Skip to main content
PMC home page
eLife logoLink to eLife
. 2022 Mar 2;11:e75540. doi: 10.7554/eLife.75540

Intrinsic timescales as an organizational principle of neural processing across the whole rhesus macaque brain

Ana MG Manea 1,2,✉, Anna Zilverstand 3,4, Kamil Ugurbil 2,5,6, Sarah R Heilbronner 1,4,5, Jan Zimmermann 1,2,4,5,7
Editors: Daeyeol Lee8, Timothy E Behrens9
PMCID: PMC8923667  PMID: 35234612

Abstract

Hierarchical temporal dynamics are a fundamental computational property of the brain; however, there are no whole brain, noninvasive investigations into timescales of neural processing in animal models. To that end, we used the spatial resolution and sensitivity of ultrahigh field functional magnetic resonance imaging (fMRI) performed at 10.5 T to probe timescales across the whole macaque brain. We uncovered within-species consistency between timescales estimated from fMRI and electrophysiology. Crucially, we extended existing electrophysiological hierarchies to whole-brain topographies. Our results validate the complementary use of hemodynamic and electrophysiological intrinsic timescales, establishing a basis for future translational work. Further, with these results in hand, we were able to show that one facet of the high-dimensional functional connectivity (FC) topography of any region in the brain is closely related to hierarchical temporal dynamics. We demonstrated that intrinsic timescales are organized along spatial gradients that closely match FC gradient topographies across the whole brain. We conclude that intrinsic timescales are a unifying organizational principle of neural processing across the whole brain.

Research organism: Rhesus macaque

Introduction

Recent work supports the idea that the neural processing is organized over multiple timescales (Cavanagh et al., 2020; Hasson et al., 2008; Honey et al., 2012; Kiebel et al., 2008; Murray et al., 2014; Raut et al., 2020; Soltani et al., 2021). Hierarchical temporal dynamics are a fundamental computational property of the brain—fast-changing sensory input is encoded by the fast dynamics of sensory areas, which is in turn integrated into temporally extended representations encoded by the slower dynamics of higher-order areas (Cavanagh et al., 2020; Raut et al., 2020; Soltani et al., 2021). Temporal integration and information processing over multiple timescales are crucial for understanding how the brain deals with the complexity of the real world (Zimmermann et al., 2018); hence, understanding whole-brain temporal dynamics is highly important.

As one moves from sensory to associative areas, neural processing reflects a hierarchy of progressively longer ‘temporal receptive windows’ (TRWs)—defined as the length of time during which an area accumulates information during task performance (Hasson et al., 2008). In other words, the range of the TRW of each area reflects its functional role, with sensory areas having short TRWs, to enable fast sensory processing, and associative areas having longer TRWs, allowing them to process information unfolding over longer periods of time. Not only does neural processing display these timescales during task performance, but we now know that temporal specialization arises from a region’s intrinsic dynamics, that is, temporal dynamics that are present even in the absence of direct sensory stimulation. This was first described by Murray et al., 2014 analyzing action potentials of single neurons in the macaque. They showed that timescales of baseline neuronal activity vary across cortical areas, revealing a hierarchical progression from fast timescales in sensory areas, to slower timescales in association areas. These intrinsic fluctuations in the neural signal, also described as intrinsic neural timescales (INTs), are thought to set the duration over which brain regions integrate inputs (Murray et al., 2014). They are hypothesized to be the result of multiple neuronal mechanisms, including local dynamics and recurrent connectivity (Murray et al., 2014; Soltani et al., 2021; Watanabe et al., 2019; Demirtaş et al., 2019).

Additional studies have replicated and extended the INT hierarchy, reinforcing the idea that brain regions exhibit unique timescales in their intrinsic activity, which reflect functional specialization in terms of temporal integration (Cavanagh et al., 2016; Cirillo et al., 2018; Fascianelli et al., 2019; Murray et al., 2014; Ogawa and Komatsu, 2010; Rossi-Pool et al., 2021; Maisson et al., 2021). The concept of INT hierarchy has even been extended to subcortical regions—specifically the basal ganglia (Nougaret et al., 2021). Moreover, a large-scale network model, based on anatomical connectivity (AC) and the strength of local recurrent connectivity, has been used to successfully reproduce the temporal hierarchy (Chaudhuri et al., 2015). This suggests that AC is closely related to INTs.

Alterations in hierarchical INT patterns can result in cognitive and behavioral abnormalities (Watanabe et al., 2019; Wengler et al., 2020; Zilio et al., 2021). For example, abnormal INT patterns have been associated with core symptoms of high-functioning individuals with Autism Spectrum Disorder (Watanabe et al., 2019). Likewise, delusions and hallucinations are associated with INT alterations at different hierarchical levels, leading to symptom-specific changes in hierarchical INT gradients (Wengler et al., 2020). Hence, investigations into INTs have the potential to clarify hierarchical dysfunction in whole-brain dynamics in neuropsychiatric disorders.

Traditionally, INTs have been investigated using invasive electrophysiology, which is not suited to uncover whole-brain dynamics. It is also challenging to collect in humans, particularly for large data sets (as would be needed to assess differences across populations). One way to overcome this problem is using functional magnetic resonance imaging (fMRI), which offers whole-brain coverage and can be performed across species and in larger sample sizes. However, it is not yet clear whether hemodynamic INTs derived from fMRI reflect similar neural phenomena as those derived from electrophysiology. To clarify this issue, electrophysiological and hemodynamic INTs must be compared directly within the same species. To that end, recent advancements in nonhuman primate (NHP) neuroimaging, based on the use of very high magnetic fields (10.5 T) for significantly boosting signal-to-noise ratios (SNRs) (Lagore et al., 2021; Yacoub et al., 2020), make it possible to study hemodynamic INTs in the macaque brain. Moreover, taking advantage of the whole-brain coverage of fMRI, we are investigating INT topography. This is important since more recent work has shown that there is significant heterogeneity within regions (Cavanagh et al., 2020; Cirillo et al., 2018; Nougaret et al., 2021), even though initially INTs were assigned to cortical regions as a whole.

In addition to INTs, gradients of functional connectivity (FC) can also describe the functional architecture of neural processing (Haak et al., 2018; Vos de Wael et al., 2020). Gradient approaches map macroscale brain features to low-dimensional manifold representations (Vos de Wael et al., 2020). Gradient mapping has the potential to uncover overlapping functional topographies by treating the high-dimensional connectivity structure as a spatial variance decomposition problem. Notably, this approach finds patterns in the high-dimensional FC similarity structure that describe the spatial direction of maximum FC variance. For example, areas in the early visual cortex are known to exhibit at least two gradual modes of functional change (i.e., eccentricity and polar angle maps), which can be successfully retrieved from resting-state FC via gradient mapping (Haak et al., 2018). The existence of multiple overlapping principles of topographic organization is likely a general property of the whole brain (Haak et al., 2018; Harvey et al., 2020; Jbabdi et al., 2013; Marquand et al., 2017; Przezdzik et al., 2019). Here, we carry out an investigation into the multidimensional functional architecture of the macaque brain by probing its spatial and temporal intrinsic dynamics. This allows us to achieve several significant goals. First, we are able to relate fMRI hemodynamic timescales to the gold standards of electrophysiology within the same species. Second, with these results in hand, we are able to show that one facet of the high-dimensional FC topography of any region in the brain is closely related to INT topography. We show that the hierarchical intrinsic dynamics of neural processing manifest into gradients of INTs and FC which follow the same gradual topographical arrangement. In sum, we reliably describe one facet of the macroscale functional organization of the brain, with respect to spatial and temporal dynamics.

Results

INT hierarchies

First, we set out to replicate and validate findings from electrophysiological INT estimation with ultrahigh field fMRI. Murray et al., 2014 initially described (in macaques) a visual cortical hierarchy of seven cortical areas: middle temporal (MT) area, lateral intraparietal (LIP) area, orbitofrontal cortex (OFC), lateral PFC (LPFC), anterior cingulate cortex (ACC); and a somatosensory hierarchy including primary somatosensory (S1) and secondary somatosensory (S2) cortex (Figure 1A–B, left). This cortical hierarchy has been replicated and further extended by assigning INTs to the dorsal premotor (PMd) cortex, frontopolar cortex (PFp), ventral and dorsal LPFC (VLPFC and DLPFC) (Cavanagh et al., 2016; Cirillo et al., 2018; Fascianelli et al., 2019; Figure 1C–D, left). Recently, INTs have been estimated in the medial wall of the PFC, including the ventromedial PFC (vmPFC), subgenual ACC (sgACC), pregenual ACC (pgACC), and dorsal ACC (dACC) (Maisson et al., 2021; Figure 1E, left). Finally, a subcortical INT hierarchy has been described for regions of the basal ganglia: external segment of the globus pallidus (GPe), subthalamic nucleus (STN), and the striatum (Nougaret et al., 2021; Figure 1F, left).

Figure 1. Intrinsic neural timescales estimated from functional magnetic resonance imaging (fMRI) and electrophysiology reveal a similar hierarchical ordering of cortical and subcortical areas.

Group-averaged (N=9) INTs were estimated for each area of interest by averaging across voxels. Each color represents a hierarchy estimated from a different electrophysiology data set (Left). The hierarchies estimated from spiking data were replicated via fMRI (Right). Somatosensory (A) and visual (B) hierarchies reported by Murray et al., 2014. Note: Each symbol represents a different data set as originally reported by the respective authors. Frontal hierarchy reported by Fascianelli et al., 2019 and Cirillo et al., 2018 (C). Frontal hierarchy reported by Cavanagh et al., 2016 (D). Medial prefrontal hierarchy reported by Maisson et al., 2021 (E). Subcortical hierarchy reported by Nougaret et al., 2021 (F). The areas were defined bilaterally for each panel to match the recording sites of the data set—the Cortical Hierarchy Atlas of the Rhesus Macaque (Jung et al., 2021) and Subcortical Atlas of the Rhesus Macaque (Hartig et al., 2021) were used when possible (see Supplementary file 1). The hemodynamic INTs were estimated as the sum of the autocorrelation function (ACF) values in the initial positive period—that is, this measurement considers both the number of lags and the magnitude of the ACF values. Hence, the INT does not have a time unit and it is used as a proxy of the timescale of that area, with higher values reflecting longer timescales. See Figure 1—figure supplement 2 for an estimate of the time constant of each area. The INT of individual areas was estimated by taking the average of the voxels’ INT values. The analysis was performed at the group level (N=9). Abbreviations: S1/S2 (primary/secondary somatosensory cortex), MT (middle temporal area), LIP (lateral intraparietal area), OFC (orbitofrontal cortex), LPFC (lateral prefrontal cortex), ACC (anterior cingulate cortex), PMd (dorsal premotor cortex), LOFC (lateral OFC), DLPFC (dorso-lateral PFC), VLPFC (ventro-lateral PFC), PFp (polar prefrontal cortex), vmPFC (ventro-medial PFC), sgACC (subgenual ACC), pgACC (pregenual ACC), dACC (dorsal ACC), GPe (external globus pallidus), STN (subthalamic nucleus). Figure 1—figure supplement 3 depicts single-subject INT stability and similarity and the group-level stability of the INT hierarchies.

Figure 1.

Figure 1—figure supplement 1. Whole-brain hemodynamic intrinsic neural timescales (INTs), lag-1 and 0-crossing lag.

Figure 1—figure supplement 1.

The hemodynamic INTs were estimated as the sum of the autocorrelation function (ACF) values in the initial positive period. This approach considers both the magnitude (quantified as the lag-1 ACF value) and the timescale (quantified as the 0-crossing lag). To examine whether one of these components is a stronger driver of our INT measurement, the lag-1 ACF value ((A); Min: 0.93 – Max: 0.96) and the 0-crossing lag ((B); Min: 5 – Max: 12) were compared to the INTs ((C); Min: 3.3 – Max: 6). The voxel-wise maps were highly correlated: Pearson’s r (A, B): 0.96; Pearson’s r (B, C): 0.97; Pearson’s r (A, C): 0.96. The maps were computed at the group level (N=9). Overall, the magnitude and timescale components of the INTs are highly related. The color bar indicates the lag-1 ACF values, the 0-crossing lag, or INTs. Since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the INT does not have a time unit and it is used as a proxy of the timescale of that area, with higher values reflecting longer timescales.
Figure 1—figure supplement 2. The time constants of hemodynamic intrinsic neural timescales (INTs).

Figure 1—figure supplement 2.

The colors are consistent with the hierarchies from Figure 1. The time constant was operationalized as the 0-crossing lag multiplied by the repetition time (TR=1.1 s). The relative ordering of the areas is consistent with that based on our INT measure. The analysis was performed at the group level (N=9). Abbreviations: S1/S2 (primary/secondary somatosensory cortex), MT (middle temporal area), LIP (lateral intraparietal area), OFC (orbitofrontal cortex), LPFC (lateral prefrontal cortex), ACC (anterior cingulate cortex), PMd (dorsal premotor cortex), LOFC (lateral OFC), DLPFC (dorso-lateral PFC), VLPFC (ventro-lateral PFC), PFp (polar prefrontal cortex), vmPFC (ventro-medial PFC), sgACC (subgenual ACC), pgACC (pregenual ACC), dACC (dorsal ACC), GPe (external globus pallidus), STN (subthalamic nucleus).
Figure 1—figure supplement 3. Stability and similarity analysis of whole-brain intrinsic neural timescales (INTs).

Figure 1—figure supplement 3.

(A) Single-subject stability and similarity of cortical (Top) and subcortical (Bottom) INTs. The stability score was estimated by averaging Fisher’s Z-transformed pairwise Pearson’s correlation coefficients across runs. The similarity score was estimated by averaging the Fisher’s Z-transformed pairwise Pearson’s correlation coefficients with all the other subjects. (B) The stability of fMRI-based INT hierarchies (see Figure 1). To estimate stability, the hierarchies were re-computed for all the possible combinations of groups of four subjects (n=126) steps. For each step, the Spearman’s rank correlation between the group (N=9, see Figure 1) and the subgroup (N=4) hierarchies was computed. The stability score was estimated by averaging the correlation coefficients across steps. Abbreviations: Murray et al., 2014; Somatosensory hierarchy; Murray et al., 2014; Visual hierarchy; Fascianelli et al., 2019; Cirillo et al., 2018; Cavanagh et al., 2016; Maisson et al., 2021; Nougaret et al., 2021. fMRI, functional magnetic resonance imaging.
Open in a new tab

The electrophysiology-based (Figure 1, left) and fMRI-based (Figure 1, right) INTs reveal similar hierarchies across the brain. By replicating these physiological hierarchies, we demonstrate that fMRI is well-suited to investigate aspects of whole-brain temporal dynamics. Minor deviations from hierarchies estimated from single-unit recordings (e.g., in STN, sgACC, and pgACC) can be attributed to several factors such as a high degree of INT heterogeneity within these regions, and differences between studies in the atlases used and size of the areas under investigation (see Figure 1—figure supplement 3 for the INT stability analysis).

