Role of Subplate Neurons in Functional Maturation of Visual Cortical Columns

See allHide authors and affiliations

Science  25 Jul 2003:
Vol. 301, Issue 5632, pp. 521-525
DOI: 10.1126/science.1084152


The subplate forms a transient circuit required for development of connections between the thalamus and the cerebral cortex. When subplate neurons are ablated, ocular dominance columns do not form in the visual cortex despite the robust presence of thalamic axons in layer 4. We show that subplate ablation also prevents formation of orientation columns. Visual responses are weak and poorly tuned to orientation. Furthermore, thalamocortical synaptic transmission fails to strengthen, whereas intracortical synapses are unaffected. Thus, subplate circuits are essential not only for the anatomical segregation of thalamic inputs but also for key steps in synaptic remodeling and maturation needed to establish the functional architecture of visual cortex.

Subplate neurons, located in the developing white matter (WM), are among the first postmitotic cortical neurons (1). They are also first to receive functional synaptic inputs from the thalamus; their axons relay this input into the cortical plate well in advance of the invasion of thalamic axons into layer 4 (1). Once thalamic axons have arrived in layer 4, subplate neurons are gradually eliminated during the period of ocular dominance column (ODC) formation (13). Thus, subplate neurons are in a key, intermediate position to control the flow of information into the developing cortex when first spontaneous (prenatal) and then visual (postnatal) activity are present (3). Early subplate ablation prevents the invasion of thalamic axons into layer 4 (4), whereas later ablation blocks the anatomical segregation of thalamic axons from the lateral geniculate nucleus (LGN) into ODCs in primary visual cortex (5).

The prevailing hypothesis about ODC formation and plasticity is that activity-dependent competition between LGN axons representing each eye leads to their selective growth or pruning (3). However, recent observations that ODCs emerge earlier than previously thought, even before the onset of patterned visual experience (2, 6), have required a revision of this hypothesis, and one suggestion is that subplate neurons are involved (2).

To determine how the subplate might influence functional development and organization of visual cortex, subplate neurons were selectively ablated at P6 to P9 in cats (7). This time is just before the onset of ODC formation (6, 8, 9) and just after the ingrowth of LGN axons to cortical layer 4 (10). To make localized and selective ablations, we either used kainic acid injections (4, 5) or an immunotoxin (11, 12) to p75, a neurotrophin receptor expressed in the neocortex only by subplate neurons at this age (1).

Response properties of cortical neurons were evaluated after subplate ablation by optical imaging and microelectrode recordings at P24 to P49 (7), when well-organized orientation maps are normally present (6, 13, 14). Orientation maps in control hemispheres or far from the ablation were highly organized (Fig. 1A, left and center), in contrast to poor organization in the highly local region surrounding the immunotoxin-ablated area (Fig. 1A, center). Kainic acid injections created even larger regions of disrupted or absent orientation maps (Fig. 1A, right). Polar maps representing both the angle and strength of tuning revealed weaker orientation selectivity in the ablated region (Fig. 1B), indicating the existence of an additional severe functional deficit. ODC maps were also degraded in ablated hemispheres (fig. S1), consistent with previous anatomical data (5).

Fig. 1.

Subplate ablation disrupts orientation maps and prevents sharpening of orientation tuning. (A) In vivo optical imaging of orientation maps in control and ablated visual cortex. Color indicates preferred orientation at each pixel. Scale bars indicate 1 mm. Each immunotoxin or kainate injection site is marked with an “x.” Compared with control hemisphere (left), which shows robust orientation domains and pinwheels, maps in subplateablated hemispheres at P48 after immunotoxin injections at P8 and at P37 after kainate at P9 are disorganized in regions receiving ablations. Immunotoxin treatments (n = 2 animals) are more focal (area above dashed line) than with kainate (n = 5 animals). Optical maps after sham injections (saline) were indistinguishable from that of the control [n = 1 animal (29)]. (B) Polar maps showing both preferred orientation (color) and tuning strength (brightness). Strong orientation-selective responses (bright colors) and pinwheels are evident in unablated areas. Orientation maps are nonselective (dark) and unorganized in ablated cortex. (C and D) Summary histograms of orientation-tuning strength (OTS) (7) of single units in ablated (C) and control (D) cortex (0, unselective; 1, perfectly selective). Compared to control (0.48 ± 0.25, n = 31), OTSs of all units (including unresponsive units, in red) were reduced by 82% in ablated cortex (0.09 ± 0.15, n = 62, P < 0.00001). Compared to those of controls (0.48 ± 0.25, n = 31), OTSs in ablated cortex of just the visually driven units (black) were reduced by 44% (0.27 ± 0.24, n = 22, P < 0.005). Insets show typical OTS examples.

