The cortical column: a structure without a function

Abstract

This year, the field of neuroscience celebrates the 50th anniversary of Mountcastle's discovery of the cortical column. In this review, we summarize half a century of research and come to the disappointing realization that the column may have no function. Originally, it was described as a discrete structure, spanning the layers of the somatosensory cortex, which contains cells responsive to only a single modality, such as deep joint receptors or cutaneous receptors. Subsequently, examples of columns have been uncovered in numerous cortical areas, expanding the original concept to embrace a variety of different structures and principles. A "column" now refers to cells in any vertical cluster that share the same tuning for any given receptive field attribute. In striate cortex, for example, cells with the same eye preference are grouped into ocular dominance columns. Unaccountably, ocular dominance columns are present in some species, but not others. In principle, it should be possible to determine their function by searching for species differences in visual performance that correlate with their presence or absence. Unfortunately, this approach has been to no avail; no visual faculty has emerged that appears to require ocular dominance columns. Moreover, recent evidence has shown that the expression of ocular dominance columns can be highly variable among members of the same species, or even in different portions of the visual cortex in the same individual. These observations deal a fatal blow to the idea that ocular dominance columns serve a purpose. More broadly, the term "column" also denotes the periodic termination of anatomical projections within or between cortical areas. In many instances, periodic projections have a consistent relationship with some architectural feature, such as the cytochrome oxidase patches in V1 or the stripes in V2. These tissue compartments appear to divide cells with different receptive field properties into distinct processing streams. However, it is unclear what advantage, if any, is conveyed by this form of columnar segregation. Although the column is an attractive concept, it has failed as a unifying principle for understanding cortical function. Unravelling the organization of the cerebral cortex will require a painstaking description of the circuits, projections and response properties peculiar to cells in each of its various areas.

Figure 1

Mountcastle's evidence for columnar segregation of skin and deep receptors in areas 1, 2 and 3 of macaque somatosensory cortex. Drawing shows multiple micropipette penetrations from different monkeys, made within 1 mm of the sagittal plane marked by ‘A’ in the surface view below. Horizontal tick marks denote single units; grey shading indicates multiunit recordings. Radial penetrations in area 2 and adjacent area 1 yielded recordings devoted entirely to a single modality, whereas oblique penetrations through areas 1 and 3 resulted in regular switches between skin and deep receptors, presumably as column boundaries were crossed. The width of columns is difficult to judge because parallel penetrations were not shown from the same animal. In this illustration, columns seem to range from 200 μm to more than 1 mm in diameter. Electrode tracks were reconstructed from surface photos of vascular landmarks, electrode carrier dial readings and gliosis in tissue sections. No lesions were made. Mountcastle remarked candidly, ‘The majority of electrode tracks were extremely difficult to find’ (Powell & Mountcastle 1959, p. 135).

Figure 2

Implanted wires in the cat's visual cortex produce no disturbance in form perception. (a) X-rays of a formalin-fixed cat brain showing segments of tantalum metal inserted months before death to sever horizontal projections. This procedure produced no loss of the ability (measured before and after surgery), to differentiate geometric figures. (b) Coronal section through the posterior lateral gyrus prepared after wire removal, showing seven holes in the primary visual cortex. Note scars (arrows) left from cheese-wiring of the tantalum through the cortex. The wires probably failed to impair visual function simply because they were placed too far apart (Sperry et al. 1955).

Figure 3

Hubel and Wiesel's ice cube tray model of the striate cortex. In oblique microelectrode penetrations, they attributed the regular shift in orientation preference to an orderly stacking of slab-like orientation columns. Orientation hypercolumns contained 18 discrete columns about 50 μm wide, each 10° apart. Ocular dominance hypercolumns consisted of a single right (R) eye column and left (L) eye column about 500 μm wide. This diagram implies that a module consists of about four orientation and ocular dominance hypercolumns, measuring approximately 2×2 mm. Electrode movement in the cortex from one module to another was thought to produce a non overlapping displacement in the location of the aggregate receptive field of neurons. Hubel & Wiesel (1979, p. 160) later proposed that the combination of a single-orientation hypercolumn and ocular dominance hypercolumn ‘can be considered an elementary unit of the primary visual cortex’ (After Hubel et al. 1976).

Figure 4

Ontogenetic columns extend from white matter into cortex. A cross-section through macaque somatosensory cortex stained for Nissl substance shows radial stacks of cells extending from the white matter into the cortical layers. These remnants of foetal development have no known functional relationship to columns in the adult cortex.

Figure 5

Formation of radial minicolumns or ontogenetic columns. (a) Progenitor cells in the ventricular zone (VZ) give rise to progeny that migrate in succession along a glial scaffold into the cortical plate (CP). These cells remain roughly aligned along their migratory trajectory, giving rise to vertical minicolumns. (b) A β-gal⁺ clone of cells derived from an asymmetrically dividing stem cell, shown at various stages of vertical migration. The probe was introduced by intraventricular injection of a retroviral probe in the monkey foetus (Rakic 1995).

