Chimpanzees (Pan troglodytes) along with bonobos (Pan paniscus) represent the closest extant relatives of humans sharing a common ancestor approximately 7–8 million years ago (Langergraber et al., 2012). Experimental and observational studies, in both the field and in captivity, have documented a range of cognitive abilities that are shared with humans such as tool use and manufacturing (Shumaker et al., 2011), symbolic thought (de and Frans, 1996), mirror self-recognition (Anderson and Gallup, 2015; Hecht et al., 2017) and some basic elements of language (Savage-Rumbaugh, 1986; Savage-Rumbaugh and Lewin, 1994; Tomasello and Call, 1997) like conceptual metaphorical mapping (Dahl and Adachi, 2013). This cognitive complexity together with similar neuroanatomical features (Zilles et al., 1989; Rilling and Insel, 1999; Gómez-Robles et al., 2013; Hopkins et al., 2014; Hopkins et al., 2017) and genetic proximity (Waterson et al., 2005) renders these species unique among non-human primates to study the evolutional origins of the human condition. In view of evolutionary neurobiology, the relatively recent divergence between humans and chimpanzees explains the striking similarities in major gyri and sulci, despite profound differences in overall brain size. Numerous studies using magnetic resonance imaging (MRI) have compared relative brain size, shape, and gyrification in humans and chimpanzees (Zilles et al., 1989; Rilling and Insel, 1999; Gómez-Robles et al., 2013; Hopkins et al., 2014; Hopkins et al., 2017).

Previous studies of brain aging in chimpanzees have reported minimal indications of atrophy (Herndon et al., 1999; Sherwood et al., 2011; Chen et al., 2013; Autrey et al., 2014). Nevertheless, Edler et al., 2017 recently found that brains of older chimpanzees’ exhibit both neurofibrillary tangles and amyloid plaques, the classical features of Alzheimer’s disease (AD). Neurodegeneration in the aging human brain includes marked atrophy in frontal and temporal lobes and decline in glucose metabolism even in the absence of detectable amyloid beta deposition, which increases the likelihood of cognitive decline and development of AD (Jagust, 2018). Given the strong association of brain atrophy and amyloid beta in humans, this phenomenon requires further investigation in chimpanzees.

Cortical asymmetry is a prominent feature of brain organization in many primate species (Hopkins et al., 2015) and was recently shown in humans in a large-scale ENIGMA (Enhancing Neuroimaging Genetics through Meta-Analysis) study (Kong et al., 2018). For chimpanzees, various studies have reported population-level asymmetries in different parts of the brain associated with higher order cognitive functions like tool-use (Freeman et al., 2004; Hopkins et al., 2008; Hopkins et al., 2017; Hopkins and Nir, 2010; Lyn et al., 2011; Bogart et al., 2012; Gilissen and Hopkins, 2013) but these results are difficult to compare within and across species, due to the lack of standardized registration and parcellation techniques as found for humans.

To date, there is no common reference space for the chimpanzee brain available to reliably associate and quantitatively compare neuro-anatomical evidence, nor is there a standardized image processing protocol for T1-weighted (T1w) brain images from chimpanzees that matches human imaging standards. With the introduction of voxel-based morphometry (Ashburner and Friston, 2000) and the ICBM (international consortium of brain mapping) standard human reference brain templates almost two decades ago (Mazziotta et al., 2001), MRI analyses became directly comparable and generally reproducible. In this study, we adapt state-of-the-art MRI (magnetic resonance imaging) processing methods to assess brain aging and cortical asymmetry in the chimpanzee brain. To make this possible, we rely on the largest openly available resource of chimpanzee MRI data: the National Chimpanzee Brain Resource (NCBR, http://www.chimpanzeebrain.org/), including in vivo MRI images of 223 subjects from 9 to 54 years of age (Mean age = 26.9 ± 10.2 years). The aim of this study is the creation of a chimpanzee template permitting automated and reproducible image registration, normalization, statistical analysis, and visualization to systematically investigate brain aging and hemispheric asymmetry in chimpanzees.

