Parietal cortical areas have generally been considered as part of the dorsal stream and, as such, only indirectly connected with inferotemporal cortex. In this report we demonstrate, by using the anterograde tracer BDA, that much of the inferior parietal lobule (IPL) has direct connections to anterior-ventral TE (TEav) around the anterior middle temporal sulcus (amts). Connections from area PG terminate in layers 1 and 5 as well as 4; and those from area PF, target layer 6 of TEav, with a small secondary focus in layer 4 of anterior-dorsal TE. Connections from areas PG and PF are relatively sparse; but those from the mid-IPL region (approximately area PFG), which terminate in layer 4, are light to moderate. In confirmation of these results, injections of retrograde tracers in TEav produce labeled neurons in the IPL. These are most numerous in layer 3 at the border of areas PG and PFG, but also occur in layer 5/6. These laminar patterns are more complex than the classical ‘feedforward’ or ‘feedback’ patterns associated with early sensory areas. Branched collaterals are common; and three of seven reconstructed axons branched to both TEav and to the lateral bank of the occipito-temporal sulcus, itself a major source of inputs to TEav. The existence of connections from the IPL preferentially to TEav and the amts provides another example where direct ‘bypass’ connec- tions operate in parallel with multiple indirect routes. It provides further evidence for the differential connectivity of subdivisions within anterior TE and is consistent with recent evidence from functional magnetic resonance imaging studies that the region around the amts may be part of a network involved in three- dimensional shape, which is distributed across both ‘what’ and ‘where’ processing streams.
Area TE in macaque monkeys receives dense direct connections from area TEO (Desimone et al., 1980; Webster et al., 1991; Distler et al., 1993; Saleem et al., 1993) and is often considered the final stage in a ventral cortical processing stream (Gross, 1994; Logothetis and Sheinberg, 1996). It is extensively documented as involved in object recognition (Gross, 1994; Logothetis and Sheinberg, 1996; Tanaka, 1996; Rolls, 2000). The anterior part of TE (TEa) has at least two subdivisions, each with differential connectivity (Desimone et al., 1980; Iwai et al., 1987; Martin-Elkins and Horel, 1992; Yukie et al., 1992; Saleem and Tanaka, 1996; Cheng et al., 1997; Saleem et al., 2000). In particular, the anterior-ventral (TEav), but not the anterior- dorsal (TEad), subdivision is preferentially connected (i) with posterior visual areas TEp and TEO, in ventral (not dorsal) locations (Martin-Elkins and Horel, 1992); (ii) with the lower (not upper) bank of the anterior superior temporal sulcus (Saleem et al., 2000); and (iii) with CA1 of the hippocampus (Martin-Elkins and Horel, 1992; Saleem and Hashikawa, 1998; Rockland and Van Hoesen, 1999; Yukie, 2000). TEav also projects more strongly to the perirhinal cortex (Saleem et al., 2000). It has been inferred that TEav, in contrast with TEad, may be more closely associated with visuospatial processes (Saleem et al., 2000). Recent functional magnetic resonance imaging (fMRI) studies (Sereno et al., 2002) provide further evidence that the region around the anterior middle temporal sulcus (amts; mainly TEav) may have a distinct specialization, and that this is related to processing of three-dimensional shapes.
A possible substrate for visuospatial interactions would be the known connections from the superior temporal sulcus (STS) or intraparietal sulcus to subdivisions of TE. However, there have also been reports of direct connections from the posterior part of the inferior partial lobule (IPL) to TEav (Iwai et al., 1992; Martin-Elkins and Horel, 1992). These studies both used large injections of retrograde tracers in TEav. In this study, we use high resolution anterograde tracers to further investigate the density and topography of connections from various parts of the IPL to TEa. We were also interested in how these direct connections might converge with other potentially important inputs to the amts and to the STS, a region heavily inter- connected with the anterior part of TE. Single axon analysis was included in order to compare terminations with classical ‘feedforward’ and ‘feedback’ patterns.
