do not account for the mood of the patient, but that some process of adaptation must take place. The proposed mechanism of action,

which also accounts for the latency of onset, is a downregulation of 5HT1A inhibitory autoreceptors on postsynaptic neurons due to the increase in synaptic serotonin59,60. Downregulation of this receptor causes a decrease in inhibitory action and thus more electrical activity in secondary serotonergic neurons, which in turn is believed to account for the therapeutic effects of SSRIs and the latency of therapeutic response60,61. See also section 1.3.3 for discussion of how the delayed onset of SSRI therapeutic effects may be mediated by more long-term structural processes. Although SSRI treatments are in many cases ineffective, they are not the silver bullet that they may appear, and more powerful tools for specifying diagnoses will perhaps allow future clinicians to better know when an SSRI can be expected to be efficacious. Hieronymus et al. makes an argument for symptom-specific efficacy (mood) rather than a systemic improvement of all symptoms the patient experiences62. Despite controversies, SSRIs remain the most common and most widely accepted first-line drug interventions against common types of depression.

1.3.2 The 5-HTT Long Polymorphic Region

It is widely accepted that there is a degree of heritability to MDD63, and that gene-environment- interactions are important factors in its development64-67. Although a number of candidate genes have been investigated68-70, the serotonin transporter promoter has received the most attention. A pivotal 2003 cohort study by Caspi et al. showed a positive interaction between a genetic polymorphism and periods of life adversity with relation to depressive symptom scoring71. The cohort was followed from age 3 to age 26, and the study has set much precedence for the genetics of depression. The genetic variants described are contained in a repeating region in the promoter of the serotonin transporter gene which typically appears in a short or a long (43 bp longer) version64,72. Several sub-variants of have been also described73,74, but most studies have focused only on the difference between the long and short promoter. The short variant correlates with lower transcriptional efficiency75,76 and many studies argue for a relationship between 5-HTTLPR genotype and susceptibility to depression64,77,78 in particular as a result of life stress, rather than an absolute relationship between the genotype and developing depression in general. In the current thesis, I investigate the relationship between this polymorphism, a history of depression and hippocampal volumetry.

1.3.3 Serotonin and hippocampal neurogenesis

A role for 5-HT in brain structural processes is underpinned by a body of evidence points to a stimulatory effect of 5-HT on hippocampal neurogenesis. It has long been known that the hippocampus receives serotonergic projections79, but the implications of this anatomy were not clear. More recent evidence comes from observations of serotonin-perturbing drugs appearing to increase hippocampal neurogenesis in the dentate gyrus granule cells31,33. Moreover, Brezun and Daszuta demonstrated in rats that ablation of serotonergic projections from the RN to the hilus of the dentate gyrus has a suppressing effect on expression of neurogenesis markers. This effect was reversed by reinnervation, implying a role of serotonin in regulating the process of neurogenesis80. It has been suggested that the delay in antidepressant effect commonly seen in SSRI-treatment may be due to neurotrophic effects81,

although this remains controversial. Being one of the regions now available in the new hippocampus subfield segmentation protocol of FreeSurfer v6.0, effects of genotypes and depression history on the granule cell layer of the dentate gyrus is assessed in this thesis.

1.4 Brain-Derived Neurotrophic Factor

The Brain-Derived Neurotrophic Factor (BDNF) is a growth factor in the neurotrophin family. It shares structural similarities with the better-known Nerve Growth Factor (NGF)82, but is a distinct protein. It is widespread among animals and is found in all vertebrates83 BDNF is found all over the central nervous system as well as in the blood, heart, and periphery like most neurotrophins. It acts as a 27 kDa dimer which is loaded into vesicles in the trans-Golgi network in preparation for activity- dependent exocytosis. Tropomyosin Receptor Kinase B (TrkB) is its most well-studied target84,85, but it also binds to the Low-affinity Nerve Growth Factor Receptor (LNGFR or p75). The significance of its binding to LNGFR is currently not known, but p75 has been shown to be involved in apoptotic regulation86,87. BDNF has a range of functions in the development and maintenance of the vertebrate nervous system. Among its functions is the differentiation, growth, survival and even apoptosis of specific neuronal populations87-90 BDNF-knockout mice show high postnatal lethality and generalized deficiencies in sensory and motor divisions88

Since BDNF is involved in the development of the brain, it is not surprising that behavioral findings have also been made. Many experiments have studied the behavioral effects of BDNF, often focusing on the modulation of the response to various types of stress91-94. Infusions of BDNF and rapid-release polymers have been shown to decrease immobility in rats in a forced swim test94. Several studies have demonstrated BDNF dynamics in striatal regions (here: Nucleus Accumbens and Ventral Tegmental Area) as a modulator of the response to chronic social defeat in mice91-93. Berton et al. also demonstrated how in vivo-knockouts of BDNF locally in the VTA affects the ability of mice to develop a social-aversion behavior that is normal after repeated social defeat, and how treatment with an SSRI had similar effects93. This suggests that social aversion is a behavior that is modulated by an interaction with BDNF in brain areas important in motivation93,95 (VTA, NA). This connection to learned behavior makes BDNF a good candidate for biological models of affective disorders resulting from life adversity (such as depression and schizophrenia).

1.4.1 A functional polymorphism in the BDNF gene correlates with depression

A Single Nucleotide Polymorphism (rs6265) in the protein-coding region of the BDNF gene has been the subject of many studies since its discovery. The gene variant contains a point mutation changing a guanine to an adenine altering the codon from GUG to AUG. Upon translation, this causes the Valine at position 66 of the BDNF protein to be changed into a Methionine. Studies suggest that the physiological effect of this mutation stems from altered release dynamics of the mutant protein96,97 rather than altered levels of transcription or altered interactions with the TrkB-receptor. Chen et al.98 reported no differences in transcription levels between val/val and met/met mice, but up ~30% less activity-dependent release of BDNF. The met allele has also been positively correlated with higher

depression scores99,100. The case of the rs6265 shows its complexity when comparing development of depression to treatment and remission. Colle et al.101 found that for SSRIs the val/val genotype patients responded better, whereas the val/met genotype patients experienced greater symptom relief from SNRI/TCA treatment. This provides a compelling argument for doing further research into personalizing medicine based on biological parameters in depression. However, the previously mentioned study can be contrasted by the findings of Yan et al.102 who saw a greater response to SSRI treatment and better remission rates in val/met carriers. It should be mentioned that these findings were in an Asian demographic, and ethnogeographical effects cannot be ruled out. In addition to its link to depression in humans, anxiety models in animals also show this genetic variant as modulating98. Although not strictly related to rs6265 mutation, a separate line of evidence points to BDNF as important for the rapid-acting antidepressant effects of small doses of ketamine103-106 (2-(2-Chlorophenyl)-2- (methylamino)cyclohexanone). This implies that antidepressant effects may in some cases rely on a functional BDNF system. The Val66Met genetic variant has also been correlated with reduced extinction of fear memories107 leading to speculation about whether the BDNF polymorphism alters the ability to learn from cues that signal safety. This could help explain why some patients with depression are more prone to experience remission, as they might be less plastic when it comes to shedding the stress of past experience. Although the allele produces 3 genotypes; val/val, val/met, and met/met, the val/met heterozygous phenotype has been of particular interest, as it is much more common than the met/met homozygous phenotype. The distribution between the Val allele and the Met allele varies depending on population, but European populations typically display about 80% Val alleles with a Met/Met genotype frequency around 4%. Importantly, implications for hippocampal function96,108,109 and volume96,110-112 has been shown in many studies, pointing to decreased volumes and decreased performance in hippocampus-dependent memory tasks in carriers of the met-allele. In the BDNF literature, this polymorphism has received much attention and it is also one of the most well- known SNPs in depression literature (see review by Hosang et al.113). For controversies on the Val66Met effect on hippocampal volumes, see meta-analysis by Harrisberger et al. Note that most of the studies used 1.5T MRI imaging, and that none of the studies use the new ex-vivo atlas described by Iglesias et al.11.

1.4.2 The BDNF and 5-HT systems interact

There are indications of an interaction between the BDNF and serotonin-systems. This is perhaps not surprising in the direction of BDNF having an effect on serotonergic cells, but the opposite seems less obvious. A 1996 study by Mary J. Eaton and Scott R. Whittemore showed that not only did exposure to BDNF increase the lifetime of serotonergic cells in vitro, but grafts of serotonergic cells injected with a BDNF-expressing vector survive longer when implanted into the rat hippocampus and cortex. Russo- neustadt et al. showed that BDNF expression in the rat hippocampus increased when given an MAOI- drug114. We must interpret this finding with care, as the MAOI (once classic drugs against depression) is not selective to what monoamines are broken down. This means that dopamine and norepinephrine signaling might also have been enhanced. In another research project, the authors also tried blocking exercise- and antidepressant-induced BDNF-elevations by using specific inhibitors of serotonin receptors (5-HT1A and 5-HT2A/2C). Serotonin inhibition block antidepressant-induced BDNF increases, but not exercise-induced, pointing to subsystem interactions between 5-HT and BDNF. Mamounas et al. infused the frontoparietal cortex of rats with BDNF and observed a local sprouting of 5-HT axons. They also showed that a BDNF infusion protected 5-HT neurons from a selective toxin targeting serotonergic cells (p-chloroamphetamine)115. This indicates that an ineffective BDNF-system might

have downstream destabilizing effects on the serotonin-system, see also review by Mattson116. It is currently not a major model in the understanding of depression, but this interaction seems natural to assess in the current thesis as it uses data on the two most studied serotonin-, and BDNF-related genes with respect to depression.

