Functional network mirrored in the prefrontal cortex, caudate nucleus, and thalamus: high-resolution functional imaging and structural connectivity

Abstract

Despite myriads of studies on a parallel organization of cortico-striatal-thalamo-cortical loops, direct evidence of this has been lacking for the healthy human brain. Here, we scrutinize the functional specificity of the cortico-subcortical loops depending on varying levels of cognitive hierarchy as well as their structural connectivity with high-resolution fMRI and diffusion-weighted MRI (dMRI) at 7 tesla. Three levels of cognitive hierarchy were implemented in two domains: second language and nonlanguage. In fMRI, for the higher level, activations were found in the ventroanterior portion of the prefrontal cortex (PFC), the head of the caudate nucleus (CN), and the ventral anterior nucleus (VA) in the thalamus. Conversely, for the lower level, activations were located in the posterior region of the PFC, the body of the CN, and the medial dorsal nucleus (MD) in the thalamus. This gradient pattern of activations was furthermore shown to be tenable by the parallel connectivity in dMRI tractography connecting the anterior regions of the PFC with the head of the CN and the VA in the thalamus, whereas the posterior activations of the PFC were linked to the body of the CN and the MD in the thalamus. This is the first human in vivo study combining fMRI and dMRI showing that the functional specificity is mirrored within the cortico-subcortical loop substantiated by parallel networks.

Keywords: cognitive hierarchy; functional specificity of the cortico-subcortical loop; high-resolution fMRI and dMRI.

Figure 1.

Schematics of the three levels of cognitive hierarchies in L2 and NL domains. A, A sentence example from the miniature version of Korean is provided. We colored letters in the sentence in the same way as in a sequence in NL to control the variability of visual input across the domains and, thus, colors have no meanings here. Participants judged the grammaticality of each of the six phrases and responded via button press. In the contextual condition, participants judged the grammar within a phrase itself without referring to other phrases. For example, the first contextual condition ( in ➁) is a correct trial where each stimulus is ordered correctly with a noun (: sister) before subject particle (). The second contextual condition ( in ➂) is incorrect; the order should be noun (: flower)–number (: one)–numeral classifier for flower ()–object particle (), whereas in this phrase, number, and numeral classifier are switched. In the episodic condition, participants responded to stimulus with respect to the discrete preceding stimulus. For example, the verb tense of a phrase ( in ➄) is indicated as being in the past by a prefinal ending for past tense () because of the prior temporal adverb (: yesterday in ➃). Additionally, each grammatical feature is ordered correctly in ➄: verb (: buy)–prefinal ending for past tense ()–final ending ()–complementizer ()–object particle (). For the branching condition, processing one stimulus is maintained in a pending state while another stimulus is being performed, and reactivated upon completion of the ongoing one. In this example, the processing of the subject in the main clause (: father in ➀) along with a subject particle () is suspended by a center embedded sentence denoted by brackets and then reactivated, influencing the verb phrase (➅) at the end of the sentence. Since the subject is father, prefinal ending for subject honorification () should be included in the verb phrase which is omitted in ➅ as verb (: like)–prefinal ending for past tense ()–final ending (). Therefore, this is an incorrect trial. The brackets in A are displayed only in this example to denote an embedded sentence but not in the learning session or in the fMRI session. Direct translation: Father liked that sister bought one flower yesterday. B, A sequence example in NL is provided. Participants were asked to judge the Fero sequence rule of each of the six chunks and to respond via button press. In the contextual condition, the chunks begin with symbols of F3 ( in in ➃, in in ➄) so that the Nero color rule of the chunks is applied to only the current ones without referring to preceding or upcoming chunks. Therefore, with the first Nero symbol in (➃), the color sequence displays blue (B)–yellow (Y)–red (R)–green (G), which is a correct trial. Since there are only three symbols in ➃, the last color (G) is omitted. Similarly, the first Nero symbol in (➄) is such that the symbols in the chunk should be colored with G–R–Y–B. However, this chunk is incorrectly colored with B–Y–R–G. In the episodic condition, the chunk starts with a symbol of F1 ( in in ➁). Therefore, the Nero color rule of the chunk is applicable to the current and the next chunk, which also leads to the color sequence G–R–Y–B in the third chunk ( in ➂). The last symbol ( in ➂) is colored with G because of a repetitive rule of color sequence when a chunk consists of more than four symbols across all the conditions. These two chunks (➁, ➂) are correct trials. In the branching condition, the first Fero symbol of the first chunk ( in in ➀) belongs to F2 so that the Nero color rule of the chunk (B–Y–R–G) also applies to a subsequent chunk starting with F2 (i.e., ➅ in this example). Therefore, in ➅ is correctly colored with B–Y–R following the color rule of the preceding chunk (➀).

Figure 2.

Behavioral results from fMRI studies of L2 and NL. Mean percentage accuracy (A) and RTs (B) are described in the contextual, episodic, and branching conditions in L2 and NL. Error bars denote SEM. *p < 0.05.

Figure 3.