Our definition of INTs combines the effects of ACF magnitude and timescale (i.e., the 0-crossing lag). Since it has previously been shown that the magnitude of the ACF values, parameterized as the lag-1 ACF value, differs across brain areas (Shinn et al., 2021), we further examined whether the observed hierarchies are driven by magnitude, timescale, or both. In line with previous results, the lag-1 ACF values systematically varied across the brain (Shinn et al., 2021). Moreover, the lag-1 ACF values and the 0-crossing lag were linearly related to INTs as operationalized in this study (see Figure 1—figure supplement 1 for voxel-wise whole-brain maps of the lag-1, 0-crossing lag, and INTs). The maps suggest that magnitude and timescale are closely related. To exclude the possibility that the observed hierarchies are driven by magnitude, rather than timescale, we re-computed them by using the 0-crossing lag as a proxy for INT. To obtain a time estimate of hemodynamic timescales, the 0-lag crossing lag was multiplied by the TR (i.e., 1.1 s). The hierarchical ordering of both cortical and subcortical areas is consistent (see Figure 1—figure supplement 2) with that based on our INT measure (see Figure 1). Hence, while the ACF magnitude is closely related to our INT index and the 0-crossing lag, temporal hierarchies are not driven by heterogeneity in the ACF magnitude.

We next investigated previously unexamined regions, making unique use of our high resolution, whole-brain coverage. Specifically, we reveal a hierarchy of INTs in the dorsal visual pathway (Goodale et al., 1994; Goodale and Milner, 1992; Figure 2). INTs gradually increase from the entry point of visual information to higher-order areas involved in visuo-motor integration (Figure 2, inset). INTs are hierarchically organized within the occipito-parietal, parieto-premotor, and parieto-prefrontal components of the dorsal pathway (Kravitz et al., 2011). Also, in line with the occipito-parietal pathway being the source of the other two pathways, it is lower in the hierarchy. Taken together, we replicate electrophysiological INT hierarchies and demonstrate the invaluable contribution of using whole-brain fMRI to extend this research by describing a previously uncharacterized fine-grained temporal hierarchy along the visual pathway as an example.

Figure 2. Functional magnetic resonance imaging (fMRI) reveals a hierarchical ordering of timescales in the dorsal pathway for visuospatial processing.

Figure 2.

Open in a new tab

Group-averaged (N=9) intrinsic neural timescales (INTs) were estimated for each area of interest by averaging across voxels. The areas were defined bilaterally based on the Cortical Hierarchy Atlas of the Rhesus Macaque (Jung et al., 2021) (see Supplementary file 2). The dorsal pathway reveals a gradual progression of INTs from sensory to association and motor cortices. The hierarchical pattern of INTs can be observed in the original description of the dorsal pathway (inset), but also in more recently described subdivisions. Every color represents a subdivision of the dorsal pathway as described in Kravitz et al., 2011. Note: Since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the INT measure does not have a time unit and it is used as a proxy of the timescale of that area, with larger values reflecting longer timescales. Abbreviations: DLPFC (dorso-lateral prefrontal cortex), VIP (ventral intraparietal area), MIP (medial intraparietal area), rIPL (rostral inferior parietal lobule), MT (middle temporal area), MST (medial superior temporal area), LIP (lateral intraparietal area), A46 (area 46), A8A (area 8A), PMd (dorsal premotor cortex), PMv (ventral premotor cortex).

INT hierarchies are closely related to FC gradients across the brain

We next examined the relationship between INTs and gradients of FC at different levels of description, starting with the frontal and parietal lobes (see Figure 3, gradients computed twice, once using within-lobe FC and again using frontal-parietal FC). In the frontal lobe, the first component (i.e., G1) shows an overall anterior-posterior organization spanning from sensory to motor areas. This aligns with previous studies showing a global anterior-posterior functional gradient in the frontal lobe (Badre, 2008; Riley et al., 2018; Tan et al., 2020). The second component (i.e., G2) shows an overall lateral-medial organization, extending from the PMv to the medial wall. The INT map was correlated with G2 for both within-area (r=0.47) and fronto-parietal (r=–0.26) connectivity. In the parietal lobe, G1 showed a medial-lateral and G2 a ventro-dorsal overall organization. The INT map was correlated with G1 for both within-area (r=0.56) and fronto-parietal (r=0.47) connectivity. (Note that the sign of the gradients is arbitrary as a byproduct of the decomposition method [Haak and Beckmann, 2020]; hence, only the magnitude but not the direction of the INT-FC gradient correlation is relevant.)

Figure 3. Functional connectivity gradients are related to intrinsic neural timescale (INT) topographies.

INT maps (Middle) and functional connectivity (FC) gradients were computed at the group level (N=9). Gradient 1 (G1) and Gradient 2 (G2) were estimated using a cosine similarity affinity computation of the within-area FC (Left) and frontal-parietal FC (Right), followed by diffusion mapping (Note: G1 and G2 are the first components that describe the axes of largest variance). (A) Frontal INTs were correlated with G2 for both within-area (r=0.47) and frontal-parietal connectivity (r=–0.26). (B) Parietal INTs were correlated with G1 for both within-area (r=0.56) and parietal-frontal connectivity (r=0.47). The color bar indicates the position along the FC gradient (Note: values are on an arbitrary scale) or INT values (Note: since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the INT does not have a time unit and it is used as a proxy of the timescale of that area, with larger values reflecting longer timescales). The value ranges differ across panels unless otherwise stated.

Figure 3.

Figure 3—figure supplement 1. Intrinsic neural timescales (INTs) are topographically organized along functional connectivity (FC) gradients across the whole cortex.

Figure 3—figure supplement 1.

(A) The first (G1; Left) and second (G2; Middle) FC gradients estimated at the whole-cortex level, and the INT map. The brain was parcellated using the Cortical Hierarchy Atlas of the Rhesus Macaque (Level 6; Jung et al., 2021) and the connectivity matrix was constructed by computing the Pearson’s correlation of the resulting time series (134×134 connectivity matrix). G1 and G2 were estimated using a cosine similarity affinity computation of this connectivity matrix, followed by diffusion mapping (Note: G1 and G2 are the first components that describe the axes of largest variance). The INT of individual parcels was estimated by taking the average of the voxels’ INT values. The analysis was performed at the group level (N=9). The INT map was highly correlated with G2 (Pearson’s r=0.70), and with a lesser extent to G1 (Pearson’s r=0.20). The color bar indicates the position along the FC gradient (Note: values are on an arbitrary scale) or INT values (Note: since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the INT does not have a time unit and it is used as a proxy of the timescale of that area, with larger values reflecting longer timescales). The value ranges differ across panels. (B) At the parcel level, there is a monotonic relationship between INT and G2 values. (C) The INT and G2 values are monotonically related within each lobe. Note: the subplots are a decomposition of the scatter plot in (B) into the four lobes.
Open in a new tab

We then selected functionally meaningful subdivisions of the frontal lobe to probe finer-grained connectivity profiles and interrogate the existence of multiple hierarchical gradients of function (see Figure 4). In all cases, the INT maps were highly correlated with one of the principal components estimated from within-area connectivity (i.e., LPFC G1: r=0.35; Motor G1: r=0.68; OFC G2: r=0.45; MPFC G1: r=0.49). For example, the motor cortex has a well-established hierarchy, with premotor regions at the top and M1 at the bottom (Binkofski and Buxbaum, 2013; Wong et al., 2015). Accordingly, the maps reflected this functional hierarchy, with M1 being at one end and the PMv at the other. The OFC shows a predominantly anterior-posterior axis of change, which agrees with anatomical findings—the caudal and lateral regions are more input-like unimodal regions while more anterior areas integrate these multimodal signals (Carmichael and Price, 1994; Carmichael and Price, 1996; Ongür and Price, 2000; Price, 1999). The lateral and medial subdivisions of the PFC revealed a predominately anterior-posterior organization. The same pattern of results was observed at the single-subject level (see Figure 4—figure supplement 2).

Figure 4. The relationship between functional connectivity (FC) and intrinsic neural timescales (INTs) in the frontal lobe.

Group-averaged INT maps (Right) and FC gradients (Left) were estimated at the group level (N=9). Gradient 1 (G1) and Gradient 2 (G2) were estimated using a cosine similarity affinity computation of the within-area followed by diffusion mapping (Note: G1 and G2 are the first components that describe the axes of largest variance). The FC gradient with the highest correlation to INTs was plotted. (A) INTs in medial prefrontal cortex were correlated with G1 (r=0.49). (B) INTs in the orbitofrontal cortex were correlated with G2 (r=0.45). (C) INTs in the motor cortex were correlated with G1 (r=0.68). (D) INTs in the lateral prefrontal cortex were correlated with G1 (r=0.35). The color bar indicates the position along the FC gradient (values are on an arbitrary scale) or INT values (Note: since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the INT does not have a time unit and it is used as a proxy of the timescale of that area, with larger values reflecting longer timescales). The value ranges differ across panels unless otherwise stated. For a more in-depth analysis of the relationship between INTs and FC, see Figure 4—figure supplement 1. For the single-subject analysis, see Figure 4—figure supplement 2.

Figure 4.

Figure 4—figure supplement 1. Intrinsic neural timescales (INTs) and functional connectivity (FC) gradients are monotonically related in the frontal lobe.

Figure 4—figure supplement 1.

For a better interpretation of the relationship between the INT maps and FC gradients, the INT/gradient values were averaged across voxels within the functional subdivisions of the areas considered in Figure 4. The functional subdivisions were defined based on the Cortical Hierarchy Atlas of the Rhesus Macaque (CHARM; Jung et al., 2021). The error bars represent ±1 Standard Error (SE; calculated as the standard deviation divided by the square root of the number of voxels). The blue circles represent INT values (left Y-axis) and the red circles represent the FC gradient value (right Y-axis). In the medial wall of the PFC, both INT and gradient values roughly change in an anterior-posterior direction with the slowest timescales found in the anterior and the fastest timescales in the posterior regions (Note: both SMA and pre-SMA were included). In the OFC, both INT and gradient values have a medio-lateral axis of change, with the slowest timescales in the MOFC and fastest timescales in the LOFC. In the lateral motor cortex, both the INT and gradient values increase in a medio-lateral way, with M1 having the fastest and PM the slowest timescales. In the lateral prefrontal cortex, the INT values are decreasing from FEF to the VLPFC while the gradient values are decreasing from FEF/VLPFC to the DLPFC. The plots are based on the INT and gradient maps derived at the group level (N=9) (Note: the plots were derived from the maps in Figure 4). Abbreviations and CHARM codes: FEF (frontal eye fields; Level 3 Code 51), DLPFC (dorsolateral prefrontal cortex; Level 3 Code 54), VLPFC (ventrolateral prefrontal cortex; Level 3 Code 64), M1 (primary motor cortex; Level 4 Code 79), PM (premotor cortex; level 4 Code 80), MOFC (medial orbitofrontal cortex; Level 3 Code 17), LOFC (lateral orbitofrontal cortex; Level 3 Code 25), ACC (anterior cingulate cortex; Level 3 Code 3), SMA (supplementary motor area; Level 3 Code 87), MCC (midcingulate cortex; Level 3 Code 11).
Figure 4—figure supplement 2. Intrinsic neural timescales (INTs) closely match functional connectivity (FC) gradients at the single-subject level.

Figure 4—figure supplement 2.

The within-area (Right, (A–F)) and cortico-striatal (Top, (G)) FC gradient, and corresponding INT map (Left) (Note: the gradient with the highest correlation to INTs is depicted). Similar to the group results, INTs were related to FC gradients across the whole brain: frontal lobe ((A), r=0.66); parietal lobe ((B), r=0.81); medial PFC ((C), r=0.62); orbitofrontal cortex ((D), r=0.51); motor cortex ((E), r=0.89); lateral PFC ((F), r=0.29); and striatum ((G), r=0.47). The color bar indicates the position along the FC gradient (Note: values are on an arbitrary scale) or INT values (Note: since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the INT does not have a time unit and it is used as a proxy of the timescale of that area, with larger values reflecting longer timescales). The value ranges differ across panels unless otherwise stated.
Open in a new tab

Although we were mainly interested in the fine-grained functional organization of individual areas, we also asked whether the INT-FC relationship holds at the whole-cortex level (see Figure 3—figure supplement 1). It is important to note that the whole-cortex estimation reflects overall inter-areal FC differences rather than the fine-grained FC patterns in smaller areas; that is because inter-areal connectivity differences are more pronounced than intra-areal differences. The whole-cortex INT map was highly correlated with the second gradient estimated from the cortical FC matrix (r=0.70). With these results in hand, we conclude that INTs and FC gradients are closely related and represent a functional principle of organization across the whole brain.

We then investigated cortical-subcortical hierarchical gradients of function. Cortico-striatal AC has been extensively studied in NHPs, however, striatal-cortical FC and INTs have not. The striatal INT map was correlated with the FC gradient estimated from cortico-striatal (r=0.38) connectivity (see Figure 5). The cortico-striatal FC topography exhibited a gradual pattern of connectivity change corresponding to the anterior-posterior, dorsal-ventral, and medial-lateral anatomical gradient of connectivity derived from tract-tracing studies in NHPs (Haber, 2016; Haber and Knutson, 2010; Friedman et al., 2002; Alexander et al., 1986). INTs followed a similar pattern. Moreover, the FC gradients were similar to those estimated from human neuroimaging data (Marquand et al., 2017). To examine the underlying FC patterns, we projected the striatal gradient onto the cortex (see Figure 5A). This analysis not only validated the meaning of the gradients, but also revealed that the cortico-striatal FC closely matched AC. To further investigate the link between FC and INTs, we reconstructed the striatal INT map by replacing the INT of every voxel with the average INT value of the cortical voxels (i.e., 10 voxels) with which they are most correlated (see Figure 5B). In accordance with the above-mentioned results, the striatal INTs and reconstructed cortico-striatal INT maps were correlated (r=0.45). At the voxel level, the INT values closely matched the average INTs of their top 10 strongest cortical connections.

Figure 5. Cortico-striatal functional connectivity (FC) is related to intrinsic neural timescales (INTs).

