The present investigation utilized tissue from the eight male rhesus monkeys (Macaca mulatta) that were the basis for our previous study (Mortazavi et al., 2016). Briefly, all animals were obtained from national primate centers and had known birth dates and health records which were screened to ensure that they were free from any history of disease or experimentation that might compromise the brain. All were part of other ongoing studies and were housed at the Laboratory Animal Science Center on the Boston University Medical Campus (BUMC). This facility is managed by a licensed veterinarian, staffed by trained and accredited laboratory animal staff and fully accredited by the Association for the Assessment and Accreditation of the Laboratory Animal Care. All procedures conformed to the National Institutes of Health guidelines and the Institute of Laboratory Animal Resources Commission on Life Sciences’ Guide for the care and use of laboratory animals (1996). All procedures were approved by the BUMC Institutional Animal Care and Use Committee.

After all behavioral testing and other procedures were completed, subjects were tranquilized with ketamine (10 mg/kg, intramuscular), deeply anesthetized with sodium pentobarbital (15 mg/kg, intravenous to effect), and euthanized by exsanguination during transcardial perfusion with 4% Krebs buffer at 4°C, followed by 4% paraformaldehyde in phosphate buffer (0.1 M, pH 7.4) at 37°C. Brains were blocked in situ, in the coronal plane just posterior to the splenium, removed from the skull, weighed, post-fixed overnight in 4% paraformaldehyde for no more than 18 h, and transferred to cryoprotectant solution to eliminate freezing artifact (Rosene et al., 1986). Blocks were flash frozen at −75°C and stored at −80°C until sectioned at 30 μm thickness into 10 repeating series of sections. Series not processed immediately were stored in cryoprotectant (15% buffered glycerol) at −80°C until removed, thawed and batched processed for immunohistochemistry (Estrada et al., 2017).

As described more fully in Mortazavi et al. (2016), sections were thawed, and rinsed three times for 5 min in 0.05 M Tris-buffered saline (TBS; pH 7.4) to remove the cryoprotectant. To quench endogenous peroxidases, sections were incubated for 30 min in 0.05 M TBS and 1% hydrogen peroxide. After three 5 min washes in 0.05 M TBS, sections were incubated for 1 h in a blocking solution of 10% Normal Goat Serum (NGS) and 0.4% Triton-X in 0.05 M TBS before incubation for 48 h at 4°C with gentle agitation in mouse anti-NeuN IgG (1:10,000; MAB377, Chemicon, Temecula, CA, USA), in 0.05 M TBS containing 2% NGS and 0.1% Triton-X. Following this incubation, the sections were again washed, and further processed by a 2 h incubation period with the secondary antibody (goat anti-mouse, 1:600; Vector, Burlingame, CA, USA), in 0.05 M TBS with 2% NGS, and 0.4% Triton-X. After washing, sections were incubated with an avidin biotinylated horseradish peroxidase complex (ABC, Vector Labs, Burlingame, CA, USA) for 1 h. After another wash cycle, sections were incubated for 7 min in sodium acetate containing 0.55 mM 3-3′-diaminobenzidine (DAB; Sigma, St. Louis, MO, USA) and 0.01% H₂O₂. Sections were washed, mounted onto gelatin-coated slides, air dried and cover-slipped with Permount mounting medium (ThermoFisher Scientific, Waltham, MA, USA).

As an extension of Mortazavi et al. (2016), this study focused on four WM regions (SFG, FR, Par and TE). We felt this provided a base comparison of associational cortical areas, from dorsal and ventral streams, and representative of anterior-posterior levels, and with different gray matter cytoarchitecture and connectivity. Analysis of additonal regions was not undertaken because of lack of tissue (for occipital) and, at this time, lack of resources to undertake a more comprehensive study between primary and association WM areas or further detailed subregions.

The nomenclature for gross anatomical, macro WM subdivisions is much less well established than that for the overlying gray matter or for defined axon tracts within the WM. For gyral WM regions, in particular, nomenclature sometimes adheres to gross anatomical distinctions (“superior temporal gyrus, STG”) but other times follows gray matter architectonic subdivisions (“area TEd”). In this study, we used “frontal region” to denote the territory from the anterior tip of the principal sulcus to the arcuate spur; the “parietal region”, the zone from the anterior tip of the intraparietal sulcus (IPS) to the posterior part of the superior temporal sulcus (STS); and the “temporal region”, from the mid-amygdala to the anterior IPS. For the temporal region, counts were subdivided so as to correspond to the STG and the gyrus ventral to the STS (TEd; Figure 1).