Posterior-to-anterior pattern of activations depending on the levels of cognitive hierarchies across L2 and NL in the PFC. Brain activations elicited by the episodic and branching conditions across L2 and NL were rendered on a canonical brain provided with MRIcron software (http://www.mccauslandcenter.sc.edu/mricro/mricron/). Activations from the contextual condition in the left hemisphere and the branching condition in the right hemisphere are not displayed here (see Table 1 for their coordinates). Bar graphs represent the mean percentage signal change within the sphere-shaped regions (radius 2 mm) centered upon the peak voxels for each condition. Error bars denote SEM. n = 19, *p < 0.05. BRAN, Branching condition; EPIS, episodic condition.

Figure 4.

Posterior-to-anterior pattern of activations depending on the levels of cognitive hierarchies across L2 and NL in the CN. Brain activations elicited by the branching and episodic conditions across L2 and NL were rendered on a sagittal view of the CN. Bar graphs represent the mean percentage signal change within the sphere-shaped regions (radius 2 mm) centered upon the peak voxels for each condition. Left, The activations related to the branching condition were observed in the anterior region of the CN (head of the CN) across the domains. Right, The activations involved in the episodic condition were observed in the posterior region of the CN (body of the CN). n = 19, *p < 0.05. Error bars denote SEM. BRAN, Branching condition; EPIS, episodic condition; LH, left hemisphere.

Figure 5.

Posterior-to-anterior pattern of activations depending on the levels of cognitive hierarchies across L2 and NL in the thalamus. Brain activations in the thalamus elicited by the branching condition and the episodic condition across L2 and NL were rendered on an axial view of the brain. Activations related to the contextual condition in the bilateral hemisphere and the episodic-related activation in the right hemisphere are not displayed here (see Table 1 for their coordinates). Bar graphs represent the mean percentage signal change within the sphere-shaped regions (radius 2 mm) centered upon the peak voxels for each condition. Left, The activations related to the branching condition were depicted in the anterior region (VA). Additionally, branching-related activations were also observed in the posterior region (MD). The overlapping regions between the domains were colored with red and blue obliquely. Right, The activations involved in the episodic condition were depicted in the more posterior region (MD). The overlapping areas between the two domains were colored with orange and green obliquely. n = 19, *p < 0.05. Error bars denote SEM. BRAN, Branching condition; EPIS, episodic condition; LH, left hemisphere.

Figure 6.

Parallel cortico-striatal pathways. A, Parallel cortico-striatal pathways between the PFC and the CN for different levels of cognitive hierarchies. Left, 3D rendering of the group average of probabilistic tractography between the seeds in the PFC and the CN for the branching and episodic conditions across L2 and NL. Brain activations used as seed regions are depicted with small spheres for the PFC and circles filled with oblique lines for the CN. The surface of the connecting tract is shown in the corresponding colors. The tractography follows parallel and segregated pathways for the different conditions. Inset, Quantitative analysis of the connection strength (CS) for all participants. The bars indicate the median CS for all conditions computed as the relative number of pathways connecting each pair of seed regions. Error bars show quartiles. Right, Group overlap of the tractography visitation map of all participants for the branching conditions (two rows at the top displayed in axial slices) and the episodic conditions (two rows at the bottom displayed in coronal slices) with slice coordinates in MNI space. B, Parallel cortico-striatal pathways between the PFC and the thalamus for different levels of cognitive hierarchies. Left, 3D rendering of the group average of probabilistic tractography between the seeds in the PFC and the thalamus for the branching and episodic conditions across L2 and NL. Brain activations used as seed regions are depicted with small spheres for the PFC and circles filled with oblique lines for the thalamus. The surface of the connecting tract is shown in the corresponding colors. The tractography follows parallel and segregated pathways for the different conditions. Inset, Quantitative analysis of the CS for all participants. The bars indicate the median CS for all conditions computed as the relative number of pathways connecting each pair of seed regions. Error bars show quartiles. Right, Group overlap of the tractography visitation map of all participants and conditions in axial slices with slice coordinates in MNI space (n = 9). BRAN, Branching condition; EPIS, episodic condition.

Figure 7.

Comparison of peak activations within the prefrontal cortex from the present study and those from Badre and D'Esposito (2007) and Koechlin et al. (1999, 2003). A schematic display of the approximate distribution of the activation foci within the prefrontal cortex is provided with spheres (8 mm radius) centered on the coordinates of peak activations in L2 from the present study (red), Badre and D'Esposito (2007; green), and Koechlin et al. (1999, ; blue). The MNI-coordinates of our study were converted into the Talairach ones conforming to the other studies. The experimental conditions and the coordinates of the peak voxels in each study are described as follows: 1, “branching control” (−36, 57, 9); 2, “context level” (−36, 50, 6); 3, branching condition (−43, 49, −4); 4, “episodic control” (−40, 32, 20); 5, episodic condition (−45, 27, 24); 6, “dimension level” (−50, 26, 21); 7, “feature level” (−37, 11, 31); 8, “contextual control” (−44, 8, 20); 9, contextual condition (−32, −2, 43); 10, “response level” (−30, −7, 63). Note that context level from Badre et al. (2007) is different from contextual level in Koechlin et al. (1999, 2003) and the contextual condition in the present study. Adapted from Badre et al. (2007).