Group-averaged INT maps and FC gradients (Left) were estimated at the group level (N=9). Gradient 1 (G1) and Gradient 2 (G2) were estimated using a cosine similarity affinity computation of the cortico-striatal FC followed by diffusion mapping (Note: G1 and G2 are the first components that describe the axes of largest variance). Only the gradient with the highest correlation to the striatal INT map was plotted. (A) The striatal FC gradient estimated from cortico-striatal connectivity (Top) and its projection onto the cortex (Bottom) (Note: the same value range was used). The FC patterns underlying the cortico-striatal gradient matched the anatomical connectivity derived from animal invasive tracing. (B) Striatal INTs (Top) and cortical INTs projection based on cortico-striatal FC (Bottom). The striatal and cortico-striatal INT maps were related (r=0.45). Striatal INTs were related to G1 estimated from cortico-striatal connectivity (r=0.38). The color bar indicates the position along the FC gradient (values are on an arbitrary scale) or INT values (Note: since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the INT does not have a time unit and it is used as a proxy of the timescale of that area, with larger values reflecting longer timescales). The value range differs across panels unless otherwise stated.

Figure 5.

Figure 5—figure supplement 1. Intrinsic neural timescales (INTs) and functional connectivity (FC) gradients are monotonically related in the striatum.

Figure 5—figure supplement 1.

For a better interpretation of the relationship between the striatal and cortico-striatal INT maps and FC gradients, the INT/gradient values were averaged across voxels within the functional subdivisions of the striatum. The functional subdivisions were defined based on the Subcortical Atlas of the Rhesus Macaque (SARM; Hartig et al., 2021). The error bars represent 2*Standard Error (SE; calculated as the standard deviation divided by the square root of the number of voxels). (A) Dividing the striatum into its major functional subdivisions revealed that for both striatal and cortico-striatal INTs and FC, there is a dorsal to ventral axis of change. The dorsal striatum had the slowest and ventral striatum the fastest timescales, which was reflected in both the striatal and cortico-striatal INTs. (B) A finer subdivision of the striatum reveals a similar ventro-dorsal but also medio-lateral axis of change across the three maps. While the striatal INTs and the FC gradient have a very good correspondence, the head of the caudate in the cortico-striatal INT map deviates from this pattern—that is, it has higher values relative to the other structures in the ventral striatum. The plots are based on the INT and gradient maps derived at the group level (N=9). Note: since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the INT does not have a time unit and it is used as a proxy of the timescale of that area, with larger values reflecting longer timescales. Abbreviations and SARM codes: Caudate (H) (head of the caudate; Level 6 Code 46); Caudate (T) (tail of the caudate; Level 6 Code 47); Nucleus Acc. (nucleus accumbens; Level 6 Code 52); Olfactory T. (olfactory tubercule; Level 6 Code 53), Dorsal Striatum (Level 4 Code 44), Ventral Striatum (Level 4 Code 51), G1 (Gradient 1).
Open in a new tab

Overall, we demonstrated that INTs are organized along spatial gradients that closely match FC gradients.

Discussion

Temporal integration is a fundamental property of neural processing (Cavanagh et al., 2016; Hasson et al., 2008; Honey et al., 2012; Soltani et al., 2021), both during task performance and in the absence of external stimulation (Cirillo et al., 2018; Hasson et al., 2008; Murray et al., 2014; Raut et al., 2020; Ito et al., 2020). However, until this point, there have been major gaps in investigating temporal integration across the brain, across species, and across populations. This is because, traditionally, temporal integration has been studied primarily using single-unit electrophysiology in NHPs (see Gao et al., 2020). Although there is some evidence that BOLD signals from fMRI may allow for whole-brain interrogation of timescales in humans (Watanabe et al., 2019; Wengler et al., 2020), such an approach had not yet been validated. Here, we focused on investigating whole-brain INT and FC topographies at different levels of description, to first validate, within-species, the study of INTs using BOLD fMRI, and then investigate the relationship between temporal and spatial dynamics of the resting brain with ultrahigh resolution fMRI across the whole macaque brain.

We have demonstrated the feasibility of estimating INTs from fMRI in the NHP brain. We did so by showing that hemodynamic INT hierarchies closely match those derived from NHP electrophysiology. While some recent human neuroimaging studies have investigated whole-brain hemodynamic INTs (Raut et al., 2020; Watanabe et al., 2019; Wengler et al., 2020), temporal hierarchies derived from electrophysiology have not been replicated via fMRI, nor within species. Furthermore, in a previous paper, INTs derived from fMRI data in human subjects were elegantly validated with EEG (Watanabe et al., 2019). The present findings are in agreement with these results, finding slower and faster INTs in frontoparietal areas and sensory-related areas, respectively. This points toward cross-species homologies in cortical INTs, hence, laying the groundwork for future translational work. Nevertheless, the step presented here is essential, because we were able to demonstrate the equivalence of the fMRI estimates of INT hierarchies with single-neuron spiking results within the same species. This finding provides validation for the complementary use of hemodynamic and electrophysiological INTs. Using fMRI in this context has several advantages. First, it can be performed on the whole brain. Second, it can easily be employed across different human populations in the future, dramatically expanding the possible uses of INT investigations.

We did observe minor deviations from electrophysiological INTs in our data. These may be in large part attributable to a high degree of INT heterogeneity within individual anatomical regions. This observation is in line with previous reports of heterogeneity within neural populations in single-unit recordings (Cirillo et al., 2018; Nougaret et al., 2021). This suggests that averaging INTs within anatomical regions might not provide a full description of macroscale organization.

It is worth noting that the current study could not resolve the exact correspondence between the time constants estimated from spiking activity and BOLD. The relationship between neural activity and the BOLD signal has been revealed via simultaneous electrophysiological-fMRI recording—that is, a spatially localized increase in the fMRI signal directly and monotonically reflects an increase in neural activity, with the mean onset time of the BOLD signal relative to the neural activation being 2 s (Logothetis et al., 2001). As a result of sensory stimulation, BOLD signal is delayed by ~2–3 s, followed by a ramp of 6–12 s to a plateau or peak value. Local field potentials (LFPs), and to a lesser extent multiunit activity, can predict the BOLD signal when convolved with the neuro-vascular impulse response function. In sum, the hemodynamic response is roughly a low-pass-filtered expression of the total neural activity; with LFPs, however, having a greater contribution to the signal. This has several implications for the interpretation of hemodynamic timescales: (1) they are more likely to reflect timescales related to LFPs rather than spiking activity; (2) the time constants estimated from spiking activity might not be linearly related to those estimated from fMRI. A further complication for the interpretability of hemodynamic timescales is that INTs have only been estimated from spiking activity, but not LFPs. The exact relationship between timescales derived from spiking activity, LFPs and the BOLD signal remains an empirical question. We hypothesize that hemodynamic timescales generally increase monotonically with those estimated from electrophysiology. In that respect, Watanabe et al., 2019 bring preliminary evidence—that is, timescales derived from electroencephalography (EEG) and fMRI have been shown to differ by 2 orders of magnitude. This is reasonable considering the BOLD time course—when convolved with the hemodynamic response function, the INTs estimated from EEG were similar in magnitude to those estimated from fMRI. Nevertheless, we show that temporal hierarchies are a general principle of organization, irrespective of the method used to estimate them.

To further demonstrate the potential of INTs estimated from fMRI, we extend the current literature by providing a finer description of the hierarchy of visual INTs compared to those previously described with electrophysiology (Murray et al., 2014). Remarkably, areas along the dorsal visual pathway follow the INT hierarchical ordering predicted by functional specialization and connectivity (Kravitz et al., 2011), with primary visual areas showing the fastest timescales and prefrontal areas showing the longest. Our findings bring further evidence that the dorsal pathway is not a unitary system, but rather subdivided into multiple parallel pathways (Binkofski and Buxbaum, 2013; Kravitz et al., 2011). This demonstrates the importance of fine-grained, large-scale descriptions of the brain for uncovering multiple hierarchies.

We investigated timescales and connectivity topographies at different levels of description to probe whether one aspect of the multidimensional functional architecture of the brain is related to temporal structure. We show that INT topographies are closely related to the overlapping FC gradient dimensions. Previous studies have demonstrated that FC gradient dimensions in sensory cortex corresponded with known organizational principles such as eccentricity and polar angle (Haak et al., 2018). Moreover, FC gradients followed M1’s somatotopic map. Our findings put forward a significantly more unified framework that generalizes over the entire brain. INTs form a topographic organizational principle ranging from sensory to associative cortex. While other organizational principles can and likely always will be represented across topographic networks of the brain, intrinsic timescales appear to be special in their general extent.

It is noteworthy that there are crucial methodological differences between the gradient and INT estimation approaches: (1) the gradient estimation pipeline consists of multiple steps to reduce the dimensionality of the data, and consequently, the amount of noise (e.g., the FC matrix was made sparse; applying the cosine similarity kernel to build an affinity matrix; and applying a nonlinear dimensionality reduction technique to estimate the gradients); (2) the gradients were estimated bilaterally. In comparison to gradients, where the spatial correlation between the voxels’ time series is just the initial step in the pipeline, the INT measure consists of the temporal autocorrelation of individual voxels. No extra steps have been taken for noise reduction. Hence, while gradients are the result of a complex processing pipeline, INTs are a relatively simple calculation. As a result, there is a higher amount of variability in the INT maps which could potentially lead to the underestimation of the FC-INT relationship. With respect to the second point, the symmetry of the gradients could be a byproduct of estimating the gradients bilaterally. Other studies e.g., Haak et al., 2018; Przezdzik et al., 2019 have derived gradients for each hemisphere individually and no major asymmetries have been reported. We estimated the gradients bilaterally because of an issue that is inherent with experimental ultrahigh fields. Specifically, RF transmit field (or B1) inhomogeneities lead to asymmetric signal intensities. We were worried that these effects might introduce asymmetries between the two hemispheres. We have reason to believe that this issue is partially being taken care of by estimating the gradients bilaterally, but the INTs are still susceptible to these effects. This could also result in the underestimation of the FC-INT relationship since there are visible asymmetries in the INT maps.

One potential limitation of the current study is using anesthetized subjects. Although obtaining high-resolution images of awake NHPs at ultrahigh fields would be ideal, it is still an enormous technical challenge that the field is working to overcome. Using our low anesthesia strategy, we were able to obtain a relatively large sample size of NHPs (because extensive training was not required). It is important to note that it was previously shown that our anesthesia protocol demonstrates no observable biases to whole-brain connectivity analysis (Yacoub et al., 2020). Moreover, it has been shown that this protocol has a relatively low impact on reproducibility (Autio et al., 2021). To our knowledge, the largest effect of vasodilative anesthetics (such as isoflurane) is an overall blunting of responses which is counteracted by ultrahigh field MRI. Finally, besides the possible effects on the hemodynamic response, anesthesia introduces a crucial difference between the current study and previous research—while in previous studies, it was not controlled for the subjects engaging in various cognitive processes, in anesthetized subjects, this confound is potentially removed. Overall, future research should replicate current results to exclude a possible effect of anesthesia.

We show that resting-state fMRI in NHPs is crucial for bridging the gap between human and animal models, but also across the various techniques of investigating the brain. Nevertheless, macaques have a disadvantage as a model organism for fMRI because of their small head size (~6% of the human brain) which makes the use of standard field strength magnets problematic for acquiring high-resolution data. As a result, collecting high-resolution, high SNR whole-brain data has been extremely difficult (Yacoub et al., 2020). We show here that one way of overcoming this problem is to use ultrahigh field strengths in combination with high channel count receiver arrays (Lagore et al., 2021; Uğurbil et al., 2019; Yacoub et al., 2020). Ultrahigh magnetic fields not only enhance SNR but also increase the BOLD contrast, especially for the microvascular BOLD contributions—which leads to mapping signals with higher precision to sites of neuronal activity (Ugurbil, 2016; Uğurbil, 2018). In conclusion, our unique approach made it possible to obtain whole-brain high-resolution data to investigate FC and INTs across both cortical and subcortical structures.

INTs, like AC and FC, are one way of characterizing the organization of the brain. Critically, combined maps of all three will require a nonhuman animal model such as the macaque, as AC cannot be directly measured in humans (Grier et al., 2020). It is clear that AC, FC, and INTs are related, although the precise nature of the relationships is somewhat speculative. For instance, vertex FC strength is related to human INTs estimated from fMRI (Raut et al., 2020). Our own results show that cortico-striatal INTs and FC gradients broadly track AC. AC establishes the pathways which the signals underpinning FC use. However, the fact that the pathways exist and are used does not tell us what types of information transformations occur along the way. Through cascaded processing, longer and longer TRWs can be built, measured as longer INTs. Thus, we expect that the space of potential INTs is constrained by both AC and FC. Future exploration of these organizational principles, including using pharmacological interventions and electrical stimulation, will likely clarify the causal relationships at play.

We show that INTs are topographically organized along FC gradients throughout the NHP brain. Although we focused our analysis on the frontal and parietal lobes, and the striatum, we also showed that a whole-brain analysis recapitulates the same conclusion: resting-state temporal and spatial dynamics are closely related. Future investigations should examine this relationship in other cortical and subcortical brain areas. A promising avenue would be to examine the FC-INT link throughout the brain across different levels of parcellation (and different types of parcellations). While we show that intrinsic timescales are related to FC, other studies have related them to other anatomical and functional markers (e.g., see Ito et al., 2020; Watanabe et al., 2019). Hence, further investigations and careful experimentation are needed to understand (1) how temporal dynamics relate to other functional and anatomical characteristics across the brain; (2) their functional relevance and translational potential.

Materials and methods

Animal preparation

We obtained data from nine adult macaque monkeys (Macaca fascicularis: five female, and Macaca mulatta, three male and one female). Weights ranged from 3.0 kg to 9.4 kg. Experimental procedures were carried out in accordance with the University of Minnesota Institutional Animal Care and Use Committee and the National Institute of Health standards for the care and use of NHPs. All subjects were fed ad libitum and pair-housed within a light and temperature-controlled colony room. Animals were not water restricted. None of the subjects had any prior implant or cranial surgery. Animals were fasted for 14–16 hr prior to imaging.

On scanning days, anesthesia was first induced by intramuscular injection of atropine (0.5 mg/kg), ketamine hydrochloride (7.5 mg/kg), and dexmedetomidine (13 µg/kg). Subjects were transported to the scanner anteroom and intubated using an endotracheal tube. Initial anesthesia was maintained using 1.0%–2% isoflurane mixed with oxygen (1 L/m during intubation and 2 L/m during scanning to compensate for the 12-m length of the tubing used). For functional imaging, the isoflurane level was lowered to 1%.