2. Aims of study

The aim of this study is to use methods of molecular biology along with the FreeSurfer hippocampal subfield segmentation to investigate the relationship between a clinical history of affective disorder, two genetic factors, and the volumes of hippocampal sub regions. The regions of focus will be those previously not available in the 5.3 version of FreeSurfer (parasubiculum, HATA, Granule cell layer of the dentate gyrus and the molecular layer of the subiculum). 2.1 Hypotheses

Hypothesis 1: There will be a difference in distribution of BDNF rs6265 and/or 5-HTTLPR long/short variant across the subject groups “rMDD” and “Healthy control”. This is based on the body of literature that points to interactions between these polymorphisms, stress, and MDD.

Hypothesis 2: The volumes of both hippocampi and hippocampal subfields will be smaller in the remitted group, this is in line with previous research38-40,117. In accordance with existing findings about subfields, the main effects are expected to be found in the CA1-CA4 (CA1, CA2/3, CA4 containing the polymorphic layer and molecular layer of DG) fields and the granule cell layer of the Dentate Gyrus31- 33,36,118.

Hypothesis 3: There will be differences in hippocampal subfield volumes across the biological metrics BDNF genotype, gender and age. This approach is exploratory and conceptually driven. The following specific sub-hypotheses are being investigated:

• Sub-hypothesis 1: BDNF genotype alone will produce an effect on subfield volumes • Sub-hypothesis 2: SERT genotype alone will produce an effect on subfield volumes • Sub-hypothesis 3: Interaction effects between the genotypes may produce additional volumetric effects.

3. Materials and methods

3.1 Recruitment of subjects

Research subjects were recruited as part of two research projects conducted by the Clinical Neuroscience Research Group at the Department of Psychology, University of Oslo (Clin.gov ID:

NCT0265862, Clin.gov ID: NCT02931487). A total of four hundred and forty-three subjects were recruited and went through genotyping. The clinical design warranted a recruitment aimed at roughly 70% women and 30% men. For the sake of this thesis, initial exclusion was made by removing all subjects that did not go through MRI-imaging. Additionally, any subject which failed to provide a DNA- sample or for which genetic analyses failed was excluded. After a final exclusion, data from one- hundred eighty subjects remitted from clinically significant depression were accepted. The clinical exclusion criteria were serious head trauma, depressive illness at the time of interview, structural abnormalities of the brain, and pregnancy at the time of scanning. In addition, seventy-seven healthy control subjects were recruited. Nine of these we excluded for the same reasons as mentioned above leading to the final demographical data summarized in table 1.

Table 1. Demographics of subject groups. All tests are across subject group only. Gender distribution is tested with a Chi-square test, and group-wise age distributions are tested with an independent t-test for the whole groups (rMDD and control). Remitted MDD Control Group difference P

Female Male Female Male

N 123 57 45 22 No 0.861 Mean age 40.5 41.0 37.7 42.8 No 0.406 SD age 13.7 12.9 12.5 14.0 No 0.326

N Total 180 67 247

Figure 6. Gender distribution between the subject groups Healthy Controls (Control) and subjects remitted from Major Depressive Disorder (rMDD). The groups were found to have equal distributions in a Chi-square test.

3.2 DNA acquisition and isolation

DNA was isolated from cheek epithelium using Isohelix SK-1S DNA Buccal Swabs. The research subjects were given instructions on proper procedure for the use of the swabs and provided the samples at their own leisure. The swabs were stored at room temperature according to the manufacturers guidelines along with a preserving silica pellet proprietary to Isohelix.

The DNA isolation was performed using Qiagen DNA Mini Kits (Qiagen, Düsseldorf, Germany) according to the buccal swab protocol, and using Isohelix Buccal Swab DNA Isolation Kits (Isohelix, Kent, UK). The Qiagen kit utilizes a spin column to ensure high purity of the end product, but provide lower yields compared to the Isohelix kit. For reasons of availability, samples 1-41 and 205-229 were collected using Qiagen Mini, whereas 42-204 and 230-445 were collected using the Isohelix kit. This was deemed acceptable due to the nature of downstream applications of the isolated DNA.

3.3 PCR of BDNF

The BDNF gene was amplified in a block cycler from the isolated DNA using primers: forward 5′-ACTCTGGAGAGCGTGAATGG- 3′ reverse 5′-ACTACTGAGCATCACCCTGGA-3′

Ideal cycling conditions were after experimentation found to be:

5 min initial denaturation (95 °C)

1 min denaturation (95 °C)

30 sec annealing (60 °C)

1 min extension (72 °C)

Repeat step 2-4 40 times

10 min final extension (72 °C)

Hold 4 °C.

This yielded a 171 bp band from the coding region of the BDNF gene which was confirmed by using agarose gel electrophoresis in a 2% gel using a 1kb ladder from Thermo Scientific.

3.4 Genotyping of BDNF

BDNF genotypes were established using restriction cutting with Eco72I (Thermo Fisher Scientific, Waltham, MA, USA) with the following conditions:

Buffer and enzyme volumes according to manufacturer Thermo Fisher

15 μl of PCR product

12 hours of incubation at 37°C

Subsequent separation in 2,5% agarose gel electrophoresis was performed at at 100V for approximately 2 hours using 1X TAE-buffer. Biotium GelRed™ (Biotium, Fremont, CA 94538, USA) was used for staining the agarose gel at a concentration of 4µl/50mL. A Thermo Scientific GeneRuler 50bp DNA Ladder (Thermo Fisher Scientific, Waltham, MA, USA.) (bands: 1000, 900, 800, 700, 600, 500, 400, 300, 250, 200, 150, 100, 50) was used to assess fragment size on each run.

The two alleles have the following sequences, with the restriction site in green, and the mutated base in red, bold, in order to show the disruption of the restriction site.

Val66 allele:

ACTCTGGAGAGCGTGAATGGGCCCAAGGCAGGTTCAAGAGGCTTGACATCATTGGCTGACACTTTCGAACACG TGATAGAAGAGCTGTTGGATGAGGACCAGAAAGTTCGGCCCAATGAAGAAAACAATAAGGACGCAGACTTGT ACACGTCCAGGGTGATGCTCAGTAGT

Met66 allele:

ACTCTGGAGAGCGTGAATGGGCCCAAGGCAGGTTCAAGAGGCTTGACATCATTGGCTGACACTTTCGAACACA TGATAGAAGAGCTGTTGGATGAGGACCAGAAAGTTCGGCCCAATGAAGAAAACAATAAGGACGCAGACTTGT ACACGTCCAGGGTGATGCTCAGTAGT

Restriction cutting of the Val66 allele produces two bands of 99 and 72 base pairs, whereas restriction cutting of the Met66 allele leaves the full 171 bp sequence intact. Genotypes were read as following:

1 (strong) band at 171 bp – Homozygous Met66

3 bands at 171, 99 and 72 bp – Heterozygous

2 (strong) bands at 99 and 72 bp – Homozygous Val66

Several samples had to be excluded due to difficulties reading from the gel even after several reruns.

Figure 7. Uncut band of 171 bp is highlighted in blue, indicating Met/Met genotype. Note weak band in well 6. Ladder: 1kb to 50bp.

Figure 8. Banding characteristic of both strands cut is highlighted in green, indicating Val/Val genotype. Heterozygote profile is highlighted in red. Ladder: 1kb to 50bp.

Table 2. Distribution of the Val/Met alleles across the subject groups. rMDD (n = 180) Control (n = 67) Val/Val Val/Met Met/Met Val/Val Val/Met Met/Met

N 114 53 13 50 14 3 Fraction 63.3 % 29.4 % 7.2 % 74.6 % 21.0 % 4.4 %

Val/Val Val/Met Met/Met

Total N 164 67 16 Total fraction 66.4 % 27.1 % 6.5 %

3.5 5-HTTLPR analysis

In order to perform downstream restriction cutting analysis of a polymorphism in the long allele of the serotonin transporter, numerous attempts at regular PCR amplification of this GC-rich region were attempted. Being unsuccessful in this endeavor, we sought the aid of Marianne Kristiansen Kringen at the Center for Psychopharmacology at Diakonhjemmet. Using a TaqMan-method for analysis, they were able to provide us with the genotype for our research subjects. Unfortunately, the amount of PCR product in these samples was too low for downstream analysis with restriction enzymes and agarose gel electrophoresis disallowing the search for SNPs in the 5-HTTLPR. We were still able to distinguish between the canonical long and short alleles, however.

Table 3. Distribution of the long/short alleles across subject groups rMDD (n = 180) Control (n = 67) S/S S/L L/L S/S S/L L/L

N 27 96 57 14 31 22 Fraction 15.0 % 53.3 % 31.7 % 20.9 % 46.3 % 32.8 %

S/S S/L L/L

Total N 41 127 79 Total fraction 16.6 % 51.4 % 32.0 %

3.6 MRI acquisition

The images used for MRI analyses were generated using a 12-channel head coil on a Philips Ingenia 3.0T (Philips, Amsterdam, Netherlands) scanner at the Intervention Centre at Oslo University Hospital. T1-weighted sequences were produced with the following parameters:

• Repetition time (TR) 3000ms • Echo time (TE) 3.61 ms • Flip angle 8° • Acquisition duration 3 minutes and 16 seconds.

The T1 files consisted of 184 transverse slices containing 1mm3 voxels. Manual elimination of obvious structural defects was performed by Luigi Maglanoc and Dani Beck to ensure that only suitable samples were run through the FreeSurfer pipeline.

