Neuroanatomy and Its impact on Structural and Functional Imaging (In Memory of Karl Zilles)

I will discuss how and why largely overlooked, or "forgotten," neuroanatomical features of the cerebral cortex are critical for understanding how the relationship between brain structure and function across spatial scales contribute to behavior and cognition. To do so, I will consider the classic dichotomy pitting primary vs. association cortices in which the former is considered to be much more orderly across spatial and functional scales compared to the latter. Nevertheless, in this educational course, students will learn that this classic dichotomy is actually not so clear. And instead, when considering largely overlooked anatomical structures that are largely hominoid-specific, the anatomical and functional organization of association cortices are much more orderly than a majority of classic and modern literature conveys. Thus, students have the rare opportunity to learn this information that (to our knowledge) is taught in less than a handful of universities throughout the world. For nearly all students, this may be the only opportunity to learn about these anatomical structures and their importance for understanding the relationship between brain structure and function in association cortices within the broad field of human brain mapping. 

The cerebral cortex is organized into horizontal layers. The number and delineability of these layers, the size and packing density of the cell bodies in each layer, and the width of each layer relative to other layers vary throughout the cortical ribbon and determine its cytoarchitectonic segregation. Differences in cytoarchitecture enable the definition of the phylogenetically older allocortex and the phylogenetically younger isocortex, which are separated by a transitional mesocortex. Each of these regions can be parcellated into numerous areas with a distinct cytoarchitecture. A further degree of complexity in cortical segregation is introduced by the fact that the densities of receptors for classical neurotransmitters are not only heterogeneously distributed throughout the cortex, but also vary across the depth of the cortical ribbon, and thus determine the neurochemical properties of cytoarchitectonic regions and layers.
Interestingly, the cytoarchitecture of a given cortical area can vary between species. For example, the primary visual cortex of humans and non-human primates, but not that of rodents, is characterized by an extremely differentiated laminar structure with a prominent inner granular layer which can be subdivided into sublayers IVa, IVb, IVcα and IVcβ. Likewise, the laminar distribution pattern of a given neurotransmitter receptor, e.g., the serotonin 5-HT1A receptor, throughout the cortex can differ between species. In humans the 5-HT1A receptor presents a bimodal distribution pattern within isocortical areas, with very high densities in layers I-III followed by strikingly low densities in layers IV and V and a second maximum (though considerably lower than that of superficial layers) in layer VI. In the rat isocortex, however, the 5-HT1A receptor is present at lower densities in the superficial than in the deep layers.
Since animal models are essential to further our understanding of the physiological mechanisms and processes subserving behavior, as well as the alterations associated with neurologic and neuropsychiatric diseases, my talk will address the similarities and differences that the macaque monkey and the rat, two of the most frequently used animal models, share with the human brain at the cyto- and receptor architectonic levels. Similarities are indicative of features which are conserved across species, and thus potentially highlight homolog regions across species. However, understanding the differences between species is equally important because they can influence the choice of the animal model to be used depending on the research goals. 

