A model for mapping between the human and rodent cerebral cortex

Abstract

Iterative interactions between functional brain imaging in humans and mechanistic research in other animals hold great promise for understanding and treating mental health and neurological disorders. One way to accelerate discovery in this domain is to develop a unified, widely accepted vocabulary for describing structural and functional localization in the mammalian brain. To facilitate generation of such a defined nomenclature we begun by creating a table for direct mapping between terms for describing cerebral cortex parcellation according to six schemes in human, three schemes in mouse, and two schemes in rat. Use, refinement, and expansion of this model could lead to the development of a standardized structural neuroscience terminology, such as those long employed successfully in fields like chemistry, electrical engineering, and the naming of species.

“The only way to recognize the general principles of mammalian cortical structure is by means of ample comparative anatomical material comprehending the whole class” —Korbinian Brodmann (1909, translation p. 166)

Functional imaging of the human brain and genetic dissection of neural circuits and behavior in other animals are complementary neuroscience domains that are beginning to yield dramatic advances in our understanding of human cognition in health and disease, along with increasing knowledge of underlying biological mechanisms. Unfortunately, research in these two domains more often than not proceeds in relative isolation to the detriment of progress in clarifying the etiology and more effective treatment of mental and neurological disorders.

One basic reason for this isolation is the use of essentially different sets of terms for describing topography, architectural features, and localization of function in the cerebral cortex of humans and other animals. The human cortex has a relatively large surface area with convolution patterns that are quite variable among individuals and even between the right and left sides of the same individual. Human brain imaging technologies in common use today have a spatial resolution in the millimeter range (much less than the naked eye at about 0.15 mm), but these methods have the great advantage of revealing global spatial patterns of dynamic functional activity at the gross anatomical level in living, unanesthetized individuals. In contrast, experimental results in other mammals can be interpreted within a spatial framework of micrometers to nanometers using a wide range of histological technologies. As a result, most human imaging results are described in gross anatomical terms (for example, with respect to lobes, gyri, and sulci), whereas cutting-edge basic research in other mammals (most often in mice and rats, as well as in a few non-human primates) is described in terms of neurons and their distribution as revealed histologically.

The most straightforward and reliable way to compare equivalent cortical regions across mammalian taxa (for example, humans with a highly variable convoluted surface, and mice and rats with a relatively smooth surface) is through histology-level mapping combined with functional analysis because direct comparisons have shown that in humans gross gyral patterns and histological parceling schemes are poorly correlated, although recent evidence suggests that relationships between cortical regionalization and folding patterns are correlated with modular gene expression patterns during embryogenesis in different mammalian taxa (Albert & Huttner, 2015). Brodmann (1909) laid the foundation for the comparative histological approach by analyzing the cortical regionalization pattern of more than 60 mammalian species based on cytoarchitecture, and then by proposing a basic plan and common set of terms that applies to mammals generally. Over a century later, the details of mammalian cortical regionalization remain an ongoing research arena with important refinements incorporated regularly, the most recent based on gene expression patterns.

Ideally, the results of human brain imaging should stimulate the design of experiments in animals to clarify underlying biological mechanisms, and the results of hypothesis testing in animal experimentation should help interpret the results of human brain imaging in terms of underlying mechanisms—with both approaches working iteratively toward the common goal of understanding the biology of the human brain. To facilitate this synergy, the time is ripe for revisiting the question of a unified terminology to describe accurately cerebral cortical regionalization in mammals generally (see also ten Donkelaar et al., 2018).

We present here a common set of terms to describe global cortical regionalization in humans and two rodents (mice and rats). It is a working model to stimulate further refinement in future versions—it is simply Version 1.0 of a modern set of terms that hypothetically could be applied to all mammals with appropriate modifications. It is surprising that a standard set of terms like this, widely accepted by the neuroscience community, has not been developed yet (Bota & Swanson, 2008). The utility of a standardized terminology in chemistry, electrical engineering, geography, and the naming of species has been accepted for centuries. This lack of a standard, defined terminology in neuroscience, and in biology generally, is a serious impediment to progress because science and engineering rely on clear and unambiguous communication (Lazebnik, 2002).

The strategy we developed emerged from systematic connectomics analysis of cortical association and commissural macroconnections in rat, where the only nearly complete dataset is available (Swanson et al., 2018). When constructing a connection table database (connectome) for network analysis it is obviously necessary to use one internally consistent set of terms to describe the origin and termination of individual connections. If other connectomes are constructed with different sets of terms (for the same or different species), those connectomes cannot be compared directly unless relationships between the sets of terms are defined clearly— unless there is a formal mapping between the terms in each set. This type of semantic mapping would be necessary, for example, to compare directly the network organization of rodent and human intracortical connections.