Initially, we created the population-based Juna.Chimp (Forschungszentrum Juelich - University Jena) T1-template, tissue probability maps (TPM) for tissue classification and a non-linear spatial registration ‘Shooting’ templates (Figure 1) in an iterative fashion at 1 mm spatial resolution. The preprocessing pipeline and templates creation were established using the freely available Statistical Parametric Mapping (SPM12 v7487, http://www.fil.ion.ucl.ac.uk/spm/) software and Computational Anatomy Toolbox (CAT12 r1704 http://www.neuro.uni-jena.de/cat/). Juna.Chimp templates, the Davi130 parcellation, gray matter (GM) masks utilized, as well as statistical maps from our analysis are interactively accessible and downloadable via the Juna.Chimp web viewer (http://junachimp.inm7.de/).

To enable more direct comparison to previous research, we manually created the Davi130 parcellation (by R.D. and S.V.), a whole brain macroanatomical annotation based on the Juna T1 template (Figure 2). The delineation of regions within the cortex was determined by following major gyri and sulci, whereby, large regions were arbitrarily split into two to three sub-regions of approximate equal size even though histological studies show that micro-anatomical borders between brain regions are rarely situated at the fundus (Sherwood et al., 2003; Schenker et al., 2010; Spocter et al., 2010; Amunts and Zilles, 2015). This process yielded 65 regions per hemisphere for a total of 130 regions for the Davi130 macro-anatomical manual parcellation (Figure 2 and Figure 2—source data 1).

Following successful CAT12 preprocessing, rigorous quality control (QC) was employed to identify individual MRI scans suitable for statistical analysis of brain aging and hemispheric asymmetry in chimpanzees. Our final sample consists of 194 chimpanzees including 130 females with an age range of 9–54 years and a mean age of 26.3 ± 9.9 years (Figure 3A). The linear regression model with GM fraction of total intracranial volume as the dependent variable and age, scanner field strength, sex, and rearing environment revealed a significant negative association between age and GM (p<0.0001) demonstrating age-related decline in overall GM density (Figure 3B). Both sex (p=0.004) and scanner field strength (p<0.0001) showed a significant effect on total GM volume. Therefore, the sample was split into male and female subjects and into 1.5T and 3T scanner, whereby, all sub-samples showed a significant age effect on GM (male: R² = 0.17, p=0.0004; female R² = 0.13, p<0.0001, 1.5T: R² = 0.19, p<0.0001; 3T: R² = 0.09, p=0.004). There were no significant sex differences of GM decline (p=0.3). The same analysis was conducted on a matched human sample from the IXI dataset (Figure 3C; https://brain-development.org/ixi-dataset/). The human sample was matched based on age, sex, and scanner field strength (n = 194, 128 females, 20–78 y/o, mean = 39.4 ± 14.0). As life span and aging processes are different between species, the human sample was matched to chimpanzees roughly by using a factor of 1.5* for age. A significant age-related decline in overall GM (p<0.0001) as well as a significant sex effect (p<0.0001) was also found in the human sample (Figure 3D). Similar to the chimpanzee sample, both males and female subjects show a significant age effect on total GM (male: R² = 0.58, p<0.0001; female: R² = 0.61, p<0.0001) but with no significant sex differences on GM decline (p=0.8). Although both species present a significant age-related GM decline, humans show a higher negative correlation between age and GM (chimpanzee: R² = 0.12; human: R² = 0.55) with less variance as compared to chimpanzees.