Degraded orientation maps can result from neurons well tuned to different orientations but not segregated into columns, from neurons poorly tuned for orientation, or simply from neurons that respond poorly to visual stimuli. To distinguish between these alternatives, ocular dominance and orientation tuning of single cortical neurons were measured with microelectrode recordings (7) in both ablated and control regions in each animal (Fig. 1A and fig. S2). Because LGN axons are abundant in layer 4 of subplate-ablated cortex (Fig. 2A) (5), we had expected to find strong visually driven responses. Instead, there were fewer visually responsive units in the ablated area (15), and they were less orientation-selective (Fig. 1, C and D). In addition, visually driven units in ablated cortex were only weakly driven by the ipsilateral eye, resulting in a strong response bias toward the contralateral eye (15). This bias is reminiscent of normal visual cortex at earlier developmental ages (6, 13), suggesting arrested development.

Fig. 2.

Immunotoxin ablation of subplate neurons prevents segregation of LGN axons into ODCs and upregulates expression of BDNF in cortex. (A) Darkfield autoradiograph of visual cortex at P82 after four injections (indicated by asterisks) of immunotoxin at P9 (two posterior injections of 0.25 mg/ml and two anterior injections of 0.5 mg/ml). ODCs were visualized by transneuronal transport after monocular injection of 3H-proline (7). (Silver grains, which label LGN axons and axon terminals, appear white in darkfield autoradiograph.) Ablation extended ∼1 mm from each injection site, consistent with reported data (11). Robust, continuous transneuronal labeling (between arrowheads) in layer 4 signifies absence of ODCs above the subplate ablation zone as compared to patchy labeling of ODCs in control regions. Control immunotoxin treatment does not disrupt ODCs (fig. S3). (B) In situ hybridization (darkfield autoradiograph) shows increased BDNF mRNA expression in cortex at P16 after four immunotoxin (1 mg/ml) injections (asterisks) into subplate at P9. Boxes indicate locations of (C) and (D). (C and D) BDNF mRNA is abnormally increased in layers 2/3 and 4 in subplate-ablated (D) but not adjacent control (C) cortex. Normally, at this early age, BDNF mRNA expression is restricted to the subplate–layer 6 border (30). Cortical layers and WM indicated on right. (E) Histology of cortical plate (CP) at P28 (cresyl stain) appears normal after immunotoxin (1 mg/ml) ablations at P7. The presence of fluorescent microspheres (coinjected with immunotoxin) in the WM is indicated (above asterisks). (F and G) Immunohistochemistry with the use of NeuN [a neuron-specific marker (Chemicon, Temecula, CA)] confirms selective loss of subplate neurons at injection sites [indicated by asterisks in (G)]. Scale bars, 1 mm [(A) and (B)], 0.2 mm [(C), (D), and (E)], and 80 μm [(F) and (G)].

Alterations in cortical organization, as measured by several anatomical and molecular correlates, were also similar after immunotoxin or kainate ablations. In the immunotoxin-ablated region, uniform transneuronal labeling showed that ODCs were absent despite a robust projection from the LGN to layer 4 (Fig. 2A), as with kainate ablations (5); ODCs were present in immediately adjacent unablated regions. Brain-derived neurotrophic factor (BDNF) mRNA (7) also increased in layers 2 to 4 of cortex located above the immunotoxinablated region (Fig. 2, B and D), as occurs after kainate ablations (16).

Although immunotoxin ablation altered geniculocortical connectivity, gene expression, and columnar functional organization of cortex, the gross histology of cortical layers residing above the ablation site was indistinguishable from that of unablated adjacent regions (Fig. 2E). Cortical p75 is almost exclusively restricted to the subplate (1), but cholinergic fibers from basal fore-brain are also known to express p75 (17). Therefore, we also verified that cholinergic fibers were abundant in immunotoxinablated regions (fig. S4). The only obvious histological effect seen after immunotoxin injection was the loss of neurons in WM, as expected because subplate neurons are normally located there (Fig. 2, F and G).

In summary, both immunotoxin and kainate treatments have similar consequences for cortical development yet act through completely different mechanisms. Thus, observed changes are specifically because of removal of subplate neurons rather than nonspecific effects of the injections (18).

One explanation for the functional deficits in visual cortex is that LGN neurons are affected by subplate ablation. When we recorded in the LGN, however, neurons projecting to the ablated site had responses and receptive field organization (Fig. 3A) appropriate for this age (19). Thus, observed defects in orientation tuning and visual responsiveness must occur beyond the LGN.