The identification of the cingulate sulcus and its nomenclature are straightforward. The dorsally adjoining gyrus, however, which we used for gyral counts in this region, was more problematic. This territory, from the spur of the arcuate sulcus posterior to the central sulcus, underlies several architectonic cortical regions, including the supplementary motor, and somatomotor functional cortical domains. In the human brain, it would correspond to the gross anatomical “SFG”, and we adopted this term for the sake of convenience.

The results in the current study are based on the same tissue as used in Mortazavi et al. (2016). ROIs were digitized at on a Zeiss Axiophot microscope using a 5× objective. These high resolution images were then stitched using the pair-wise stitch function in Fiji (Schindelin et al., 2012), and further zoomed for counting purposes, to a total magnification of 120–150×. Two to four sections per animal, per region generated a total of 147 counting frames in the sulcus and 125 in the gyrus (per region: 57 in SFG, 72 in FR, 75 in Par and 68 in TE).

Polygons of area 0.21 mm² were created with the Freehand selection tool in ImageJ (NIH, Bethesda, MD, USA). Polygons approximated a semi-circle or curved cone for gyral crowns, and crescent for sulcal depths (Figure 2). Neuron counts were carried out by using the cell counter tool in ImageJ and entered into Excel sheets. Counts were carried out by one investigator (AM), and partially replicated by two or three other scorers (AK, LG, SR, or KSR), with a quality control criterion that counts differed by less than 10%. As in Mortazavi et al. (2016), all neurons within the bounded ROI were counted. As noted in other quantitative studies of WMNs (e.g., Yang et al., 2011), designed-based stereological counting is not easily applied to the relatively borderless WM expanse.

Density measurements were made from four association areas; namely, FR, SFG, Par and TE. These represented a broad sample spanning the length of the cerebral hemisphere and areas with different gray matter cytoarchitecture and connectivity. Given the functional importance of the TE region, we further distinguished two subdivisions in this region, corresponding to the STG, between the lateral and superior temporal sulci, and TEd (ventral to the STS; Saleem and Logothetis, 2007). The border of layer 6 and WM was judged by a fall-off in neuron density and a frequent change in preferred cellular orientation, from perpendicular to parallel to the overlying pia surface. To minimize the risk of including the deepest portions of layer 6, we systematically placed the superficial border of the ROI slightly beneath the qualitatively judged deep boundary of layer 6 (see Figure 2).

To evaluate changes in WMN density proceeding away from the gryral crown, in the two TE subdivisions, we used a counting frame in the shape of a narrow rectangle (0.2 mm × 5.0 mm). One edge of this rectangular counting frame was placed at the gyral crown, and four underlying zones (each 1.25 mm) were demarcated medially toward the sulcal depth. The rectangle was placed approximately in the middle of the gyral core, so as to avoid possible overlap with domains associated with gyral walls.

In our previous investigation, we found no change in WMN numbers between young adult and aged monkeys (Mortazavi et al., 2016). Hence in the current study, we pooled the old and young groups for all analyses. In contrast to the earlier study, the combined data from the larger sample were distributed normally as determined by a Shapiro Wilk test of normality, enabling the use of parametric statistics in the current study.

Statistical analysis was carried out using a within subjects design, and Bonferroni adjustments were made for multiple comparisons. To determine whether density of WMNs was different in the gyral crown vs. the sulcal depth, a one-way within ANOVA was used. To determine regional differences in the density of WMNs, a two-way repeated measures within-subjects ANOVA was used (two locations: gyrus, sulcus × four regions: FR, SFG, Par and TE). To determine if there was a statistically significant change in the WMN density in moving from the gyral crown towards deep WM, a repeated-measures ANOVA was used. All data are reported as mean ± SEM, and alpha was set at 0.05.