Accordingly, we have constructed Supporting Information Table S1 that directly compares 11 cortical parcellation schemes from human (6 schemes), mouse (3 schemes), and rat (2 schemes), with respect to the basic mammalian plan proposed by Brodmann. There are, of course, many other parcellation schemes for all or specialized domains of the cerebral cortex in these, and many other, mammalian species, including NeuroNames (Bowden et al., 2012) and others (see, for example, Boccara et al., 2015); Shepherd et al., 2019). We have chosen to begin this project by narrowing its focus to widely recognized parcellations in the three most commonly studied mammals: human, mouse, and rat—in which the majority of current research is being conducted.

Two basic principles guided construction of Supporting Information Table S1. First, perhaps the best documented atlas of the rat brain (Swanson, 2018), in its 4^th edition, was used, and compared to a systematic and compatible terminology recently documented for the human brain (Swanson, 2015a). And second, all of the parcellation schemes were compared directly to Brodmann’s (1909, 1910) system of cortical regionalization. Foundational to Brodmann’s comparative approach was his recognition of 11 cortical regions common to all mammals, with particular species displaying more or less differentiated areal parcellation within each region.

For the human cortex, we have listed Brodmann’s original parcellation (Brodmann, 1909, 1910; Šimić & Hof, 2014), an updated account of Brodmann compiled by the authors, a recent detailed parcellation based on parameters determined with multimodal MRI parameters (Glasser et al., 2016), a standard macroscopic parcellation of gyri (Swanson, 2015a), the automated gyrus-based parcellation scheme of FreeSurfer (Desikan et al., 2006), and the Allen Human Brain Atlas—Brain Explorer 2 (Ding et al., 2016). Two standard histological atlases of the rat were compared, the 7^th edition of Paxinos and Watson (2014) and the 4^th edition of Swanson (2018). Finally, three histological atlases of the mouse were compared: the 4^th edition of Paxinos and Franklin (2013), the Allen Mouse Brain Atlas (Dong, 2008), and the Hof et al. atlas (2000).

The comparative table presented here is simply a starting point and model for attempting direct comparisons of cortical regionalization for different interpretations in the same taxonomic group, or for comparisons across human, mouse, and rat. On the other hand, it does correlate the most popular parcellation schemes in human, mouse, and rat, currently the three most intensely studied mammalian taxa. Because Supporting information Table S1 is presented in Excel format, it is easily modified by users who wish to change our interpretation of the comparisons, or to add new parcellation schemes for human, rat, and mouse, or any other taxonomic group of interest. Evidently, in this context, new, or refined parcellations or ontologic relationships may emerge from analyses of gene expression in discrete areas across species and from studies combining neuroanatomy, physiology, and behavior (e.g., Barthas & Kwan, 2017; Cembrowski et al., 2018).

Standardization of structural neuroscience terminology remains an unrealized goal, but one that is highly desirable in this era of big data and formal network analysis—in particular when databases are coupled with inference engines. For this coupling, it is essential to have a complete, internally consistent, defined vocabulary for database tables. To date, the only complete, internally consistent, hierarchically arranged, defined vocabularies for mammalian brain parts are for human (Swanson, 2015a), rat (Swanson, 2018), and mouse (Dong, 2008), and these three schemes were designed to be mutually compatible. They form the backbone of Table S1.

As consensus matures in the future, Table S1 may be useful for helping develop a standardized terminology for nervous system parts with broad community utilization. Developing a standardized terminology is a complex issue that has been considered in detail elsewhere (see Bota & Swanson, 2008, Swanson 2015a). But in short, the best standardized terminologies include a set of defined terms that is internally consistent and that completely, systematically, clearly, and usefully describes a given domain (such as the cerebral cortex, the nervous system, the cardiovascular system, and so on). Defining terms is, of course, especially difficult, and first and foremost should be based on evidence and not authority. There are two evidence-based approaches, monothetic and polythetic (Bailey, 1994). Monothetic definitions, which are fundamentally deductive, are based on one criterion, for example gyral pattern, cytoarchitecture, discrete distribution of a specific connection, or a particular gene expression pattern, whereas polythetic definitions, which are fundamentally empirical, are based on all available, critically reviewed evidence that taken together yields unique clusters (entities like cortical areas, or cortical neuron types) in parameter space (Bota & Swanson 2008, Fig. 2b). Modern approaches to defining nervous system parts, like cortical areas, commonly and appropriately are polythetic and take into account all relevant structural, functional, and genomic results.

In the end, broad community utilization of a standardized terminology for cortical regionalization, or for any other part of nervous system organization, will depend on at least three factors. First, the standard terminology itself must be useful, compared to alternate schemes. Second, the standard terminology must be available in an open-access, user-friendly format on the web (see Swanson, 2015b). And third, the standard terminology must be versioned to incorporate new knowledge, and this versioning must be supervised by expert knowledge. Finally, modern experience in other domains suggests that successful implementation of standard terminologies involves the oversight of international committees.

Supplementary Material

Supp TableS1

Table S1 (an Excel worksheet). Comparison of cerebral cortical parceling schemes in mammal generally, and in human, rat, and mouse specifically.