Demographics of the participants is given in Table 1. Participants with extensive visual field (VF) defects due to advanced open-angle glaucoma (GL; n = 4) or to retinitis pigmentosa (RP; n = 3; RP2 also had secondary glaucoma, see Table 1) and age-matched visually healthy controls (HC) with normal vision [best-corrected decimal visual acuity ≥ 1.0 (Bach, 1996); n = 7] took part in the study. Written informed consents and data usage agreements were signed by all participants. The study was conducted in adherence to the tenets of the Declaration of Helsinki and was approved by the ethics committee of the University of Magdeburg.

Standard automated perimetry (SAP) was performed using 24–2 Swedish Interactive Threshold Algorithm protocol (SITA-Fast; Goldmann size III white-on-white stimuli; Humphrey Field Analyzer 3; Carl Zeiss Meditec AG; Jena, Germany). For three participants, VF to the central 30° was tested using another perimeter (OCTOPUS^® Perimeter 101, Haag-Streit International, Switzerland; dG2; dynamic strategy; Goldmann size III). For the patient cohort (except 1; GL3), fixation stability was determined with a fundus-controlled microperimeter (MP-1 microperimeter, Nidek, Padua, Italy) during fixation of a central fixation target. Fixations were tracked with 25 Hz and the proportion of fixations falling within the central 2° radius was determined using built-in MP1 analysis (Table 1).

All MRI and fMRI data were collected on a 3 Tesla Siemens Prisma scanner (Erlangen, Germany). In order to allow for an unrestricted view of the entire projection screen, we used only the lower section of a 64-channel head coil, resulting in a 34-channel coil covering most of the brain. fMRI scans parallel to the AC-PC line were acquired using a T2^∗-weighted BOLD gradient-EPI sequence (TR | TE = 1,500 | 30 ms and voxel size = 2.5³ mm³). A total of 160 fMRI time series images (volumes) were obtained for each run after the removal of the first 8 volumes by the scanner itself to allow for steady magnetization. The fMRI parameters were the same for the pRF mapping data, except for the number of volumes (136). One high-resolution whole brain anatomical T1-weighted scan (MPRAGE, 1 mm isotropic voxels, TR | TI | TE = 2,500 | 1,100 | 2.82 ms) was collected for each participant to allow for cortical visualization of fMRI responses. An inversion recovery EPI sequence (TR | TI | TE = 4,000 | 1,100 | 23 ms) with spatial coverage (FOV) and resolution identical to the T2^∗ EPI was obtained to aid in the alignment of structural and functional data.

Gray-white matter boundaries in the T1-weighted anatomical images were segmented using Freesurfer¹. ITK-gray was used to manually inspect the Freesurfer segmentation and correct for possible segmentation errors². A 3-D rendering of the cortical surface was generated by reconstruction of the segmented boundaries (Wandell et al., 2000). Within and between fMRI scans, head motion artifacts were corrected using AFNI³. For each participant, motion-corrected fMRI time series of the repetitions of each stimulation condition (i.e., PV, OBT, FDT, and pRF mapping data) were averaged into separate groups with MATLAB-based Vistasoft tools (mrVista⁴). The inversion recovery EPI was aligned spatially with the anatomical scan in two steps; first manually with rxAlign function in mrVista and then automatically using Kendrick Kay’s alignment toolbox⁵. The obtained alignment matrix was used to align the fMRI images with the anatomy.

Firstly, as a validation and replication of the existing literature, we examined the scope of aberrant cortical responses in the deafferented visual cortex of a reference group of three participants with RP. Consistent with the results from Masuda and colleagues (Masuda et al., 2010) in their cohort of three RP participants primary visual cortex (V1), we found task-dependent activity in the peripheral LPZ in V1 (OBT-PV: t = 16.8; p = 0.012), and non-significant trends in V2 (OBT-PV: t = 6.2; p = 0.075) and V3 (OBT-PV: t = 3.7; p = 0.198) (see Figure 2, Supplementary Figure 1, and Supplementary Table 1).