Qualitative observations suggest that WMNs are distinctly more numerous in the gyral crowns than sulcal depths (Figure 3). To confirm the apparent difference, the density of WMNs in the gyral crown and sulcal depths of our four cortical regions (FR, SFG, Par and TE) was quantified. Analysis showed an average of 1.5× more WMNs in gyral crowns (75 ± 4 WMNs per 0.21 mm²) than in sulcal depths (about 47 ± 3 WMNs per 0.21 mm²; F_(1,7) = 58.41, p < 0.05).

To assess region-specific differences in density of WMNs, the two-way ANOVA indicated that the global average difference summarized in Figure 3 is maintained at the region-specific level (Figure 4); and in all four regions, gyral WMN density was greater than that in the sulcus (F_(1,12) = 24.93, p < 0.05). For gyral locations, WMN density ranged from a low of 57 ± 4 neurons per counting frame in the SFG region to 93 ± 6 WMNs per counting frame in TE (TEd = 95 ± 5; STG = 81 ± 5). For FR and Par, we found 81 ± 5 and 80 ± 6 WMNs per counting frame, respectively. For sulcal locations, WMN density showed the same trend; that is, least for the SFG/cingulate region (31 ± 4 WMNs per counting frame), intermediate for FR (41 ± 3 WMNs) and Par (54 ± 2 WMNs), and highest for TE (63 ± 3 WMNs; TEd = 58 ± 4, STG = 67 ± 4). Additionally, post hoc analysis showed that there are significantly more WMNs in both TE and Par (both sulcal and gyral) than in FR. While the ratio of sulcal/gyral WMNs varied in the different regions (FR (0.59), SFG (0.59), Par (0.73) and TE (0.70)), the differences were not statistically significant (Figure 5).

Qualitative observations of all four regions suggest that WMNs become progressively sparser with distance from the gyral crown (Figure 6). As a first step, we compared the density of WMNs individually in STG and TEd. Both regions had fewer WMNs in the sulcal location. The sulcal/gyral proportions were slightly different, with a ratio of 83% for STG and 60% for TEd, though this was not statistically significant (Figure 7A). For the two temporal subregions, we quantified WMN density from the gyral crown toward the sulcal depth by tabulating WMN number along a rectangular counting frame (0.2 mm wide by 5.0 mm in length), divided into four adjacent segments or zones (0.2 mm × 1.25 mm each in; Figure 6A). As summarized in Figure 7, we found for zone 1, immediately near the gyral crown, 70 ± 2 WMNs in TEd and 63 ± 4 in STG. We also found a consistent reduction from zones 1 and 2, where comparable numbers in zone 2 are, for TEd, 26 ± 3 neurons and, for STG, 28 ± 3 (per 0.25 mm²). There is a small reduction from zone 2 to 3 but otherwise numbers are stable from zone 3 to 4; namely, for zone 3, 12 ± 1 neurons in TEd and 22 ± 2 in STG; and for zone 4, 10 ± 1 in TEd and average 21 ± 3 in STG. Statistical analysis confirmed that the drop in WMN density was significant in both STG and TEd, proceeding from the gyral crown medially, towards deep WM (F_(3,16) = 98.464, p < 0.05). Further, post hoc analysis showed that there are significantly more WMNs in zones 3 and 4 of STG as compared to TEd (Figure 7B).

Individual tissue sections show a less regular pattern, in that the number of WMNs can remain constant over zones 2–4, or the number in zone 4 can even be slightly higher, as assessed within our rectangular counting frame. This is at least partially explained by an irregular pattern in the arrangement of WMNs.

In gyral locations, the arrangement of WMNs is not uniform, but rather exhibits distinct, usually circular neuron-free zones (Figure 8). These were relatively common, and were observed in all four regions. Circular zones ranged from 120 μm to 250 μm in diameter. Quantification of size differences will require further investigation in strict serial sections or in thick tissue slabs.