3.7 FreeSurfer reconstruction and hippocampal subfield segmentation

T1 weighted scans of all brain volumes underwent the FreeSurfer V5.3 (http:// surfer.nmr.mgh.harvard.edu/) pipeline reconstruction in order to be available for downstream segmentation. This was tested both small scale in a private Ubuntu environment as a conceptual test of the software package with real data, but due to the computing power intensive process of reconstruction the full dataset was acquired by requesting access to the Abel Supercluster owned by the University of Oslo and the Norwegian metacentre for High Performance Computing (NOTUR), operated by Department for Research Computing at USIT, the University of Oslo IT-department (http://www.hpc.uio.no/). The actual uploading of raw data and bash-script to Abel was performed by Dani Beck and Luigi Maglanoc.

The hippocampal subfield segmentation is an automated procedure available in FreeSurfer, and V6.0 of this specific tool outputs twelve separate regions of the hippocampal formation as well as the whole hippocampus. This is done bilaterally, leading to twenty-seven unique metrics when total IntraCranial Volume (ICV) is included. All subfields and ICV are provided by the software in cubic millimeters (mm3). See the previous section on FreeSurfer and hippocampal segmentation for details on how the ex-vivo atlas for the procedure has been generated. In addition to providing quantitative volumetric data, the subfield segmentation can be displayed in the FreeView software which is part of FreeSurfer. This allows for visual inspection and gives a qualitative view of what FreeSurfer does.

Figure 9. Output of coronal view from FreeView. Hippocampal subfields are color coded. Fimbria and hippocampal tail subfields not shown.

Figure 10. Saggital view produced in FreeView. Hippocampal subfields are color coded. Parasubiculum subfield is not shown.

Figure 11. Axial inferior-bound series. The image has intrinsic tilt which shows both hippocampi with a slight spatial displacement for a given section.

Figure 12. Caudal-coronal view, rostrally bound series. Grey-white contrasts correspond to grey matter/white matter boundaries. The cerebellum and brain-stem structures (pons) can be seen inferiorly to the hippocampi. 3.8 Statistical calculations

Statistical calculations and production of graphs were carried out in Microsoft Excel or in IBM (Armonk, New York, USA) SPSS statistics processor. Unless otherwise specified, an alpha level of < 0.05 was considered significant for all analyses. Graphs were generated in GraphPad Prism (La Jolla, CA 92037, USA) unless otherwise specified. All analyses of covariance used ICV and age as covariates (rationale shown below) and were performed in IBM SPSS statistics processor. Chi-square and t-tests were performed in IBM SPSS statistics processor. Hypothesis 1 was tested with distribution statistics for genotypes and subject groups using Chi-square tests for distribution and to assess Hardy-Weinberg equilibria. Hypothesis 2 was tested using univariate and multivariate GLMs using covariates established in 4.1, see discussion for determination of α-levels.

4. Results

4.1 Preliminary statistics and determination of covariates

In order to determine covariates for downstream analyses, a linear regression for the sum of 2 hippocampal volumes (right + left) as a function of ICV was run (R = 0.35, F(1, 245) = 132, p < 0.001) (see figure 15) implying a strong correlation. This was also true for both hippocampi separately ((Right 2 2 hippocampus R = 0.38, F(1, 245) = 152, p < 0.001),(Left hippocampus R = 0.35, F(1, 245) = 133, p < 0.001)). 2 Age alone did not have a significant effect on total hippocampal volumes (R = 0.008, F(1, 245) = 3.072, p 2 = 0.08) but had a weak effect on the left hippocampus volume (R = 0.016, F(1, 245) = 4.899, p = 0.028). Based on these trend-level findings and previous literature (see review by De Flores et al.27), age was included as a covariate in later analyses. All these effects were linear with no deviation from linearity detected (Runs test, p>0.4 in all cases). Gender had a significant effect on total hippocampal volume of both hippocampi, but it did not survive correction for ICV in a GLM using ICV as a covariate (F(1, 245) = 0.047, p = 0.828)(see figure 13). In other words, males had generally larger hippocampal volumes, but only due to their generally larger ICVs. Expressed as fraction of ICV, on the other hand, females had larger hippocampi (n.s) (see figure 14). Notably, analyzed without any covariates, neither of these measures (absolute hippocampus volume or fraction of ICV) differed between control and remitted groups. This finding led to the assumption that there is a basic volumetric difference as a function of gender which is in reality an effect of differences in ICV. In addition to age, ICV is thus included as a covariate in the following analyses of variance (Section 4.3).

Figure 13. Total hippocampal volume in healthy controls (control) and subjects remitted from major depressive disorder (rMDD). Significance disappears when controlling for total intra-cranial volume (see Figure 14). Data presented as mean ± S.E.M.

Figure 14. Hippocampus (HC) volume in healthy controls (control) and subjects remitted from major depressive disorder (rMDD) as a percentage of total intra-cranial volume (ICV). Data presented as mean ± S.E.M. With exception of the hippocampal fissure (bilaterally), all subfields and each individual hippocampus displayed normal distribution at p > .200 using a Kolmogorov-Smirnov test of normality. For this reason, the hippocampal fissure was excluded from further analyses. Before further testing, bilateral symmetry of subfield volumes was assessed by way of a paired-samples t-test for all subfields across the two brain hemispheres.

Table 4. Paired-samples t-test for hippocampal subfields and whole hippocampus across the entire dataset. Positive values indicate that the right subfield is larger, negative values (in italics) indicate that the left subfield is larger.

Subfield ∆Mean ± S.E.M. t (246) P Cohen’s d

Whole hippocampus 86.52 ± 10.7 8.110 < 0.001 0.52

CA1 27.87 ± 3.7 7.475 < 0.001 0.48 CA2/3 24.64 ± 1.8 13.270 < 0.001 0.84 CA4 16.55 ± 1.4 11.916 < 0.001 0.76 DG GC 14.85 ± 2.7 9.096 < 0.001 0.58 Subiculum -5.01 ± 2.0 -2.463 0.014 -0.16 Presubiculum -18.89 ± 1.8 -18.89 < 0.001 -0.69 Parasubiculum -2.84 ± 0.7 - 3.865 < 0.001 -0.25 ML subiculum 14.85 ± 2.7 5.543 < 0.001 0.35 Hippocampal tail 5.08 ± 3.1 1.620 0.106 0.10 Fimbria 8.51 ± 1.1 7.832 < 0.001 0.5 HATA 0.49 ± 0.8 0.584 0.560 0.037

Figure 15. Linear regression of total hippocampal volumes over ICV. R2 = 0.348. Genders are labeled by color and shape (Females = Black dot, Males = grey square) in order to illustrate clustering and variance for the two groups.

4.2 Population genetics: 5-HTTLPR/BDNF(Val66Met) and depression

To test the power of the two genotypes with respect to predicting a history of depression, a series of distribution analyses were carried out. The two alleles for each of the genes were tested against the assumption of a Hardy-Weinberg equilibrium. This tests the actual distribution of alleles against a mathematical expectation in order to determine whether the population appears to be evolving with respect to the particular alleles. Central assumptions in the Hardy-Weinberg model are:

• Random Mating • No mutation, migration or natural selection • No genetic drift • Infinitely large population size

The allelic distribution for BDNF rs6265 for the whole dataset was found to significantly deviate from a Hardy-Weinberg equilibrium (p2+2pq+q2) in a Chi-square test (χ2 = 5.453, N = 247 df = 1, p < 0.025) (See discussion for implications). Within the subject groups “rMDD” and “healthy controls” this did not hold true and they displayed HW equilibrium ((χ2 = 3.553, N = 180, df = 1, p > 0.05) and (χ2 = 1.630, N = 67, df = 1, p > 0.1) respectively).

The 5-HTTLPR genotype was found to display HW equilibrium characteristics for the whole dataset (χ2 = 0.701, N = 247 df = 1, p > 0.1). Within the two subject groups “rMDD” and “healthy controls”, evidence of HW equilibrium was also found ((χ2 = 1.699, N = 180, df = 1, p > 0.1) and (χ2 = 0.251, N = 67, df = 1, p > 0.1) respectively).

In order to test assumptions of genotype distribution as a specific predictor of having a history of depression (Hypothesis 1), a Chi-square test was run for each genotype against subject group. There were no significant differences in distribution of SERT long/short between the two groups (Pearson χ2 = 1.517, N = 247, df = 2, p = .468) (figure 16). Similarly, Chi-square was run for the BDNF Val66Met SNP, but no significant effect of group was found (Pearson χ2 = 2.821, N = 247, df = 2, p = .244) (figure 17).

Figure 16. Distributions of the three genotypes for the Serotonin Transporter promoter (L/L, S/L, S/S) displayed across the groups Healthy Controls (Control) and remitted from Major Depressive Disorder (rMDD).

Figure 17. Distributions of the three genotypes for the BDNF gene prodomain (Val/val, val/met, met/met) displayed across the groups Healthy Controls (Control) and remitted from Major Depressive Disorder (rMDD).

4.3 Effects of depression history on subfield volumes

The neurotrophic model of depression indicates that total hippocampal volumes may change in relation to the development of MDD. Hypothesis 2 was approached in three steps: comparison of total hippocampal (right, left and the sum of both) volumes across the rMDD and Controls groups, multivariate analysis to investigate whether specific subfields appear to be affected, and a series of univariate analyses covering all subfields separately.

4.3.1 Group and gender effects on total hippocampal volume

Analyses of the total hippocampal volume as well as the whole hippocampus on each side were performed as separate two-way ANCOVAS with the factors subject group and gender, using ICV and age as covariates. No significant effects were found across subject groups or genders, or in interactions between subject group and gender. Summary of results are found in table 5.