Figure 6

Different columns are activated by the same stimulus orientation. (a) Optical images of ferret primary visual cortex showing difference images produced by a texture of vertically moving bars oriented at 45°. The inset is shown at higher power below, with iso-orientation contours overlaid from a grating angle map. (b) The same region, after stimulation with horizontally moving bars oriented at 45°. The pattern of activation is quite different, although the stimulus bars were presented at the identical orientation. Red dots are placed to facilitate comparison between (a) and (b) (Basole et al. 2003).

Figure 7

Ocular dominance columns in macaque striate cortex flatmount. (a) Drawing of the left cortex of a normal macaque monkey showing ocular dominance columns supplied by the left (black) and right eye (white). Columns are present everywhere except in the representation of the monocular crescent (MC) and the blind spot (asterisk). The grey sliver extending from 1 to 4° corresponds to a shallow sulcus where the columns could not be reconstructed. (b) CO montage of layer 4C after right eye enucleation, showing the column pattern used to produce (a). (c) Autoradiograph prepared from alternate sections after injection of [³H]proline into the remaining left eye. There is a nearly perfect match between (b) and (c), validating CO histochemistry as a method for labelling ocular dominance columns (Sincich & Horton 2003a).

Figure 8

Monocular core zones and binocular border strips in primary visual cortex. Schematic of layer 4C in normal macaque striate cortex, showing the ocular dominance column borders marked with brackets. Each ocular dominance column consists of a monocular core zone containing CO patches and a binocular border strip. The transition between the two zones, although marked with a dashed line, is gradual. Their CO content is equal, giving layer 4C a homogenous appearance in normal animals. After monocular enucleation, CO activity is lost in the core zones and border strips of the missing eye's ocular dominance columns (below left). After lid suture, CO activity is reduced in the closed eye's ocular dominance columns and in the open eye's border strip regions, creating a pattern of thin dark columns alternating with wide pale columns (below right). Their contrast is low because eye lid suture has less effect on CO activity than enucleation (Horton & Hocking 1998).

Figure 9

Variable spectrum of ocular dominance columns in squirrel monkeys. The top row shows CO montages of layer 4C, after enucleation of one eye, from flatmounts of the left cortex of four normal animals. From left to right are examples of large, intermediate, fine and nearly absent columns. In animals with fine or nearly absent columns, a pattern is visible corresponding to the representation of retinal blood vessels (arrow). The blind spot representation is pale in three cases following enucleation of the left eye and dark in one case after enucleation of the right eye. Below each montage is a drawing of the column pattern prepared by Fourier filtering the CO montage. MC, monocular crescent; *, blind spot. Boxed regions in the lower panels are shown in figures 12 and 20 at higher power (Adams & Horton 2003a).

Figure 10

Regional, mirror-symmetric variation in the expression of ocular dominance columns. The ocular dominance columns were labelled in these flatmounts of the left and right cortex from a normal squirrel monkey by removing one eye. CO montages of layer 4C reveal well-segregated columns beyond an eccentricity of 8°. The columns disappear in the macular representation, but are present within the foveal representation (*). Ophthalmological and histological examination of the eyes showed no evidence of retinitis pigmentosa, nor any abnormality that might erase the columns in this regional fashion. What function could ocular dominance columns serve in such an animal? (Adams & Horton 2003a).

Figure 11

CO patches in V1 and CO stripes in V2 are revealed by many different labelling techniques. Here, they are shown by histochemical processing for another metabolic enzyme, β-nicotinamide adenine dinucleotide, reduced form (β-NADH). (a) CO patches in layers 2 and 3 of V1 form rows that follow the pattern of ocular dominance columns (arrows). (b) Adjacent section processed for β-NADH. The pattern of patches is identical (arrows). (c) CO-stained section from dorsal V2, with thick (large arrows) and thin (small arrows) stripes separated by pale stripes. (d) Adjacent section processed for β-NADH, showing the same pattern of stripes. Sections were processed for 2 h in a solution of 100 mg NADH (Sigma, grade III), 45 mg nitroblue tetrazolium and 12 ml DMSO in 100 ml of PBS (Horton unpublished work).

Figure 12

Conflicting designs: patches and ocular dominance columns in macaque and squirrel monkey. (a) CO patches in layer 3 of a normal macaque. (b) Montage of the [³H]proline-labelled ocular dominance columns in layer 4C. (c) Each row of patches runs down the middle of each ocular dominance column. Arrows in histological images mark vessels used for alignment. (d ) Patches in layer 3 of a squirrel monkey. (e) Ocular dominance columns in layer 4C, labelled by CO after monocular enucleation. This region corresponds to the box in figure 9, lower left. (f) Random relationship between patches and ocular dominance columns in the squirrel monkey (Adams & Horton unpublished work).