At least some, if not all, WMNs are connectionally integrated into gray matter circuitry (Clancy et al., 2001; Tomioka and Rockland, 2007; Suárez-Solá et al., 2009; von Engelhardt et al., 2011). Thus, the differential density of WMNs at the gyral and sulcal locations can be taken to imply a differential influence of these neurons in relationship to the corresponding gyral and sulcal cortical gray matter. There could be different numbers of synapses, different postsynaptic targets, or different axonal collateralization patterns at gyral crowns and sulcal depths. For example, retrograde tracing experiments demonstrate that many WMNs at the gyral crest project corticothalamically (for monkey frontal cortex: Figures 10, 12 in Giguere and Goldman-Rakic, 1988). The smaller number of WMNs in the sulcal depth may mean that there are fewer corticothalamic projections from sulcal locations. Further, if corticothalamic WMNs share features of the overlying layer 6 corticothalamic (type 1) neurons (Rockland, 1996), these potentially would have collaterals to the reticular nucleus of the thalamus, the thalamus and the overlying cortical gray matter (in layer 4). Interestingly, recent studies in mouse report that the subpopulation of complexin-3 positive subplate neurons projects both to cortical thalamorecipient layers and to the thalamus, based on immunohistochemical localization of the presynaptic terminals (Hoerder-Suabedissen et al., 2016; Viswanathan et al., 2016).

A small population of WMNs, seemingly in a gyral location, contributes to long distance cortico-cortical projections (Mohedano-Moriano et al., 2015 and (identified as GABAergic), Tomioka and Rockland, 2007).

Sulcal and gyral specialization (“folding”) has been actively discussed in the context of cortical development (see Lewitus et al., 2013; Zilles et al., 2013; Sun and Hevner, 2014; Striedter et al., 2015). Recent evidence suggests the location of folds and fissures (i.e., gyral crown and sulcal depth) is mirrored by regional variation in progenitor cell proliferation (de Juan Romero et al., 2015); and comparison of gyral and sulcal progenitor transcriptomes demonstrates “thousands of genes” differentially expressed in prospective folds (gyri) or fissures (sulci; reviewed in Fernández et al., 2016). In brains of young ferrets (P6), particular combinations of genes that are modularly expressed in the outer subventricular zone correspond to the location of folds and fissures, and the combinations are thought to differ depending on the specific gyrus or sulcus (Fernández et al., 2016). If WMNs have the same expression differential, this would suggest that they develop in concert with overlying cortical neurons but if they have a different pattern it would raise the possibility that they constitute a unique functional entity. Since, moreover, WMNs occur in lissencephalic brains, it seems more likely that any contribution to cortical folding is contributory rather than key.

An increased density of WMNs has consistently been associated with schizophrenia (Eastwood and Harrison, 2003; Connor et al., 2011; Kostović et al., 2011; Yang et al., 2011; Joshi et al., 2012), and there are reports of an increased frontal density of WMNs in fetal alcohol syndrome in macaque (Burke et al., 2009). In these conditions, less information is available about changes in the density of WMNs in relation to gyral/sulcal locations; but one postmortem investigation of epileptic brains reports an overall increased density of WMNs (Figure 3 in Loup et al., 2009). Age-related and Alzheimer’s-related changes occur in the cholinergic system, and these may differentially effect gyral, intra-gyral, and sulcal regions (Mesulam, 2013). Both plaques and tangles are reported as more prominent in sulcal depths than gyral crowns (Arendt et al., 2016). A common discussion point has been that superficial WM is particularly vulnerable due to a unique environment of late-myelinating U-fibers and late-myelinating oligodendrocytes (Haroutunian et al., 2014; Phillips et al., 2016a,b). The lower density of WMNs in the sulcal depths may be a contributing factor if, for example, these serve a protective role via neurovascular or synaptic processes (Rockland and Nayyar, 2012).

Finally, as observed in Reveley et al. (2015), the differential neuropil environment of sulci and gyri, which includes the dendritic and axonal meshwork of WMNs, impacts on the accuracy of tracking algorithms in DTI and fMRI protocols. In the in vivo imaging field, the segmentation of gray and WM, and in particular the accurate determination of gray matter/WM borders can have major consequences; for example, MR studies commonly assess thickness of cortical gray matter and WM in disease and normal development and aging. Currently, there is no one algorithm acknowledged as the most accurate because segmentation in MR is largely limited by partial volume effects (volume averaging of voxels; Yeh and Tseng, 2013; Yeh et al., 2013; Tohka, 2014; Rullmann et al., 2016). The microstructural organization of WMNs, across individual sulcal and gyral locations, underlines the continued need for better algorithms that are more sensitive to the complexities and ambiguities of anatomical segmentation.