Table 5. Separate linear models testing for gender and subject group effects as well as interactions between gender and subject group. 2 F(1,245) P ηp Left hippocampus

Gender 0.124 0.726 0.001 Subject group 0.405 0.525 0.002 Gender * subject group 0.402 0.526 0.002

Right hippocampus

Gender 0.026 0.873 >0.001 Subject group 0.006 0.937 >0.001 Gender * subject group 0.882 0.349 0.004

Total hippocampal volume

Gender 0.181 0.671 0.001 Subject group 0.097 0.755 >0.001 Gender* subject group 0.853 0.357 0.004

To investigate subject group effects on subfields two separate factorial MANCOVAs were carried out; one for each hemisphere. The factors subject group and gender were used with ICV and age as covariates. For the left subfields, subject group showed a significant effect and gender showed trend- level significance (P = 0.051). No interaction effect of gender and subject group was observed. In the right subfields, subject group showed a highly significant effect (p < 0.001) but no effect was evident for gender. No interaction effect of gender and subject group was observed.

Table 6. Results of two multivariate GLMs for hippocampal subfields in each hemisphere. The F value for Pillai’s Trace was interpreted to test multivariate significance. 2 V F(11, 231) P ηp Left subfields Gender 0.08 1.822 0.051 0.080 Subject group 0.131 3.170 0.001 0.131 Gender * subject group 0.038 0.827 0.613 0.038

Right subfields Gender 0.043 0.938 0.505 0.043 Subject group 0.139 3.401 <0.001 0.139 Gender * subject group 0.025 0.529 0.882 0.025

Follow-up ANCOVAS were performed for subject group effects on subfield volume in each hemisphere using ICV and age as covariates. Results are summarized in Table 6 and Table 7.

Table 7. Results of a Generalized Linear Model ran for 11 out of 12 hippocampal subfields in the left hemisphere. Volumes are given in mm3 ± S.D. Left subfields rMDD Healthy Controls ANCOVA

2 (n = 180) (n = 67) F(1, 245) P ηp

Whole hippocampus 3631.64 ± 375.6 3643.65 ± 305.8 0.097 0.76 <0.001

CA1 681.53 ± 85.2 686.42 ± 67.9 0.221 0.64 0.008

CA2/3 247.60 ± 34.2 247.69 ± 34.9 0.105 0.75 <0.001

CA4 271.22 ± 33.7 277.31 ± 27.8 3.528 0.06 0.014

DG GC 329.85 ± 40.8 336.95 ± 33.4 3.864 0.050 0.016

Subiculum 450.52 ± 56.0 453.42 ± 49.2 0.489 0.46 0.002

Presubiculum 337.66 ± 41.6 328.01 ± 37.2 1.664 0.2 0.007

Parasubiculum 69.62 ± 11.7 72.84 ± 9.4 5.589 0.019 0.023

ML subiculum 549.95 ± 58.0 563.33 ± 63.1 4.176 0.042 0.017

Hippocampal 550.47 ± 66.5 543.78 ± 54.8 0.322 0.6 0.001 tail Fimbria 73.18 ± 20.1 66.53 ± 18.00 7.391 0.007 0.03

HATA 68.59 ± 12.7 69.66 ± 10.8 0.902 0.34 0.004

Table 8. Results of a Generalized Linear Model ran for 11 out of 12 hippocampal subfields in the right hemisphere. Volumes are given in mm3 ± S.D. Right subfields rMDD Healthy Controls ANCOVA

2 (n = 180) (n = 67) F(1, 245) P ηp

Whole hippocampus 3724.89 ± 392.3 3712.09 ± 324.2 0.006 0.937 <0.001

CA1 712.32 ± 90.0 706.44 ± 77.7 0.306 0.58 0.001

CA2/3 273.26 ± 33.8 268.95 ± 32.2 0.289 0.6 0.001

CA4 288.53 ± 33.2 291.83 ± 30.3 1.237 0.27 0.005

DG GC 345.43 ± 40.7 349.82 ± 37.5 1.702 0.19 0.011

Subiculum 446.30 ± 57.1 446.27 ± 46.1 0.035 0.86 <0.001

Presubiculum 318.51 ± 39.3 309.82 ± 32.3 2.756 0.1 0.011

Parasubiculum 66.99 ± 11.1 69.43 ± 9.4 2.117 0.15 0.009

ML subiculum 565.23 ± 64.2 577.05 ± 55.9 2.747 0.099 0.011

Hippocampal 555.82 ± 74.3 548.11 ± 62.9 0.625 0.43 0.003 tail Fimbria 82.07 ± 20.8 74.03 ± 17.5 7.838 0.006 0.031

HATA 69.97 ± 13.9 67.74 ± 12.1 1.489 0.224 0.006

Except for the fimbria (see figure 18 & 19), all subfields display larger volumes in the control group.

Figure 18. Left fimbria volumes for groups Healthy Controls (Control) and subjects remitted from Major Depressive Disorder (rMDD). Bars represent mean ± S.E.M. An effect of gender seen on absolute volumes is not present when ICV is used as covariate.

Figure 19. Right fimbria volumes for groups Healthy Controls (Control) and subjects remitted from Major Depressive Disorder (rMDD). Bars represent mean ± S.E.M. An effect of gender seen on absolute volumes is not present when ICV is used as covariate.

Figure 20. Left parasubiculum in groups Healthy Controls (Control) and subjects remitted from Major Depressive Disorder (rMDD). Covariates used were ICV and age. Bars represent mean ± S.E.M.

Figure 21. Right parasubiculum in groups Healthy Controls (Control) and subjects remitted from Major Depressive Disorder (rMDD). Covariates used were ICV and age. Bars represent mean ± S.E.M.

4.4 Biological metrics effects on subfield volumes

4.4.1 Sub-hypotheses 1 and 2

To test sub-hypothesis 1 and 2, a factorial MANCOVA was run for each hemisphere, testing for the effects of 5-HTTLPR genotype and BDNF rs6265 genotype for all eleven included subfields. ICV and age were included as covariates. Both tests met the criteria Box’s test of equality of covariance matrices. In addition to testing each individual genotype, interaction effects of the two genotypes were tested in the same multivariate model. The rationale for testing interactions between these genes comes from the fact that the serotonin and BDNF systems interact biologically (see Section 1.4.2). In addition, the findings in the hippocampus with relation to these systems (see Section 1.4.2 and 1.3.3) makes an effect in the hippocampus seem plausible. The rationale for splitting the analyses by hemisphere comes from the initial findings of a generally increased size of the right subfields which was consistent across gender and subject group. Evidence of simple genotype effects or interactions was not found in the multivariate model. Results are summarized in table 7.

Table 9. Interpretation of Wilk’s Lambda for two factorial multivariate GLMs assessing the 11 included subfields. Factors are BDNF and SERT genotypes and their interaction. Included as covariates are ICV and age.

V F df P ηp2 Left hemisphere

BDNF genotype .062 .661 22, 454 .877 .031 SERT genotype .891 1.223 22, 454 .223 .056 BDNF*SERT .825 1.013 44, 916 .451 .047

Right hemisphere

BDNF genotype .876 1.401 22, 454 .107 .064 SERT genotype .919 .883 22, 454 .619 .041 BDNF*SERT .837 .940 44, 916 .586 .044

5. Discussion

Support for Hypothesis 1 was not found during statistical investigations. No evidence of an uneven distribution of SERT or BDNF was found across the groups Healthy Controls and rMDD. Support of Hypothesis 2 was found with the observation that there are several subfields that appear to display

different volumes between the subject groups Healthy Controls and rMDD. Support for Hypothesis 3 was not found, as no subfield volumetric effects were found as a function of the genotypes or their interactions. The following section discusses these claims in more detail. 5.1 Hippocampal asymmetry

During the preliminary analyses establishing covariates and mapping out the large-scale trends, right- left asymmetry was discovered across the full dataset (see table 4). This applies both to the whole hippocampus and to most of the subfields except for the hippocampal tail and hippocampus- amygdala-transition-area (HATA). If one were to make a multiple comparisons correction such as the Bonferroni correction (in this case modifying α to α = 0.0045), the p-value for the subiculum (p = 0.014) would no longer be significant. Due to the clear-cut significance values of the other 8 subfields (all surviving the most stringent correction) and the qualitative approach of this specific investigation, a multiple comparisons correction was deemed unnecessary to discuss the results. Hippocampal right- left asymmetry is a common finding, and has been shown to be present even at the premature neonatal stage119. It is known from studies in the human amygdala120 and cerebral cortex121,122 that there is at least a partial functional lateralization of certain brain functions. Studies of split-brain patients, and studies of language production and understanding has also contributed to the notion of functional lateralization in the human brain. In the rat hippocampus, Sakaguchi and Sakurai found lateralization of ventral hippocampal function with respect to anxiety123 indicating right hemisphere dominance in high-anxiety situations. Quantification of asymmetry was not performed as part of the present study, but a possible prospect would be to see whether the structural asymmetry remained the same across groups of people with mental illnesses or specific genetics. Structural lateralization might be important for the development of, or protection against, depression. This data would be best paired with some functional data such as fMRI, which could shed light on the relationship between structural lateralization and functional lateralization in the hippocampus.