Region-based morphometry analysis was applied to test for local effect of age on GM. Linear regression analyses identified 55 of 130 brain regions in the Davi130 parcellation across both hemispheres that were significantly associated with age after family-wise error (FWE) correction for multiple testing (Figure 4 and Figure 4—source data 1). Specifically, GM decline with age was found bilaterally in the superior frontal gyrus (SFG), posterior middle frontal gyrus (pMFG), posterior inferior frontal gyrus (pIFG), lateral orbitofrontal cortex (lOFC), middle and inferior precentral gyrus (PrCG), cingulate gyrus (ACC, MCC, PCC), posterior superior temporal gyrus (pSTG), anterior middle temporal gyrus (aMTG), precuneus (PCun), and lingual gyrus (LG) as well as unilaterally in the right anterior insula (aIns) and middle inferior frontal gyrus (mIFG), in addition to the left superior precentral gyrus (sPrCG), anterior transverse temporal gyrus (aTTG), posterior transverse temporal gyrus (pTTG), paracentral lobule (PCL) and the area around the calcarine sulcus (Calc) within the cerebral cortex. Subcortically, age-related GM decline was found in the bilateral putamen (Pu), caudate nucleus (CN), and the nucleus accumbens (NA), as well as in the superior cerebellum (CerVI, CerIV, CerVA, Cer VB, CerIV and right CrusII). Finally, to test for more fine grained effects of aging independently of our macroanatomical parcellation, the same sample was analyzed with VBM revealing additional clusters of GM that are significantly affected by age in chimpanzees (Figure 5) after FWE correction using threshold-free cluster enhancement (TFCE) (Smith and Nichols, 2009). On top of the regions identified by region-wise morphometry, we found extensive voxel-wise effects throughout the orbitofrontal cortex (OFC), inferior temporal gyrus (ITG), transverse temporal gyrus (TTG), frontal operculum (FOP), parietal operculum (POP), postcentral gyrus (PoCG), supramarginal gyrus (SMG), angular gyrus (AnG), and in parts of the superior parietal lobule (SPL), superior occipital gyrus (sOG), and in inferior parts of the cerebellum.

Hemispheric asymmetry of the chimpanzee brain was assessed for each cortical Davi130 region with a total of 68% (44/65) exhibiting significant cortical asymmetry after FWE correction (Figure 6 and Figure 6—source data 1). The majority of regions were found with greater GM volume in the right hemisphere (n = 32) as compared to the left (n = 12). In the left hemisphere, we found more GM in the SFG, pMFG, insula, anterior TTG, and PCun within the cortex. Rightward cortical asymmetry was located in the anterior MFG, middle and posterior IFG, medial OFC, cingulate gyrus, amygdala, STG, MTG, posterior TTG, anterior and posterior fusiform gyrus (FFG), FOP, POP, middle PrCG, middle and inferior PoCG, SMG, AnG, Calc, as well as the middle occipital gyrus. Within the basal ganglia, leftward GM asymmetry was observed in the Pu, nucleus accumbens (NA), basal forebrain nucleus (BF), and globus pallidus (GP), while, rightward asymmetry in the caudate nucleus (CN) and thalamus (Th). The cerebellum exclusively showed rightward GM asymmetry in the posterior cerebellar lobe (CerIX, CerVIII, CrusI, CrusII, CerVI). The hemispheric asymmetry did not show a decipherable pattern.

As a common reference space for the analysis of chimpanzee brain data, we created the Juna.Chimp template, constructed from a large heterogeneous sample of T1w MRI’s from the NCBR. The Juna.Chimp template includes a reference T1-template, along with probability maps of brain and head tissues accompanied by a Geodesic Shooting template for the publicly available SPM12/CAT12 preprocessing pipeline to efficiently segment and accurately spatially normalize individual chimpanzee T1w images. The T1-template and TPM can also be used as the target for image registration with other popular software packages, such as FSL (https://fsl.fmrib.ox.ac.uk/fsl) or ANTs (http://stnava.github.io/ANTs/). Furthermore, our processing pipeline and templates can be utilized for data-driven approaches to create connectivity-based and structural covariance parcellations of the chimpanzee brain (Alexander-Bloch et al., 2013; Eickhoff et al., 2015).

