In some primates such as the macaque there is evidence for the existence of at least two networks of interconnected frontal cortical areas centered on the lateral orbitofrontal cortex (LOFC) and medial frontal and anterior cingulate cortex (ACC) (Ongur and Price, 2000
). This difference in anatomical connectivity may underlie differences in function between ACC and LOFC (Rushworth et al., 2007a
). Some of the first evidence for the existence of the two circuits in the macaque came from a comparison of the relative positions of zones within the MD thalamus projecting to different frontal regions (Ray and Price, 1993
). The same study and others (Barbas et al., 1991; Giguere and Goldman-Rakic, 1988; Goldman-Rakic and Porrino, 1985
) suggested the existence of a third region of prefrontal cortex, lateral prefrontal cortex (including ventrolateral prefrontal cortex, VLPFC, and dorsal prefrontal cortex, DPFC), with a distinct MD connection pattern; lateral, parvicellular parts of MD (MDpc) project most strongly to lateral prefrontal cortex. Ray and Price (1993)
also reported the existence of a fourth region within the MD thalamus that projected mainly to just a limited dorsolateral part of the lateral prefrontal cortex (DLPFC), centered on area 46. The aim of the present study was to test for evidence of topographic differences in human MD–prefrontal connections. Any evidence for such a separation would be consistent with the existence of differences in anatomical connectivity between prefrontal systems in the human brain resembling those found in other primates.
Previous studies have suggested that DWI–DT can be used to distinguish the connections of different thalamic nuclei (Behrens et al., 2003a; Johansen-Berg et al., 2005
) but whether the technique is sensitive to differences in connectivity across sub-regions of a single thalamic nucleus has not been investigated. The second and related aim of the present study was, therefore, to test whether DWI–DT could be used to provide evidence for regional variations in connectivity within a single thalamic nucleus — MD.
Using DWI–DT we found clear evidence for topographic difference in connectivity with four frontal regions within the MD nucleus in a macaque. Regional differences in connectivity resembled those previously reported in tracer injection and tract tracing studies. LOFC had a greater probability of interconnection with a relatively medially situated and dorso-ventrally oriented strip of MD thalamus — a pattern that resembled that seen in the macaque (Barbas et al., 1991; Giguere and Goldman-Rakic, 1988; Goldman-Rakic and Porrino, 1985
). VLPFC/DPFC had a high probability of interconnection with more lateral MD ( and ) while a DLPFC region in the principal sulcus that was probably mainly constituted of area 46 had a high probability of connection with a dorsal MD region situated between the first two projection zones. The overall pattern resembled that previously reported in tract tracing investigations in the macaque (Ray and Price, 1993
). A fourth region, ACC, was most strongly interconnected with a relatively caudal and lateral part of MD ( and ). A similar pattern of connectivity between MD and ACC has been reported (Giguere and Goldman-Rakic, 1988
In humans, DWI–DT produced evidence of MD connections with LOFC that resembled those established in macaque in tract tracing studies (Barbas et al., 1991; Giguere and Goldman-Rakic, 1988; Goldman-Rakic and Porrino, 1985
) and using DWI–DT in a macaque brain in the first part of the current study. As in the macaque there was evidence of connections between LOFC and the medial aspect of MD in the human subjects. A previous DWI–DT study also highlighted other similarities between the estimated connection pattern for human LOFC and the pattern established in the macaque; Croxson et al. (2005)
reported that the uncinate fascicle had a higher probability of connection with the LOFC than with medial frontal cortex/ACC, or lateral prefrontal cortex in both macaque and human subjects. In the macaque, temporal and perirhinal cortical areas are strongly interconnected with LOFC (Carmichael and Price, 1995a; Kondo et al., 2003, 2005; Lavenex et al., 2002
) via the uncinate fascicle (Ungerleider et al., 1989
). These connections may underlie a role for the lateral OFC in assigning reinforcement significance to objects represented by networks of neurons in temporal and perirhinal regions (Murray et al., 2007; Rushworth et al., 2007a
). As in the macaque, evidence was found for ACC and LOFC connections with distinct MD zones in the human brain. The separation is consistent with the existence in the human brain of distinct anatomical circuits centered on the ACC and LOFC resembling those seen in other mammal species including rats and macaques (Ongur and Price, 2000; Ray and Price, 1993
). In rats and macaques the ACC and LOFC circuits are implicated in the control of distinct aspects of reward-guided behavior (Behrens et al., 2007b; Rushworth et al., 2007a
). While the ACC is implicated in reward-guided selection of action, the detection of reward prediction errors, and the generation of novel action strategies, the OFC is concerned with the representation of reward outcomes and with the process of associating stimulus and object representations with reward. The ACC connection region in humans was found in a relatively caudo-dorsolateral part of MD as was also the case in the macaque DWI–DT data and as has been reported previously in the macaque (Giguere and Goldman-Rakic, 1988; McFarland and Haber, 2002
). Previous DWI–DT studies have highlighted other similarities between the likely connections of human ACC/medial frontal cortex and the established connections of macaque ACC/medial frontal cortex. For example, DWI–DT studies have suggested that some parts of human ACC/medial frontal cortex are connected with motor and premotor regions (Beckmann et al., 2009
) as is the case in the macaque (Barbas and Pandya, 1987; Hatanaka et al., 2003; Luppino et al., 2003; Rizzolatti et al., 1998; Takada et al., 2004