Over the past century, neuroscientists have sought to understand the brain, which is a highly complex organ, through genomics, transcriptomics, and analyses of cells, architecture, functions, and connectivity, and to explain the roles of these metrics in the behaviors of living humans and non-human primates (NHPs). These metrics are spanning a spatial scale of at least 106 in the brain ­ cells, multi-omics are localized at micron level. Cortical regions up to several centimeters in size have been shown by neuroimaging studies to be functional areas, as well as synchronous activity in directly and/or indirectly connected regions several tens of centimeters apart. The shape and cortical folding of the brain vary across individuals within species, as does its size and shape across species.
Obtaining accurate multiscale, multimodal brain maps at the full spatial scale is challenging. Brain mapping and parcellation of cortical areas began with illustrations and classification of cellular and myelin architectonic features, i.e., cytoarchitecture and myeloarchitecture (Brodmann 1905, Vogt & Vogt 1919, von Economo & Koskinas 1925, Flechsig 1920; Hopf 1955). (Brodmann 1905; von Economo & Koskinas 1925). But spatial scale calculations and registration were performed manually on paper. Microscopic brain mapping is being facilitated by the recently introduced techniques of cell typing, receptor mapping (Zilles & Amunts 2009; Palomero-Gallagher 2020) and spatial transcriptomics (Hawrylycz et al., 2012; Bakken et al., 2021). On the other hand, early neuroimaging research used magnetic resonance imaging (MRI) and computational approaches for volume registration with functional mapping; in particular, functional MRI data were mapped to 3D coordinates (Neurosynth, Yarkoni et al., 2011). More recently, multimodal 2D and 3D cartographic data have been mapped to the cortical surface or subcortical architecture (Glasser et al., 2016). This approach requires specialized tuning of 3-Tesla MRI data, with preprocessing involving surface reconstruction to accurately map multimodal data to a novel brain coordinate system (‘greyordinates’). The multimodal data include an MRI-based myelin map derived by dividing T1-weighted values by T2-weighted volumes (Glasser & Van Essen 2011), as well as resting-state functional connectivity (Glasser et al., 2016), diffusion tractography (Sotiropoulos et al., 2013) and neurite maps (Fukutomi et al., 2018). A similar strategy has been applied to non-human primates including chimps (Glasser et al., 2014), macaques (Donahue et al., 2016; Autio et al., 2020), marmosets (Hayashi et al., 2021; Ose et al., 2022) and night monkeys (Ikeda at al., in press) using optimized protocols and pipelines. Studies of cortical myelination in various primate species have provided deep insight into the cortical evolution of functional organization, particularly in early areas (Van Essen et al., 2019). Resting-state functional connectivity studies of primates provide rich information on cortical evolution, particularly in association areas (Vincent et al., 2007; Hayashi et al., 2021; Yokoyama et al., 2022). Future studies should aim to address questions related to cellular features, transcriptomics and tracer mapping and how they relate to macroscopical functions, structures, and connectivity obtained in MRI. Validation of MRI-based neurometrics and volume (3D) and surface (2D) registrations are also important to understand the evolution of the primate brains. Recent progress in myelin layer mapping and diffusion neurite mapping for NHPs will be also discussed during the lecture. 

Classic descriptive embryology was based on microscopic observations of serial sections of human embryos that were stained with histochemistry techniques, like Nissl staining. Those observations described the progressive development of tissues and organs and traced back the origin of adult body parts to their corresponding embryological primordia. Serial sections were used to elaborate 3D reconstructions of embryos at different stages of development that are still reproduced in contemporary embryology and anatomy textbooks. Descriptive studies of the neural tube in human and other vertebrate embryos identified five dilatations or vesicles from the caudal myelencephalon to the rostral telencephalon. These studies also identified four plates that extended longitudinally across the neural tube from rostral to caudal: floor, alar, and roof plates extended across the entire neural tube, but, rostral to the mesencephalon, the basal plate was supposed to be absent; instead, in the ventricular surface of the diencephalon descriptive embryologists identified four dorsoventral “columnar” bulges that originated the epithalamus, the dorsal thalamus, the ventral thalamus, and the hypothalamus. Vesicles, plates, and columns were supposed to be homologous histogenetic units of the neural tube across vertebrate species.
The observations of descriptive embryology of the neural tube served to enunciate the Columnar Model of Charles H. Herrick and Hartwig Kuhlenbeck. This model was not seriously questioned until the last decade of the 20th century when the Prosomeric Model of Luis Puelles was proposed to explain the embryological development and evolution of the central nervous system (CNS). Contrary to the Columnar Model, the Prosomeric Model is based on observations of vertebrate embryos stained in toto with in situ hybridization techniques that label the expression of morphogenetic genes. This approach is called genoarchitectonics and has been paramount in contemporary studies of developmental biology to provide a causal basis for the definition and naming of the developmental parts of the CNS. Modern genoarchitectonic studies show that the neural tube is regionalized into sectors that are specified in primary events of patterning induced by organizers. The organizers are groups of cells that secrete morphogenetic proteins to the extracellular space. These proteins diffuse around the organizer and modify the expression of genes in responding neuroepithelial cells that are induced to commit themselves to one among several predetermined fates. Primary events of patterning in the neural plate and early neural tube result in the specification of radial histogenetic domains, which are defined based on distinct molecular profiles identified by genoarchitectonic techniques and are considered fundamental morphological units (FMUs) of a bauplan of the CNS that is common to all vertebrates, including humans.
More precisely, genoarchitectonic studies identified several neuromeres (rhombomeres, mesomeres, and prosomeres) across the entire rostrocaudal extent of the neural tube and showed that the basal plate extends into the diencephalon and contributes to the development of the hypothalamus. According to the Prosomeric Model, neuromeres and plates intersect in the neural plate and form a mosaic of FMUs that conserve the same neighborhood (topological) relations in all vertebrates so far examined. Thus, in all vertebrates, the thalamus originates from four FMUs in the alar plate of prosomere 2, the reticular nucleus of the thalamus from one FMU in the alar plate of prosomere 3, and the hypothalamus from several FMUs in the floor, basal, and alar plates of prosomeres hp1, hp2, and the acroterminal domain.
The definition of the constituent parts of the vertebrate CNS, including humans, based on their origin in FMUs opens new perspectives for basic and clinical neuroscience. For instance, FMUs allow for proposing homology hypotheses of brain parts across vertebrate species, which is key for translating to humans the experimental data obtained in laboratory animals. Tracing the origin of brain parts to their corresponding FMUs will also point at selective vulnerabilities of neural circuits to developmental disruption. However, there is no atlas of the adult human brain that defines the position and limits of brain parts according to their origin in FMUs. Such an atlas could not be based on data obtained from genoarchitectonic studies of early human embryos because they are scarce. On the other hand, the preservation of invariant topological relations between FMUs across vertebrates allows for extrapolating data obtained in early mice embryos to the brain of human adults. In this communication we will show how to extrapolate the invariant topological relations between FMUs and their adult derivatives in mice to the adult human brain. The aim of this approach is tracing FMU of origin of each constituent part of the adult human brain. 