Subjects were placed onto a custom-built coil bed with integrated head fixation by placing stereotactic ear bars into the ear canals. The position of the animal corresponds to the sphinx position. Experiments were performed with animals freely breathing. An initial bolus injection of 1.5 µg/kg fentanyl was administered IV followed by a continuous administration of 3 µg/kg/hr using a syringe pump. Rectal temperature (~99.6F), respiration (10–15 breaths/min), end-tidal CO2 (25–40), electrocardiogram (70–150 bpm), and SpO2 (>90%) were monitored using an MRI compatible monitor (IRADIMED 3880 MRI Monitor, USA). Temperature was maintained using a circulating water bath as well as chemical heating pads and padding for thermal insulation.

Data acquisition

All data were acquired on a passively shielded 10.5 T, 88-cm diameter clear bore magnet coupled to Siemens gradients (‘SC72’ body gradients operating at a slew rate of 200 mT/m/s, and 70 mT/m maximal strength), and electronics (Magnetom 10.5 T Plus) (Siemens, Erlangen, Germany). Within the gradient set and the bore-liner, the space available for subject insertion had a 60-cm diameter.

The 10.5-T system operates on the E-line (E12U) platform which is directly comparable to clinical platforms (3 T Prisma/Skyra, 7 T Terra). As such, the user interface and pulse sequences were identical to those running on clinical platforms. A custom in-house built and designed RF coil with an 8-channel transmit/receive end-loaded dipole array of 18-cm length (individually) combined with a close-fitting 16-channel loop receive array head cap, and an 8-channel loop receive array of 50×100 mm2 size located under the chin (Lagore et al., 2021). The size of 14 individual receive loops of the head cap was 37 mm with two larger ear loops of 80 mm—all receiver loops were arranged in an overlapping configuration for nearest neighbor decoupling. The resulting combined 32 receive channels were used for all experiments and supported threefold acceleration in phase encoding direction. The coil holder was designed to be a semi-stereotaxic instrument holding the head of the animal in a centered sphinx position via customized ear bars. The receive elements were modeled to adhere as close to the surface of the animal’s skulls as possible. Transmit phases for the individual transmit channels were fine-tuned for excitation uniformity for one representative mid-sized animal and the calculated phases were then used for all subsequent acquisitions. Magnetic field homogenization (B0 shimming) was performed using a customized field of view with the Siemens internal 3D mapping routines. Multiple iterations of the shims (using the adjusted FOV shim parameters) were performed, and further fine adjustment was performed manually on each animal. Third-order shim elements were ignored for these procedures.

In all animals, a B1+ (transmit B1) fieldmap was acquired using a vendor provided flip angle mapping sequence and then power calibrated for each individual. Following B1+ transmit calibration, 3–5 averages (23 min) of a T1 weighted magnetization prepared rapid acquisition gradient echo protocol (3D MP-RAGE) were acquired for anatomical processing (TR=3300 ms, TE=3.56 ms, TI=1140 ms, flip angle=5°, slices=256, matrix=320×260, acquisition voxel size=0.5×0.5×0.5 mm3). Images were acquired using in-plane acceleration GRAPPA=2. A resolution and FOV matched T2 weighted 3D turbo spin-echo sequence (variable flip angle) was run to facilitate B1 inhomogeneity correction.

Before the start of the functional data acquisition, five images were acquired in both phase-encoding directions (R/L and L/R) for offline EPI distortion correction. Five to six runs of fMRI time series, each consisting of 700 continuous 2D multiband (MB)/Simultaneous Multi SLice (SMS) EPI (Moeller et al., 2010; Setsompop et al., 2012; Uğurbil et al., 2013) functional volumes (TR=1110 ms; TE=17.6 ms; flip angle=60°, slices=58, matrix=108×154; FOV=81×115.5 mm2; acquisition voxel size=0.75×0.75×0.75 mm3) were acquired. Images were acquired with a left-right phase encoding direction using in plane acceleration factor GRAPPA=3, partial Fourier=7/8th, and MB or simultaneous multislice factor=2 (i.e., the number of simultaneously excited slices=2). Since macaques were scanned in sphinx positions, the orientations noted here are what is consistent with a (head first supine) typical human brain study (in terms of gradients) but translate differently to the actual macaque orientation.

Image preprocessing

Image processing was performed using a custom pipeline relying on FSL (Jenkinson et al., 2012), ANTs (Avants et al., 2014; Avants et al., 2011), AFNI (Cox, 1996), and a heavily modified CONN (Whitfield-Gabrieli and Nieto-Castanon, 2012) toolbox. Images were first motion corrected using mcflirt (registration to the first image). Motion parameters never exceeded 0.5 mm or 0.4° rotation except in one run of one animal, which was discarded (the animal experienced anesthesia-induced emesis during the last run). Images were slice-time corrected and EPI distortion corrected using topup. High magnetic fields and large matrix sizes are commonly associated with severe EPI distortions. Anatomical images were nonlinearly warped into the ‘National Institute of Mental Health Macaque Template’ (NMT) (Seidlitz et al., 2018) template using ANTs and 3DQwarp in AFNI. The distortion correction, motion correction, and normalization were performed using a single sinc interpolation. Images were spatially smoothed (FWHM=2 mm), linear detrended, denoised using a linear regression approach including heart rate and respiration, as well as a five-component nuisance regressor of the masked white matter and cerebrospinal fluid and band-pass filtering (0.008–0.09 Hz) (Hallquist et al., 2013).

Intrinsic neural timescales

Based on previous studies (Murray et al., 2014; Watanabe et al., 2019), INTs were calculated by estimating the magnitude of the autocorrelation of the preprocessed functional data. At the single-subject level, the INT of each voxel was quantified as follows. For each run, the ACF of each voxel’s fMRI signal was estimated. The ACF is the correlation of yt (time series) and yt+k (lagged time series), where k represents the time lag in steps of TR. Next, the sum of ACF values was calculated in the initial positive period. The upper limit of this period was set at the point where the ACF first becomes negative as the time lag increases. The result was used as an index of the INT of the brain region. At the single-subject level, the number of INT maps was dependent on the number of functional runs. The maps were not smoothed, because individual variation in FC gradients and INTs has been shown to have functional significance (Przezdzik et al., 2019; Watanabe et al., 2019).

To examine the correspondence between fMRI-based and electrophysiology-based INT hierarchies, the values of individual voxels were averaged within areas. The areas were defined based on the Cortical Hierarchy Atlas of the Rhesus Macaque (CHARM) (Jung et al., 2021) and Subcortical Atlas of the Rhesus Macaque (SARM) (Hartig et al., 2021) to closely match the corresponding electrophysiological recording sites (see Supplementary file 1). In some cases, since the CHARM parcellation did not match the recording sites, the regions of interest were drawn.

Since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the hemodynamic INTs do not have a time unit and are used as a proxy of the timescale of that area, with higher values reflecting longer timescales. We have additionally provided the estimated time constants associated with the INT hierarchies that have been previously described with electrophysiological recordings by multiplying the 0-crossing lag by the TR (see Figure 1—figure supplement 2 for an estimate of the time constants).

Stability and similarity

The intra-subject stability was quantified by calculating the pairwise correlation (i.e., Pearson’s r) between the runs. The individual stability score was computed by averaging the Fisher’s Z-transformed correlation coefficients across pairwise comparisons. For every subject, the maps were averaged across runs to quantify the inter-subject similarity. The individual similarity score was computed by averaging the subject’s Fisher’s Z-transformed correlation coefficients with all the other subjects.

To quantify the stability of the hierarchies, they were re-computed for groups of four subjects. The resulting ranking was correlated (i.e., Spearman’s rank correlation) with the group-level ordering of the cortical/subcortical regions. To estimate the stability score, this procedure was repeated for all the possible combinations of subjects (i.e., n=126). The stability score of any hierarchy was computed by averaging the correlation coefficients across steps (i.e., n=126).

FC gradients

We computed FC gradients using BrainSpace, a toolbox for the identification, visualization, and analysis of macroscale gradients in neuroimaging data sets (Vos de Wael et al., 2020).

The input matrix

The input matrix consisted of the voxel-wise FC between or within areas. The Pearson correlation coefficient between the time-series of every pair of voxels within and/or between areas was computed. This resulted in a 2D matrix of correlation coefficients, with every voxel being represented by an n-dimensional row vector (with n reflecting the number of connections). The connectivity matrices were computed for each subject and run.

Generating gradients

The procedure for generating FC gradients was thoroughly described in Vos de Wael et al., 2020. The input matrix was made sparse (10% sparsity) and the cosine similarity kernel was applied to build the affinity matrix. Next, diffusion mapping was applied using the BrainSpace basic parameters. The resulting gradients were plotted on the cortical surface of the NMT template (Seidlitz et al., 2018) using AFNI (Cox, 1996) and SUMA (Saad and Reynolds, 2012; Saad et al., 2004). The gradients were generated bilaterally. We only report the gradient most correlated with the INT map, assuming that ensuing gradients reflect other principles of functional organization.

Cortical projection

To probe the relationship between the striatal FC gradients and the cortex, the striatal FC gradients were projected onto the cortex according to the highest-correlated voxels (Haak et al., 2018). To reduce noise and improve interpretability, the resulting maps were smoothed using AFNI (Cox, 1996).

Projecting INTs based on FC

To examine the relationship between INTs and FC between areas (e.g., between cortex and striatum), cortical INTs were projected onto a striatal map based on cortico-striatal FC. For every striatal voxel, the top 10 highest functionally correlated cortical voxels were identified. Next, the INTs associated with these cortical voxels were extracted. Finally, the average associated cortical INT value was computed for each striatal voxel and represented as a map.

Acknowledgements

The authors thank Steve Jungst, Gregor Adriany, and Russell Lagore for continuing support with our coils and hardware. The authors thank Jen Holmberg, Brenna Knaebe, and Mrunal Zambre, for support with animal care and data acquisition. This work was supported by NIH grants P41 EB027061 (to JZ and KU), R01 MH118257 (to SRH), R56 EB031765 (to JZ), R01 MH128177 (to JZ), from the Digital Technologies Initiative (to JZ), from the Minnesota Institute of Robotics (to JZ), two Young Investigator Awards from the Brain & Behavior Research Foundation to SRH and AZ, a P30DA048742 (to JZ, SRH, and AZ), and a UMN AIRP award to JZ, SRH, and AZ.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Ana MG Manea, Email: manea006@umn.edu.

Daeyeol Lee, Johns Hopkins University, United States.

Timothy E Behrens, University of Oxford, United Kingdom.

Funding Information

This paper was supported by the following grants:

  • National Institutes of Health P41 EB027061 to Kamil Ugurbil, Jan Zimmermann.

  • National Institutes of Health R01 MH118257 to Sarah R Heilbronner.

  • National Institutes of Health R56 EB031765 and R01 EB031765 to Jan Zimmermann.

  • National Institutes of Health R01 MH128177 to Jan Zimmermann.

  • Digital Technologies Initiative to Jan Zimmermann.

  • Minnesota Institute of Robotics to Jan Zimmermann.

  • Brain & Behavior Research Foundation Young Investigator Awards to Anna Zilverstand, Sarah Heilbronner.

  • National Institutes of Health P30DA048742 to Anna Zilverstand, Sarah Heilbronner, Jan Zimmermann.

  • UMN AIRP award to Anna Zilverstand, Sarah Heilbronner, Jan Zimmermann.

Additional information

Competing interests

No competing interests declared.

No competing interests declared.

Author contributions

Conceptualization, Formal analysis, Investigation, Methodology, Software, Visualization, Writing - original draft, Writing – review and editing.

Conceptualization, Funding acquisition, Writing – review and editing.

Funding acquisition, Writing – review and editing.

Conceptualization, Funding acquisition, Investigation, Resources, Writing – review and editing.

Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Resources, Software, Supervision, Writing - original draft, Writing – review and editing.

Ethics

Experimental procedures were carried out in accordance with the University of Minnesota Institutional Animal Care and Use Committee and the National Institute of Health standards for the care and use of nonhuman primates. Protocol IDs: 2005-38127A 2005-38135A 1911-37623A.

Additional files

Supplementary file 1. The areas were defined based on a cortical and subcortical atlas.

The areas in Figure 1 were defined based on the Cortical Hierarchy Atlas of the Rhesus Macaque (CHARM) (Jung et al., 2021) and Subcortical Atlas of the Rhesus Macaque (SARM) (Hartig et al., 2021) to closely match the corresponding electrophysiological recording sites. For some of the cortical regions of interest, the CHARM parcellation did not match the recording sites. As a result, the areas were defined according to the respective descriptions (Note: this is the case whenever “Custom” is indicated in the table). Abbreviations: S1/S2 (primary/secondary somatosensory cortex), MT (middle temporal area), LIP (lateral intraparietal area), OFC (orbitofrontal cortex), LPFC (lateral prefrontal cortex), ACC (anterior cingulate cortex), PMd (dorsal premotor cortex), LOFC (lateral OFC), DLPFC (dorso-lateral PFC), VLPFC (ventro-lateral PFC), PFp (polar prefrontal cortex), vmPFC (ventro-medial PFC), sgACC (subgenual ACC), pgACC (pregenual ACC), dACC (dorsal ACC), GPe (external globus pallidus), STN (subthalamic nucleus).

elife-75540-supp1.docx (14.2KB, docx)
Supplementary file 2. Areas defined based on a cortical atlas.

The areas in Figure 2 were defined based on the Cortical Hierarchy Atlas of the Rhesus Macaque (CHARM) (Jung et al., 2021) to match the hierarchies described by Kravitz et al., 2011. Abbreviations: DLPFC (dorso-lateral prefrontal cortex), VIP (ventral intraparietal area), MIP (medial intraparietal area), rIPL (rostral inferior parietal lobule), MT (middle temporal area), MST (medial superior temporal area), LIP (lateral intraparietal area), A46 (area 46), A8A (area 8 A), PMd (dorsal premotor cortex), PMv (ventral premotor cortex).

elife-75540-supp2.docx (14.4KB, docx)
Transparent reporting form

Data availability

The functional connectivity gradient maps and the timescale maps have been uploaded to figshare. Functional connectivity gradients: https://doi.org/10.6084/m9.figshare.19189331 Intrinsic neural timescales: https://doi.org/10.6084/m9.figshare.19197026.

The following dataset was generated:

Manea AMG, Zilverstand A, Uǧurbil K, Heilbronner SR, Zimmermann J. 2021. Resting-state functional connectivity gradients and intrinsic neural timescales. figshare.