Materials and Methods
Surgery and Tracer Injection
Ten adult rhesus monkeys (Macaca mulatta) of both sexes were used in this study. Seven monkeys received injections of biotinylated dextran amine (BDA) in different portions of the inferior parietal lobe (Fig. 1A) and one monkey (R51) in area TEO (Fig. 1D). Two other animals had large pressure injections of retrograde tracer in anterior TE. In one animal, Fast Blue (FB) and cholera toxin subunit B (CTB) were injected in TEad. In another, FB and Fluoro Ruby (FR) were injected in the amts, in its lower (FB) and upper (FR) banks and margin (Fig. 1D). Surgery was carried out under sterile conditions after the animals were deeply anesthetized with barbiturate anesthesia (25 mg/kg Nembutal i.v. after a tranquilizing dose of 11 mg/kg ketamine i.m.). Procedures were in accordance with institutional guidelines, as specified in approved Animal Care and Use Forms (University of Iowa) and protocols reviewed by the Experimental Animal Committee (RIKEN Institute, Japan). Cortical areas of interest were localized by direct visualization, subsequent to craniotomy and durotomy, in relation to sulcal landmarks (i.e. the intraparietal and superior temporal sulci). Injections were made by pressure for BDA [10% in 0.0125 M phosphate-buffered saline (PBS), 0.5–2.0 μl per injection; Molecular Probes, Eugene, OR], FB (4% in 0.1M PBS, 0.5–1.0 ml per injection; Sigma, St Louis, MO), FR (10% in 0.1M PBS, 0.5–1.5 ml per injec- tion; Molecular Probes) and CTB (1% in 0.1M PBS, 0.5 ml per injection; List Biological Laboratories, Campbell, CA).
Three additional cases (T3, T4 and T5) were available with iontophoretic injections of PHA-L in ventral areas TEO and TE (Fig. 1D). These had been used in previous studies (Wellman and Rockland, 1997; Rockland and Van Hoesen, 1999) and detailed methods are given elsewhere.
Animals were allowed to recover and survived 18–29 days after injec- tions. They were then re-anesthetized, given an overdose of Nembutal (75 mg/kg) and perfused transcardially, in sequence, with saline, 4% paraformaldehyde and chilled 0.1 M phosphate buffer (PB) with 10, 20, and 30% sucrose. Brains were cut serially in the coronal plane by frozen microtomy (at 50 mm thickness) and processed histologically for BDA, as described previously (Ding et al., 2000). Tissue was reacted for 20–24 h in avidin–biotin complex (ABC Elite kits; Vector Laboratories, Burlingame, CA) at room temperature (one drop of reagent per 7 ml of 0.1 M PBS). In the final step, BDA was demonstrated by 3,3′-diamino- benzidine tetrahydrochloride (DAB) histochemistry with the addition of 0.5% nickel-ammonium sulfate. Processing for CTB was carried out using primary antibody, anti-CTB (1:6000 in 0.1 M PBS, with addition of 0.5 % Triton-X and 2% normal goat serum, ‘PBXG’), rabbit biotinylated anti-goat secondary antibody (1:200 in 0.1 M PBXG) and ABC complex, as described for BDA. FR was reacted following similar steps after reacting with anti-FR primary antibody (in rabbit, 1:6000; Molecular Probes). For fluorescent tracers, tissue was collected in 0.1 M PB and mounted as the brain was being sectioned.