Additionally, we provide the manually segmented, macro-anatomical Davi130 whole-brain parcellation comprising 130 cortical, sub-cortical and cerebellar brain regions, which enables systematic extraction of volumes-of-interest from chimpanzee MRI data. The image processing pipeline and Davi130 parcellation were used to investigate ageing and interhemispheric asymmetry in the chimpanzee brain. The Davi130 parcellation was realized by utilizing macroscopic gyral and sulcal features such as peaks, fundi, and bends to represent the chimpanzee brain in a reduced dimensional space based on macro-anatomical landmarks. Despite the evidence that true microanatomical borders between brain areas rarely coincide with macro-anatomical patterns (Amunts and Zilles, 2015), macroscopic brain parcellations like Desikan-Killany human atlas (Desikan et al., 2006) have successfully been utilized in many studies furthering our understanding of brain structure, function, and disease (Kong et al., 2020; van den Heuvel et al., 2020). The Davi130 parcellation of the chimpanzee brain serves two main purposes. First, regions in Juna.Chimp template space enable increased interpretability and reproducibility of morphometric analyses and comparability between studies, even retrospectively. Second, our manual subdivision reduces the statistical problem of multiple testing for mass univariate approaches like VBM to uncover subtle brain - behavior relationships. Furthermore, our macroscopic parcellation mitigates the curse of dimensionality for multivariate machine learning methods to be applied to the relatively small samples like the NCBR.

We found clear evidence of global and local GM decline in the aging chimpanzee brain even though previous research into age-related changes in chimpanzee brain organization has shown little to no effect (Herndon et al., 1999; Sherwood et al., 2011; Chen et al., 2013; Autrey et al., 2014). (Herndon et al., 1999; Sherwood et al., 2011; Chen et al., 2013; Autrey et al., 2014). This can be attributed on the one hand to the larger number of MRI scans available via the NCBR including 30% of older subjects with 55 individuals over 30 and 12 over 45 years of age, which is crucial for modelling the effect of aging (Chen et al., 2013; Autrey et al., 2014). On the other hand, state-of-the-art image processing enabled the creation of the species-specific Juna.Chimp templates, which largely improves tissue segmentation and registration accuracy (Ashburner and Friston, 2000). Non-linear registration was also improved by the large heterogeneous sample utilized for the creation of the templates encompassing a representative amount of inter-individual variation. We used the well-established structural brain imaging toolbox CAT12 to build a reusable chimpanzee preprocessing pipeline catered towards analyzing local tissue-specific anatomical variations as measured with T1w MRI. The Davi130-based region-wise and the voxel-wise morphometry analysis consistently showed localized GM decline in lateral frontal cortex, lOFC, precentral gyrus, cingulate gyrus, PCun, medial parietal and occipital cortex, the basal ganglia, and superior cerebellum. The VBM approach additionally produced evidence for age effects in bilateral mOFC, PoCG, inferior temporal regions, inferior and superior lateral parietal cortex, sOG, and throughout the cerebellum. These additional effects can be expected, as VBM is more sensitive to GM changes due to aging (Kennedy et al., 2009). The multiple brain regions revealing GM decline reported here in both approaches have also been shown to exhibit GM atrophy during healthy aging in humans (Good et al., 2001b; Kennedy et al., 2009; Crivello et al., 2014; Minkova et al., 2017). Additionally, there was no significant difference in the age-related decline between humans and chimpanzee (Figure 3), even though a larger negative correlation with less variance was found in the matched human sample, which demonstrates a commonality in the healthy aging process of chimpanzees that was thought to be specific to humans. In general, GM atrophy in chimpanzees occurs across the entire cortex, sub-cortical regions, and cerebellum, however, certain local areas decline at a relative extended rate within the frontal, temporal, and parietal lobes (Fjell et al., 2014). Several Davi130 regions within the frontal lobe (SFG, MFG, IFG, and lOFC) have been previously reported in corresponding human loci in relation to GM volume decline due to aging (Kennedy et al., 2009; Crivello et al., 2014; Fjell et al., 2014; Minkova et al., 2017). Furthermore, aging effects in temporal (STG) and parietal (PoCG, AnG, PCun) regions in chimpanzees have additionally been revealed in analogous human areas (Good et al., 2001b; Kennedy et al., 2009; Crivello et al., 2014; Fjell et al., 2014; Minkova et al., 2017). The same is true for the superior occipital gyrus and caudate nucleus (Crivello et al., 2014). Similar age-related total GM decline along with presentation in homologous brain areas suggests common underlying neurophysiological processes in humans and chimpanzees due to shared primate evolution.