5.2 Distribution of genotypes and HW disequilibrium

The Hardy-Weinberg principle states that the allele and genotype frequencies of a population will remain constant in the absence of factors that select for one allele or genotype. Contained in the principle is a set of assumptions about such factors that must be met in order for the mathematical approach to provide a valid result. The BDNF rs6265 variant did not display Hardy-Weinberg proportions (see 4.2) which can be interpreted to mean that one or more of the assumptions have been violated. Allele distribution across gender showed uniformity which should compensate for a putative effect of the roughly 68% females in the dataset. Another important assumption is random mating, i.e. the genotype does not affect choice of partner. Homozygosity can be inflated by inbreeding and drive the population into non-equilibrium. No data provided in this thesis can address this question, and so genotype effects on mating stands as a possible contributor. Although BDNF genotype was not found to correlate with a history of depression, there is a literature which indicates that it may contribute to mood disorders39,99-101. That this in turn may affect reproductive success is entirely plausible, but without further investigations it is left entirely to speculation. Another speculative point is that some studies have found lowered BDNF-levels in the brains of suicide subjects124,125. A meta- review found an increased risk of suicide as a function of the met-allele in ethnic sub-populations, but no overall increase in suicide risk126. A Hardy-Weinberg equilibrium assumes that genetic drift is

negligible. This assumption cannot be tested with these data, and so this assumption must also stand unchallenged. Being intrinsically random, we must expect genetic drift to also occur in human populations, although the degree varies with each gene and population depending on evolutionary pressure. Petryshen et al. studied the distribution of these alleles in a wide variety of populations and found that the Met allele is the most common in many Asian populations, slightly less common in European populations, and almost non-existent in many African populations127. This can either point to the origins of the mutation, or to an ecological criterion for its selection effect, i.e. it might be beneficial in one environment and not in another. In general, there is little to indicate dramatic effects of this SNP on reproductive success and natural selection, but a Hardy-Weinberg disequilibrium along with the rest of the literature warrants further research into this allelic variant (see section 5.3).

5.3 Power of BDNF and SERT genotypes in predicting depression history

Hypothesis 1 posits that the BDNF Val66Met SNP might be predictive of being diagnosed with depression throughout life. The same prediction is made for the L/S promoter variant alleles of the 5- HTTLPR. As seen in section 4.2, there was no significant difference in distribution of these two genotypes between the rMDD and control groups. This implies that knowing an individual’s genotype for one of these genes carries no isolated predicting power towards whether they have suffered from MDD. Likelihood of having both “risk variants” across groups was tested post-hoc, but no trends in co- distribution were observed (data not shown). Fan et al. found an increased rate of depression in a post- earthquake population carrying the Val66Met genotype100. Since no detailed life-history data has been collected for the subjects used to generate this data, comparing them to people who have experience trauma is problematic. Ribeiro et al. found an association between the met-allele and depression in a Mexican-American subject group99. As outlined by Petryshen et al.127, distributions of this mutation vary across different populations. It is conceivable that another ethnic group may respond differently to a particular genetic variation, or that undocumented life-events account for an interaction effect that can lead to depression. Since such data lacks for both this study, and the one previously outlined, comparing them becomes difficult. Further data on life-histories would be needed, and would in general align much more with the general observation that depression and life-adversity are connected.

To contain the scope of this thesis to biological data, quantitative information on the clinical depression syndrome has not been included. Such data has been collected, and the full dataset allows for quantitative analyses of depressive symptomology (Beck and Hamilton inventories) and its relation to genetics. The groups rMDD and control are meant to represent total of biological variables that separate those who develop depression from healthy people. This way of approaching the problem greatly reduces the power of the study by omitting many control variables that should have been investigated in a more biological approach to depression. Extra variables such as number of depressive episodes and severity of depressive episodes would aid in the formation of subgroups for a more specific investigation. Other measures of life stress such as glucocorticoid levels over time (see future prospects) could also provide data which would be meaningful in conjunction with these genetic data. A final problem with this approach is that the history of depression was not confirmed through medical records upon recruitment as the grouping is based on self-reporting.

5.4 Predictive power of depression history on hippocampal subfield volumes

Table 5 shows that a multivariate GLM for gender and subject group failed to demonstrate any overall effects on hippocampal volumes. This was true for each separate hippocampus and the total volume of the two. A follow-up analysis of subfield effects in a multivariate GLM found strong evidence of an overall effect on subfields by subject group in each hemisphere (p ≤ 0.001). The test of the specific subfields through ANCOVA analyses provided one statistically significant effect for the right hippocampus (Fimbria), and statistically significant effects in four areas in the left hemisphere: Dentate gyrus molecular layer, parasubiculum, molecular layer of subiculum, and fimbria. The fimbria is the only subfield in which a bilateral effect of depression history was displayed. Additionally, the fimbria is on average larger in the rMDD group (figure 18 & 19), whereas the other subfields display the expected trend of volume decrease (exemplified in figure 20 & 21). In addition to the effects seen in the fimbria, several of the areas which are new to the FreeSurfer 6.0 segmentation protocol displayed statistically significant effects of subject group. These were the molecular layer of the subiculum, the granule cell layer of the dentate gyrus, and the parasubiculum. The molecular layer of the subiculum and the parasubiculum are regions about which very little is known, and any functional predictions from volume changes are speculative at best. Perhaps most interesting is the borderline finding of a significant group effect on the granule cell layer of the dentate gyrus. As reviewed in 1.3.3, there are reasons to believe that changes in the neurogenesis patterns of the GCDG might be involved in the etiology of depression31-33, see also review by Mahar61. This finding has strong conceptual roots, but the borderline significance level even before α-correction warrants caution.

The seeming contradiction of no overall effects on hippocampal volumes, but effects in several subfields can partially be explained by the fact that two of the areas in which significant effects were demonstrated are very small. Fimbria (different on both sides) and parasubiculum (left side different across the two groups) are small, and their contribution to the overall variance in the whole hippocampus comparisons is small (Table 5). Their effect on pooled variance might not be enough to push the total variance to a level which is detectably different between the subject groups in a GLM. This underlines the necessity for doing the two different tests; the multivariate model for the whole hippocampi is less powerful but also more conservative and less likely to produce type 1 errors. Likewise, the multivariate test for overall subfield effects is more robust than the series of ANCOVAs, but provides less specific information and much less power. Doing a series of ANCOVAs intuitively inflates the risk of making a type 1 error, and there are several reasons why a multiple comparisons correction of α-level was not performed. First off, accepting the nominal values is part of getting a qualitative feedback about the viability of a new segmentation protocol which is to be explored further. Findings by Hayasaka and MacQueen10,40 show that hippocampus volumes appear plastic within a time span of weeks to months. This poses a challenge to the current dataset due to there being no control for how long ago any depressive episodes have taken place. The choice of hypotheses (all hypothesis 2 tests) is a very likely source of error, and the original clinical study is not ideally designed to accommodate the hypotheses in this investigation. For instance, a formulating the hypothesis for only the areas novel to FreeSurfer v6.0 would make sense, but could dramatically alter the final α-level. Bonferroni correction is common for analyses like these 128,129 and would create an α-level of 0.0045. A more conceptually focused approach might have outlined the differences in only the 4 new subfields and thus experienced a much milder α-adjustment than these two 11-subfield hypotheses would. A final important point about omitting α-correction for this analysis is the idea of Disciplined analysis as outlined by Lemuel A Moyé130. The separate subfield analysis by ANCOVAs is part of the prospective

design and primary outcomes, and must thus be treated differently from post-hoc/exploratory approaches to extracting hypotheses from the data (see hypothesis 3). For this reason, the data for hypothesis 2 have been accepted at nominal values. In summary, these findings must be viewed with caution, and further investigations into FreeSurfer as a tool to answer questions like the ones asked in this thesis are necessary. See section 6 for suggestions of such approaches.

5.5 Predictive power of genotypes on subfield volumes

Summarized in table 9 are the results of two multivariate GLMs run for the BDNF genotypes and 5- HTTLPR genotypes against all subfield volumes during an exploratory approach. This model revealed no effects of genotypes, and no interaction effects, on the subfield volumes. Follow-up ANCOVAs similarly produced no statistically significant effects (table not shown) even before a multiple comparisons correction. Failing to reject the null hypothesis indicates that the BDNF and 5-HTTLPR genotypes do not predict hippocampal subfield volume differences across the whole of the dataset. As with many exploratory hypotheses, there was a speculative background for hypothesizing that there would be an effect on this level. The rationale for the hypothesis was the documented relationship between BDNF/SERT systems and depression on the one hand, and these systems and neurotrophic effects on the other.

This approach, being exploratory and not prospectively defined, has a number of problems. Firstly, it has not emerged from the literature that the BDNF SNP or the SERT polymorphism exert their effects through the hippocampus and so the hypothesis speculative. For BDNF specifically, the allelic variation appears to correlate with altered activity-dependent-release dynamics97, which may or may not coincide with any effects BDNF might have in the hippocampus108,111. As a general principle in candidate gene studies and central in biological teachings about mutations; single-nucleotide- polymorphisms and other small genetic changes are likely to produce small effects on higher organizational levels. This is, of course, not always the case, but canonical examples such as the hemoglobin-gene substitution responsible for sickle cell anemia are rare. In an age where genetic mapping is becoming more and more accessible, the zeitgeist of GWAS studies is ever more present and relevant. With genome-wide approaches available, candidate gene studies can appear akin to gambling unless there is a serious precedent for studying these specific genes. Finally, a statistical approach to this type of exploration requires great rigor and care.

6. Conclusions and future perspectives

There appears to be little power in the different genotypes for BDNF Val66Met and 5-HTTLPR L/S when it comes to predicting an individual’s likelihood of having a history of clinically significant Major Depressive Disorder unless paired with hitherto unknown data. A history of depression, however, appears alongside volumetric characteristics in certain hippocampal subfields which are not present in a demographic with no such history. To my knowledge, this effect is observed for the first time with the novel segmentation protocol found in FreeSurfer v6.0 on this dataset. No evidence of BDNF Val66Met or 5-HTTLPR L/S effects on hippocampal subfield volumes was found.