Next, we explored task-dependent LPZ responses in participants with advanced VF defects due to glaucoma. For a qualitative assessment, the fMRI-responses in the visual cortex of a representative participant with glaucoma (GL1) and a healthy control (HC1) with simulated peripheral VF-defect are shown in Figure 3 for three different stimulation conditions (PV, OBT, FDT). In the healthy control, LPZ responses were largely similar for all stimulation conditions, i.e., negative coherence_ps and BOLD modulations. In contrast, LPZ responses in the glaucoma patient resulted in positive coherence_ps and BOLD modulations for OBT, while negative coherence_ps and negative BOLD modulations, as for the control, were obtained only for PV and FDT. Taken together, the response signatures for the participant with glaucoma resembled those reported for RP (Supplementary Figure 1).

In a quantitative assessment, the above task dependence of the LPZ-responses in glaucoma was further confirmed at the group level. Each participant’s mean phase-specified coherence (coherence_ps) was extracted from NPZ and LPZ voxels for each of the three stimulation conditions. The mean coherence_ps of the three participant groups is depicted for V1 in Figure 2. For all participant groups, in NPZ we observed a strong positive coherence_ps, which was enhanced for OBT. In contrast, in LPZ the negative coherence_ps for PV turned positive for OBT in the glaucoma and RP group, while it remained negative, albeit reduced, for the control group. This differential effect between patient and control groups in LPZ was also directly evident from the inspection of the average single-cycle BOLD time series in LPZ. As depicted in Figure 4 for the group averages, only for glaucoma and RP did negative responses during stimulation (first 12 s) for PV shift to positive responses for OBT. As a consequence the BOLD-peak for OBT shifts from the second block (12–24 s) in healthy controls to the first block (first 12 s) in RP and glaucoma.

The significance of the task effect on V1 coherence_ps and its difference across participant groups were assessed with two-way repeated measures ANOVA [between-subject factor: group (glaucoma, RP, control); within-subject factor: task (PV, OBT, FDT)] for the LPZ and NPZ separately. The effect of task was significant for both LPZ and NPZ (p < 0.001; see Supplementary Table 1), but the effect of group was significant only for the LPZ (p < 0.05). No significant interactions (task × group) were observed. Post hoc tests (corrected) were performed to identify the significant comparisons as indicated in Figure 2A. To assess the task-dependence of LPZ responses in glaucoma, the comparisons of the conditions OBT and PV in the control and glaucoma group are of particular relevance: for glaucoma there was a significant difference (t = 5.4; p = 0.024), but not for the controls. This effect was also significant in glaucoma V3 (t = 10.2; p = 0.006), while the same trend in V2 (t = 2.8; p = 0.069) failed to reach significance (Figures 2B,C). We further tested in glaucoma for any hierarchical effects of task-dependence in V1, V2, and V3 by calculating the effect size (computed as Cohen’s d) of the observed task-related coherence_ps modulations in LPZ. We observed a stronger effect size in V1 (d = 0.70) and relatively lower effect size for V2 (d = 0.50) and V3 (d = 0.51). However, a repeated measures one-way ANOVA with visual areas (V1–V3) as the within-subject factor did not show any significant difference on the task-dependent effects across the visual areas [F_{(2, 6)} = 1.1; p = n.s.]. Finally, it was assessed whether the FDT-condition might serve as a better reference condition than PV by resulting in more negative coherence_ps values in LPZ. However, this was not the case, modulations for FDT fell short of that for PV, an effect that reached significance in glaucoma for V1 and V2.

Considering that our patients (both glaucoma and RP) demonstrated significant task-dependent LPZ responses, we employed an exploratory approach to investigate the relationship of the magnitude of difference in coherence_ps (i.e., OBT minus PV) with patient specific clinical characteristics. Specifically, the clinical measures included visual acuity (VA), disease duration, percentage of visual field loss (VF_loss), mean deviation (MD), foveal sensitivity. VF_loss was computed from the SAP-based VFs as the ratio of number of non-responsive test points vs. the total number of test points and considered only the regions of the VF representing our fMRI stimulus size. The relationship of the percentage of differential coherence_ps in LPZ with these clinical measures was tested with a simple linear regression model [R²(adjusted), p < 0.05]. We did not find a significant association of the task-dependent responses with any of the afore-mentioned clinical correlates, except for MD and task-effect in the visual area V2. Nonetheless, based on the magnitude of R², a negative trend in the task-effect with both MD (V1: 0.29; V2: 0.49) and the percentage of VF_loss (V1: 0.18; V2: 0.38) in V1 and V2 was indicated, but not for V3.

