Our study concerns the mechanisms that underlie functional imaging of sensory areas of cortex using hemodynamic-based methods such as optical imaging of intrinsic signals, functional magnetic resonance imaging and positron emission tomography. In temporal cortex of chinchilla, we have used optical imaging of intrinsic signals evoked by acoustic stimulation to define the functionally responsive area and then made (scanning electron microscopy) observations of the corresponding capillary networks prepared by corrosion cast methods. We report that intrinsic signals associated with auditory cortex correlate directly with discrete capillary beds. These capillary beds, within the cortical surface layers, are distributed across the cortex in a non-uniform fashion. Within cortex both the arterial supply and the capillary network contain various flow control structures. Our study suggests a causal relationship between the metabolic demands of local neuronal activity and both the density of the capillary network and the placement of the control structures. Such relationships will affect the ultimate spatial resolution obtainable by hemodynamic-based functional brain imaging studies. These relationships will also affect quantitative comparisons of activity levels in different areas of cortex.
Functional brain imaging techniques such as functional magnetic resonance imaging (fMRI), positron emission tomography (PET) and optical imaging of intrinsic signals rely on the apparently close coupling between metabolic demand of activated neurons and local changes (typically increases) in blood supply (Roy and Sherrington, 1890; Lou et al., 1987). Numerous studies have focused on the exact relationships between nerve-cell activity and the mechanisms of local blood flow modulation (Lazorthes et al., 1968; Black et al., 1987; Toga, 1987; Sirevaag et al., 1988; Edelman and Gally, 1992; Iadecola, 1993; Argandoña and Lafuente, 1996, 2000; Estrada and DeFelipe, 1998). Investigations concerned with improving functional imaging methods (particularly fMRI) have explored the possibility of capturing the earliest (and most local) hemodynamic/metabolic events associated with neural activity, thus achieving an improved spatial resolution (Malonek and Grinvald, 1996; Logothetis et al., 1999; Kim et al., 2000). In any case, all these imaging techniques have a limiting factor which is defined by the physical relationship between blood capillaries and the neurons of interest. Essentially all techniques that rely on changing hemodynamics, or blood oxygen level dependent (BOLD) mechanisms, are bound in their functional (spatial) resolution by the density of blood capillary vasculature and the fineness of hemodynamic control within those vascular beds.
In the present study we have explored these relationships in detail. We have optically imaged intrinsic signals (Grinvald et al., 1986; Frostig et al., 1990; Bonhoeffer and Grinvald, 1996) evoked by acoustic stimulation in auditory areas of temporal cerebral cortex (Harrison et al., 1998; Harel et al., 2000). We have then made corrosion casts of the vascular anatomy in the same areas for viewing by scanning electron microscopy (SEM). We find that for core areas of auditory cortex, the spatial pattern of intrinsic signals (relating to hemodynamic change) relates directly to the physical position and density of capillary beds, and to the distribution of the myogenic valves that control blood flow to these capillary beds. Our findings have important implications for a wide range of medical imaging techniques which aim to imply neuronal function from hemodynamic change.
Materials and Methods
Functional imaging and subsequent corrosion cast preparation was performed on five adult chinchillas (Chinchilla laniger). In the ketamine-anesthetized subject, intrinsic signal imaging of temporal cortex was used to define active regions of auditory cortex. This was immediately followed by plastic injection of cerebral blood vessels (via aortic perfusion). Following plastic polymerization, and digestion of tissue, these (whole brain) corrosion casts were studied using SEM. All procedures were carried out following the guidelines of the Canadian Council on Animal Care, and with the approval of the local animal care committee. Experimental details are as follows.
For optical imaging of intrinsic signals, a craniotomy was made over the temporal region of the brain extending ventrally to zygoma and dorsally to the parietal frontal bones. Body temperature was maintained at 36°C; animals are maintained under light ketamine anesthesia (15 mg/kg; i.m.). Our animals are respiring spontaneously. Positive pressure pumping causes considerable mechanical movements of the brain which are a problem in optical imaging. Other groups imaging visual cortex in the cat (Bonhoeffer and Grinvald, 1996) have time-locked image capture to respiratory pumping cycle, but this has not proved useful in our experience with the chinchilla. Without pumping we have, theoretically, less control over the physiological condition of our subjects; however, our experience has been that there is no evidence of hypoxia for 6–8 h light ketamine anesthesia. The dura was excised, the cortical surface was covered with silicone oil and a glass coverslip placed over the area to produce an optically flat interface. The cortical surface was evenly illuminated with 540 nm wavelength light. Images of the cortex were acquired with a CCD camera and macro lens assembly (Imager 2001, Optical Imaging Inc. Germantown, NY). Activity in auditory cortex was evoked by a 3 s duration broadband noise at 80 dB SPL presented to the ear contralateral to the imaged cortex. Imaging was carried out during and after acoustic stimulation for a period of up to 10 s (Harrison et al., 1998; Harel et al., 2000).