Fig. 3.

Subplate ablation alters thalamocortical but not intracortical field potentials. (A) Typical LGN receptive fields (P48) that retinotopically overlapped the visual cortex receiving immunotoxin ablations at P8. Receptive-field size and organization are within normal range for these ages (19). (B) In vivo cortical field potentials after electrical stimulation of LGN at matching retinotopic locations (vertical arrows mark stimulus onset). Fields are smaller in ablated compared with control regions throughout all layers (layer 4 is between green arrows). (C) Field recordings in vitro from ablated versus control hemispheres in cortical slices at P30 (ablated at P8) (7). WM stimulation evokes large field potentials in layer 2/3 and layer 4 in control slices (n = 16), whereas only small field potentials are present in these layers in ablated slices (n = 24). Stimulus artifacts are removed. (D) First-order current-source density (CSD) calculated from field potentials recorded in slices after WM stimulation. In control slices (left), a large monosynaptic sink is present at ∼1.5-ms latency in layer 4 (green arrow). Large, longer latency sinks are visible in layers 2/3 and 5/6. In contrast, no layer 4 sinks are present in ablated hemispheres (right) (CSD variance of 820 ± 1600 μV2/mm4 versus 3.4 ± 2.5 μV2/mm4, P < 0.002, Mann-Whitney). (E) Electrical stimulation of layer 4 in slices from ablated or control hemispheres produces monosynaptic field potentials in layer 2/3, demonstrating intact transmission from layer 4 to layer 2/3 even in ablated regions (n = 15).

To examine whether the cortical defect originates at thalamocortical synapses, we electrically stimulated the LGN while we recorded evoked field potentials at a retinotopically matching cortical location in vivo. Field potentials in subplate-ablated regions were smaller in all layers, particularly layer 4, than those in control regions (Fig. 3B), consistent with only weak activation of thalamocortical synapses. These results were confirmed by in vitro recordings from slices of visual cortex (7). Field potentials recorded in layers 2/3 and 4, evoked by WM stimulation, were vastly reduced in the ablated regions (Fig. 3C). The size of the monosynaptic response in layer 4 was significantly smaller, implying very weak thalamocortical synapses (Fig. 3D). In contrast, evoked potentials in layer 2/3 after layer 4 stimulation were close to normal in amplitude and waveform in ablated regions (Fig. 3E).

Histological observations and in vivo recordings indicate that layer 4 neurons are present after subplate ablation. Whole-cell patch recordings performed on layer 4 neurons in slices (8) confirmed that neurons from ablated regions fired action potentials to current injections (Fig. 4A), had normal measures of intrinsic membrane properties (Fig. 4B), and appeared to have normal dendritic morphology (Fig. 4A, inset). The attenuated field potentials after subplate ablation could reflect a lack of spatial organization and neuronal synchrony in subplate-ablated cortex or an underlying attenuation of synaptic currents. Direct measurement of the size of excitatory synaptic inputs to layer 4 neurons showed that excitatory postsynaptic currents evoked by electrical WM stimulation (eEPSCs) were reduced by 80% in subplate-ablated slices compared to those of controls (Fig. 4C; see also fig. S5).

Fig. 4.

Subplate ablation alters thalamocortical synaptic efficacy in layer 4 neurons. (A) Whole-cell recording of a layer 4 neuron in a subplate-ablated cortical slice (7). Traces show responses to multiple current injections. Inset shows a biocytin-filled layer 4 neuron with normal morphology located in ablated region. (B) Neurons from control (“c”) as compared to ablated (“a”) slices have similar resting potential (–63.9 ± 5.2 mV for control, n = 24, as compared to –65.4 ± 7.3 mV for ablated, n = 20; P > 0.1) and input resistance (217.8 ± 65.4 MΩ for control, n = 24, as compared to 225.3 ± 125.9 MΩ for ablated, n = 20; P > 0.1). (C) Cumulative amplitude histograms show that WM stimulation evokes eEPSCs in ablated or control slices. The eEPSC amplitude (inset) is 80% smaller in ablated slices (29.3 ± 54.6 pA for ablated, n = 22, as compared to 148.3 ± 102.9 pA for control, n = 22; P < 0.0001, double asterisks). Adjacent traces show mean eEPSCs of representative neurons from each population. eEPSC amplitudes from saline and control-immunotoxin slices are indistinguishable from control (fig. S5). (D) Darkfield autoradiograph of in situ hybridization for GluR1 mRNA in control (left) or ablated (right) cortex. Scale bar, 0.2 mm. Densitometry (histogram) shows that layer 4 GluR1 mRNA levels in ablated regions are reduced to 35% of that of controls (P < 0.005, double asterisks, n = 4 animals). GluR1 mRNA levels are unchanged after layer 4 kainate injections (n = 2 animals) (fig. S6) or saline injections into the subplate [n = 1 animal (29)]. (E) Cumulative histograms and bar graphs (inset) show that mEPSC amplitudes are similar (8.6 ± 2.4 pA for ablated, n = 13, compared to 8.0 ± 2.6 pA for control, n = 13; P > 0.1). (F) Cumulative mEPSC frequency histograms and bar graph (inset) show that mEPSC frequency is 3.6-fold higher in ablated slices (2.34 ± 1.34 Hz for ablated, n = 13, compared to 0.65 ± 0.08 Hz for control, n = 13; P < 0.001, double asterisks).