Areas in the inferotemporal cortex (TE) were identified by correlation with surface landmarks [i.e. the amts, pmts, occipito-temporal sulcus (OTS) and rhinal sulcus] and by architectonic analysis of selected histological sections stained for cell bodies (Bonin and Bailey, 1947; Yukie et al., 1990; Suzuki and Amaral, 1994; Saleem and Tanaka, 1996). Several different systems have been proposed for subdividing inferotemporal cortex (Tamura and Tanaka, 2001; Lavenex et al., 2002). Our classi- fication closely follows the nomenclature proposed by Yukie and his colleagues (Yukie et al., 1990) and adopted by Saleem and Tanaka (Saleem and Tanaka, 1996); see also Suzuki and Amaral (Suzuki and Amaral, 1994). In this classification, TE is divided into four subregions in relation to the amts: TEad (anterior-dorsal), TEav (anterior-ventral), TEpd (posterior-dorsal) and TEpv (posterior-ventral). In defining the border between TEav and the perirhinal cortex (area 36), we followed the criteria of Saleem and Tanaka (Saleem and Tanaka, 1996); see also Suzuki and Amaral (Suzuki and Amaral, 1994). Area TEav is identified by a clear separation between layers 5 and 6. Area 36 is identified by the differentiation of layer 3 into 3A and 3B, and by the presence of densely stained large pyramidal cells in layer 5. The precise border between perirhinal cortex and TEav can be difficult to delineate, but the location of the projection sites relative to the amts and the rhinal sulcus further aided in area identification (Fig. 2). The border between TEav and TEad is placed at the upper lip of amts in accordance with the criteria for distinguishing between TE1 and TE2 (Seltzer and Pandya, 1978), as adopted by Saleem and Tanaka (Saleem and Tanaka, 1996); that is, layer 5 is less cell-dense in TEad than TEav.
The inferior parietal lobule (IPL) has also been subdivided in several different ways (Brodmann, 1909; Economo, 1929; Pandya and Seltzer, 1982; Lewis and Van Essen, 2000; Cavada, 2001). Our results were most consistent with the map of Bonin and Bailey (Bonin and Bailey, 1947), as modified by Pandya and Seltzer (Pandya and Seltzer, 1982). This map designates the most anterior sector of the IPL (area 7b of Brodmann) as PF and the most posterior sector as PG (area 7a of Brodmann). A transitional subdivision, called PFG, is recognized between areas PF and PG. Within this framework, injections P1 and P2 can be considered as localized to area PF; injection P3, as at the border of PF and PG (at least partly in PFG); injections P5 and P6, as in area PG; and injections P4 and P7, as at least partly in PG (P7 may extend into area Opt and P4 had slight involvement of the adjoining lateral sulcus in its most posterior portion). None of the injections extended into the intraparietal sulcus (IPS).
Analysis focused on areas TEad and TEav. Other areas, which receive parietal cortical connections, were inspected mainly for consistency with previous results (Desimone et al., 1980; Cavada and Goldman-Rakic, 1989; Anderson et al., 1990; Martin-Elkins and Horel, 1992; Yukie et al., 1992; Cusick et al., 1995; Seltzer et al., 1996). Global projection foci in TEa were scanned and plotted onto enlarged camera lucida drawings of partial section outlines at 0.4 mm intervals. The locations of labeled terminals were projected onto contour lines of each section and then transposed onto a two-dimensional unfolded map, from the lower bank of STS to the lateral lip of the rhinal sulcus. The inferior horn of the lateral ventricle served as the most useful reference point in making the alignment. Results are presented in the ‘flat map’ format (Fig. 2), as well as schematically on representative coronal section outlines (Figs 1B and 2). In order to assess the density of foci, bouton counts were carried out from the densest case (P3), at 1000×, within a standardized 50 × 50 × 50 μm ‘counting box’. One section was selected from P3 through the densest portion of the projection foci in TEav, STS and OTS; and, for comparison, similar counts were performed of the OTS projections in P4, which appeared to be particularly dense. Three fields from each section were counted. As a separate indicator of projection density, the number of retrogradely filled neurons was counted, in the IPL and in the STS (Fig. 1D), in a 500 × 500 × 50 μm region.
Serial section reconstruction of individual axons was accomplished with the aid of a camera lucida microscope attachment as described previously (Ding et al., 2000). Photomicrographs were taken with an Olympus DP50 digital camera mounted on an Olympus BX60 microscope. Images were saved in TIFF format and imported into Adobe Photoshop 5.5. Image brightness, contrast and color were adjusted if necessary to reproduce the original histological data and final prints were made with a Pictrography 3500 printer.