). Such connections could explain the greater importance of ACC/medial frontal cortex, as opposed to LOFC, in learning the reinforcement significance of actions in several species (Ostlund and Balleine, 2007; Rudebeck et al., 2008; Rushworth et al., 2007a
). Other connections of ACC/medial frontal cortex, with areas such as the hypothalamus and amygdala, may underlie its importance in social cognition in rats, monkeys, and humans (Beckmann et al., 2009; Behrens et al., 2009, 2008; Rudebeck et al., 2006, 2007
Evidence for connections with DPFC and VLPFC, prefrontal regions involved with memory retrieval, conditional learning, language, and action selection (Murray et al., 2000; Passingham et al., 2000; Petrides, 2005
), was found in lateral MD, again matching predictions from macaque. The MD region with a high probability of connection with the DLPFC in the middle frontal gyrus, in between the VLPFC and DPFC, was situated between the MD regions interconnected with LOFC on the one hand and VLPFC and DPFC on the other hand. This pattern is reminiscent of the DLPFC connection pattern in the macaque seen in the present study and in previous reports (Ray and Price, 1993
). Again there is increasing evidence for important similarities between the connection patterns of lateral frontal regions and temporal and parietal cortical areas via the extreme capsule and branches of the superior longitudinal fascicle in macaques and humans (Croxson et al., 2005; Frey et al., 2008; Rilling et al., 2008
In our approach, we use macroscopic anatomy to define cortical targets for DWI–DT, while in macaque, projections from MD have been established with respect to cytoarchitectonically defined regions (Ray and Price, 1993
). In our study on living human brains, cytoarchitectonic data on an individual level are not available. There is an approximate relationship between macroscopic features such as sulci and gyri and cytoarchitectonic anatomy but it is not always exact (Amunts et al., 1999; Fischl et al., 2008
). Current developments in MRI may eventually overcome the problem of defining individual cytoarchitectonic borders in the human brain (Walters et al., 2007
The choice of registration method to account for individual differences in brain size or anatomy might influence the accuracy of the registration of brains to one another (Klein et al., 2009
), and hence has some impact on the degree to which parcellation appears similar across subjects. While a non-linear approach can be expected to offer a benefit in registration accuracy, our results demonstrate that linear registration is capable of preserving individual connectivity characteristics of MD sub-portions.
In comparison with tracer injection studies that can be carried out only in animals, DWI–DT suffers from several limitations (Johansen-Berg and Behrens, 2006
). DWI–DT has a lower spatial resolution and it cannot identify the polarity of connections. In the case of MD–prefrontal connections it is known from tracer injection studies that regional variation in connectivity patterns is most apparent when the thalamocortical, rather than corticothalamic connections, are considered (McFarland and Haber, 2002
). Haber and McFarland reported that while there are corticothalamic projections that reciprocate thalamocortical projections there are, in addition, corticothalamic projections to more extensive regions of thalamus. Because the DWI–DT technique used here is sensitive to relative differences in connection probability, however, it is still able to identify regional variations in MD–prefrontal connectivity.
Importantly, DWI–DT can only be used to establish the probability
of interconnection between areas rather than providing definitive evidence for the presence of fiber terminals and synapses. A major potential source of bias in tractography is any of a number of strong fiber systems that connections between MD and the frontal lobe would need to traverse before reaching their target. Callosal fibers, the cingulum bundle and parts of the superior longitudinal fascicle could disrupt DWI–DT investigation of connections between MD and the frontal lobe, particularly medial areas such as the ACC (Schmahmann et al., 2007
). Nevertheless, the similarity between the DWI–DT and tract tracing results suggests that the current algorithm, which is able to identify multiple orientations of fibers in any voxel (Behrens et al., 2007a
), was able to identify and follow the major pathways between MD and frontal cortex. It is, however, important to remember that for the human data acquired here (with 60 diffusion encoding directions), it is not possible to resolve more than two different fiber populations in any voxel (Behrens et al., 2007a
). Despite these disadvantages DWI–DT is important because it makes it possible to investigate whether neuroanatomical connection patterns found in experimental animals, such as monkeys, are also likely to be present in humans.
In summary, DWI–DT confirms that the principal features of connectivity between MD and anterior cingulate and prefrontal cortex in humans are comparable to those found in macaque using DT–DWI or tract tracing techniques. There is a clear separation between MD regions with a higher probability of interconnection with ACC or LOFC consistent with the possibility that there are distinct anatomical circuits centered on ACC and LOFC concerned with distinct aspects of reward-guided behavior. Importantly this is the first study to provide insight into the topography of projections of an individual thalamic nucleus, and demonstrates the value of DWI–DT as a technique to study human anatomy. While tracer techniques remain the gold standard of establishing connectivity between brain structures, DWI–DT's non-invasive nature enables its use for whole brain studies in human subjects. DWI–DT can be used to examine the degree to which anatomical interconnectivity patterns established in animal models are likely to pertain to the human brain (Behrens et al., 2003a; Croxson et al., 2005; Ramnani et al., 2006; Rushworth et al., 2006; Tomassini et al., 2007