Functional differences in spatial processing along an anterior-posterior axis: are there implications for the retrosplenial cortex?

Abstract

The retrosplenial cortex has recently received attention from the neuroscience community for its role in spatial processing and involvement in diseases such as Alzheimer's. Here, we discuss a recent study by Silson et al. (Silson EH, Gilmore AW, Kalinowski SE, Steel A, Kidder A, Martin A, Baker CI. J Neurosci 39: 705–717, 2019.) that reported functionally specific activation within this region during scene perception and (mnemonic) construction. We then propose considerations for future experiments such as adopting standardized methodology and terminology that may improve the interpretation of retrosplenial cortex function within the broader literature.

 The retrosplenial cortex (RSC) may, in part, underlie higher-order executive cognitive processes crucial for mammalian species to successfully orient and navigate within their environment, yet its precise function remains poorly understood. The RSC is located in Brodmann’s areas 29 and 30 (Brodmann and Garey 2005) among the cingulate cortex with cortical connections projecting to the (para)hippocampi, the anterior thalamic nuclei, and the parietal and prefrontal cortices. In humans, RSC function has been implicated in various cognitive processes including scene perception (MacEvoy and Epstein 2007), spatial processing (Epstein 2008), and autobiographical memory retrieval (Maguire 2001). Moreover, results from recent studies suggest the RSC may be a potentially important region in the development of maladaptive neurophysiological changes associated with prodromal Alzheimer's disease (AD) (Minoshima et al. 1997). A comprehensive understanding of the RSC’s function and adjoining regions may improve parcellation of these regions, and, in turn, better localization of regions that contribute to cognitive deficits seen in AD.

Experimental studies involving rodents have shown that the anterior and posterior regions of the RSC may underlie distinct spatial processes. For instance, lesion studies have indicated that more posterior regions of the RSC may be important for visually guided spatial memory and navigation, whereas anterior regions may be significant for internally directed navigation (see Vann et al. 2009 for review). Electrophysiology results further support the heterogeneity of anterior-posterior RSC function by reporting that only neurons within the anterior RSC are activated due to location, direction, and movement stimuli (Vann et al. 2009). While collectively these animal studies suggest anterior and posterior RSC may underlie specific aspects of spatial behavior, it remains unclear whether this heterogeneous organization persists within the human RSC.

A recent study published in the Journal of Neuroscience (Silson et al. 2019) used functional magnetic resonance imaging (fMRI) to investigate whether there exist separable regions within the RSC involved in processing perceptual and mnemonic information. The authors hypothesized that perceptual-driven scene information is processed in a more posterior portion of the RSC the authors termed the “medial place area” (MPA), whereas memory-based scene information is processed in a more anterior, resting-state connectivity-defined region (CON). To test their hypotheses, the authors developed a double-dissociation experimental paradigm, where 19 young adults (13 female) completed two task-based fMRI experiments: a mnemonic scene construction and a scene perception task. Task-related activations were then correlated with two regions of interest (ROIs) the authors defined using functional (MPA) and connectivity-based (CON) measurements. During the mnemonic task, participants were presented with a commonplace word or phrase and cued either to remember a unique autobiographical event or to imagine a potential future event and to reflect on this throughout the duration of the 10-s trial. The scene perception task was used as a localizer to functionally define the MPA ROI. During this task, participants were alternately presented with images of scenes or faces and asked to respond when the same image appeared sequentially. The MPA for each participant was defined as the region with the greater response to scenes than faces, averaged across two scan runs. Their results indicate the MPA was consistently located in the ventral and posterior bank of the parietal-occipital sulcus (POS) across all subjects but importantly did not extend into the retrosplenial cortex proper. The CON ROI was located largely anterior to the MPA, although there was approximately a 39% and 43% spatial overlap between the two regions in each hemisphere, respectively. To address this overlap, Silson et al. (2019) restricted their ROI analyses to regions that were distinctly CON and MPA in each participant.

Consistent with their hypotheses, Silson et al. (2019) reported that memory-based constructions of past/future events evoked greater responses in the CON region than MPA, with a left-hemispheric bias. They observed responses to mnemonic and perceptual stimuli in both regions, but the pattern of responses suggested that the fundus of the POS may be a transitional neuroanatomical landmark for these representations. Based on their results, the authors posit that finer parcellations of functional subregions within the RSC are needed for more precise labeling.