In biology, structure and function are inextricably linked. Much of neuronal function is determined by anatomical connections—the brain’s ‘wiring diagram.’ Moreover, most brain disorders are understood to be problems not confined to the cells of a particular
region, but distributed through the communication among multiple brain regions. Thus, they are essentially connectionist disorders. My laboratory’s ultimate goal is to build the wiring diagram of the human brain. However, this is not a straightforward task: the gold standard for assessing anatomical connectivity, tract-tracing, is not possible in humans, and requires combining across brains to infer whole-brain connectivity. Other methods are either post-mortem (label-free optical imaging) and/or non-invasive (diffusion MRI); however, these have their own challenges with resolution and accuracy.

I will highlight how deliberately cross-species and cross-modal pipelines can help us achieve more accurate wiring diagrams of the human brain. These diagrams can aid us in neuromodulation and provide us with the neuroanatomical underpinnings of complex behaviors and resting-state fMRI results. Finally, I will describe the use of connectivity as a defining metric of brain regional similarity across species, including rodents. The coming years will require comparing connectivity maps across nonhuman animals and humans to establish translational value. 

Development of cortical tissue during infancy and childhood is critical for the emergence of typical brain functions in cortex. Moreover, understanding cellular and molecular mechanisms underlying cortical tissue development has important implications for the diagnosis of neurodevelopmental disorders and delays. This lecture has three learning goals. Using the visual cortex as a test bed, first I will introduce concepts of multimodal quantitative magnetic resonance imaging (qMRI) and discuss how this method has enriched our ability to track earliest structural developments in primary sensory and higher-order visual cortex. I will also discuss how scanning protocols and analysis pipelines can be developed to study gray matter development in tiny infant brains as well as in child and adult brains. Second, we will learn how quantitative MR metrics can serve as proxies for mechanisms like cortical myelination by examining recent efforts in combining post-mortem, histological data with MR measures to obtain ground truth for various biological processes. Third, I will discuss how open-source, transcriptomic gene database such as the Brain Span Atlas can provide an excellent means to examine the genetic landscape of cortical development during various stages of human life, including prenatal, postnatal, and childhood development. Overall, this lecture will underscore the importance of using multimodal measurements of microstructural and morphological changes of brain tissue and provide an overview of the tools that can be used to uncover the developmental trajectories of biological mechanisms in different brain regions across the human lifespan.