Global Pattern of Labeled Terminals
Of the seven brains with injections in the IPL, all except two, near the IPS in its posterior portion, resulted in projections to TEa (Fig. 1B). Four injection sites (cases P2, 3, 4 and 7) produced terminations in TEav, along the posterior two-thirds of the amts, in its lower bank or adjoining cortex. Terminations were sparse except for those originating from the mid-portion, around the border of areas PG and PFG. These, of light to moderate density, formed foci that measured ~0.70 mm DV × 2.4 mm AP (in P3) and 0.40 mm DV × 1.2 mm AP (in P2). To provide some standard of density, bouton counts were made of terminations in case P3 from TEav and from denser foci in the STS and OTS. Mean numbers were, respectively, 55, 93 and 73 boutons per 50 μm3 counting box (see Materials and Methods). The densest projection, to the OTS in case P4, had 165 boutons per 50 mm3. The overall extent of the projections in the STS and OTS is larger (3.0–5.0 mm) than for projections in TEav.
A fifth injection (P1), in area PF, resulted in terminations in the upper bank of the amts, in TEav, and also produced a weak, secondary focus (~0.30 mm DV × 0.25 mm AP) in what appeared to be TEad just dorsal to the amts. There was some indication of a topographic mapping of the projection foci, such that anterior-to-posterior portions of the IPL had an inverse, dorsal-to-ventral mapping in relation to the amts (Fig. 2). One posterior injection (P7), however, projected diffusely along the lateral bank, fundus and medial bank of the amts.
The laminar termination patterns in TEav from these five cases were complex. The densest projections, from the border of areas PG and PFG (case 3), were mainly to layer 4 (Fig. 3). From more posterior regions (PG), sparse terminations occurred in layers 1 (Fig. 4A–C) and 5, as well as layer 4. From area PF, terminations were concentrated in layer 6 in TEav (Fig. 4D,E), but in layer 4 in TEad. Sporadic retrogradely labeled neurons occurred in area TE in some cases (particularly in P2, P3 and P4).
In order to confirm the BDA results, injections of the retrograde tracers FB, FR, and CTB were made in portions of TEa in two monkeys (Fig. 1D). An injection of FB, mainly in the lower lip of the amts (TEav) and ventrally adjoining cortex, resulted in labeled neurons in the mid-portion of the IPL, in area PG, possibly extending into area PFG (Figs 1C and 5A,B). The absence of labeled neurons in area PF may be explained by the location of the FB injection, which did not involve the upper bank of the amts. Dorsally in the IPL, labeled neurons were mainly in layer 3, but more ventrally and in its mid-portion, neurons were bistratified in layers 3 and 5/6. In the IPL, a total of 677 labeled neurons were counted through 43 sections, of which 615 were in the supra- and 62 in the infragranular layers. Per 500 × 500 μm field in the densest region in area PFG, 11 neurons were counted in the IPL. (Every section, collected in a 1-in-3 series, for an interval spacing of 150 mm, was scanned. The field of labeled neurons extended for 5.4 mm AP.) A denser field in the depth of the STS, by comparison, had 59 neurons in the same sized field.
In accordance with Martin-Elkins and Horel (Martin-Elkins and Horel, 1992), dense retrograde labeling occurred in ventral TEO around the OTS (mainly in layers 3 and 6), in perirhinal cortex (in layers 3, 6 and, sparsely, layer 5) and in the STS (in layers 3, 5 and 6; Fig. 5C,D). After injections in TEad (FR, FB and CTB), no labeled neurons were observed in the IPL, consistent with our anterograde tracer experiments and with previous investigations.