The results of this study provide potentially important insights regarding localized cortical activations during spatial perception and mnemonic construction and, importantly, highlight the field’s need for adopting universal nomenclature within the broader RSC literature. In the remainder of this article, we briefly discuss Silson et al. (2019)’s methodological choices and their implications for comparison across studies and propose considerations for future experiments that may expand their findings. First, we review how the authors’ results are consistent with other neuroimaging findings that support the idea that mnemonic and perceptual spatial information is processed in possibly distinct areas and that, collectively, this large body of work could benefit from future studies establishing a clear and concise delineation of these separate functional subregions. We then discuss how the authors defined the ROIs used in their experiment and evaluate the extent to which this may affect generalizability to results from related studies as well as interpretation of RSC function. Finally, we discuss our understanding of cognitive and neuropathological changes associated with spatial processing deficits in diseases such as AD and how a clearer understanding of the RSC’s role in spatial processing could aid in localizing biomarkers associated with prodromal AD.

First, the broad scope of the authors’ findings aligns with those of other recent neuroimaging studies that have reported similar functional-based parcellations along an anterior-posterior axis in nearby cortical regions. For example, a recent study by Hutchison et al. (2015) examined the medial parieto-occipital cortex (mPOC), a broad region that includes the RSC and has been implicated in tasks such as spatial navigation, reaching and grasping, and higher-order visual processing. Their group collected resting-state fMRI from human and nonhuman primates and found that in both species the mPOC could be reliably divided along an anterior-posterior gradient that extends across the POS and that humans specifically demonstrated unilateral activation.

However, a remaining question of Silson et al.’s (2019) report regards the authors’ conclusions that there may exist a functional heterogeneity between anterior and posterior regions within the RSC. We suggest that, due to the ambiguous location of the anterior and posterior ROIs used in their experiment, the neural underpinnings of the observed anterior-posterior responses within the RSC remain unclear. As only a portion of the CON (and, importantly, none of the MPA) ROI was located within RSC proper, it remains unclear whether anatomically distinct regions within the RSC were driving the observed responses or whether these responses arose from anterior and posterior regions of the medial parietal cortex. Indeed, throughout the literature the RSC/parietal-occipital sulcus regions have been defined using cytoarchitechtonic, resting-state connectivity, and tasked-based functional data, creating a nebulous definition of these regions (see Epstein 2008 for review), which potentially confounds its boundary with surrounding cortical (e.g., parietal) structures. Because the RSC is anatomically constrained to Brodmann’s areas 29 and 30 (Brodmann and Garey 2005), future analyses could apply a Brodmann's atlas to neuroimaging data to verify the ROIs evaluated are located within the RSC proper and possibly further support Silson et al.’s (2019) assertion that finer parcellations of these subregions are needed for more precise labeling.

Another question that arises from Silson et al.’s (2019) report regards the difference in methodology used to define the two ROIs used in their experiment. For instance, the CON ROI was defined as the group-averaged region with stronger resting-state fMRI connectivity with anterior (versus posterior) parahippocampal place area (i.e., the ROI was not specific to each subject). Using group-averaged activation to quantify a sample-based ROI may be potentially problematic because blood oxygen level-dependent activation can vary considerably across scanning sessions and individuals (Swallow et al. 2003), whereas averaging across runs and using subject-specific data can lead to greater reliability in ROI localization. While the authors did use subject-specific data averaged across scans to define the MPA, the intra- and intersubject variance of this ROI was unreported such that it remains unknown whether the MPA region was in a consistent location and size between scans, as well as across the group. Reporting this information could avoid questions regarding data reliability, quality control, etc., and inform future neuroimaging studies during the experimental design process. Furthermore, the high degree of overlap between the ROIs and their responsiveness to both mnemonic and perceptual stimuli in the authors’ results may point to a processing continuum as you move along the anterior-posterior axis, rather than distinct, separable regions.

