The posterior parietal cortex contributes to visuomotor processing for saccades in blindsight macaques

Patients with damage to the primary visual cortex (V1) lose visual awareness, yet retain the ability to perform visuomotor tasks, which is called “blindsight.” To understand the neural mechanisms underlying this residual visuomotor function, we studied a non-human primate model of blindsight with a unilateral lesion of V1 using various oculomotor tasks. Functional brain imaging by positron emission tomography showed a significant change after V1 lesion in saccade-related visuomotor activity in the intraparietal sulcus area in the ipsi- and contralesional posterior parietal cortex. Single unit recordings in the lateral bank of the intraparietal sulcus (lbIPS) showed visual responses to targets in the contralateral visual field on both hemispheres. Injection of muscimol into the ipsi- or contralesional lbIPSs significantly impaired saccades to targets in the V1 lesion-affected visual field, differently from previous reports in intact animals. These results indicate that the bilateral lbIPSs contribute to visuomotor function in blindsight.

For the analysis, all centrifugal saccades from the correct trials were collected and grouped into 4 types by saccade direction (saccade to intact/intact-to-be or affected/affected-to-be visual field) and period (pre-lesion or post-lesion). We tested whether the saccadic reaction times (SRTs) and saccade endpoints were significantly different, especially when the monkeys made saccades to the target in their affected visual field (left visual field for monkey C, right visual field for monkey T, after lesioning). Both monkeys exhibited longer SRTs after lesioning regardless of the direction of saccades compared to those before lesioning (significant main effect of lesioning by 2-way analysis of variance on rank; monkey C: F(1, 20320) = 1904.99, p < 10 -10 ; monkey T: F(1, 19724) = 4726.53, p < 10 -10 , a). The interaction effect was significant in monkey C (F(1, 20320) = 264.99, p < 10 -10 ), but not in monkey T (F(1, 19724) = 3.51, p = 0.06). Saccades to the affected field had significantly shorter SRTs in monkey C, but longer SRTs in monkey T, compared to saccades directed toward the intact visual field (p < 10 -10 , post-hoc Mann-Whitney test with Bonferroni's correction). Both monkeys showed significant effect of direction of saccade, lesioning and interaction effects on saccade endpoints (2-way analysis of variance on rank; monkey C: F(1, 20320) = 134.28 (direction of saccade), 323.7 (lesioning), 2872.52 (interaction), p < 10 -10 for all; monkey T: F(1, 19724) = 10.16 (direction of saccade), 219.17 (lesioning), 24.03 (interaction), p < 0.005 for all; b), but there was no consistent tendency for saccade to the affected field among two monkeys, other than the endpoint variability (c). Asterisk (*) indicates significant difference (p <

Supplementary Figure 3. Whole brain PET analysis: individual subjects.
Brain areas with a significant relationship to task condition analyzed separately for individual monkeys for each period (pre-and post-lesion). The left and right sides of the brain were flipped to match the side of the lesion (the lesioned side is presented on the right side in this figure). Red-yellow and blue-light blue show a significant positive and negative relationship to task condition, respectively. Only clusters larger than 200 voxels are shown (uncorrected p < 0.01). a and b show the results of monkey T in the pre-and post-lesion periods, respectively. c and d show the results of monkey C in the pre-and post-lesion periods, respectively.
We also analyzed individual monkeys separately to study further the brain areas that are related to task condition. The data were analyzed in the same manner described previously, except that the 4D data were aligned separately for each monkey using rigid body transformation. contralesional (blue) lbIPS. The mean of each group of lbIPS neurons is indicated as a filled box for the latency (p < 0.05 by Welch's non-paired t-test. ipsilesional lbIPS: n = 7, contralesional lbIPS: n = 9. Error bars indicate SD) and a filled circle for the visual response magnitude (p > 0.05 by non-paired t-test. ipsilesional lbIPS: n = 7, contralesional lbIPS: n = 9. Error bars indicate SD) with the corresponding color.