To further investigate the potential pathways between the IPL and temporal areas, several other sets of projections were analyzed. One was from the seven parietal injection sites to the STS. We confirmed that the upper bank of the STS, which is interconnected with TEad (Saleem and Tanaka, 1996), receives dense connections, mainly in layer 4, from all the injection sites. In our material, the lower bank of the STS, which is inter- connected with TEav (Saleem and Tanaka, 1996), also receives rather dense connections, mainly in layer 4, from cases P1, P4, P6 and P7 and less dense connections from cases P2, P3 and P5 (Fig. 6). With our tracers, these appeared denser than in previous studies (Cavada and Goldman-Rakic, 1989; Cusick et al., 1995; Seltzer et al., 1996). Thus, both TEad and TEav may receive indirect IPL connections through the STS.
Another indirect pathway from the IPL to TE is through ventral occipito-temporal cortex (TEO and TEp). Projections from all seven IPL injection sites were in fact verified to these regions, especially in the lateral bank of the OTS, but extending into the medial bank at anterior levels (Fig. 7). Laminar termination was variable, depending on the AP level of the OTS. Posteriorly, terminations were mainly directed to layers 1, 4, 5 and 3.
Finally, we analyzed four cases with injections in different parts of area TEO or TEp (Fig. 1D). Again as previously reported (Desimone et al., 1980; Martin-Elkins and Horel, 1992; Yukie et al., 1992), we found that projections from more ventral parts of TEO and TEp (n = 3) were selectively directed to TEav (mainly in layer 4; Fig. 8). The densest projections (from T4) formed a focus measuring ~0.5 mm DV × 1.5 mm AP. These were at about the same AP level as the terminations from the IPL (i.e. posterior two-thirds of amts), but were somewhat denser (cf. Figs 3 and 8). The lack of projections to TEad from case R51 (with a more lateral injection in TEO) may reflect the relatively small size and location of the injection site. Although near the inferior occipital sulcus, the injection was ventral to several other injection sites reported in the literature (Webster et al., 1991; Saleem et al., 1993).
In three cases (T3, T4 and R51, but not T5), we also observed moderate projections to the lower bank of STS, mainly in layer 4. As these foci appeared to overlap with projections from the IPL at the border of the middle and anterior portions of the STS, this may constitute an additional convergence node for connections from temporal and parietal regions.
Single Axon Reconstruction
In order to investigate the parietal-to-temporal projections at a finer level, seven axons were analyzed by serial section reconstruction. Five of the seven were relatively simple and terminated in the supragranular layers or in layer 1 exclusively. One of these, axon 4-1 (Fig. 9A), closely resembled ‘feedback’ axons projecting from area V2 to V1 (Rockland and Virga, 1989). That is, the main axon ascended to the pia, made a right angle turn, and traveled in layer 1 for ~1.0 mm, with terminations concentrated over its distal 0.60 mm (95 boutons were counted in total). Three other axons (of which one is illustrated) ascended to layer 1, but gave rise to only a small terminal tuft. All of these axons had collateral branches in layers 3 and 5 (total number of boutons were, respectively, 71, 95 and 102; Fig. 9B). The fifth axon had a larger terminal cluster in layers 1–3, with two collaterals in layer 4 (bouton count = 286; Fig. 9C). The main cluster measured ~0.40 mm DV × 0.50 mm AP. For all five of these axons, some terminal specializations also occurred along the main axon trunk in the subjacent layers. For axon 4-1, these were most pronounced in layer 4.
Two of these axons were followed posterior, to the region of the OTS, where branches were given off to an overlying dense focus of terminations (Fig. 9A,B). Axon 4-5 was also followed in the white matter, but this time continued to the vicinity of the entorhinal cortex (Fig. 9C).