Very recently, it has been shown that stress hormone levels increase with age in chimpanzees, a process previously thought to only occur in humans, which can cause GM volume decline (Emery Thompson et al., 2020). This further strengthens the argument that age-related GM decline is also shared by humans closest relative, the chimpanzee. Furthermore, Edler et al., 2017 found Alzheimer’s disease-like accumulation of amyloid beta plaques and neurofibrillary tangles located predominantly in prefrontal and temporal cortices in a sample of elderly chimpanzees between 37 and 62 years of age. As the aggregation of these proteins is associated with localized neuronal loss and cortical atrophy in humans (La Joie et al., 2012; Lladó et al., 2018), the age-related decline in GM volume shown here is well in line with the findings by Jagust, 2016 associating GM atrophy with amyloid beta. These findings provide a biological mechanism for accelerated GM decrease in prefrontal, limbic, and temporal cortices found in chimpanzees. In contrast, elderly rhesus monkeys show GM volume decline without the presence of neurofibrillary tangles (Alexander et al., 2008; Shamy et al., 2011). Taken together, regionally specific GM atrophy seems to be a common aspect of the primate brain aging pattern observed in macaque monkeys, chimpanzees, and humans. To make a case for the existence of Alzheimer’s disease in chimpanzees, validated cognitive tests for Alzheimer’s-like cognitive decline in non-human primates are needed, to test for direct associations between cognitive decline with tau pathology and brain atrophy.

To further analyze the possible moderator effects on aging, we considered the historical composition of the NCBR sample, with respect to the rearing environment. The majority of elderly chimpanzees over 40 years old (23/26) were born in the wild and captured at a young age, whereas only very few chimpanzees under 40 were wild born (5/168). The capture, separation from their mothers, and subsequent transport to the research centers can be considered a traumatic event with possible lasting effects on brain development and morphology (Bremner, 2006). In captivity, different chimpanzee-rearing experiences, either by their mother or in a nursery, has been shown to affect brain morphology (Bogart et al., 2014; Bard and Hopkins, 2018). The same should be expected in comparison of captive and wild-born chimpanzees. The disproportionate distribution of rearing and early life experiences likely influences our cross-sectional analyses of the effect of aging on GM volume. However, we have some reason to be confident that the aging effect shown here is not solely driven by these factors as rearing environment was added as a covariate to all age regression models and the VBM age regression model with and without rearing as a covariate are almost identical (Figure 5 and Figure 5—figure supplement 1 respectively). Moreover, the GM decline we found is extensive, widespread, and also present in chimpanzees under 30 years of age (p<0.0001), where 99% are captive born (143/144).