In an exploratory analysis, we probed for the presence of structural changes of the visual cortex following visual field deficits in patients compared to the controls. Freesurfer derived cortical thickness estimates, measured as the distance between the gray/white boundary and the pial surface (Fischl and Dale, 2000) was used as the indicator for structural integrity. For each participant and visual area (V1–V3) voxel-wise cortical thickness was extracted. Based on eccentricity estimates from a participant-specific anatomy driven retinotopic atlas (Benson et al., 2014), 12 ROIs of 2° eccentricity bins (0–24°) were created for each visual area and the mean cortical thickness was computed for each ROI. To investigate for differences in cortical thickness across groups a repeated measure ANOVA with participant group (patients and controls) as between-subject factor and eccentricity (0–24°) and visual areas (V1–V3) as within-subject factors was performed. The subject-specific global cortical thickness was included as a covariate to account for participant variability. While there was an overall trend for smaller cortical thickness in the patient group [cortical thickness controls vs. patients (mm) mean ± SEM: 1.76 ± 0.03 vs. 1.65 ± 0.02 (V1); 1.99 ± 0.05 vs. 1.93 ± 0.05 (V2); 2.31 ± 0.07 vs. 2.08 ± 0.05 (V3)], this did not reach significance. No other significant main effects or interaction were found. In addition, we tested for any potential association between the duration of visual field deficits and cortical thickness across the patients and did not find any significant relationship [R²: 0.08 (V1); 0.12 (V2); 0.20 (V3)].

The potential of remapping in V1 in acquired defects is often inferred from adult animal models (Kaas et al., 1990; Gilbert and Wiesel, 1992; Giannikopoulos and Eysel, 2006). While some degree of developmental V1 plasticity has been reported for congenital vision disorders (Baseler et al., 2002), the nature and magnitude of a large-scale reorganization in adulthood is questionable (Wandell and Smirnakis, 2009) and still warrants quantitative ascertainment. The differential comprehension in the above literature on LPZ activation primarily arises from the variable definitions of cortical reorganization and remapping (Morland, 2015). While the proponents of plasticity speculate that the mere presence of abnormal LPZ responses is sufficient evidence, the critiques point out the need for such responses to be non-explainable by the normal visual cortex organization following visual field loss. In the context of the latter definition of cortical reorganization, from our data, bottom-up large-scale reorganization appears an unlikely cause of the reported LPZ activation in glaucoma, as it would lead to LPZ responses irrespective of the condition, i.e., task. While this supports top-down effects as a cause of the task-related LPZ responses, these do not appear to strictly follow the inverse visual hierarchy, i.e., a decreasing effect size from V3 to V1, as might be expected for top-down modulations. In fact, the differential activation (reflected by Cohen’s d) we observed in V1 was not exceeded by those in V2 and V3 and did not reach statistical significance in a respective ANOVA. A stronger effect size in the extra-striate areas (i.e., V3 > V2 > V1) would have added further evidence to a top-down hypothesis. Further research is necessary to decipher the nature of the top-down modulations. One rewarding avenue to pursue for this purpose is paved by the advent of MRI at submillimeter resolution (Ress et al., 2007; Yakupov et al., 2017; Fracasso et al., 2018; Kashyap et al., 2018). It opens the possibility to recover information on the directionality of information flow in the cortex via laminar imaging (Dumoulin et al., 2018; Lawrence et al., 2019) that allows to dissociate activations in cortical input and output layers. Consequently, future studies measuring layer-specific functional activity in the visual cortex might unravel the missing pieces of information, i.e., origin and directionality of task-related LPZ activations, to validate or invalidate existing theories on the aberrant cortical activity observed in patients with de-afferented visual cortex.

A subset of glaucoma patients continues advancing toward blindness regardless of disease management. While restoration therapies might offer treatment options for this patient entity, it has been suggested that vision loss associated changes at the level of visual cortex might be a reason for treatment failure in such cases (Gupta and Yücel, 2007; Davis et al., 2016; Nuzzi et al., 2018). Remarkably, the existence of cortical responses in the de-afferented visual cortex, as demonstrated in the present study, suggests that the LPZ is still to some degree operational. These finding are of particular importance for current promising initiatives to restore the visual input to the cortex (Roska and Sahel, 2018), which include retinal gene (Ashtari et al., 2015; Aguirre, 2017; Russell et al., 2017) and stem-cell therapy (Schwartz et al., 2015; Jutley et al., 2017; Mead et al., 2018; Wang et al., 2020), and retinal (da Cruz et al., 2016; Edwards et al., 2018) and visual cortex implants (Beauchamp et al., 2020). Our findings demonstrate the relevance of fMRI-investigations of the functionality of the visual cortex for the preparation of such demanding vision restoration procedures, e.g., via the individualized prediction of their clinical effectiveness.