Figure 13

Alignment of singularities (pinwheel centres) and CO patches. (a) Optical imaging map of orientation preference in macaque striate cortex. The black lines represent the borders of ocular dominance columns. (b) Iso-orientation contours from (a) superimposed on an optical imaging map of the ocular dominance columns. Note that orientation singularities tend to be situated in the centres of ocular dominance columns. (c) Comparison between iso-orientation contours and CO patches. In some instances, singularities and patches coincide, but the correlation is far from perfect. There seem to be more singularities than patches, although determining their exact numbers is difficult because neither structure is always well demarcated (Data from Blasdel 1992b with permission).

Figure 14

Patchy horizontal connections in the visual cortex. After biocytin injection into layers 2 and 3 of squirrel monkey striate cortex, clusters of axon terminals surround the injection site. Their distribution is elliptical, with the long axis aligned with the preferred orientation recorded at the injection site. Patches of terminals are sometimes visible in deeper layers. When present, they form columns by virtue of their registration with patches of terminals in layers 2 and 3 (Sincich & Blasdel 2001).

Figure 15

Dovetailed, columnar projections from V1 and pulvinar to area V2. The projections from V1 to V2 are divided by cytochrome oxidase. Patches project to thin stripes and interpatches project to pale and thick stripes. The strongest V1 output is to pale stripes, whereas the pulvinar innervates more heavily thin and thick stripes. Controversy abounds regarding the receptive field properties of cells in the V2 stripes.

Figure 16

In V2 of many species, it is impossible to recognize two classes of CO stripes: thick and thin. In other species (e.g. squirrel monkeys), they can always be differentiated clearly. Remarkably, even within a species, expression of V2 stripes can be whimsical. In some macaques, thick and thin stripes are obvious, whereas in other macaques, they cannot be distinguished. Failure to visualize thin and thick stripes in some macaques is not an artefact of poor histology. This section shows an example of abysmal histology with uneven section thickness, chatter and tissue cracking, yet a tripartite pattern of CO stripes is still obvious in V2. After flatmounting and staining for CO in area V2 of more than 20 macaques, we have concluded that the variation in stripe expression is real.

Figure 17

A barrel is an isomorphic representation of a whisker, not a cortical column. (a) Flatmount section through the somatosensory cortex of a ground squirrel (Spermophilus beccheyi) processed for CO to show the barrelfield. (b) Masson trichrome-stained section through the vibrissae of a rat. A blood-filled sinus (arrow) surrounds each whisker, which may alter mechanical transduction of whisker motion.

Figure 18

Ocular dominance columns form without eyes. (a) In control ferrets, biotinylated dextran amine was injected into the lateral geniculate nucleus, yielding ocular dominance columns in coronal sections through the visual cortex. Serial sections were aligned and drawn to provide a 2D view of the columns. (b) Enucleation was performed at birth, before geniculate afferents innervate the cortex. Despite this manipulation, ocular dominance columns were present after tracer injection into the lateral geniculate nucleus at 2 months of age (Crowley & Katz 1999).

Figure 19

Random retinal activity waves cannot generate the patterns formed by ocular dominance columns. Despite variability in column expression among squirrel monkeys, the patterns in the left and right hemispheres are always mirror-image symmetric (see figure 10). Their formation from retinal patterns of neuronal activity would require spontaneous waves (here, scaled from ferret Feller et al. 1996 to human retina) that occur in a symmetric fashion across the vertical meridian (dashed line) of one eye. Alternatively, symmetric column patterns could result if the waves in the temporal retina of one eye and the nasal retina of the other eye were coordinated for all visual field loci. Because the waves are random and independent between the two eyes, neither requirement is feasible.

Figure 20

A reciprocal exchange of territory accompanies the formation of angioscotoma representations in animals with ocular dominance columns, but does not occur in those lacking columns. (a) Boxed region from figure 9, showing a dark angioscotoma from an animal with fine columns. It is hardly visible in the raw image at this scale, emphasizing that CO patterns in tissue sections often reflect only a small difference in optical density. The optical density plotted below shows the mean value of each pixel column. Numbers refer to level of grey (0–255) and parentheses indicate optical density of neutral density filters used for calibration (Horton & Hocking 1998). Frame below shows the same image, normalized to full grey scale range, to make the angioscotoma more readily visible. Note the pale zones flanking the angioscotoma, reflected in the inverted Mexican hat profile of the optical density plot. These surrounding zones of opposite contrast are served exclusively by the other eye, whose geniculocortical afferents have been driven out of the cortex corresponding to the angioscotoma representation. (b) A dark angioscotoma representation, from an animal with essentially no ocular dominance columns, without flanking zones dedicated entirely to the other eye (Adams & Horton 2002).