Supplementary Figure 6. Presaccadic activity of ipsilesional lbIPS neurons.
Two examples of ipsilesional lbIPS neurons recorded in monkey C (a and b) and monkey U (c and d), which showed presaccadic activity in the step saccade task. a and c show the spike rastergram. b and d show the averaged spike density functions. In case of the left panels, the data are aligned to the target onset. In case of the right panels, the data are aligned to the saccade onset. Red dots and lines indicate the neural activity when the target was presented in their response field (affected visual field), while blue dots and lines indicate the neural activity when the target was presented outside the response field (intact visual field). In case of the neuron in a and b, the peak of phasic activity starting immediately before the saccade onset was sharper and higher when the data are aligned to the saccade onset. In case of the neuron in c and d, the second peak emerged when the data are aligned in the saccade onset and started before the saccade onset. Therefore, both neurons could be considered to exhibit the presaccadic activity. Differences between each distribution of direction error were tested by the Steel-Dwass test (by R package: NSM3 or "http://aoki2.si.gunma-u.ac.jp/R/src/Steel-Dwass.R", encoding="euc-jp"). In this test, the dataset of each target direction in the affected visual field was compared with other 5 datasets (3 target directions in the intact visual field and affected visual field, we focused on the comparison with 3 target directions in the intact field. We conducted the test on 3 target groups for each 2 eccentricities (10 and 15°) in 2 animals, that is, totally 12 target groups. It was found that direction errors of saccades towards 6/12 target groups in the affected visual field were larger than all of three target groups in the intact visual field with same eccentricity (*▲: Steel-Dwass test, p < 0.05).
In one case with *▽ the direction error in the affected visual field was smaller than those on the intact side. f, h, j and l. Distribution of saccade direction error in the affected visual field was compared with those in the intact visual field also using the IQR and its 95% confidence intervals. IQR (blue dot) and the 95% confidence interval (black line) of each dataset in a, b, c and d, are plotted. #▲ placed right to the IQR on the affected side indicated that the IQR was significantly larger than all upper edges of 95% confidence interval of three target groups on the intact side (10/12). All these results fit our previous study which showed saccades became less accurate after V1 lesion (Yoshida et al, 2008) and our current datasets obtained in the round saccade task (see Supplementary Figure 2 The PPI searched the areas whose rCBF values are correlated with that of a seed region (e.g., ipsi-or contralesional IPS) and associated by the other factors (e.g., task and lesion).
Seed regions of ipsi-and contralesional IPS were defined by the 5x5x5 voxels around local maxima of Post > Pre contrast in task relationship (as shown in Fig.4c-d). The firstlevel PPIs were analyzed using FEAT, separately for each subject, period, and seed region.
Regional CBFs were regressed by variables of task condition (1, 2, 3, 4, or 6) and rCBF values in seed region (ipsi-or contralesional IPS), and an interaction between task condition and seed rCBF. We also used a regressor for global signal to remove its effect. statistically determined in Fig.7 and Supplementary Figures 7 and 8 using the Silverman's test. Here, we focused on the on-target saccades. The saccades in the mode judged as being remote from the target location by the Silverman's test in multimodal case were considered as "off-target" saccades and the remaining saccades were judged as "ontarget". If the distribution was unimodal, all the saccades were regarded as on-target.
Supplementary Table 2 and 3 show the results comparing the on-target saccade endpoint errors before and after the muscimol injection (Supplementary Table 2: direction error, Supplementary Table 3: eccentricity error), focusing on on-target saccades. Here, even after excluding those off-target saccades, the direction and the eccentricity of saccades became less accurate after the inactivation in 4 out of 16 experiments and 6 out of 16 experiments, respectively (indicated by * in Supplementary Table 2 and Supplementary   Table 3). The effects were observed in 4 experiments in monkey C (direction error: contC1 and contC2: eccentricity error: ipsiC1 and contC3), and 4 experiments in monkey U (both errors: ipsiU4 and contU2; eccentricity error: ipsi U3 and contU3).