To prepare corrosion casts of the cerebral vasculature, the ascending aorta was perfused with 50 ml heparinized phosphate buffered saline (PBS) followed by 20 ml of Batson's #17 resin. Complete polymerization of the resin took ~12 h, after which the brain was dissected from the cranium. Tissue was macerated in 40% KOH at 50°C for 24 h with intermittent distilled water rinses. The plastic cast was air dried, mounted onto a stub with colloidal silver paste and sputter coated with gold. The cast was examined in an Hitachi S570 SEM at 10 keV.
Definition of Functional Area
Figure 1 shows the time course and spatial position of the intrinsic signal in response to a broadband sound stimulus. The figure shows thresholded images from one experimental subject recorded at 0.5 s intervals during acoustic stimulation (upper data series) and in a no-stimulus condition (lower data series). The outline of the maximum intrinsic signal, captured at 1–1.5 s after stimulus onset (third image of the stimulus trial), is superimposed on the grayscale image of the cortical surface in the upper right-hand panel. The response is typical of that previously reported (Harrison et al., 1998; Harel et al., 2000). The intrinsic signal area corresponds with regions of ‘core’ auditory cortex (AI, AII and anterior auditory field) within which single units respond very actively to acoustic stimulation. Figure 2 shows this relationship for one subject in which we have made concurrent intrinsic signal imaging (1 kHz, 80 dB) and single unit electrophysiological recordings. Point symbols are at sites where neurons are acoustically driven; × symbols show no-response sites.
The relationship between cortical surface and the intrinsic signal boundary from Figure 1 is reproduced in the left panel of Figure 3. The right panel shows an SEM image of the same area after corrosion cast preparation, and again the boundary of the intrinsic signal area is indicated. Note that both images are flat representations of curved structures at slightly different orientations, thus the intrinsic signal outlines have similar but not identical shapes.
The general organization of the cortical layer vasculature is that both major arterial supply and venous drainage is at the surface of cortex, and can be directly visualized as seen in the left image of Figure 3. The pial arteries give rise to arterioles which course down perpendicular to the surface; some vessels pass directly down to deeper cortical areas with little or no branching, and others feed a subsurface capillary system. There is venous return from such capillary beds back to the cortical surface.
Our first important observation is that the capillary bed distribution across (temporal) cortex is not uniform. Figure 4 shows areas with a very dense capillary network separated by regions with few or no capillaries from four different subjects. Our second observation is that certain of these dense capillary beds closely fit the boundary of the intrinsic signal area. In the example shown (Figure 3, left) this is particularly evident on the posterior (caudal) boundary of the intrinsic signal region. Figure 5 shows a higher magnification image of this area. The upper inset shows the cortical surface view of the identical region with the intrinsic signal boundary superimposed. The major vein and artery are labeled (A and B respectively) on both images. The capillary bed on the left side of the SEM image is clearly associated with the measured intrinsic signal from primary auditory cortex; indeed it is difficult not to conclude that the intrinsic signal results from hemodynamic changes in that capillary bed. The dense capillary mesh to the right in this SEM image produced no acoustically evoked intrinsic signal. Between these two high-density capillary areas is a region with no capillary bed. We do not believe that this uneven distribution is a preparation artifact for a number of reasons. First, we know that capillary beds are well perfused when the resin is seen to fill the venous side of the system. Secondly, close inspection of the arterioles adjacent to this gap does not reveal evidence of incomplete plastic infiltration of side branching vessels. In addition, similar patterns of heterogeneous capillary bed distribution were observed in all subjects.