Hemispheric asymmetry was found in 68% (44/65) of all regions of the Davi130 parcellation, reproducing several regional findings reported in previous studies using diverse image processing methods as well as uncovering numerous novel population-level asymmetries. Previous studies utilizing a region-wise approach based on hand-drawn or atlas derived regions to analyze asymmetry in cortical thickness also reported leftward asymmetry of the insula (Hopkins et al., 2017) and rightward lateralization of cortical thickness of the PCC (Hopkins et al., 2017) as well as STG, MTG, and SMG (Hopkins and Avants, 2013). Previous VBM findings also revealed leftward asymmetry in the anterior SFG (Hopkins et al., 2008) along with rightward lateralization of the MFG, PrCG, PoCG, mOG, and CrusII (Hopkins et al., 2008; Hopkins and Avants, 2013). In the current study, new regions of larger GM volume in the left hemisphere were found in frontal (pMFG, mSFG, pSFG), temporal (aTTG), and parietal (PCun) cortices as well as in the basal ganglia (BF,GP, Pu). Novel rightward asymmetries could also be seen in the frontal (IFG, mOFC, FrOP), limbic (CC, Amy), temporal (pTTG, FFG), parietal (POP, AnG), and occipital (Calc) cortices besides the basal ganglia (Th, CN) and the cerebellum (CrusI, CerIX, CerVI, CerVIII).

The Davi130s’ region pTTG which contains the planum temporale (PT), presented significant rightward lateralization, while previous studies of the PT have shown leftward asymmetry in chimpanzee GM volume, surface area (Hopkins and Nir, 2010), and cytoarchitecture (Zilles et al., 1996; Gannon et al., 1998; Spocter et al., 2010). A possible reason for the divergence in this finding is that the anterior border of PT (Hopkins and Nir, 2010) lies several millimeters posterior from the anterior posterior split of the Davi130 TTG. Additionally, the left lateral sulcus in the Juna.Chimp template appears to proceed further posteriorly and superiorly compared to the right, which is consistent with previous findings in asymmetrical length of the POP in chimpanzees (Gilissen and Hopkins, 2013) and Sylvian fissure length in old world monkeys (Lyn et al., 2011; Marie et al., 2018). Population-level asymmetries in the pIFG in chimpanzees were documented almost two decades ago by Cantalupo and Hopkins, 2001, who reported a leftward asymmetry in pIFG volume in a small sample of great apes. In subsequent studies, this result could not be replicated when considering GM volume (Hopkins et al., 2008; Keller et al., 2009) or cytoarchitecture (Schenker et al., 2010). We also failed to find a leftward asymmetry in GM volume for the pIFG, in contrary to asymmetries found in humans (Amunts et al., 1999; Uylings et al., 2006; Keller et al., 2009).

A substantial amount of regions presenting significant inter-hemispheric differences of local morphology in chimpanzees has also been shown in humans (Good et al., 2001a; Plessen et al., 2014; Kong et al., 2018). Specifically, leftward lateralization has been found in human analogous regions of the SFG and insula utilizing GM thickness and volume in voxel-wise and atlas derived region-wise approaches (Good et al., 2001a; Takao et al., 2011; Plessen et al., 2014; Kong et al., 2018). Rightward asymmetry in the Davi130 regions IFG, STG, MTG, AnG, mOG, Calc, in addition to the thalamus and lateral cerebellum is documented in the human brain also using both VBM and surface measures (Good et al., 2001a; Takao et al., 2011; Plessen et al., 2014; Kong et al., 2018). Gross hemispheric asymmetry in humans follows a general structure of frontal rightward and occipital leftward asymmetry known as the ‘Yakovlevian torque’ (Toga and Thompson, 2003). This general organizational pattern of asymmetry was not apparent in the chimpanzee (Li et al., 2018).

The NCBR offers the largest and richest openly available dataset of chimpanzee brain MRI scans acquired over a decade with 1.5T and 3T MRI at two locations, capturing valuable inter-individual variation in one large heterogeneous sample. To account for the scanner effect on GM estimation, field strength was modeled as a covariate of no interest for analyzing the age effect on GM volume. The focus of this study was the analysis of GM volume, even though the CAT12 image processing pipeline enables surface projection and analysis. Consequently, the next step will be the application of CAT12 to analyze cortical surface area, curvature, gyrification, and thickness of the chimpanzee brain, to include behavioral data and the quantitative comparison to humans and other species, as cortical surface projection permits a direct inter-species comparison due to cross-species registration.