References

  1. Alexander GE, DeLong MR, Strick PL. Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annual Review of Neuroscience. 1986;9:357–381. doi: 10.1146/annurev.ne.09.030186.002041. [DOI] [PubMed] [Google Scholar]
  2. Autio JA, Zhu Q, Li X, Glasser MF, Schwiedrzik CM, Fair DA, Zimmermann J, Yacoub E, Menon RS, Van Essen DC, Hayashi T, Russ B, Vanduffel W. Minimal specifications for non-human primate MRI: Challenges in standardizing and harmonizing data collection. NeuroImage. 2021;236:118082. doi: 10.1016/j.neuroimage.2021.118082. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Avants BB, Tustison NJ, Wu J, Cook PA, Gee JC. An open source multivariate framework for n-tissue segmentation with evaluation on public data. Neuroinformatics. 2011;9:381–400. doi: 10.1007/s12021-011-9109-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Avants BB, Tustison NJ, Stauffer M, Song G, Wu B, Gee JC. The Insight ToolKit image registration framework. Frontiers in Neuroinformatics. 2014;8:44. doi: 10.3389/fninf.2014.00044. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Badre D. Cognitive control, hierarchy, and the rostro-caudal organization of the frontal lobes. Trends in Cognitive Sciences. 2008;12:193–200. doi: 10.1016/j.tics.2008.02.004. [DOI] [PubMed] [Google Scholar]
  6. Binkofski F, Buxbaum LJ. Two action systems in the human brain. Brain and Language. 2013;127:222–229. doi: 10.1016/j.bandl.2012.07.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Carmichael ST, Price JL. Architectonic subdivision of the orbital and medial prefrontal cortex in the macaque monkey. The Journal of Comparative Neurology. 1994;346:366–402. doi: 10.1002/cne.903460305. [DOI] [PubMed] [Google Scholar]
  8. Carmichael ST, Price JL. Connectional networks within the orbital and medial prefrontal cortex of macaque monkeys. The Journal of Comparative Neurology. 1996;371:179–207. doi: 10.1002/(SICI)1096-9861(19960722)371:2<179::AID-CNE1>3.0.CO;2-#. [DOI] [PubMed] [Google Scholar]
  9. Cavanagh SE, Wallis JD, Kennerley SW, Hunt LT. Autocorrelation structure at rest predicts value correlates of single neurons during reward-guided choice. eLife. 2016;5:e18937. doi: 10.7554/eLife.18937. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Cavanagh SE, Hunt LT, Kennerley SW. A diversity of intrinsic timescales underlie neural computations. Frontiers in Neural Circuits. 2020;14:615626. doi: 10.3389/fncir.2020.615626. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Chaudhuri R, Knoblauch K, Gariel MA, Kennedy H, Wang XJ. A large-scale circuit mechanism for hierarchical dynamical processing in the primate cortex. Neuron. 2015;88:419–431. doi: 10.1016/j.neuron.2015.09.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Cirillo R, Fascianelli V, Ferrucci L, Genovesio A. Neural Intrinsic Timescales in the Macaque Dorsal Premotor Cortex Predict the Strength of Spatial Response Coding. IScience. 2018;10:203–210. doi: 10.1016/j.isci.2018.11.033. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Cox RW. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Computers and Biomedical Research, an International Journal. 1996;29:162–173. doi: 10.1006/cbmr.1996.0014. [DOI] [PubMed] [Google Scholar]
  14. Demirtaş M, Burt JB, Helmer M, Ji JL, Adkinson BD, Glasser MF, Van Essen DC, Sotiropoulos SN, Anticevic A, Murray JD. Hierarchical Heterogeneity across Human Cortex Shapes Large-Scale Neural Dynamics. Neuron. 2019;101:1181–1194. doi: 10.1016/j.neuron.2019.01.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Fascianelli V, Tsujimoto S, Marcos E, Genovesio A. Autocorrelation Structure in the Macaque Dorsolateral, But not Orbital or Polar, Prefrontal Cortex Predicts Response-Coding Strength in a Visually Cued Strategy Task. Cerebral Cortex (New York, N.Y. 2019;29:230–241. doi: 10.1093/cercor/bhx321. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Friedman DP, Aggleton JP, Saunders RC. Comparison of hippocampal, amygdala, and perirhinal projections to the nucleus accumbens: combined anterograde and retrograde tracing study in the Macaque brain. The Journal of Comparative Neurology. 2002;450:345–365. doi: 10.1002/cne.10336. [DOI] [PubMed] [Google Scholar]
  17. Gao R, van den Brink RL, Pfeffer T, Voytek B. Neuronal timescales are functionally dynamic and shaped by cortical microarchitecture. eLife. 2020;9:e61277. doi: 10.7554/eLife.61277. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Goodale MA, Milner AD. Separate visual pathways for perception and action. Trends in Neurosciences. 1992;15:20–25. doi: 10.1016/0166-2236(92)90344-8. [DOI] [PubMed] [Google Scholar]
  19. Goodale MA, Meenan JP, Bülthoff HH, Nicolle DA, Murphy KJ, Racicot CI. Separate neural pathways for the visual analysis of object shape in perception and prehension. Current Biology. 1994;4:604–610. doi: 10.1016/s0960-9822(00)00132-9. [DOI] [PubMed] [Google Scholar]
  20. Grier MD, Zimmermann J, Heilbronner SR. Estimating Brain Connectivity With Diffusion-Weighted Magnetic Resonance Imaging: Promise and Peril. Biological Psychiatry. Cognitive Neuroscience and Neuroimaging. 2020;5:846–854. doi: 10.1016/j.bpsc.2020.04.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Haak KV, Marquand AF, Beckmann CF. Connectopic mapping with resting-state fMRI. NeuroImage. 2018;170:83–94. doi: 10.1016/j.neuroimage.2017.06.075. [DOI] [PubMed] [Google Scholar]
  22. Haak KV, Beckmann CF. Understanding brain organisation in the face of functional heterogeneity and functional multiplicity. NeuroImage. 2020;220:117061. doi: 10.1016/j.neuroimage.2020.117061. [DOI] [PubMed] [Google Scholar]
  23. Haber SN, Knutson B. The reward circuit: linking primate anatomy and human imaging. Neuropsychopharmacology. 2010;35:4–26. doi: 10.1038/npp.2009.129. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Haber SN. Corticostriatal circuitry. Dialogues in Clinical Neuroscience. 2016;18:7–21. doi: 10.31887/DCNS.2016.18.1/shaber. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Hallquist MN, Hwang K, Luna B. The nuisance of nuisance regression: spectral misspecification in a common approach to resting-state fMRI preprocessing reintroduces noise and obscures functional connectivity. NeuroImage. 2013;82:208–225. doi: 10.1016/j.neuroimage.2013.05.116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Hartig R, Glen D, Jung B, Logothetis NK, Paxinos G, Garza-Villarreal EA, Messinger A, Evrard HC. The Subcortical Atlas of the Rhesus Macaque (SARM) for neuroimaging. NeuroImage. 2021;235:117996. doi: 10.1016/j.neuroimage.2021.117996. [DOI] [PubMed] [Google Scholar]
  27. Harvey BM, Dumoulin SO, Fracasso A, Paul JM. A Network of Topographic Maps in Human Association Cortex Hierarchically Transforms Visual Timing-Selective Responses. Current Biology. 2020;30:1424–1434. doi: 10.1016/j.cub.2020.01.090. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Hasson U, Yang E, Vallines I, Heeger DJ, Rubin N. A hierarchy of temporal receptive windows in human cortex. The Journal of Neuroscience. 2008;28:2539–2550. doi: 10.1523/JNEUROSCI.5487-07.2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Honey CJ, Thesen T, Donner TH, Silbert LJ, Carlson CE, Devinsky O, Doyle WK, Rubin N, Heeger DJ, Hasson U. Slow cortical dynamics and the accumulation of information over long timescales. Neuron. 2012;76:423–434. doi: 10.1016/j.neuron.2012.08.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Ito T, Hearne LJ, Cole MW. A cortical hierarchy of localized and distributed processes revealed via dissociation of task activations, connectivity changes, and intrinsic timescales. NeuroImage. 2020;221:117141. doi: 10.1016/j.neuroimage.2020.117141. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Jbabdi S, Sotiropoulos SN, Behrens TE. The topographic connectome. Current Opinion in Neurobiology. 2013;23:207–215. doi: 10.1016/j.conb.2012.12.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Jenkinson M, Beckmann CF, Behrens TEJ, Woolrich MW, Smith SM. FSL. NeuroImage. 2012;62:782–790. doi: 10.1016/j.neuroimage.2011.09.015. [DOI] [PubMed] [Google Scholar]
  33. Jung B, Taylor PA, Seidlitz J, Sponheim C, Perkins P, Ungerleider LG, Glen D, Messinger A. A comprehensive macaque fMRI pipeline and hierarchical atlas. NeuroImage. 2021;235:117997. doi: 10.1016/j.neuroimage.2021.117997. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Kiebel SJ, Daunizeau J, Friston KJ. A hierarchy of time-scales and the brain. PLOS Computational Biology. 2008;4:e1000209. doi: 10.1371/journal.pcbi.1000209. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Kravitz DJ, Saleem KS, Baker CI, Mishkin M. A new neural framework for visuospatial processing. Nature Reviews. Neuroscience. 2011;12:217–230. doi: 10.1038/nrn3008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Lagore RL, Moeller S, Zimmermann J, DelaBarre L, Radder J, Grant A, Ugurbil K, Yacoub E, Harel N, Adriany G. An 8-dipole transceive and 24-loop receive array for non-human primate head imaging at 10.5 T. NMR in Biomedicine. 2021;34:e4472. doi: 10.1002/nbm.4472. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Logothetis NK, Pauls J, Augath M, Trinath T, Oeltermann A. Neurophysiological investigation of the basis of the fMRI signal. Nature. 2001;412:150–157. doi: 10.1038/35084005. [DOI] [PubMed] [Google Scholar]
  38. Maisson DJN, Cash-Padgett TV, Wang MZ, Hayden BY, Heilbronner SR, Zimmermann J. Choice-relevant information transformation along a ventrodorsal axis in the medial prefrontal cortex. Nature Communications. 2021;12:4830. doi: 10.1038/s41467-021-25219-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Marquand AF, Haak KV, Beckmann CF. Functional corticostriatal connection topographies predict goal directed behaviour in humans. Nature Human Behaviour. 2017;1:0146. doi: 10.1038/s41562-017-0146. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Moeller S, Yacoub E, Olman CA, Auerbach E, Strupp J, Harel N, Uğurbil K. Multiband multislice GE-EPI at 7 tesla, with 16-fold acceleration using partial parallel imaging with application to high spatial and temporal whole-brain fMRI. Magnetic Resonance in Medicine. 2010;63:1144–1153. doi: 10.1002/mrm.22361. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Murray JD, Bernacchia A, Freedman DJ, Romo R, Wallis JD, Cai X, Padoa-Schioppa C, Pasternak T, Seo H, Lee D, Wang X-J. A hierarchy of intrinsic timescales across primate cortex. Nature Neuroscience. 2014;17:1661–1663. doi: 10.1038/nn.3862. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Nougaret S, Fascianelli V, Ravel S, Genovesio A. Intrinsic timescales across the basal ganglia. Scientific Reports. 2021;11:21395. doi: 10.1038/s41598-021-00512-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Ogawa T, Komatsu H. Differential temporal storage capacity in the baseline activity of neurons in macaque frontal eye field and area V4. Journal of Neurophysiology. 2010;103:2433–2445. doi: 10.1152/jn.01066.2009. [DOI] [PubMed] [Google Scholar]
  44. Ongür D, Price JL. The organization of networks within the orbital and medial prefrontal cortex of rats, monkeys and humans. Cerebral Cortex (New York, N.Y. 2000;10:206–219. doi: 10.1093/cercor/10.3.206. [DOI] [PubMed] [Google Scholar]
  45. Price JL. Networks within the orbital and medial prefrontal cortex. Neurocase. 1999;5:231–241. doi: 10.1080/13554799908402728. [DOI] [Google Scholar]
  46. Przezdzik I, Faber M, Fernandez G, Beckmann CF, Haak KV. The functional organisation of the hippocampus along its long axis is gradual and predicts recollection. Cortex; a Journal Devoted to the Study of the Nervous System and Behavior. 2019;119:324–335. doi: 10.1016/j.cortex.2019.04.015. [DOI] [PubMed] [Google Scholar]
  47. Raut RV, Snyder AZ, Raichle ME. Hierarchical dynamics as a macroscopic organizing principle of the human brain. PNAS. 2020;117:20890–20897. doi: 10.1073/pnas.2003383117. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Riley MR, Qi XL, Zhou X, Constantinidis C. Anterior-posterior gradient of plasticity in primate prefrontal cortex. Nature Communications. 2018;9:3790. doi: 10.1038/s41467-018-06226-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Rossi-Pool R, Zainos A, Alvarez M, Parra S, Zizumbo J, Romo R. Invariant timescale hierarchy across the cortical somatosensory network. PNAS. 2021;118:e2021843118. doi: 10.1073/pnas.2021843118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Saad ZS, Reynolds RC, Argall B, Japee S, Cox RW. SUMA: an interface for surface-based intra- and inter-subject analysis with AFNI. 2004 2nd IEEE International Symposium on Biomedical Imaging.2004. [Google Scholar]
  51. Saad ZS, Reynolds RC. SUMA. NeuroImage. 2012;62:768–773. doi: 10.1016/j.neuroimage.2011.09.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Seidlitz J, Sponheim C, Glen D, Ye FQ, Saleem KS, Leopold DA, Ungerleider L, Messinger A. A population MRI brain template and analysis tools for the macaque. NeuroImage. 2018;170:121–131. doi: 10.1016/j.neuroimage.2017.04.063. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Setsompop K, Gagoski BA, Polimeni JR, Witzel T, Wedeen VJ, Wald LL. Blipped-controlled aliasing in parallel imaging for simultaneous multislice echo planar imaging with reduced g-factor penalty. Magnetic Resonance in Medicine. 2012;67:1210–1224. doi: 10.1002/mrm.23097. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Shinn M, Hu A, Turner L, Noble S, Achard S, Anticevic A, Scheinost D, Constable RT, Lee D, Bullmore ET, Murray JD. Spatial and Temporal Autocorrelation Weave Human Brain Networks. bioRxiv. 2021 doi: 10.1101/2021.06.01.446561. [DOI]
  55. Soltani A, Murray JD, Seo H, Lee D. Timescales of Cognition in the Brain. Current Opinion in Behavioral Sciences. 2021;41:30–37. doi: 10.1016/j.cobeha.2021.03.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Tan PK, Tang C, Herikstad R, Pillay A, Libedinsky C. Distinct Lateral Prefrontal Regions Are Organized in an Anterior-Posterior Functional Gradient. bioRxiv. 2020 doi: 10.1101/2020.12.16.423034. [DOI] [PMC free article] [PubMed]
  57. Uğurbil K, Xu J, Auerbach EJ, Moeller S, Vu AT, Duarte-Carvajalino JM, Lenglet C, Wu X, Schmitter S, Van de Moortele PF, Strupp J, Sapiro G, De Martino F, Wang D, Harel N, Garwood M, Chen L, Feinberg DA, Smith SM, Miller KL, Sotiropoulos SN, Jbabdi S, Andersson JLR, Behrens TEJ, Glasser MF, Van Essen DC, Yacoub E, WU-Minn HCP Consortium Pushing spatial and temporal resolution for functional and diffusion MRI in the Human Connectome Project. NeuroImage. 2013;80:80–104. doi: 10.1016/j.neuroimage.2013.05.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Ugurbil K. What is feasible with imaging human brain function and connectivity using functional magnetic resonance imaging. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences. 2016;371:20150361. doi: 10.1098/rstb.2015.0361. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Uğurbil K. Imaging at ultrahigh magnetic fields: History, challenges, and solutions. NeuroImage. 2018;168:7–32. doi: 10.1016/j.neuroimage.2017.07.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Uğurbil K, Auerbach E, Moeller S, Grant A, Wu X, Van de Moortele PF, Olman C, DelaBarre L, Schillak S, Radder J, Lagore R, Adriany G. Brain imaging with improved acceleration and SNR at 7 Tesla obtained with 64-channel receive array. Magnetic Resonance in Medicine. 2019;82:495–509. doi: 10.1002/mrm.27695. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Vos de Wael R, Benkarim O, Paquola C, Lariviere S, Royer J, Tavakol S, Xu T, Hong SJ, Langs G, Valk S, Misic B, Milham M, Margulies D, Smallwood J, Bernhardt BC. BrainSpace: a toolbox for the analysis of macroscale gradients in neuroimaging and connectomics datasets. Communications Biology. 2020;3:103. doi: 10.1038/s42003-020-0794-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Watanabe T, Rees G, Masuda N. Atypical intrinsic neural timescale in autism. eLife. 2019;8:e42256. doi: 10.7554/eLife.42256. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Wengler K, Goldberg AT, Chahine G, Horga G. Distinct hierarchical alterations of intrinsic neural timescales account for different manifestations of psychosis. eLife. 2020;9:e56151. doi: 10.7554/eLife.56151. [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Whitfield-Gabrieli S, Nieto-Castanon A. Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks. Brain Connectivity. 2012;2:125–141. doi: 10.1089/brain.2012.0073. [DOI] [PubMed] [Google Scholar]
  65. Wong AL, Haith AM, Krakauer JW. Motor Planning. The Neuroscientist. 2015;21:385–398. doi: 10.1177/1073858414541484. [DOI] [PubMed] [Google Scholar]
  66. Yacoub E, Grier MD, Auerbach EJ, Lagore RL, Harel N, Adriany G, Zilverstand A, Hayden BY, Heilbronner SR, Uğurbil K, Zimmermann J. Ultra-high field (10.5 T) resting state fMRI in the macaque. NeuroImage. 2020;223:117349. doi: 10.1016/j.neuroimage.2020.117349. [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Zilio F, Gomez-Pilar J, Cao S, Zhang J, Zang D, Qi Z, Tan J, Hiromi T, Wu X, Fogel S, Huang Z, Hohmann MR, Fomina T, Synofzik M, Grosse-Wentrup M, Owen AM, Northoff G. Are intrinsic neural timescales related to sensory processing? Evidence from abnormal behavioral states. NeuroImage. 2021;226:117579. doi: 10.1016/j.neuroimage.2020.117579. [DOI] [PubMed] [Google Scholar]
  68. Zimmermann J, Glimcher PW, Louie K. Multiple timescales of normalized value coding underlie adaptive choice behavior. Nature Communications. 2018;9:3206. doi: 10.1038/s41467-018-05507-8. [DOI] [PMC free article] [PubMed] [Google Scholar]