Vascular Control System
Figure 6 shows the dense capillary mesh associated with auditory cortex. This is an area within that corresponding to the maximum intrinsic signal response to the noise stimulus. Arteries and their branches (A) can be recognized by their round profile and the elongated imprints of vessel wall epithelial cells (Hodde et al., 1997); resin casts of veins (V) tend to be flattened and with a smoother surface. The rich vascular network has clear properties that indicate that it is not passively gorged with blood, but rather has a number of controlling sites. Most obvious in the corrosion cast preparations is evidence of the position of smooth muscle bands, strategically placed along arteriole trunks and at branching points (asterisks), to control blood flow to the capillary networks. In these corrosion casts we do not observe directly the smooth muscles but rather the cast of the endothelial spaces that surround the muscle, sometimes referred to as plastic strips (Anderson and Anderson, 1978; Castenholz, 1983a, b; Reina-De La Torre et al., 1998). The larger arteries show evidence of patchy muscle banding. It is likely that this incomplete pattern is due to the limitations of the corrosion casting technique, which can only show spaces that have a continuum with the inside of blood vessels.
The main artery (A) gives off three types of collateral vessel. The first type (labeled 1) courses over the cortical surface for some distance, giving rise to other collateral vessels before either penetrating the cortical surface itself or dividing into terminal branches which eventually penetrate the surface. The second type (2) arises from the side or undersurface of the artery and branches out both collaterally and terminally, giving rise to many pre-capillary arterioles. The third type (3) is an intracortical artery which penetrates deep within cortex without contributing to the superficial capillary network. Most arterioles (e.g. all those shown in Fig. 6) show some evidence of myogenic banding, and we have an impression that those vessels feeding capillary beds have more such structures (compare arteriole 2 to arteriole 3); further quantitative analysis is needed. We have noted that these smooth muscle zones are often associated with arteriole branch points and can serve either to constrict downstream vessel walls or to distort areas next to side branches so as to alter blood flow down them.
Pre-capillary arterioles frequently possess a distinct form of smooth muscle banding as seen in the upper panels of Figure 7. A single bifurcating structure encircles the arteriole. These are often seen in close association with collateral capillaries. Downstream from the arterioles we see evidence of a second set of vascular control structures, the vascular pericytes (Castenholz, 1980, 1983a; Sims, 1986; Tilton, 1991; Rodriguez-Baeza et al., et al., 1998). Most often with corrosion casting, pericytes do not become resin filled but rather leave their signature depressions surrounding the capillary (Kojimahara and Ooneda, 1980; Castenholz, 1983a; Motti et al., 1986; Pannarale et al., 1996). Examples of these often sphincter-like depressions are seen in the lower panels of Figure 7 (arrows). In capillary beds of auditory cortex these structures are usually situated at or close to capillary branching points.
The capillary mesh itself is a complicated series of loops. In Figure 8 we have traced two of these loops from the arterial supply to the venous drain. Multiple stereo-pair micrographs were used to follow the course of these capillaries. First we note that most, if not all, capillaries in this micrograph are interconnected. The path of only two capillary loops are highlighted. A pre-capillary arteriole (red) gives rise to a collateral capillary (green) which divides into two terminal capillaries, one short (orange) and the other longer (cyan). The short (orange) capillary connects with three other capillary loops before entering a post-capillary venule (blue) as a collateral vessel ~35 μm above its origin. The longer capillary (cyan) follows a more tortuous route circling the venule and connecting with nine other capillary loops before entering the same post-capillary venule, also as a collateral vessel. Approximately 15 μm above this point the venule is joined by another post-capillary venule these join with others to become the venous branch shown in Figure 6.
Corrosion Cast Technique
A number of studies have used corrosion casting to explore the capillary networks of cerebral cortex (Motti et al., 1986; Reina-De La Torre et al., 1998; Rodriguez-Baeza et al., 1998). Many of the features that we observe in chinchilla auditory cortex are similar to previous descriptions, including the sphincter-like grooves caused by vascular pericytes (Kojimahara and Ooneda, 1980). Rodriguez-Baeza and his group have recently reported on corrosion cast SEM studies in human cerebrum (Reina-De La Torre et al., 1998; Rodriguez-Baeza et al., 1998), and the general arrangement of vasculature is similar to that reported for other mammalian species. These authors draw special attention to the variations in vascularity according to depth, including a lack of capillarity in the superficial (pial) layer which we also observe. An increase in capillary density associated with cytoarchitectonic distinct whisker barrels has been shown (Cox et al., 1993; Woolsey et al., 1996).