A natural next step is to add clinical and biological data on life stress and adversity in conjunction with quantitative data on the depression syndrome to produce more specific predictions about subfield volumetrics. The full dataset contains a cortisol metric designed to investigate the effect of depression on the Cortisol Awakening Response, as well as Beck and Hamilton inventories on depression symptoms and a medication status/history of each subject. These were left out for the sake of limiting the scope of the investigations to the genetical and life-history data which were initially the most relevant. fMRI data has also been collected and could provide further information on the implications of subfield changes. It seems reasonable to believe that hippocampal subfield changes may result in altered activity-patterns, and this could be approached with fMRI. In addition to the new hippocampal subfield segmentation protocol, FreeSurfer can also make heatmaps of cortical thickness for large datasets, and it can be used to do longitudinal studies on hippocampal development. Both of these methods may offer additional information upon which to base future studies.

At the time of writing this thesis, MRI remains the most powerful non-invasive technique for making structural assessments of the living brain in human beings. The 6.0 version of the FreeSurfer software suit shows promise as a means of investigating volumetric differences in the hippocampus and its subfields. Further studies into the effects of biological and environmental factors that affect brain structure will surely be aided by this powerful tool.

References

1. Bliss TVP, Lømo T. Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J Physiol. 1973;232:331- 356. 2. Hubel DH, Wiesel TN. The period of susceptibility to the physiological effects of unilateral eye closure in kittens. J Physiol. 1970;206:419-436. 3. Hakamata Y, Lissek S, Bar-Haim Y, et al. Attention Bias Modification Treatment: A meta- analysis towards the establishment of novel treatment for anxiety. Biol Psychiatry. 2010;68(11):982-990. 4. Beevers CG, Clasen PC, Enock PM, Schnyer DM. Attention Bias Modification for Major Depressive Disorder: Effects on Attention Bias, Resting State Connectivity, and Symptom Change. J Abnorm Psychol. 2015;124(3):463-475. 5. WHO MDD fact sheet (http://www.who.int/mediacentre/factsheets/fs369/en/). 6. WebMD Depression fact sheet.pdf. http://www.webmd.com/depression/guide/what-is- depression#1. 7. Jonassen R, Landro NI. Project description: Secondary prevention of depression applying an experimental Attentional Bias Modification procedure. 8. Disner SG, Beevers CG, Haigh EA, Beck AT. Neural mechanisms of the cognitive model of depression. Nat Rev Neurosci. 2011;12(8):467-477. 9. Malykhin NV, Coupland NJ. Hippocampal neuroplasticity in major depressive disorder. Neuroscience. 2015;309:200-213.

10. Hayasaka S, Nakamura M, Noda Y, et al. Lateralized hippocampal volume increase following high-frequency left prefrontal repetitive transcranial magnetic stimulation in patients with major depression. Psychiatry Clin Neurosci. 2017. 11. Iglesias JE, Augustinack JC, Nguyen K, et al. A computational atlas of the hippocampal formation using ex vivo, ultra-high resolution MRI: Application to adaptive segmentation of in vivo MRI. Neuroimage. 2015;115:117-137. 12. Sandrone S, Bacigaluppi M, Galloni MR, et al. Weighing brain activity with the balance: Angelo Mosso's original manuscripts come to light. Brain. 2014;137(Pt 2):621-633. 13. Birkner A, Tischbirek CH, Konnerth A. Improved deep two-photon calcium imaging in vivo. Cell Calcium. 2017;64:29-35. 14. Stucht D, Danishad KA, Schulze P, Godenschweger F, Zaitsev M, Speck O. Highest Resolution In Vivo Human Brain MRI Using Prospective Motion Correction. PLoS One. 2015;10(7):e0133921. 15. Kandel ER, Schwartz JH, Jessell TM, Siegelbaum SA, Hudspeth AJ. Principles of Neural Science Fifth Edition. In:2013. 16. O'keefe J, Dostrovsky J. The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat. Brain Res. 1971;34(1):171-175. 17. O'keefe J. Place Units in the Hippocampus of the Freely Moving Rat. Experimental Neurology. 1976;51:78-109. 18. Fyhn M, Molden S, Witter MP, Moser EI, Moser MB. Spatial Representation in the Entorhinal Cortex. Science. 2004;305(5688):1258-1264. 19. Hafting T, Fyhn M, Molden S, Moser MB, Moser EI. Microstructure of a spatial map in the entorhinal cortex. Nature. 2005;436(7052):801-806. 20. Scoville WB, Milner B. Loss of recent memory after bilateral hippocampal lesions. J Neurol Neurosurg Psychiat. 1957;20(11):11-21. 21. Roberts BM, Libby LA, Inhoff MC, Ranganath C. Brain activity related to working memory for temporal order and object information. Behav Brain Res. 2017. 22. Yonelinas AP, Hopfinger JB, Buonocore MH, Kroll NEA, Baynes K. Hippocampal, parahippocampal and occiptial-temporal contributions to associative and item recognition memory: an fMRI study. Neuroreport. 2000;12(2):359-363. 23. Karlsson MP, Frank LM. Awake replay of remote experiences in the hippocampus. Nat Neurosci. 2009;12(7):913-918. 24. Goshen I, Brodsky M, Prakash R, et al. Dynamics of retrieval strategies for remote memories. Cell. 2011;147(3):678-689. 25. Lee JQ, Zelinski EL, McDonald RJ, Sutherland RJ. Heterarchic reinstatement of long-term memory: A concept on hippocampal amnesia in rodent memory research. Neurosci Biobehav Rev. 2016;71:154-166. 26. Tae WS, Kim SS, Lee KU, Nam EC, Kim KW. Validation of hippocampal volumes measured using a manual method and two automated methods (FreeSurfer and IBASPM) in chronic major depressive disorder. Neuroradiology. 2008;50(7):569-581. 27. de Flores R, La Joie R, Chetelat G. Structural imaging of hippocampal subfields in healthy aging and Alzheimer's disease. Neuroscience. 2015;309:29-50. 28. Eriksson PS, Perfilieva E, Björk-Eriksson T, et al. Neurogenesis in the adult human hippocampus. Nat Med. 1998;4(11):1313-1317. 29. Zhang J, Griemsmann S, Wu Z, et al. Connexin43, but not connexin30, contributes to adult neurogenesis in the dentate gyrus. Brain Research Bulletin. 2017. 30. Spalding KL, Bergmann O, Alkass K, et al. Dynamics of hippocampal neurogenesis in adult humans. Cell. 2013;153(6):1219-1227. 31. Boldrini M, Hen R, Underwood MD, et al. Hippocampal angiogenesis and progenitor cell proliferation are increased with antidepressant use in major depression. Biol Psychiatry. 2012;72(7):562-571.

32. Boldrini M, Santiago AN, Hen R, et al. Hippocampal Granule Neuron Number and Dentate Gyrus Volume in Antidepressant-Treated and Untreated Major Depression. Neuropsychopharmacology. 2013;38(6):1068-1077. 33. Boldrini M, Underwood MD, Hen R, et al. Antidepressants increase neural progenitor cells in the human hippocampus. Neuropsychopharmacology. 2009;34(11):2376-2389. 34. Sheline YI, Sanghavi M, Mintun M, Mokhtar HG. Depression Duration But Not Age Predicts Hippocampal Volume Loss in Medically Healthy Women with Recurrent Major Depression. The Journal of Neuroscience. 1999;19(12):5034-5043. 35. Sheline YI, Wang PW, Mokhtar HG, Csernansky JG, Vannier MW. Hippocampal atrophy in recurrent major depression. Proc Natl Acad Sci U S A. 1996;93:3908-3913. 36. Han KM, Won E, Sim Y, Tae WS. Hippocampal subfield analysis in medication-naive female patients with major depressive disorder. J Affect Disord. 2016;194:21-29. 37. Chen MC, Hamilton J, Gotlib IH. Decreased Hippocampal Volume in Healthy Girls at Risk of Depression. Arch Gen Psychiatry. 2010;67(3):270-276. 38. MacMaster FP, Mirza Y, Szeszko PR, et al. Amygdala and Hippocampal Volumes in a Familial Early Onset Major Depressive Disorder. Biol Psychiatry. 2008;63(4):385-390. 39. Colle R, Segawa T, Chupin M, et al. Early life adversity is associated with a smaller hippocampus in male but not female depressed in-patients: a case-control study. BMC Psychiatry. 2017;17(1):71. 40. MacQueen GM, Yucel K, Taylor VH, Macdonald K, Joffe R. Posterior hippocampal volumes are associated with remission rates in patients with major depressive disorder. Biol Psychiatry. 2008;64(10):880-883. 41. Keller J, Shen L, Gomez RG, et al. Hippocampal and amygdalar volumes in psychotic and nonpsychotic unipolar depression. Am J Psychiatry. 2008;165(7):872-880. 42. Rydmark I, Wahlberg K, Ghatan PH, et al. Neuroendocrine, cognitive and structural imaging characteristics of women on longterm sickleave with job stress-induced depression. Biol Psychiatry. 2006;60(8):867-873. 43. Schmaal L, Yucel M, Ellis R, et al. Brain Structural Signatures of Adolescent Depressive Symptom Trajectories: A Longitudinal Magnetic Resonance Imaging Study. J Am Acad Child Adolesc Psychiatry. 2017;56(7):593-601 e599. 44. Cole J, Costafreda SG, McGuffin P, Fu CH. Hippocampal atrophy in first episode depression: a meta-analysis of magnetic resonance imaging studies. J Affect Disord. 2011;134(1-3):483- 487. 45. Frazer A, JG H. Serotonin Receptors. In: Siegel GJ, Agranoff BW, Albers RW, et al., editors. Basic Neurochemistry: Molecular, Cellular and Medical Aspects. 6th edition. Philadelphia: Lippincott-Raven; 1999. Available from: https://www.ncbi.nlm.nih.gov/books/NBK28234/. 1999. 46. Törk I. Anatomy of the Serotonergic System. Annals of New York Academy of Sciences. 1990. 47. Maddaloni G, Bertero A, Pratelli M, et al. Development of Serotonergic Fibers in the Post- Natal Mouse Brain. Front Cell Neurosci. 2017;11:202. 48. Bolliet V, Ali MA. Immunohistochemical study of the development of serotoninergic neurons in the brain of the brook trout Salvelinus fontinalis. Brain Behav Evol. 1992;40(5):234-249. 49. Halliday GM, Tork I. Serotonin-like immunoreactive cells and fibres in the rat ventromedial mesencephalic tegmentum. Brain Res Bull. 1989;22(4):725-735. 50. Glennon RA, Malgorzata D, Westkaemper RB. Serotonin Receptor Subtypes and Ligands. In: Psychopharmacology : the fourth generation of progress. New York: Raven Press; 1995. 51. Kandel ER, Schwartz JH, Jessell TM, Siegelbaum SA, Hudspeth AJ. Small-Molecule transmitters are actively taken up into vesicles. In: Kandel Principles of Neural Science Fifth Edition. McGraw-Hill Education; 2012:295-298. 52. Daniel W, Syrek M, Hauch A. The effect of selective serotonin reuptake inhibitors (SSRIs) on the pharmacokinetics and metabolism of perzine in the rat. JPP. 2001;53:449-461.