Editor's evaluation

Daeyeol Lee 1

Neural activity measured in both electrophysiological and functional neuroimaging experiments are often temporally correlated, and the timescales of such correlation in ongoing neural activity, or intrinsic neural timescales, show a hierarchical pattern across the cortical surface. The present study establishes a close link between these timescales and functional connectivity in the brains of non-human primates, suggesting that temporal autocorrelation is an important organizing feature of large-scale neural activity.

Decision letter

Editor: Daeyeol Lee1
Reviewed by: Daeyeol Lee2

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Intrinsic timescales as an organizational principle of neural processing across the whole rhesus macaque brain" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Daeyeol Lee as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Timothy Behrens as the Senior Editor.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1. Effects of anesthesia. Compared to previous human neuroimaging studies and NHP electrophysiological experiments, a significant weakness of this study is that the experiments were performed in anesthetized animals. This is even more problematic, since the analyses in this study mostly focus on the fronto-parietal network, which might be more affected by anesthetics. Although it has been shown that the FC can be studied in anesthetized animals, how specific methods of anesthesia might influence by the measures of INT should be discussed more thoroughly.

2. INT measure. They authors put forth the INT measure as related to intrinsic timescale. INT is defined as the integrated area under the autocorrelation function (ACF) up until the first point where the ACF goes below zero. This is different than how intrinsic timescale has been measured in single-neuron spike trains or in prior fMRI studies. While a longer timescale would be expected to increase INT, the problem is that INT (as an integrated area) combines effects of autocorrelation timescale and autocorrelation amplitude.

– It would be insightful to visualize INT properties at the whole-brain or whole-cortex level (instead of only a single lobe), including (i) INT values themselves, (ii) the lag-one autcorrelation value reflecting autocorrelation amplitude, and (iii) the zero-crossing lag time used to compute INT.

– A highly relevant paper (which is not currently cited) is Ito et al., (2020) NeuroImage, "A cortical hierarchy of localized and distributed processes revealed via dissociation of task activations, connectivity changes, and intrinsic timescales". Figure 5 of Ito shows a cortex-wide map of intrinsic timescale as defined in single-neuron studies (i.e. fitting time constant of decay). Figure 6 then shows this is related to cortical hierarchy as reflected in the T1w/T2w map (which in principle could be tested by the authors here too). Ito's analysis was performed on the parcellated timeseries, not the voxel level as in the present study, which is a notable methodological difference.

– That INT combines effects of timescale and amplitude would not be a problem if the autocorrelation amplitude does not vary across brain regions. However, it appears that it does for whatever reasons (neural and/or in fMRI measurement such as SNR). A relevant preprint is by Shinn et al., (2021) bioRxiv, “Spatial and temporal autocorrelation weave human brain networks”. In human cortex, again using parcellated timeseries, Figure 1F there shows systematic variation across cortical parcels of the lag-1 autocorrelation value. In the present study, it is currently unknown whether INT is reflecting regional differences in autocorrelation timescale (as interpreted), amplitude (not considered), or both.

– Contribution of autocorrelation amplitude to INT may potentially explain why a cortex-wide map of INT does not follow an expected hierarchy as much the more restricted views within one lobe as the current manuscript focuses. For instance, Figure 1 shows that INT values for somatosensory cortex (Figure 1A) are larger than association regions (Figure 1B). Is this potentially due to autocorrelation amplitude being larger in somatosensory cortex?

– Perhaps some smoothing or parcellation would be required to better tease apart autocorrelation timescale from autocorrelation amplitude.

– A highly relevant paper (which is not currently cited) is Ito et al., (2020) NeuroImage, "A cortical hierarchy of localized and distributed processes revealed via dissociation of task activations, connectivity changes, and intrinsic timescales". Figure 5 of Ito shows a cortex-wide map of intrinsic timescale as defined in single-neuron studies (i.e. fitting time constant of decay). Figure 6 then shows this is related to cortical hierarchy as reflected in the T1w/T2w map (which in principle could be tested by the authors here too). Ito's analysis was performed on the parcellated timeseries, not the voxel level as in the present study, which is a notable methodological difference.

– That INT combines effects of timescale and amplitude would not be a problem if the autocorrelation amplitude does not vary across brain regions. However, it appears that it does for whatever reasons (neural and/or in fMRI measurement such as SNR). A relevant preprint is by Shinn et al., (2021) bioRxiv, "Spatial and temporal autocorrelation weave human brain networks". In human cortex, again using parcellated timeseries, Figure 1F there shows systematic variation across cortical parcels of the lag-1 autocorrelation value. In the present study, it is currently unknown whether INT is reflecting regional differences in autocorrelation timescale (as interpreted), amplitude (not considered), or both.

– Contribution of autocorrelation amplitude to INT may potentially explain why a cortex-wide map of INT does not follow an expected hierarchy as much the more restricted views within one lobe as the current manuscript focuses. For instance, Figure 1 shows that INT values for somatosensory cortex (Figure 1A) are larger than association regions (Figure 1B). Is this potentially due to autocorrelation amplitude being larger in somatosensory cortex?

– Perhaps some smoothing or parcellation would be required to better tease apart autocorrelation timescale from autocorrelation amplitude.

3. Does the calculation of fMRI-based neuronal time constants obscure the unit of time? True comparison with ephys data is not possible without clarifying the relationship of the two quantities compared. In the ephys measurements time constants are in units of seconds and often below 1s. In contrast, BOLD response has a sluggish time course (tens of seconds) due to the properties of the hemodynamic response function. The smoothing of spiking and field-potential activity with the HRF introduces substantial auto-correlation in BOLD and is expected to reduce our ability to distinguish small differences of time constants discovered with ephys. Because the analyses in this paper do not explore the complications caused by the slow and noisy BOLD measurements, it is impossible to know if the observed temporal hierarchy has the same nature and origin as those reported with ephys. It would be useful to perform additional analyses and modeling that clarifies this missing link. If that is not possible, at the very least I would recommend explicit reporting of the units of time constants based on BOLD in the figures, and discussing if the differences of BOLD time constants across regions match the differences of spiking activity time constants in previous publications. Also, throughout the paper, figures should clearly indicate the unit for INT.

4. Functional connectivity gradients: Figures 3 and 4 rely on functional connectivity gradients calculated within a single lobe, against which INT topography is correlated. On such a restricted geometry as a single lobe, a functional connectivity gradient may be reflecting a simpler property, namely the geometry of the restricted cortical sheet. In other words, given the sheet geometry of the frontal lobe, does an anterior-posterior topography fall out naturally as the first gradient (e.g. with distance-dependent falloff) and medial-lateral as second gradient? If so, it is difficult to strongly interpret these results as linking INT to functional connectivity when the gradient is a generic consequence of the sheet geometry. In human neuroimaging such functional gradients are typically calculated at the whole-cortex level which reveals less trivial topographies (e.g. Margulies et al., 2016, PNAS). These results and interpretations should be considered in light of this concern.

5. There is no strong relationship between within area FC and INT maps in figure 4. Unlike the FC maps that seem to monotonically change along a cortical axis, the INT maps seem to have non-monotonic changes. For example, the INT map for OFC has maximum values both in the anterior-medial and caudal-lateral regions. Similar discrepancies between the INT and FC maps are apparent for PFC. Also, unlike the FC maps, the INT maps seem asymmetrical in the two hemispheres. Unfortunately, the visualization and stats are too dense to probe what is going on. Deeper analyses and explanation are necessary (please see the next comment for a potential solution).

6. Many fine level discrepancies are also observed between striatal INTs and cortico-striatal INTs. In the current figures, it is impossible for readers to know if the variations and discrepancies are within the range expected from noise in the data. The alternative is that some of the variations are meaningful but ignored by the authors to keep the narrative simple. Either way, the authors are encouraged to improve their presentation method and analyses in order to provide adequate information for readers. One possibility is to use old-fashioned scatter plots with error bars, where individual data points are different locations within the striatum and the axes are striatal INTs and cortico-striatal INTs. The authors can choose how they'd like to divide striatum into distinct sub-regions (e.g., a simple 3D grid or functional divisions). Similar plots would be useful for figure 4, and more generally anywhere that the authors report a correlation between INTs and FCs. 7.

7. The authors should provide more detailed information about how the hierarchy shown in the x-axis of Figure 1 was computed, so that this can be replicated by other investigators. Some schematic diagram for cortical hierarchy might be also helpful.

8. The paper would have more value if the same analyses were carried out in other brain areas, such as the temporal cortex, that are not currently covered. If such results are not included, the authors should at least provide a reason why the paper focuses only on the frontal and parietal regions.

eLife. 2022 Mar 2;11:e75540. doi: 10.7554/eLife.75540.sa2

Author response

Essential revisions:

1. Effects of anesthesia. Compared to previous human neuroimaging studies and NHP electrophysiological experiments, a significant weakness of this study is that the experiments were performed in anesthetized animals. This is even more problematic, since the analyses in this study mostly focus on the fronto-parietal network, which might be more affected by anesthetics. Although it has been shown that the FC can be studied in anesthetized animals, how specific methods of anesthesia might influence by the measures of INT should be discussed more thoroughly.

The reviewers raise an important question and concern, and we would like to apologize for this omission. In a previous paper from our lab (Yacoub et al., 2020) we were able to demonstrate that our low isoflurane + dexamethasone (Autio et al., 2021)ng autocorrelation amplitude, and (iii) the zero-crossing la anesthesia demonstrates no observable biases to whole brain connectivity analysis, and stimulation experiments in our lab (unpublished) demonstrate non altered hemodynamic response function properties. In fact, hemodynamic properties seem to be significantly more affected by the common use of contrast agents such as ferromolecules (Pelekanos et al., 2020). In our experience, the largest effect of vasodilative anesthetics (such as isoflurane) is an overall blunting of responses which the use of ultrahigh magnetic fields counteracts. We have added the following paragraph to the discussion of the manuscript for transparency:

“One potential limitation of the current study is using anesthetized subjects. Although obtaining high resolution images of awake NHPs at ultrahigh fields would be ideal, it is still an enormous technical challenge that the field is working to overcome. Using our low anesthesia strategy, we were able to obtain a relatively large sample size of NHPs (because extensive training was not required). It is important to note that it was previously shown that our anesthesia protocol demonstrates no observable biases to whole-brain connectivity analysis (Yacoub et al., 2020). Moreover, it has been shown that this protocol has a relatively low impact on reproducibility (Autio et al., 2021). To our knowledge, the largest effect of vasodilative anesthetics (such as isoflurane) is an overall blunting of responses which is counteracted by ultrahigh field MRI. Finally, besides the possible effects on the hemodynamic response, anesthesia introduces a crucial difference between the current study and previous research – while in previous studies it was not controlled for the subjects engaging in various cognitive processes, in anesthetized subjects, this confound is potentially removed. Overall, future research should replicate current results to exclude a possible effect of anesthesia.”