Two other axons terminating in TEav were more complicated and had multiple arbors in layer 4. Axon 4-6 had a single large arbor (‘principal arbor’) in layer 4, which was 0.95 mm DV × 0.6 mm AP. Four additional smaller foci could be distinguished (Figs 10 and 11). These were separated by intervals of 0.3–0.5 mm, center-to-center. The full extent of the axon, measured as a hollow square in the gray matter, was ~1.8 mm DV × 3.35 mm AP (bouton count ≥ 314). (This exceeds the denser core of the projection focus, the dimensions of which are given in the first section of the Results.) Axon 4-7 (not illustrated) also had a principal arbor in layer 4 (0.6 mm DV × 0.25 mm AP). Another, smaller cluster was concentrated in layer 3, ~0.5mm posterior to the principal arbor. Two thin branches leading to layer 4 suggested a likely third focus, but were too faint to follow. The total extent of this axon, which is somewhat incomplete, was 1.25 mm DV × 2.05 mm AP in the gray matter (bouton count ≥ 147).
Branched segments could easily be found in the white matter (Fig. 12); that is, at least one branched profile could be seen per section. In addition to the seven serially recon- structed axons, one long axon segment (not illustrated) was followed from layer 6 of TEav (case P4) for 2.0 mm ventral in the white matter. One collateral entered the gray matter of the perirhinal cortex and continued anterior, to the border of perirhinal and entorhinal cortex, where it became too faint to follow. Three other axon segments (not illustrated) terminated in TE (one from P4, two from P3) and also sent branches toward the rhinal sulcus.
Input to anterior TE from parietal cortex is generally considered to be indirect; for example, via parietal-recipient parts of the STS, prefrontal cortex, or area TEO (Cavada and Goldman-Rakic, 1989; Morel and Bullier, 1990; Baizer et al., 1991; Webster et al., 1991, 1994; Seltzer and Pandya, 1994; Seltzer et al., 1996). Direct connections are known to occur, but primarily between area LIP and TEad (Webster et al., 1994). Two previous studies, however, using large retrograde tracer injections, have reported direct connections, consistent with our results, from parts of the inferior parietal lobule to TEav (Iwai et al., 1992; Martin-Elkins and Horel, 1992).
There are two likely explanations for why these connections have not been emphasized. First, there is the issue of tracer sensitivity. BDA is more effective for visualizing light con- nections than either WGA-HRP or autoradiography. Anderson et al. (Anderson et al., 1990) in fact illustrate (their fig. 18), but do not remark on, light projections from area 7b, corresponding to our case P1, to what looks like layer 6 of the upper bank of the amts. A second issue is location. Most of the projections in this investigation are within the amts, in a location that might well not ordinarily be reached by injections of retrograde tracer in TE. Thus, even large retrograde tracer injections in lateral TE might not produce label in the IPL. Thirdly, projections from the more anterior and posterior IPL (except for the middle portion, PG/PFG) are relatively sparse and could easily be missed. Given the moderate to sparse density of the projections we have demonstrated, these are unlikely to be a major sole source of parietal-temporal interactions. What influence they may have (potentiating? inhibitory?) in conjunction with other, denser connections will be important to determine. What are the properties of the parent neurons in the IPL, the identity of the postsynaptic populations in TEav and connectional strength or efficacy?
If TEav is selectively involved in recognition memory (Saleem et al., 2000), it may be that the parietal connections contribute to aspects such as topokinetic memory (Berthoz, 1997). Consistent with memory-related functions, five (P3–P7) of the seven injection sites in the IPL also sent projections to CA1 of the hippocampus (Rockland and Van Hoesen, 1999). Moreover, TEav, but not TEad, receives direct connections from CA1 (Yukie, 2000; Insausti and Munoz, 2001; Zhong and Rockland, 2002).
It may also be important that parietal areas are associated with spatial representations which encode locations and objects of interest in several egocentric reference frames (Colby and Goldberg, 1999). In a recent fMRI study in monkeys, the amts has specifically been implicated in a network of areas that are co-activated during tasks having to do with three-dimensional shape processing (Sereno et al., 2002). Direct connections from the IPL would be a plausible substrate. Another fMRI study, in humans, reported robust and consistent somatosensory activation in occipito-temporal cortex and suggested that object-related brain areas may hold both visual and haptic representations of objects (Amedi et al., 2001).