53. Catterson M, Preskorn S. Pharmacokinetics of Selective Serotonin Reuptake Inhibitors: Clinical Relevance. Pharmacology & Toxicology. 1996;78:203-208. 54. Fournier JC, DeRubeis RJ, Hollon SD, et al. Antidepressant Drug Effects and Depression Severity: A Patient-Level Meta-analysis. J American Medical Association. 2010;303(1):47-53. 55. Kirsch I, Deacon BJ, Huedo-Medina TB, Scoboria A, Moore TJ, Johnson BT. Initial Severity and Antidepressant Benefits: A Meta-Analysis of Data Submitted to the Food and Drug Administration. PLoS Medicine. 2008;5(2):260-268. 56. Olfson M, Marcus SC, Shaffer D. Antidepressant Drug Therapy and Suicide in Severely Depressed Children and Adults. Arch Gen Psychiatry. 2006;63:865-872. 57. Whittington CJ, Kendall T, Fonagy P, Cottrell D, Cotgrove A, Boddington E. Selective serotonin reuptake inhibitors in childhood depression: systematic review of published versus unpublished data. The Lancet. 2004;363(9418):1341-1345. 58. Celada P, Puig MV, Amargós-Bosch M, Adell A, Artigas F. The therapeutic role of 5-HT1A and 5-HT2A receptors in depression. Rev Psychiatr Neurosci. 2004(4):252-265. 59. Stahl S. 5HT1A receptors and pharmacotherapy. Is serotonin receptor down-regulation linked to the mechanism of action of antidepressant drugs? Psychopharmacol Bull. 1994;30(1):39- 43. 60. Stahl SM. Stahl's essential psychopharmacology: Neuroscientific basis and practical applications. Cambridge: Cambridge University Press; 2008. 61. Mahar I, Bambico FR, Mechawar N, Nobrega JN. Stress, serotonin, and hippocampal neurogenesis in relation to depression and antidepressant effects. Neurosci Biobehav Rev. 2014;38:173-192. 62. Hieronymus F, Emilsson JF, Nilsson S, Eriksson E. Consistent superiority of selective serotonin reuptake inhibitors over placebo in reducing depressed mood in patients with major depression. Mol Psychiatry. 2016;21(4):523-530. 63. Sullivan PF, Neale MC, Kendler KS. Genetic Epidemiology of Major Depression: Review and Meta-Analysis. Am J Psychiatry. 2000;157:1552-1562. 64. Caspi A, Sugden K, Moffitt TE, et al. Influence of Life Stress on Depression: Moderation by a Polymorphism in the 5-HTT Gene. Science. 2003;301:386-389. 65. Haeffel GJ, Getchell M, Koposov RA, et al. Association Between Polymorphisms in the Dopamine Transporter Gene and Depression. J Association for Psychological Science.19(1):62-69. 66. Wells KB, Golding JM, Burnam MA. Psychiatric Disorder in a Sample of the General Population With and Without Chronic Medical Conditions. Am J Psychiatry. 1988:976-981. 67. Saveanu RV, Nemeroff CB. Etiology of depression: genetic and environmental factors. Psychiatr Clin North Am. 2012;35(1):51-71. 68. Kang HJ, Voleti B, Hajszan T, et al. Decreased expression of synapse-related genes and loss of synapses in major depressive disorder. Nat Med. 2012;18(9):1413-1417. 69. Choi MJ, Lee HJ, Lee HJ, et al. Association between major depressive disorder and the - 1438A/G polymorphism of the serotonin 2A receptor gene. Neuropsychobiology. 2004;49(1):38-41. 70. McMahon FJ, Buervenich S, Charney D, et al. Variation in the Gene Encoding Serotonin 2A Receptor Is Associated with Outcome of Antidepressant Treatment. Am J Hum Genet. 2006;78. 71. Caspi A, Sugden K, Moffitt TE, et al. Influence of life stress on depression: moderation by a polymorphism in the 5-HTT gene. Science. 2003;301. 72. Heils A, Teufel A, Petri S, et al. Allelic Variation of the Human Serotonin Transporter Gene Expression. Journal of Neurochemistry. 1996;66(6):2621-2624. 73. Wendland JR, Martin BJ, Kruse MR, Lesch KP, Murphy DL. Simultaneous genotyping of four functional loci of human SLC6A4, with a reappraisal of 5-HTTLPR and rs25531. Mol Psychiatry. 2006;11:224-226.

74. Nakamura M, Ueno S, Sano A, Tanabe H. The human serotonin transporter gene linked polymorphism (5-HTTLPR) shows ten novel allelic variants. Mol Psychiatry. 2000;5:32-38. 75. Lesch KP, Bengel D, Heils A, et al. Association of Anxiety-Related Traits with a Polymorphism in the Serotonin Transporter Gene Regulatory Region. Science. 1996;274(5292):1527-1531. 76. Sugden K, Tichopad A, Khan N, Craig IW, D'Souza UM. Genes within the serotonergic system are differentially expressed in human brain. BMC Neurosci. 2009;10:50. 77. Otte C, McCaffery J, Ali S, Whooley MA. Association of a serotonin transporter polymorphism (5-HTTLPR) with depression, perceived stress, and norepinephrine in patients with coronary disease: the Heart and Soul Study. Am J Psychiatry. 2007;164(9):1379-1384. 78. Baune BT, Hohoff C, Mortensen LS, Deckert J, Arolt V, Domschke K. Serotonin transporter polymorphism (5-HTTLPR) association with melancholic depression: a female specific effect? Depress Anxiety. 2008;25(11):920-925. 79. Gage FH, Thompson RG. Differential distribution of norepinephrine and serotonin along the dorsal-ventral axis of the hippocampal formation. Brain Res Bull. 1980;5(6):771-773. 80. Brezun JM, Daszuta A. Serotonergic reinnervation reverses lesion-induced decreases in PSA- NCAM labeling and proliferation of hippocampal cells in adult rats. Hippocampus. 2000;10(1):37-46. 81. Mahar I, Tan S, Davoli MA, et al. Subchronic Peripheral Neuregulin-1 Increases Ventral Hippocampal Neurogenesis and Induces Antidepressant-Like Effects. PLOS ONE. 2011;6(10):e26610. 82. Binder DK, Scharfman HE. Brain-Derived Neurotrophic Factor. Growth Factors. 2004;22(3):123-131. 83. Tettamanti G, Cattaneo AG, Gornati R, de Eguileor M, Bernardini G, Binelli G. Phylogenesis of brain-derived neurotrophic factor (BDNF) in vertebrates. Gene. 2010;450(1-2):85-93. 84. Haniu M, Montestuque S, Bures EJ, et al. Interactions between Brain-derived Neurotrophic Facor and the TRKB Receptor. J Biological Chemistry. 1997;272(40):25296-25303. 85. Naylor RL, Robertson AG, Allen SJ, et al. A discrete domain of the human TrkB receptor defines the binding sites for BDNF and NT-4. Biochem Biophys Res Commun. 2002;291(3):501-507. 86. Mukai J, Hachiya T, Shoji-Hoshino S, et al. NADE, a p75NTR-associated cell death executor, is involved in signal transduction mediated by the common neurotrophin receptor p75NTR. J Biol Chem. 2000;275(23):17566-17570. 87. Michaelsen K, Zagrebelsky M, Berndt-Huch J, et al. Neurotrophin receptors TrkB.T1 and p75NTR cooperate in modulating both functional and structural plasticity in mature hippocampal neurons. Eur J Neurosci. 2010;32(11):1854-1865. 88. Ernfors P, Kucera J, Lee K, Loring J, Jaensch R. Studies on the physiological role of brain- derived neurotrophic factor and neurotrophin-3 in knockout mice. Int J Dev Biol. 1995;39:799-807. 89. Yoshii A, Constantine-Paton M. BDNF induces transport of PSD-95 to dendrites through PI3K- AKT signaling after NMDA receptor activation. Nat Neurosci. 2007;10(6):702-711. 90. Bartkowska K, Paquin A, Gauthier AS, Kaplan DR, Miller FD. Trk signaling regulates neural precursor cell proliferation and differentiation during cortical development. Development. 2007;134(24):4369-4380. 91. Der-Avakian A, Mazei-Robison MS, Kesby JP, Nestler EJ, Markou A. Enduring deficits in brain reward function after chronic social defeat in rats: susceptibility, resilience, and antidepressant response. Biol Psychiatry. 2014;76(7):542-549. 92. Wook Koo J, Labonte B, Engmann O, et al. Essential Role of Mesolimbic Brain-Derived Neurotrophic Factor in Chronic Social Stress-Induced Depressive Behaviors. Biol Psychiatry. 2016;80(6):469-478. 93. Berton O, McClung CA, DiLeone RJ, et al. Essential Role of BDNF in the Mesolimbic Dopamine Pathway in Social Defeat Stress. Science. 2006;311:864-868.