2. INT measure. They authors put forth the INT measure as related to intrinsic timescale. INT is defined as the integrated area under the autocorrelation function (ACF) up until the first point where the ACF goes below zero. This is different than how intrinsic timescale has been measured in single-neuron spike trains or in prior fMRI studies. While a longer timescale would be expected to increase INT, the problem is that INT (as an integrated area) combines effects of autocorrelation timescale and autocorrelation amplitude.

– It would be insightful to visualize INT properties at the whole-brain or whole-cortex level (instead of only a single lobe), including (i) INT values themselves, (ii) the lag-one autcorrelation value reflecting autocorrelation amplitude, and (iii) the zero-crossing lag time used to compute INT.

– A highly relevant paper (which is not currently cited) is Ito et al., (2020) NeuroImage, "A cortical hierarchy of localized and distributed processes revealed via dissociation of task activations, connectivity changes, and intrinsic timescales". Figure 5 of Ito shows a cortex-wide map of intrinsic timescale as defined in single-neuron studies (i.e. fitting time constant of decay). Figure 6 then shows this is related to cortical hierarchy as reflected in the T1w/T2w map (which in principle could be tested by the authors here too). Ito's analysis was performed on the parcellated timeseries, not the voxel level as in the present study, which is a notable methodological difference.

– That INT combines effects of timescale and amplitude would not be a problem if the autocorrelation amplitude does not vary across brain regions. However, it appears that it does for whatever reasons (neural and/or in fMRI measurement such as SNR). A relevant preprint is by Shinn et al., (2021) bioRxiv, "Spatial and temporal autocorrelation weave human brain networks". In human cortex, again using parcellated timeseries, Figure 1F there shows systematic variation across cortical parcels of the lag-1 autocorrelation value. In the present study, it is currently unknown whether INT is reflecting regional differences in autocorrelation timescale (as interpreted), amplitude (not considered), or both.

– Contribution of autocorrelation amplitude to INT may potentially explain why a cortex-wide map of INT does not follow an expected hierarchy as much the more restricted views within one lobe as the current manuscript focuses. For instance, Figure 1 shows that INT values for somatosensory cortex (Figure 1A) are larger than association regions (Figure 1B). Is this potentially due to autocorrelation amplitude being larger in somatosensory cortex?

– Perhaps some smoothing or parcellation would be required to better tease apart autocorrelation timescale from autocorrelation amplitude.

The possible confounding effect of a systematic regional variation of ACF amplitude is an important consideration that we have overlooked and thank the reviewers for pointing out. To further examine this issue: (1) we plotted whole-brain INTs, lag-1, and 0-crossing lag (see Figure 1 Supplement 1); (2) correlated the maps; (3) re-computed the hierarchies by using the 0crossing lag multiplied by the TR (i.e., to obtain a time estimate; see Figure 1 Supplement 2). In this study, while the amplitude as reflected by the lag-1 ACF value varies across the brain, it is not dissociated from the timescale as reflected by the 0-crossing lag. The three voxel-wise maps are highly correlated (Pearson’s r > 0.90) and re-computing the hierarchies by only using the 0-crossing lag does not change the relative “temporal” position of the various areas. Notably, the variations in the lag-1 ACF value were small, with values ranging from 0.92 to 0.95 for subcortical and 0.92 and 0.97 for cortical voxels. In consequence, it is not surprising that the hierarchies are not impacted by removing the amplitude of the ACF values from the INT measure. We thus extend the findings by Shinn et al., (2021) by showing that the lag-1 ACF value is closely related to our INT measure as well. Compared to solely using the lag-1 ACF value, our INT measure can be used to estimate the time constants of the various areas as reflected by BOLD, an added benefit in our opinion.

In contrast to previous studies, we decided to use this INT measure for two reasons: (1) it has already been validated via simultaneous EEG-fMRI recordings by Wanatabe and Masuda (2019); (2) we avoided curve fitting to reduce the adverse effects of the low sampling rate of fMRI recordings.

To clarify these important points to the reader, we added Supplement 1 and 2 to Figure 1. Moreover, we added the following paragraph to the Results section:

“Our definition of INTs combines the effects of ACF magnitude and timescale (i.e., the 0-crossing lag). Since it has previously been shown that the magnitude of the ACF values, parameterized as the lag-1 ACF value, differs across brain areas (Shinn et al., 2021), we further examined whether the observed hierarchies are driven by magnitude, timescale, or both. In line with previous results, the lag-1 ACF values systematically varied across the brain (Shinn et al., 2021). Moreover, the lag-1 ACF values and the 0-crossing lag were linearly related to INTs as operationalized in this study (see Figure 1 Supplement 1 for voxel-wise whole-brain maps of the lag-1, 0-crossing lag and INTs). The maps suggest that magnitude and timescale are closely related. To exclude the possibility that the observed hierarchies are driven by magnitude, rather than timescale, we recomputed them by using the 0-crossing lag as a proxy for INT. To obtain a time estimate of hemodynamic timescales, the 0-lag crossing lag was multiplied by the TR (i.e., 1.1s). The hierarchical ordering of both cortical and subcortical areas is consistent (see Figure 1 Supplement 2) with that based on our INT measure (see Figure 1). Hence, while the ACF magnitude is closely related to our INT index and the 0-crossing lag, temporal hierarchies are not driven by heterogeneity in the ACF magnitude.”

The reviewers bring up another interesting point about T1/T2(*) weighted anatomical images and their potential relationship to our findings. While we have acquired T2 weighted images in most of our animals they are unfortunately not purely T2 but rather a variable flip angle approach. This approach was taken due to technical difficulties porting standard sequences to our 10.5T instrument. We have recently made additional progress on allowing “purer” T2 TSE images, but for this dataset, we do not have the correct modalities available. We hope we can provide that in the future. We think there are multiple interesting parameters that should be examined in the future for their relationship with INTs, including T1/T2(*) as well as other neuroimaging and histologically derived measures.

– A highly relevant paper (which is not currently cited) is Ito et al., (2020) NeuroImage, "A cortical hierarchy of localized and distributed processes revealed via dissociation of task activations, connectivity changes, and intrinsic timescales". Figure 5 of Ito shows a cortex-wide map of intrinsic timescale as defined in single-neuron studies (i.e. fitting time constant of decay). Figure 6 then shows this is related to cortical hierarchy as reflected in the T1w/T2w map (which in principle could be tested by the authors here too). Ito's analysis was performed on the parcellated timeseries, not the voxel level as in the present study, which is a notable methodological difference.

– That INT combines effects of timescale and amplitude would not be a problem if the autocorrelation amplitude does not vary across brain regions. However, it appears that it does for whatever reasons (neural and/or in fMRI measurement such as SNR). A relevant preprint is by Shinn et al., (2021) bioRxiv, "Spatial and temporal autocorrelation weave human brain networks". In human cortex, again using parcellated timeseries, Figure 1F there shows systematic variation across cortical parcels of the lag-1 autocorrelation value. In the present study, it is currently unknown whether INT is reflecting regional differences in autocorrelation timescale (as interpreted), amplitude (not considered), or both.

– Contribution of autocorrelation amplitude to INT may potentially explain why a cortex-wide map of INT does not follow an expected hierarchy as much the more restricted views within one lobe as the current manuscript focuses. For instance, Figure 1 shows that INT values for somatosensory cortex (Figure 1A) are larger than association regions (Figure 1B). Is this potentially due to autocorrelation amplitude being larger in somatosensory cortex?

– Perhaps some smoothing or parcellation would be required to better tease apart autocorrelation timescale from autocorrelation amplitude.

3. Does the calculation of fMRI-based neuronal time constants obscure the unit of time? True comparison with ephys data is not possible without clarifying the relationship of the two quantities compared. In the ephys measurements time constants are in units of seconds and often below 1s. In contrast, BOLD response has a sluggish time course (tens of seconds) due to the properties of the hemodynamic response function. The smoothing of spiking and field-potential activity with the HRF introduces substantial auto-correlation in BOLD and is expected to reduce our ability to distinguish small differences of time constants discovered with ephys. Because the analyses in this paper do not explore the complications caused by the slow and noisy BOLD measurements, it is impossible to know if the observed temporal hierarchy has the same nature and origin as those reported with ephys. It would be useful to perform additional analyses and modeling that clarifies this missing link. If that is not possible, at the very least I would recommend explicit reporting of the units of time constants based on BOLD in the figures, and discussing if the differences of BOLD time constants across regions match the differences of spiking activity time constants in previous publications. Also, throughout the paper, figures should clearly indicate the unit for INT.

This is a very important question and hopefully future research will shed light on the exact relationship between timescales estimated from spiking activity, LFPs and BOLD. Watanabe and Masuda (2019) have shown that hemodynamic INTs differ by two orders of magnitude from those estimated from EEG. We hypothesize a similar relationship, but it is very unlikely that the hemodynamic INTs reflect only what has been previously estimated based on spiking activity. For transparency, we are adding the following paragraph to the Discussion:

“It is worth noting that the current study could not resolve the exact correspondence between the time constants estimated from spiking activity and BOLD. The relationship between neural activity and the BOLD signal has been revealed via simultaneous electrophysiological-fMRI recording-i.e., a spatially localized increase in the fMRI signal directly and monotonically reflects an increase in neural activity, with the mean onset time of the BOLD signal relative to the neural activation being 2s (Logothetis, Pauls, Augath, Trinath, and Oeltermann, 2001). As a result of sensory stimulation, BOLD signal is delayed by ~ 2-3 s, followed by a ramp of 6-12 s to a plateau or peak value. Local field potentials (LFPs), and to a lesser extent multi-unit activity (MUAs), can predict the BOLD signal when convolved with the neuro-vascular impulse response function. In sum, the hemodynamic response is roughly a low-pass-filtered expression of the total neural activity; with LFPs however, having a greater contribution to the signal. This has several implications for the interpretation of hemodynamic timescales: (1) they are more likely to reflect timescales related to LFPs rather than spiking activity; (2) the time constants estimated from spiking activity might not be linearly related to those estimated from fMRI. A further complication for the interpretability of hemodynamic timescales is that INTs have only been estimated from spiking activity, but not LFPs. The exact relationship between timescales derived from spiking activity, LFPs and the BOLD signal remains an empirical question. We hypothesize that hemodynamic timescales generally increase monotonically with those estimated from electrophysiology. In that respect, Watanabe and Masuda (2019) bring preliminary evidence – i.e., timescales derived from electroencephalography (EEG) and fMRI have been shown to differ by two orders of magnitude. This is reasonable considering the BOLD time course-when convolved with the hemodynamic response function, the INTs estimated from EEG were similar in magnitude to those estimated from fMRI. Nevertheless, we show that temporal hierarchies are a general principle of organization, irrespective of the method used to estimate them.”

Moreover, we have estimated the hemodynamic time constants by multiplying the 0-crossing lag by the TR. With those results in hand, we have re-computed the INT hierarchies from Figure 1 (see Figure 1 Supplement 2) to get an estimate for the areas that have been previously investigated with electrophysiology. With respect to the unit for INT, since the hemodynamic INTs were estimated as the sum of the ACF values in the initial positive period (our measure of INT considers both the number of lags and the magnitude of the ACF values), the INT does not have a time unit and it is used as a proxy of the timescale of that area, with higher values reflecting longer timescale. (See Figure 1 Supplement 2 for an estimate of the time constant of each area). To make this clear throughout the manuscript, we added the following paragraph to the Methods section and in the figure captions throughout the manuscript.

“Since the hemodynamic INTs were estimated as the sum of ACF values in the initial positive period, this measurement considers both the number of lags and the magnitude of the ACF values. As a result, the INT does not have a time unit and it is used as a proxy of the timescale of that area, with higher values reflecting longer timescales. We have nonetheless estimated the time constants associated with the INT hierarchies that have been previously described with electrophysiological recordings by multiplying the 0-crossing lag by the TR (see Figure 1 Supplement 2 for an estimate of the time constants).”

4. Functional connectivity gradients: Figures 3 and 4 rely on functional connectivity gradients calculated within a single lobe, against which INT topography is correlated. On such a restricted geometry as a single lobe, a functional connectivity gradient may be reflecting a simpler property, namely the geometry of the restricted cortical sheet. In other words, given the sheet geometry of the frontal lobe, does an anterior-posterior topography fall out naturally as the first gradient (e.g. with distance-dependent falloff) and medial-lateral as second gradient? If so, it is difficult to strongly interpret these results as linking INT to functional connectivity when the gradient is a generic consequence of the sheet geometry. In human neuroimaging such functional gradients are typically calculated at the whole-cortex level which reveals less trivial topographies (e.g. Margulies et al., 2016, PNAS). These results and interpretations should be considered in light of this concern.

Thank you for this insightful comment and question. Although previous studies have applied gradient approaches to the whole-brain, we avoided this approach for two reasons: (1) this is usually achieved by considerably downsampling the data, which would erase the spatial resolution provided by the ultra-high field; (2) the result would reflect the overall inter-areal connectivity differences, since inter-areal connectivity patterns are generally more pronounced than intra-areal patterns, rather than the fine-grained connectivity patterns within the individual areas. With respect to the first point, downsampling the data would potentially blur import functional details. It has been previously shown that gradients estimated for smaller regions of interest uncover important functional details across individuals – and even more, these gradients cand predict performance (Przeździk, Faber, Fernández, Beckmann, and Haak, 2019). With respect to the second point, we were not interested in capturing large-scale modes of connectivity change, but rather the functional organization of individual areas.

To address the comment raised by the reviewers, we parcellated the brain using the 6th level (most fine grained) of the CHARM (Jung et al., 2021), and we calculated the FC matrix using the resulting timeseries. This resulted in a 134x134 FC matrix. This was used as input to our pipeline of estimating gradients. While the first gradient reflected an overall anterior-posterior axis of connectivity change, the second gradient captured the change in whole-brain INTs. (Pearson’s r = 0.70; see Figure 3 Supplement 1). We therefore conclude that the whole-brain result supports our initially proposed interpretation.