The presence and magnitude of LPZ task-dependent responses are likely also related to the patient’s clinical characteristics and their variability (Ferreira et al., 2019). Testing this, we observed a non-significant trend for the task-elicited LPZ responses to be negatively associated with the VF loss, i.e., the smaller the deficit the larger the response. This might suggest that the task-dependent LPZ-activations do not depend on the presence of extensive scotomas. Given the limited sample size in the present study, this explorative observation disserves following up in future studies. Nevertheless, the findings emphasize the benefits of patient stratification strategies for the investigation of disease related changes in the visual cortex.

As a primary limitation in the study, we acknowledge the small sample size of glaucoma patients, which was still fully sufficient to identify the relevant effects. It should be noted, that our target patients were those with strongly advanced VF defects. Consequently, as glaucoma is an age-progressive disorder, such patients are mostly in their later stages of life and likely to have at least one MRI-related contraindication, which makes them a rare cohort for fMRI investigations. Studies investigating the dynamics of the observed task-dependent LPZ responses in patients with different stages of the pathology are limited (Ferreira et al., 2017, 2019) which should be addressed by future research and more importantly aim to uncover the mechanisms underlying such responses with submillimeter laminar fMRI imaging.

Our definition of the lesion projection zone (LPZ) was based on negative modulations observed with full-field visual stimulation in accordance with previous studies (Masuda et al., 2010; 2008), but not on explicit spatial VF mapping. In consideration to previous studies (Plank et al., 2017) reporting changes in BOLD response modulation with stimulus type (for e.g., faces vs. checkerboard), it is to be noted that there might be a marginally differential characterization of LPZ with a full-field stimulus (in this case, gratings) compared to a spatial mapping stimulus. Importantly, in the context of population receptive field properties, distinctions in the cortical representation of visual field has been even reported for different spatial mapping-based approaches (Alvarez et al., 2015). As we tested for the presence of task-dependent responses in advanced glaucoma with substantial VF loss, full-field stimulation eliciting robust fMRI responses was preferred over a precise spatial delineation of the LPZ with spatially specific stimuli. Given the fact, irrespective of the stimulus type, keeping it consistent across the different conditions sufficed for our objective, the choice of stimulus could be rationalized. Nevertheless, a comparative analysis of LPZ definitions with different stimulation approaches might be of relevant interest to understand the dynamics of LPZ boundaries.

Some of our patients (GL2, RP2, and RP3) had quite low visual acuity and one of these patients (RP3) was predicted to have unreliable fixation with fundus-controlled perimetry, and the quantitative eye tracking data might have been informative. Despite this limitation, the qualitative monitoring and evaluation of the stimulated eye during scanning indicated all our participants to have good fixation of the stimulus in the scanner. In addition, given that the intact VF of our low visual acuity patients was largely constricted to central few degrees, we believe the responsive central visual field to be well within the stimulus window for potential small or moderate eye movements, thereby mitigating any possible eye-movement related influence in the fMRI response.

We also acknowledge a potential confound in our study with one RP patient also diagnosed with secondary glaucoma. Considering that, RP is not the primary disease under investigation in this study and used as a reference cohort to demonstrate the replicability and reproducibility of previous reports of task-dependent LPZ responses, we did not see a reason to exclude this patient in our analysis.

It should also be noted that our characterization of artificial scotoma was based on a generalized pattern of VF deficits as observed in peripheral vision disorders and not aiming at the simulation of our patient cohort’s idiosyncratic individual VF deficits. Although we acknowledge benefits of the latter approach, as our primary aim was to look at the task-dependent effects at a group level rather than individual patient-to-control comparison, the employed approach in the present study has the benefit of better within-group comparability. In addition, future research should also explore alternative methods to efficiently simulate artificial scotomas which closely correspond to patient like VF deficits, for e.g., via retinal bleaching (Magnussen et al., 2001; Gaffney et al., 2013) with high luminance levels and temporarily desensitizing the locations intended to represent an artificial scotoma.