Because thalamocortical axons are present in layer 4 (for example, Fig. 2A) and LGN neurons have normal visual responses, it is possible that the attenuation in thalamocortical transmission is because of a postsynaptic defect. In fact, mRNA expression of the dominant AMPA receptor subunit, GluR1, is selectively reduced by 65% within layer 4 in the ablated area (Fig. 4D), implying that the maturation of layer 4 synapses is impaired. In marked contrast to the decreased eEPSC amplitudes after subplate ablation (Fig. 4C), the amplitudes of the spontaneous miniature EPSCs (mEPSCs) are unchanged in layer 4 neurons (Fig. 4E), and their frequency actually increased (Fig. 4F). The observed increase in spontaneous activity in the face of decreased thalamic inputs is consistent with the presence of a homeostatic mechanism controlling overall synaptic input (20). Taken together, recordings from layer 4 in vivo and in vitro indicate that layer 4 neurons are present and can receive intracortical synaptic input (as evident from the mEPSCs) but that there has been a selective loss of synaptic drive from the thalamus.

The major finding of this study is that subplate ablation leads to an impairment in synaptic transmission of visually driven activity from the LGN into cortical circuits. Although LGN axons are present in layer 4, the cortex becomes essentially uncoupled from the thalamus, as assessed both in vivo and in vitro. In addition, functional ODCs and orientation maps do not develop [Supporting Online Material (SOM) Text]. Although the majority of neurons were unresponsive to visual stimuli, some cells did have robust visual responses but nevertheless were not orientation-selective. This global absence of functional tuning of cortical maps after subplate ablation is related to the failure of thalamocortical synapses to strengthen and the abnormally low expression in layer 4 neurons of key molecular components mediating glutamatergic transmission, such as GluR1.

How can subplate ablation lead to the observed functional decoupling of thalamus from cortex? Although increased feed forward inhibition could mask thalamocortical EPSCs, our observation of increased mEPSC frequency and absence of large inhibitory postsynaptic currents makes this hypothesis unlikely. We favor the hypothesis that depolarizing input from subplate neurons is needed for the activation of previously “silent” thalamocortical synapses (21) because subplate neurons supply excitatory input to cortical neurons (22, 23). In this case, some form of learning rule, known to be present at thalamocortical synapses (24, 25), is disrupted in the absence of subplate neurons, preventing the progressive strengthening of thalamic synapses. The observation here that GluR1 mRNA is decreased after subplate ablation implies that excitatory drive on all cortical neurons, including inhibitory neurons, is also diminished.

It is remarkable that eye segregation fails even though LGN receptive fields are normal and both LGN axons and layer 4 neurons are present in subplate-ablated cortex. Our observations constrain mechanisms of patterning of connectivity in cortical circuits. It has been suggested that thalamocortical connectivity during the critical period depends on cortical circuits outside of layer 4 (26). In contrast, our findings show that patterning of thalamocortical connectivity before onset of the critical period crucially depends on proper LGN input to the cortex, as regulated by subplate neurons. Furthermore, Crowley and Katz have proposed that eye-specific molecular cues set up the initial pattern of ODCs during development (2, 27), with activity being required only at later ages. If so, then our results argue that these hypothetical molecular cues in cortex reside not in layer 4 or LGN axon terminals, both of which are present in the subplate-ablated cortex, but instead in the subplate. Alternatively, subplate neurons could mediate very early activity-dependent interactions driven by correlated spontaneous activity relayed from the retina to the cortex by LGN neurons, which is also known to be present in corticogeniculate circuits (3, 28). Regardless of which alternative proves to be the case, our observations here demonstrate that circuits formed by subplate neurons are needed both for strengthening and for selective remodeling of thalamocortical excitatory synaptic connections that underlie the development of columnar organization in cortex.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S6

References and Notes

Stay Connected to Science

Navigate This Article