94. Sirianni RW, Olausson P, Chiu AS, Taylor JR, Saltzman WM. The behavioral and biochemical effects of BDNF containing polymers implanted in the hippocampus of rats. Brain Res. 2010;1321:40-50. 95. Bourdy R, Barrot M. A new control center for dopaminergic systems: pulling the VTA by the tail. Trends Neurosci. 2012;35(11):681-690. 96. Chen ZY, Bath K, McEwen B, Hempstead B, Lee F. Impact of genetic variant BDNF (Val66Met) on brain structure and function. Novartis Found Symp. 2008;289:180-195. 97. Chen ZY, Ieraci A, Teng HK, et al. Sortilin Controls Intracellular Sorting of Brain-Dervied Neurotrophic Factor to the Regulated Secretory Pathway. J Neurosci. 2005;25(26):6156-6166. 98. Chen ZY, Jing D, Bath KG, et al. Genetic Variant BDNF (Val66Met) Polymorphism Alters Anxiety-Related Behavior. Science. 2006;314(5796):140-144. 99. Ribeiro L, Busnello JV, Cantor RM, et al. The brain-derived neurotrophic factor rs6265 (Val66Met) polymorphism and depression in Mexican-Americans. Neuroreport. 2007;18(12):1291-1293. 100. Fan M, Li RH, Hu MS, et al. Association of Val66Met polymorphism at brain derived neurotrophic factor gene with depression among Chinese adolescents after Wenchuan earthquake: An 18months longitudinal study. Physiol Behav. 2017;179:16-22. 101. Colle R, Gressier F, Verstuyft C, et al. Brain-derived neurotrophic factor Val66Met polymorphism and 6-month antidepressant remission in depressed Caucasian patients. J Affect Disord. 2015;175:233-240. 102. Yan T, Wang L, Kuang W, et al. Brain-derived neurotrophic factor Val66Met polymorphism association with antidepressant efficacy: a systematic review and meta-analysis. Asia Pac Psychiatry. 2014;6(3):241-251. 103. Garcia LS, Comim CM, Valvassori SS, et al. Acute administration of ketamine induces antidepressant-like effects in the forced swimming test and increases BDNF levels in the rat hippocampus. Prog Neuropsychopharmacol Biol Psychiatry. 2008;32(1):140-144. 104. Zanos P, Moaddel R, Morris PJ, et al. NMDAR inhibition-independent antidepressant actions of ketamine metabolites. Nature. 2016;533(7604):481-486. 105. Haile CN, Murrough JW, Iosifescu DV, et al. Plasma brain derived neurotrophic factor (BDNF) and response to ketamine in treatment-resistant depression. Int J Neuropsychopharmacol. 2014;17(2):331-336. 106. Lepack AE, Fuchikami M, Dwyer JM, Banasr M, Duman RS. BDNF release is required for the behavioral actions of ketamine. Int J Neuropsychopharmacol. 2014;18(1). 107. Soliman F, Glatt CE, Bath KG, et al. A genetic variant BDNF polymorphism alters extinction learning in both mouse and human. Science. 2010;327(5967):863-866. 108. Cao B, Bauer IE, Sharma AN, et al. Reduced hippocampus volume and memory performance in bipolar disorder patients carrying the BDNF val66met met allele. J Affect Disord. 2016;198:198-205. 109. Lamb YN, Thompson CS, McKay NS, Waldie KE, Kirk IJ. The brain-derived neurotrophic factor (BDNF) val66met polymorphism differentially affects performance on subscales of the Wechsler Memory Scale - Third Edition (WMS-III). Front Psychol. 2015;6:1212. 110. Hajek T, Kopecek M, Hoschl C. Reduced hippocampal volumes in healthy carriers of brain- derived neurotrophic factor Val66Met polymorphism: meta-analysis. World J Biol Psychiatry. 2012;13(3):178-187. 111. Pezawas L, Verchinski BA, Mattay VS, et al. The brain-derived neurotrophic factor val66met polymorphism and variation in human cortical morphology. J Neurosci. 2004;24(45):10099- 10102. 112. Maass A, Duzel S, Brigadski T, et al. Relationships of peripheral IGF-1, VEGF and BDNF levels to exercise-related changes in memory, hippocampal perfusion and volumes in older adults. Neuroimage. 2016;131:142-154.

113. Hosang GM, Shiles C, Tansey KE, McGuffin P, Uher R. Interaction between stress and the BDNFVal66Met polymorphism in depression: a systematic review and meta-analysis. BMC Medicine. 2014;12(1):7. 114. Russo-Neustadt AA, Beard RC, Huang YM, Cotman CW. Physical activity and antidepressant treatment potentiate the expression of specific brain-derived neurotrophic factor transcripts in the rat hippocampus. Neuroscience. 2000;101(2):305-312. 115. Mamounas LA, Blue ME, Siuciak JA, Altar CA. Brain-derived neurotrophic factor promotes the survival and sprouting of serotonergic axons in rat brain. J Neurosci. 1995;15(12):7929-7939. 116. Mattson MP, Maudsley S, Martin B. BDNF and 5-HT: a dynamic duo in age-related neuronal plasticity and neurodegenerative disorders. Trends in Neurosciences. 2004;27(10):589-594. 117. Koolschijn PC, van Haren NE, Lensvelt-Mulders GJ, Hulshoff Pol HE, Kahn RS. Brain volume abnormalities in major depressive disorder: a meta-analysis of magnetic resonance imaging studies. Hum Brain Mapp. 2009;30(11):3719-3735. 118. Treadway MT, Waskom ML, Dillon DG, et al. Illness progression, recent stress, and morphometry of hippocampal subfields and medial prefrontal cortex in major depression. Biol Psychiatry. 2015;77(3):285-294. 119. Thompson DK, Wood SJ, Doyle LW, Warfield SK, Egan GF, Inder TE. MR-determined hippocampal asymmetry in full-term and preterm neonates. Hippocampus. 2009;19(2):118- 123. 120. Lanteaume L, Khalfa S, Regis J, Marquis P, Chauvel P, Bartolomei F. Emotion induction after direct intracerebral stimulations of human amygdala. Cereb Cortex. 2007;17(6):1307-1313. 121. Cantalupo C, Hopkins W. Asymmetric Broca's area in great apes: A region of the ape brain uncannily similar to one linked with speech in humans. Nature. 2001;29(414):6863. 122. Ross ED, Monnot M. Neurology of affective prosody and its functional-anatomic organization in right hemisphere. Brain Lang. 2008;104(1):51-74. 123. Sakaguchi Y, Sakurai Y. Left-right functional asymmetry of ventral hippocampus depends on aversiveness of situations. Behav Brain Res. 2017;325(Pt A):25-33. 124. Dwivedi Y, Rizavi H, Conley R, Roberts R, Tamminga C, Pandey G. Altered Gene Expression of Brain-Derived Neurotrophic Factor and Receptor Tyrosine Kinase B in Postmortem Brain of Suicide Subjects. Arch Gen Psychiatry. 2003;60:804-815. 125. Pandey GN, Ren X, Rizavi HS, Conley RR, Roberts RC, Dwivedi Y. Brain-derived neurotrophic factor and tyrosine kinase B receptor signalling in post-mortem brain of teenage suicide victims. International Journal of Neuropsychopharmacology. 2008;11(8):1047-1061. 126. Gonzalez-Castro TB, Salas-Magana M, Juarez-Rojop IE, Lopez-Narvaez ML, Tovilla-Zarate CA, Hernandez-Diaz Y. Exploring the association between BDNF Val66Met polymorphism and suicidal behavior: Meta-analysis and systematic review. J Psychiatr Res. 2017;94:208-217. 127. Petryshen TL, Sabeti PC, Aldinger KA, et al. Population genetic study of the brain-derived neurotrophic factor (BDNF) gene. Mol Psychiatry. 2010;15. 128. Li X, Li D, Li Q, et al. Hippocampal subfield volumetry in patients with subcortical vascular mild cognitive impairment. Sci Rep. 2016;6:20873. 129. Foo H, Mak E, Chander RJ, et al. Associations of hippocampal subfields in the progression of cognitive decline related to Parkinson's disease. Neuroimage Clin. 2017;14:37-42. 130. Moyé L. Disciplined analyses in clinical trials: the dark heart of the matter. Statistical Methods in Medical Research. 2008;17:253-264.