Direct and Indirect Connectional Loops
The combination of sparse direct and dense indirect con- nections is a common feature of cortical architectures, and even of connectional loops in general [see, for example, Nambu et al. (Nambu et al., 2002), concerning the ‘hyperdirect’ cortico- subthalamo-pallidal pathway]. In the early visual system, a small number of direct ‘bypass’ projections have been reported from V1 to V4 and from V2 to TEO. The direct connections have been interpreted as important in timing mechanisms (Nakamura et al., 1993), but the actual significance is not known. In another, more recent example, a small number of neurons have been reported in auditory association cortex that project directly to primary visual cortex, especially in the calcarine fissure (Rockland and Ojima, 2001; Falchier et al., 2002). Presumably, the direct pathway operates in conjunction with multiple other, more indirect pathways between these regions, although both the specific architecture and functions require further elucidation.
As one attempt to evaluate connectional density, it may be useful to compare our results with those in other retrograde tracer experiments of direct ‘bypass’ connections. Falchier et al. (Falchier et al., 2002), after retrograde tracer injections in area V1 in the calcarine fissure, report 236 labeled neurons, from 34 sections, in auditory areas; 408 in the superior temporal polysensory area; and 1300 in the STS complex (their case M88). In other words, the weaker direct connections in auditory areas originate from about one-fifth the number of neurons as the denser indirect connections from the STS. This is within range of our results (where there were 11 neurons per counting box in the IPL and 59 in the STS). The neuron number in auditory areas is less than the total number of neurons (677) that we counted in the IPL. Similarly, the number of labeled neurons in the IPL is greater than the number of neurons given by Nakamura et al. (Nakamura et al., 1993) as projecting from V2 to TEO (for four cases: 33, 89, 218 and 225).
In addition to direct connections from the IPL to TEav, there are several other, indirect routes for potential interaction between parietal and temporal areas. One is through the STS. Previous studies have emphasized dense connections from the IPL to the upper bank and fundus of the STS (Seltzer and Pandya, 1978, 1994; Cavada and Goldman-Rakic, 1989; Anderson et al., 1990; Seltzer et al., 1996), regions preferentially connected with TEad (Saleem et al., 2000). Some connections from the posterior (Seltzer et al., 1996) and anterior (Cavada and Goldman-Rakic, 1989) IPL have been noted to the lower bank as well, but our material indicates that these terminations are also relatively dense. As TEav is connected with the lower bank of the STS (Saleem et al., 2000), we conclude that both TEad and TEav may be receiving indirect parietal connections through the STS. In the case of TEav, however, these are combined with direct inputs as well.
Another route for indirect parieto-temporal interactions is through the ventral part of occipito-temporal cortex (TEO and TEp). This region receives dense connections from the posterior bank of the intraparietal sulcus (Cavada and Goldman-Rakic, 1989). According to our observations, all seven IPL injections projected ventrally, especially in the lateral bank of the OTS. This region, in turn, projects both to TEav and, in a reciprocal loop, back to the IPL (Iwai et al., 1992; Rockland, 2003). The integrity of this network—linking the IPL, TEav and ventral occipito-temporal cortex—is supported by the pattern of collateral branching. That is, three of our seven reconstructed axons from area PG had branches to TEav as well as the lateral bank of the OTS.
A third route is through the CA1 sector of the hippocampus. This region receives direct connections from several parts of the IPL (Saleem and Hashikawa, 1998; Rockland and Van Hoesen, 1999; Yukie, 2000), from the amts (personal observation) and from adjoining parts of ventromedial TE (Rockland and Van Hoesen, 1999; Yukie, 2000; Insausti and Munoz, 2001; Zhong and Rockland, 2002). The connections with ventromedial TE are reported to be reciprocal (Yukie, 2000).
While the effectiveness and significance of these convergent routes are currently unknown, it nevertheless seems that there are multiple opportunities for ‘cross-talk’ between the classical ventral and dorsal processing streams (Iwai et al., 1987; Martin-Elkins and Horel, 1992; Saleem and Tanaka, 1996; Roe and Ts’o, 1997; Hendry and Reid, 2000; Saleem et al., 2000). Continued physiological studies also provide evidence of ‘cross-talk’. For example, many inferotemporal cells selectively respond to depth profiles (Janssen et al., 2000); and it has been suggested that the range of visual information submodalities used in the dorsal and ventral pathways may overlap, since they are useful both for guidance of movements and for identification of objects (Tanaka, 2000). Task dependency and cognitive requirements also influence visual processing in dorsal and ventral streams (Fias et al., 2002). Interestingly, recent re- examination of connections from V1 to V2 has provided new evidence for ‘cross-talk’, even in the early visual pathway (Sincich and Horton, 2002).
It is well-known that laminar patterns can provide useful hints about the functional relationships of interconnected areas (Maunsell and Van Essen, 1983). According to the classical criteria used for early visual areas, the most prominent IPL connections, from area PG/PFG to layer 4, could be considered ‘feedforward’; those from area PF, to layer 6, could be ‘feedback’; and those from PG to layers 1, 4 and 6 would be mixed. Reciprocal projections, from the OTS to the mid-IPL, are to layer 1 and 6 (near the lateral sulcus) or in a columnar pattern (Rockland, 2003). One explanation of the complex laminar patterns is that the terminations are part of subtotal projections. With greater convergence and density, these might coalesce in a columnar pattern. A second consideration is that both the IPL and anterior TE consist of interconnected subfields (Pandya and Seltzer, 1982; Janssen et al., 2000; Lewis and Van Essen, 2000; Sereno et al., 2002). Thus, the laminar termination patterns may reflect complex higher order connectivity, where the short- range connections between parietal areas interact and converge with extrinsic connections to TEav in laminar-specific manners to recruit particular neuronal subpopulations.
In fact, beyond the early sensory areas, laminar termination patterns become increasingly difficult to interpret. Terminations in the upper layers of entorhinal cortex, for example, are not ‘feedback’ (Wellman and Rockland, 1997; Rockland, 2002) and subtle differences have been noted in the pattern of ‘feedforward’ TEO to TE terminations, compared with the ‘feedforward’ terminations from TE to perirhinal cortex (Saleem and Tanaka, 1996). The issue is further complicated when individual axons branch to several targets. As we have demon- strated, branched connections occur between TEav and the OTS, and between TEav and the posterior rhinal sulcus, areas that would be classified at different levels of the standard hierarchy (Felleman and Van Essen, 1991).
Electrophysiological and optical imaging studies have demon- strated a modular organization in area TEad, where neurons with similar response properties occur in clusters ~0.5 mm in diameter (Tanaka, 1996). Terminal clusters of ~0.5 mm have also been demonstrated anatomically, both for intrinsic connections within IT (Fujita and Fujita, 1996) and extrinsic connections from area TEO (Saleem et al., 1993).
Less information is available concerning the functional architecture of TEav. Physiological studies show some differ- ences between TEad and TEav. That is, cells in TEav are found to be more difficult to activate and they respond more strongly to more colorful objects (Tamura and Tanaka, 2001). Our single axon data would be consistent with some kind of columnar organization in TEav as the terminations consist of multiple foci, spaced 0.3–0.5 mm apart. Moreover, the largest arbors in layer 4 measure 0.95 × 0.60 mm, a size roughly consistent with that of functional clusters in TEad. Terminations in layer 1 are ~1.0 mm in length, perhaps crossing over multiple columns.
We would like to thank Michiko Fujisawa for expert assistance with manuscript preparation and Dr Manabu Tanifuji for helpful discussion and commentary on the manuscript. The work was supported by research funding from RIKEN Brain Science Institute. Four of the parietal injections had been used in previous studies at the University of Iowa.
Address correspondence to: Kathleen S. Rockland, Laboratory for Cortical Organization and Systematics, RIKEN Brain Science Institute, 2-1 Hirosawa, Wako-shi, Saitama 351-0198, Japan. Email: email@example.com.