The clear evidence of smooth muscle banding on arteriole walls and of capillary vascular pericytes has been described in detail by other authors (Castenholz, 1983a; Motti et al., 1986; Reina-De La Torre et al., 1998; Rodriguez-Baeza et al., 1998), although not directly in the context of functional imaging. In the case of smooth muscle banding on arterioles, it is assumed that the plastic cast is of endothelial space surrounding smooth muscle cells which is connected to the vessel lumen. As for the vascular pericytes, we occasionally observe casts of these structures which are similar to those described by others (Castenholz, 1983a; Rodriguez-Baeza et al., 1998); however, most often in our preparations we see only the deep grooves that they make in capillary walls. It should be noted that some of these grooves have also been interpreted as evidence of peristaltic vasomotion (Toribatake et al., 1997). In any case, the function of these vaso-surrounding structures is now widely assumed to be related to the fine control of blood flow in cortical capillary beds. We have not yet made a quantitative analysis of the distribution of ‘vascular control points'; we observe them ubiquitously. However, we hypothesize that their density could be different in cortical areas which require a refined hemodynamic control.
Our working hypothesis is that the capillary density of any brain area develops in direct relationship to the metabolic demand of local neurons. Thus while primary sensory areas will have dense capillary nets, some non-sensory or association areas may not. This notion has previously been put forward by a number of authors who have explored the relationship between neuronal activity and capillary density (Lazorthes et al., 1968; Black et al., 1987; Toga, 1987; Sirevaag et al., 1988; Argandoña and Lafuente, 1996).
Functional Brain Imaging
In many ways, the close relationship between the shape of the intrinsic signal region and the distribution of the capillary network should really not be much of a surprise. Intrinsic signals arise from hemodynamic effects that have their origins in capillary networks (Grinvald et al., 1986; Frostig et al., 1990). However, this statement of the obvious has some important implications for all those imaging techniques that capitalize on function-related hemodynamic change (including fMRI, PET, intrinsic signal imaging, Doppler blood flow measurement, etc.). Importantly it means that there exist certain cortical areas in which the density of blood vasculature is higher, and these areas are more likely to light up in various functional tasks than less vascularized regions. These areas are likely to include primary sensory areas, motor cortex, and perhaps in humans, speech-related regions such as Wernicke's and Broca's areas. The corollary is that many areas, because of an insufficient blood capillary network, may not produce a function related signal.
We postulate that, due to a variation in density of vascular beds and their controlling points in different areas of cortex, the spatial resolution of imaging methods will depend on which area is under investigation. Thus functional imaging of a highly vascularized area such as primary auditory cortex will reveal local activity patterns at a considerably higher spatial resolution than in less vascularized (less active) areas. In the chinchilla auditory cortex, we have previously been able to demonstrate (using optical imaging of intrinsic signals) tonotopic maps with octave spaced frequency resolution (Harrison et al., 1998; Harel et al., 2000). This represents a functional resolution of ~400 μm. On the basis of the present anatomical study, we might estimate the limit to this spatial resolution to be of the order of 100–150 μm based on the approximate distance between (pre) capillary flow control valves and start of venous drainage (see Fig. 6). There are intrinsic signal imaging and fMRI data from striate cortex of the visual system which have revealed orientation columns with a dimension in the 50–100 μm range (Grinvald et al., 2000), and we therefore predict that the capillary network in primary visual cortex may be of finer dimensions or have a greater abundance of controlling sphincters to allow finer hemodynamic control.
The general notion that different areas of the cerebral cortex have different densities of fine vasculature according to the metabolic demands of neural activity, and/or that the fineness of hemodynamic control differs from one region to the next, has a significant implication for BOLD or other blood flow-based imaging studies of neural function. Most obviously it means that quantitative comparisons between signals coming from different areas have to be made with caution. A worse-case scenario is that perhaps only certain high-activity regions of cortex are endowed with sufficient capillary network density/control to produce signals for functional imaging, and that some areas of association cortex, for example, may never generate such a response. Further detailed mapping studies of blood pathways and associated control points in various cortical areas are needed to estimate the ultimate spatial resolution of hemodynamic based functional imaging methods such as fMRI and PET.
This research was supported by the Canadian Institutes of Health Research (CIHR), and the Masonic Foundation of Ontario.
- cerebral cortex
- auditory area
- casts, surgical
- blood capillaries
- diagnostic imaging
- temporal lobe
- acoustic stimulation
- functional magnetic resonance imaging
- blood supply, arterial
- mri of brain with functional imaging
- functional imaging
- doppler hemodynamics
- fluid flow