In addition to the addition of Figure 3 Supplement 1, we add the following paragraph to the Results section:

“Although we were mainly interested in the fine-grained functional organization of individual areas, we asked whether the INT-FC relationship holds at the whole-cortex level (see Figure 3 Supplement 1). It is important to note that the whole-cortex estimation reflects overall inter-areal FC differences rather than the fine-grained FC patterns in smaller areas; that is because inter-areal connectivity differences are more pronounced than intra-areal differences. The whole-cortex INT map was highly correlated with the second gradient estimated from the cortical FC matrix (r = 0.70). With these results in hand, we conclude that INTs and FC gradients are closely related and represent a functional principle of organization across the cortex.”

5. There is no strong relationship between within area FC and INT maps in figure 4. Unlike the FC maps that seem to monotonically change along a cortical axis, the INT maps seem to have non-monotonic changes. For example, the INT map for OFC has maximum values both in the anterior-medial and caudal-lateral regions. Similar discrepancies between the INT and FC maps are apparent for PFC. Also, unlike the FC maps, the INT maps seem asymmetrical in the two hemispheres. Unfortunately, the visualization and stats are too dense to probe what is going on. Deeper analyses and explanation are necessary (please see the next comment for a potential solution).

We thank the reviewers for this comment and agree with the assessment. It is important to note that there are crucial differences between the gradient and INT estimation approaches: (1) the gradient estimation pipeline consists of several steps that significantly reduce noise (e.g., the connectivity matrix was made sparse; applying the cosine similarity kernel to build an affinity matrix; and applying a non-linear dimensionality reduction technique to estimate the gradients); (2) the gradients were estimated bilaterally. With respect to the first point, the INT estimation in comparison was done for each voxel individually and no follow-up noise reduction steps have been applied. Previous studies have smoothed the resulting INT maps, but we have decided not to do so since it could erase important functional details. Hence, while the gradients are the result of several processing steps, INTs are a relatively straightforward measure. As a result, the gradient maps appear smooth while the INT maps may appear more variable from voxel to voxel. With respect to the second point, the symmetry of the gradients (although not always present, e.g., see Figure 4C) could be a result of estimating the gradients bilaterally. Other studies (Haak, Marquand, and Beckmann, 2018; Przeździk et al., 2019) have derived the gradients for each hemisphere individually and no major asymmetries have been reported. Nevertheless, this could be a future avenue for further investigating the relationship between INTs and FC gradients. We have decided to estimate the gradients bilaterally because of some of the issues that are inherent with ultrahigh fields. RF transmit field (or B1) inhomogeneities is one of the major drawbacks of high field MRI, even with our high number of transmit channels. The inhomogeneity is a cause for regions of increase and decreased signal intensity in the data. We were worried that these effects might introduce asymmetries between the two hemispheres. We have reason to believe that this issue is partially being taken care of by estimating the gradients bilaterally, but the INTs are still susceptible to these effects. As a result, we believe that the correspondence between INTs and FC gradients might in fact be underestimated in this study. In future studies, we hope to perform full B1 shimming to counteract these shortcomings, but for safety reasons (energy deposition was only modeled for fixed phases of our coil), we did not have that ability before.

For transparency and better interpretation, we added supplementary plots that show that the overall direction of FC and INT change is similar when the values are averaged according to functional subdivisions of the areas under consideration (see Figure 4 Supplement 1; Figure 5 Supplement 1). Additionally, we have worked on the presentation of the INT and FC gradient maps (see the new versions of Figure 3, Figure 4, Figure 4 Supplement 2). We are also adding the following paragraph to the discussion of the manuscript:

“It is noteworthy that there are crucial methodological differences between the gradient and INT estimation approaches: (1) the gradient estimation pipeline consists of multiple steps to reduce the dimensionality of the data, and consequently, the amount of noise (e.g., the FC matrix was made sparse; applying the cosine similarity kernel to build an affinity matrix; and applying a non-linear dimensionality reduction technique to estimate the gradients); (2) the gradients were estimated bilaterally. In comparison to gradients, where the spatial correlation between the voxels’ time series is just the initial step in the pipeline, the INT measure consists of the temporal autocorrelation of individual voxels. No extra steps have been taken for noise reduction. Hence, while gradients are the result of a complex processing pipeline, INTs are a relatively simple calculation. As a result, there is a higher amount of variability in the INT maps which could potentially lead to the underestimation of the FC-INT relationship. With respect to the second point, the symmetry of the gradients could be a byproduct of estimating the gradients bilaterally. Other studies (e.g., Haak et al., 2018; Przeździk et al., 2019) have derived gradients for each hemisphere individually and no major asymmetries have been reported. We estimated the gradients bilaterally because of an issue that is inherent with experimental ultrahigh fields. Specifically, RF transmit field (or B1) inhomogeneities lead to asymmetric signal intensities. We were worried that these effects might introduce asymmetries between the two hemispheres. We have reason to believe that this issue is partially being taken care of by estimating the gradients bilaterally, but the INTs are still susceptible to these effects. This could also result in the underestimation of the FC-INT relationship since there are visible asymmetries in the INT maps.”

6. Many fine level discrepancies are also observed between striatal INTs and cortico-striatal INTs. In the current figures, it is impossible for readers to know if the variations and discrepancies are within the range expected from noise in the data. The alternative is that some of the variations are meaningful but ignored by the authors to keep the narrative simple. Either way, the authors are encouraged to improve their presentation method and analyses in order to provide adequate information for readers. One possibility is to use old-fashioned scatter plots with error bars, where individual data points are different locations within the striatum and the axes are striatal INTs and cortico-striatal INTs. The authors can choose how they'd like to divide striatum into distinct sub-regions (e.g., a simple 3D grid or functional divisions). Similar plots would be useful for figure 4, and more generally anywhere that the authors report a correlation between INTs and FCs.

We thank the reviewer for this comment and refer to our response to Comment 5. To improve interpretability, we added Figure 5 Supplement 1 and the paragraph from Response 5 (this addition to the Discussion is in response to both Comment 5 and 6). Additionally, we have worked on the presentation of the INT and FC gradient maps (see the new versions of Figure 3, Figure 4, Figure 4 Supplement 2).

7. The authors should provide more detailed information about how the hierarchy shown in the x-axis of Figure 1 was computed, so that this can be replicated by other investigators. Some schematic diagram for cortical hierarchy might be also helpful.

We apologize to the reviewers for not having been more explicit in our description. The hierarchies in Figure 1 are based on the studies that have already been cited – i.e., they are hierarchies derived from intrinsic neural timescales estimated from ephys data (Murray et al., 2014; Fascianelli et al., 2019; Cirillo et al., 2018; Cavanagh et al., 2018; Maisson et al., 2021). We added a more extensive description of how we calculated the values for each individual area. Although it could be useful to have a schematic hierarchy, we have refrained from adding one for several reasons: (1) we estimated INTs for both cortical and subcortical structures; (2) to date, there is no established timescale hierarchy at the whole-cortex level (the data is scarce for subcortical structures); (3) Figure 1 was intended as a demonstration of the INT correspondence between ephys and fMRI.

8. The paper would have more value if the same analyses were carried out in other brain areas, such as the temporal cortex, that are not currently covered. If such results are not included, the authors should at least provide a reason why the paper focuses only on the frontal and parietal regions.

We thank the reviewers for their suggestions and agree that further investigations into the relationship between INT and FC gradients (e.g., in the temporal and occipital lobes, or other subcortical structures) are necessary. However, in this paper, we have decided to limit our analyses to the parietal and frontal lobes, and the striatum for two reasons: (1) in the electrophysiology literature, INTs have been almost exclusively investigated within these areas; (2) the expertise of our team is mainly concerned with these parts of the brain.

To address these suggestions, we added the following paragraph to the Discussion of our manuscript:

“We show that INTs are topographically organized along FC gradients throughout the NHP brain. Although we focused our analysis on the frontal and parietal lobes, and the striatum, we also showed that a whole brain analysis recapitulates the same conclusion: resting-state temporal and spatial dynamics are closely related. Future investigations should examine this relationship in other cortical and subcortical brain areas. A promising avenue would be to examine the FC-INT link throughout the brain across different levels of parcellation (and different types of parcellations) While we show that intrinsic timescales are related to FC, other studies have related them to other anatomical and functional markers (Ito, Hearne, and Cole, 2020; Watanabe, Rees, and Masuda, 2019). Hence, further investigations and careful experimentation are needed to understand (1) how temporal dynamics relate to other functional and anatomical characteristics across the brain; (2) their functional relevance and translational potential.”

References:

Autio, J. A., Zhu, Q., Li, X., Glasser, M. F., Schwiedrzik, C. M., Fair, D. A., . . . Vanduffel, W. (2021). Minimal specifications for non-human primate MRI: Challenges in standardizing and harmonizing data collection. Neuroimage, 236, 118082. doi:10.1016/j.neuroimage.2021.118082

Haak, K. V., Marquand, A. F., and Beckmann, C. F. (2018). Connectopic mapping with resting-state fMRI. Neuroimage, 170, 83-94. doi:10.1016/j.neuroimage.2017.06.075

Ito, T., Hearne, L. J., and Cole, M. W. (2020). A cortical hierarchy of localized and distributed processes revealed via dissociation of task activations, connectivity changes, and intrinsic timescales. Neuroimage, 221, 117141.

Jung, B., Taylor, P. A., Seidlitz, J., Sponheim, C., Perkins, P., Ungerleider, L. G., . . . Messinger, A. (2021). A comprehensive macaque fMRI pipeline and hierarchical atlas. Neuroimage, 235, 117997. doi:https://doi.org/10.1016/j.neuroimage.2021.117997

Lagore, R. L., Moeller, S., Zimmermann, J., DelaBarre, L., Radder, J., Grant, A., . . . Adriany, G. (2021). An 8-dipole transceive and 24-loop receive array for non-human primate head imaging at 10.5 T. NMR in Biomedicine, 34(4), e4472.

Logothetis, N. K., Pauls, J., Augath, M., Trinath, T., and Oeltermann, A. (2001). Neurophysiological investigation of the basis of the fMRI signal. nature, 412(6843), 150-157.

Pelekanos, V., Mok, R. M., Joly, O., Ainsworth, M., Kyriazis, D., Kelly, M. G., . . . Kriegeskorte, N. (2020). Rapid event-related, BOLD fMRI, non-human primates (NHP): choose two out of three. Scientific Reports, 10(1), 1-13.

Przeździk, I., Faber, M., Fernández, G., Beckmann, C. F., and Haak, K. V. (2019). The functional organisation of the hippocampus along its long axis is gradual and predicts recollection. Cortex, 119, 324-335. doi:https://doi.org/10.1016/j.cortex.2019.04.015

Shinn, M., Hu, A., Turner, L., Noble, S., Achard, S., Anticevic, A., . . . Bullmore, E. T. (2021). Spatial and temporal autocorrelation weave human brain networks. bioRxiv.

Ugurbil, K. (2016). What is feasible with imaging human brain function and connectivity using functional magnetic resonance imaging. Philosophical Transactions of the Royal Society B: Biological Sciences, 371(1705), 20150361.

Uğurbil, K. (2018). Imaging at ultrahigh magnetic fields: History, challenges, and solutions. Neuroimage, 168, 7-32.

Uğurbil, K., Auerbach, E., Moeller, S., Grant, A., Wu, X., Van de Moortele, P. F., . . . Radder, J. (2019). Brain imaging with improved acceleration and SNR at 7 Tesla obtained with 64channel receive array. Magnetic Resonance in Medicine, 82(1), 495-509.

Watanabe, T., Rees, G., and Masuda, N. (2019). Atypical intrinsic neural timescale in autism. eLife, 8, e42256. doi:10.7554/eLife.42256

Yacoub, E., Grier, M. D., Auerbach, E. J., Lagore, R. L., Harel, N., Adriany, G., . . . Zimmermann, J. (2020). Ultra-high field (10.5 T) resting state fMRI in the macaque. Neuroimage, 223, 117349. doi:https://doi.org/10.1016/j.neuroimage.2020.117349

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Data Citations

    1. Manea AMG, Zilverstand A, Uǧurbil K, Heilbronner SR, Zimmermann J. 2021. Resting-state functional connectivity gradients and intrinsic neural timescales. figshare. [DOI]

    Supplementary Materials

    Supplementary file 1. The areas were defined based on a cortical and subcortical atlas.

    The areas in Figure 1 were defined based on the Cortical Hierarchy Atlas of the Rhesus Macaque (CHARM) (Jung et al., 2021) and Subcortical Atlas of the Rhesus Macaque (SARM) (Hartig et al., 2021) to closely match the corresponding electrophysiological recording sites. For some of the cortical regions of interest, the CHARM parcellation did not match the recording sites. As a result, the areas were defined according to the respective descriptions (Note: this is the case whenever “Custom” is indicated in the table). Abbreviations: S1/S2 (primary/secondary somatosensory cortex), MT (middle temporal area), LIP (lateral intraparietal area), OFC (orbitofrontal cortex), LPFC (lateral prefrontal cortex), ACC (anterior cingulate cortex), PMd (dorsal premotor cortex), LOFC (lateral OFC), DLPFC (dorso-lateral PFC), VLPFC (ventro-lateral PFC), PFp (polar prefrontal cortex), vmPFC (ventro-medial PFC), sgACC (subgenual ACC), pgACC (pregenual ACC), dACC (dorsal ACC), GPe (external globus pallidus), STN (subthalamic nucleus).

    elife-75540-supp1.docx (14.2KB, docx)
    Supplementary file 2. Areas defined based on a cortical atlas.

    The areas in Figure 2 were defined based on the Cortical Hierarchy Atlas of the Rhesus Macaque (CHARM) (Jung et al., 2021) to match the hierarchies described by Kravitz et al., 2011. Abbreviations: DLPFC (dorso-lateral prefrontal cortex), VIP (ventral intraparietal area), MIP (medial intraparietal area), rIPL (rostral inferior parietal lobule), MT (middle temporal area), MST (medial superior temporal area), LIP (lateral intraparietal area), A46 (area 46), A8A (area 8 A), PMd (dorsal premotor cortex), PMv (ventral premotor cortex).

    elife-75540-supp2.docx (14.4KB, docx)
    Transparent reporting form
    elife-75540-transrepform1.pdf (217.3KB, pdf)

    Data Availability Statement

    The functional connectivity gradient maps and the timescale maps have been uploaded to figshare. Functional connectivity gradients: https://doi.org/10.6084/m9.figshare.19189331 Intrinsic neural timescales: https://doi.org/10.6084/m9.figshare.19197026.

    The following dataset was generated:

    Manea AMG, Zilverstand A, Uǧurbil K, Heilbronner SR, Zimmermann J. 2021. Resting-state functional connectivity gradients and intrinsic neural timescales. figshare.

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES