Blindsight depends on the lateral geniculate nucleus

Abstract

Injury to the primary visual cortex (V1) leads to the loss of visual experience. Nonetheless, careful testing shows that certain visually guided behaviours can persist even in the absence of visual awareness. The neural circuits supporting this phenomenon, which is often termed blindsight, remain uncertain. Here we demonstrate that the thalamic lateral geniculate nucleus (LGN) has a causal role in V1-independent processing of visual information. By comparing functional magnetic resonance imaging (fMRI) and behavioural measures with and without temporary LGN inactivation, we assessed the contribution of the LGN to visual functions of macaque monkeys (Macaca mulatta) with chronic V1 lesions. Before LGN inactivation, high-contrast stimuli presented to the lesion-affected visual field (scotoma) produced significant V1-independent fMRI activation in the extrastriate cortical areas V2, V3, V4, V5/middle temporal (MT), fundus of the superior temporal sulcus (FST) and lateral intraparietal area (LIP) and the animals correctly located the stimuli in a detection task. However, following reversible inactivation of the LGN in the V1-lesioned hemisphere, fMRI responses and behavioural detection were abolished. These results demonstrate that direct LGN projections to the extrastriate cortex have a critical functional contribution to blindsight. They suggest a viable pathway to mediate fast detection during normal vision.

Figure 1

Experimental setup. a. The upper part shows a side view on the right hemisphere of an inflated macaque brain. An area of ~ 400 mm² of gray matter in the opercular part of primary visual cortex (V1) representing the visual field between ~2° and 7° has been surgically aspirated and is shown in black. Extrastriate areas, the subject of analysis in this study, are hidden in the sulci surrounding V1, including the lunate (LS), inferior occipital (IOS), and superior temporal (STS) sulci. To facilitate the visual examination of extrastriate cortex, the occipital lobe (dashed circled area in the upper panel) was cut and flattened (lower panel). b. Axial sections of monkey 1’s (upper panel) and monkey 2’s (lower panel) occipital lobes at the position indicated by the green dashed line in panel a. The outward borders of white and gray matter are highlighted by green and orange dotted lines, respectively. The lesions are evident by the absence of gray matter (red line markers). c. To compare visually elicited responses in extrastriate cortex in the presence versus absence of V1 input, rotating checkerboard stimuli were spatially restricted (2° diameter) and presented either inside (upper panel) or outside (lower panel) the scotoma (part of the visual field affected by the V1 lesion, indicated here by red circles).

Figure 2

Visual processing in V1 lesioned monkeys. a. Functional activation map of macaque 1’s non-lesioned visual cortex to 85 cycles of visual stimulation outside the scotoma (figure 1C, lower panel). The map has been horizontally flipped for easier comparison with the lesioned hemisphere. White dotted and solid lines show the position of the vertical and horizontal meridian representations, respectively, derived from independent retinotopic mapping experiments (supplementary figures 1, 2) to reveal the functional boundaries of extrastriate areas . b. Activation map of macaque 1’s lesioned hemisphere to 85 visual stimulation cycles inside the scotoma (figure 1C, upper part). The position of the stimulus inside the scotoma was effective, in that lesion surrounding V1 cortex with intact gray matter was not activated. In the absence of V1 input, areas V2/V3, V4 and V5/MT continue to be visually responsive. c. Behavioral performance of monkey 1 in detecting visual stimuli (0.2° diameter) presented inside (red line) or outside (green line) the scotoma at different luminance contrast levels compared to a constant gray background. On one third of the trials no stimulus was presented and the monkey was rewarded for maintaining central fixation (blue line). Data represent mean ± sem from five experiments. d. Functional activation map of monkey 2’s non-lesioned hemisphere to 95 visual stimulation cycles. e. Activation map of monkey 2’s lesioned hemisphere to 95 visual stimulation cycles inside the scotoma. f. Behavioral performance of monkey 2 for detecting visual stimului inside (red line) versus outside (green line) the scotoma or during catch trials (blue line). Data represent mean ± sem from five experiments. Although both monkeys display a large visual deficit, visual information continues to be processed to some extent as performance improves with stimulus contrast.

Figure 3

Role of the LGN in driving V1-independent visual processing. a. Inactivation of the LGN was achieved by injecting the GABA-A agonist THIP (methods). The drug was co-injected (total volume 2 µl, methods) with the diamagnetic MR contrast agent Gadolinium (Gd) to visualize the injection in MR images. Here, a coronal section through the posterior part of monkey 1’s LGN (~ AP +7 mm) is shown. Injection of Gd resulted in a localized increase in the intensity of the MR signal with a diameter of ~3 mm (red arrow). Reproducible injections across experiments were achieved by permanently implanting a MR-compatible cannula (yellow arrow). b. Functional activation map of macaque 1’s left, lesioned hemisphere to visual stimulation inside the scotoma (35 stimulation cycles) during inactivation of the LGN. LGN inactivation results in the elimination of V1-independent visual responses (figure 2 b). c. Monkey 1’s performance in detecting visual targets at different luminance contrasts. Data represent mean ± sem performance from three experiments with THIP injections into the LGN. The injections eliminated the monkey’s ability to detect a target inside the scotoma. d. Inactivation of macaque 2’s posterior LGN. e. Activation map of macaque 2’s right, lesioned hemisphere to visual stimulation inside the scotoma (60 stimulation cycles) during LGN inactivation. f. Monkey 2’s performance for correctly detecting targets during LGN inactivation. Data represent the mean ± sem performance in three experiments with THIP injections.

Figure 4

Quantitative summary of mean fMRI activation levels in extrastriate areas under normal conditions (V1 and LGN intact, blue bars), in the absence of V1 input (lesion, red bars), and in the absence of input from V1 and the LGN (les. + inj., green bars). Data in the upper part were collected from monkey 1 and correspond to the mean ± sem t-statistic obtained during 85 visual stimulation cycles without LGN inactivation and 60 stimulation cycles with LGN inactivation. Data in the lower part come from experiments with monkey 2, in which the mean ± sem t-statistic has been computed over 95 stimulation cycles without LGN inactivation and 20 stimulation cycles with LGN inactivation. Note that on average across areas and monkeys, fMRI activation in extrastriate areas is reduced by ~80% when V1 input is missing. Additional LGN inactivation reduces activity by more than 95% compared to normal levels.