While the authors’ findings clearly demonstrate a differential response in spatial processing within the RSC and medial parietal cortex, analyzing subregions within each respective ROI highlights a potentially problematic issue of generalizability to other data sets, although this is not a unique predicament to Silson et al. (2019). In fact, in a meta-analysis of spatial-cognitive processing in the RSC, it was found that many previous fMRI studies omitted regions belonging to the RSC or included neighboring regions (i.e., the definition of the RSC varied widely across studies depending on the functional measures examined) (Dillen et al. 2014). In brief, the CON and MPA regions used in this study may not be easily transferred to other laboratories or researchers, and the ambiguity imbued therein creates a challenge when comparing results between studies, which may impede a broader understanding of the overall function of these areas. It will be important for future studies to distinguish connectivity-, functionally, and anatomically derived sources of ROI boundaries (i.e., to develop a standard RSC mask that can be adopted among research laboratories) to create a consistent lexicon within the broader RSC and medial parietal neuroimaging literature.

Finally, if there are clearly defined regions underlying mnemonic and perceptual processing within the RSC, it could help clarify the role RSC dysfunction plays in the visuospatial impairments observed in AD and may potentially provide an interventional target for preventative or palliative therapies in prodromal AD. While the hippocampus has long been a major focus of research on the pathology of AD, results from recent animal and human studies suggest that the RSC may play a larger role in amyloid accumulation associated with prodromal AD than previously thought. Briefly, the beta-amyloid (Aβ) hypothesis posits that these stages may progress due to the accumulation of Aβ protein fragments, found prominently in medial and lateral parietal cortices, precuneus, frontal regions, and hippocampus (Rodrigue et al. 2009). Both clinical and academic research often uses combinations of behavioral or cognitive testing and neuroimaging data to quantify AD progression, as well as to discriminate between mild cognitive impairment and prodromal and later stages of AD. The focus of drug treatments (and other therapies) for AD has largely shifted toward targeting disease progression at the prodromal stage, which relies on quantification of AD biomarkers such as Aβ burden. Importantly, one putative early clinical sign of AD is a rapid decline in visuospatial function, where scores on clinical visuospatial tests correlate with cortical metabolic changes observed in people with AD (Melrose et al. 2013) and can reliably differentiate mild cognitive impairment from AD (Malloy et al. 2003). Results from other behavioral, electrophysiological, and neuroimaging studies relating visuospatial construction and perception to AD have led some researchers to view it as an early biomarker of prodromal AD (see Mandal et al. 2012 for review), implicating the RSC and other nearby parietal areas as potential therapeutic targets for AD intervention. In fact, Poirier et al. (2011) report that within mouse models, the RSC may develop Aβ precursor proteins prior to other cortical regions typically associated with Aβ accumulation in AD (e.g., hippocampus). Similarly, several clinical studies in humans have reported that the earliest hypometabolic activity in patients with prodromal AD occurs within the RSC and the adjacent posterior cingulate cortex (Minoshima et al. 1997).

Again, the nonuniform definition of the RSC (and other nearby regions) and lack of a definitive neuroimaging atlas for the region complicate translating results across studies. If an anterior-posterior division in functional processing within the RSC/medial parietal areas exists, clearly delineating parcellation by functionality is critical for creating a localized region to focus early detection of neuroimaging biomarkers and therapeutic treatments for AD. Future work is needed to fully understand how the RSC is involved in processing spatial information (such as perception, mnemonic construction, and navigation) to help in this parcellation and to provide insight into the location of neurodegenerative visuospatial changes seen in prodromal AD. A future study that examines the relationship between an individual’s visuospatial function (e.g., visual perception and construction) and neuroimaging measures of the RSC (e.g., cortical volume, Aβ deposition) across different stages of AD may implicate the RSC as an important target for therapeutic intervention of AD pathology.

Perspectives

Consistent with previous work, Silson et al. (2019) provide empirical evidence that implicate the medial parietal and retrosplenial cortices in processing distinct streams of scene-related information in humans and that the fundus of the parietal-occipital sulcus may serve as possible transitional point for these functions. Moreover, activations in left-hemispheric anterior regions of the RSC were only observed during the mnemonic task. Their results highlight that the broader RSC literature would benefit from future studies that employ standardized methodology and terminology to improve the specificity and interpretation of RSC function.

GRANTS

J. Lingo VanGilder is supported in part by the National Institute on Aging at the National Institutes of Health (F31 AG062057) and the Achievement Rewards for College Scientists Foundation (Spetzler Scholarship).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.L.V. and P.V. conceived and designed research; J.L.V. and P.V. drafted manuscript; J.L.V. and P.V. edited and revised manuscript; J.L.V. and P.V. approved final version of manuscript.