Time course of visual plasticity following adult-onset deafness

Abstract

When deprived of a sensory modality, the brain often compensates with supranormal performance in other intact systems. While this compensatory plasticity is typically attributed to early sensory loss, plasticity following adult-onset sensory loss remains poorly understood despite its clinical relevance. In many patients, adult-onset hearing loss precedes treatment by cochlear implantation, yet little is known about the neural changes occurring before this intervention. The present study examines this transitional stage using a well-established adult cat model to examine visual plasticity after hearing loss in adulthood. We employed motion-onset visually evoked potentials (VEPs), a technique validated in our previous studies, to examine compensatory neural changes over time. VEPs are widely used in human neurophysiology and offer a translational bridge between basic science and clinical research. Over a 400-day period post adult-onset deafness, we observed gradual amplification in VEP signal power and P1 amplitude, alongside shortened peak latency. Our findings provide evidence that adult-onset deafness can induce compensatory visual plasticity and highlights VEPs as a promising biomarker for tracking such changes. This exploratory study establishes a platform for future research examining post-deafness intervention, such as cochlear implantation.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-39490-8.

Introduction

The brain exhibits a remarkable capacity to adapt following sensory loss. By studying this phenomenon, we have learned numerous insights about normal and abnormal brain development^1–4. Clinically, hearing loss is highly prevalent^5,6, and many patients elect to pursue interventions such as cochlear implantation to improve communication and quality of life. While most clinical assessments focus on the recovery phase after such treatment, the period between hearing loss and intervention remains understudied. Understanding the neural adaptations that occur during this transitional phase is essential, as it may influence long-term outcomes of hearing restoration.

When one sensory modality is deprived, the remaining systems often exhibit enhanced performance. This enhancement may arise from increased reliance on the intact senses for daily tasks^7,8 and/or from compensatory reorganization. Such reorganization can occur through cross-modal plasticity, where brain regions primarily dedicated the deprived modality become more receptive to inputs from other senses, either through repurposing of modality-specific areas^9,10, and/or through the enhancement of neurons with pre-existing multisensory capacity¹¹. Alternatively, intramodal plasticity may occur, where sensory loss can result in functional enhancements of the brain areas responsible for the intact modality^7,8.

It is well established that the neural plasticity required for the normal development of the auditory system, particularly for functions such as spoken language acquisition¹², must occur within a specific postnatal window, often referred to as the critical period (typically between 0–3 years of age in humans)¹³. As a result, studies of compensatory plasticity in humans have often focused on sensory loss during early development^14–19, with animal model researchers also selecting timepoints that approximate the human critical period²⁰. In contrast, far less is known about the nature and extent of compensatory plasticity following sensory loss later in life. However, emerging evidence from adult-deafened ferrets has shown that a greater proportion of neurons in the auditory cortex respond to somatosensory input compared to hearing controls, suggesting that deafness can induce cross-modal plasticity even in adulthood²¹.

One notable visual enhancement observed in early and congenital deafness is in the performance of motion detection, which has been demonstrated psychophysically in congenitally deaf cats^22,23 as well as in early and congenitally deaf humans¹⁵. Moreover, studies in humans utilizing a moving dot stimulus and functional magnetic resonance imaging (fMRI) revealed that early deafness led to the processing of visual stimuli within the deaf auditory cortex^24,25. Further fMRI studies demonstrate that other visual moving stimuli, like sign language, can also activate auditory cortex in deaf participants²⁶. The superior temporal cortex, which is normally associated with auditory processing²⁷, has also been found to be activated in response to various non-moving visual stimuli and somatosensory stimuli in deaf individuals^{14,16,17,28–34}. These findings all demonstrate the remarkable adaptative ability of auditory cortex to process visual information following hearing loss.

Visually evoked potentials (VEPs) are a technique where electrical potentials are recorded from large groups of neurons in the cortex after a visual stimulus is presented³⁵. While VEPs are primarily used to assess the functional integrity of the visual pathway, changes in waveform can reflect cortical adaptations, making them a useful tool for inferring visual plasticity following sensory loss^35–41. VEPs are commonly triggered using pattern reversal, pattern onset/offset, and/or flash stimuli^38,42, but due to the prominence of motion detection enhancements in deaf participants, motion-onset VEPs were emphasized in the present study. Motion-onset VEPs demonstrated strong peaks and little variability both within and between subjects^36,43 and have been used in previous studies on plasticity in congenitally deaf humans^18,19.

In a substantial proportion of clinical cases, hearing loss occurs later in life⁶, after the critical period for auditory system development has passed. Unlike individuals with congenital or early-onset deafness, these patients typically undergo normal auditory development before experiencing sensory loss. These older groups of patients are also often prescribed cochlear implant (CI) treatment, but the degree of neural plasticity during the period of deafness may have important implications for CI outcomes. Specifically, limited compensatory plasticity may indicate that auditory cortical function is preserved, which could facilitate adaptation to electrical hearing after cochlear implantation⁴⁴. Several human studies have reported that cross-modal plasticity could contribute positively to speech perception and language outcomes following cochlear implantation^45–48, and a study on cats suggests that auditory responsiveness could be regained after implantation despite prior reorganization⁴⁹. Nevertheless, other studies have also shown that extensive cross-modal recruitment of auditory cortex by visual input may still hinder auditory recovery with a cochlear implant, even as it supports alternative visual communication strategies such as lip-reading or sign language^44,50.

Unfortunately, assessing visual plasticity before and after hearing loss is extremely challenging in clinical settings, largely because hearing loss often occurs unpredictably and gradually, making it nearly impossible to obtain pre-loss baseline measures. Furthermore, neurophysiological assessments of cortical reorganization are limited by the time and accessibility of participants. Therefore, we proposed to investigate this issue by making observations using an animal model. A characterization of VEP changes after deafness would serve as a critical foundation for future studies examining VEPs across the course of cochlear implantation.

To investigate this window for compensatory plasticity, we employed motion-onset VEPs, which we have extensively validated on cats in our laboratory as a sensitive measure of cortical activity^39,40,51,52. VEPs are particularly well-suited for the longitudinal tracking of sensory changes and are already widely used in human neurophysiology, making them an ideal translational tool. Our previous studies have demonstrated that motion-related VEPs robustly reflect plasticity in perinatally deafened cats⁴⁰, with consistent and interpretable waveform features that respond to changes in visual processing. The current investigation extends this approach to adult-onset deafness, leveraging VEPs to capture compensatory reorganization in a way that bridges both basic science and clinical application. We hypothesized that compensatory plasticity would emerge after adult-onset deafness in cats, and that this neural reorganization would be reflected by a continued amplification of VEPs over the time course of the study. Consistent with the cross-modal plasticity found in the previous adult-onset hearing loss studies^11,21, we observed significantly amplified VEPs both short term (first 3–7 months) and long term (up to 12 months) after deafening. The results of this study suggest that plasticity is still prominent even in adult-onset deafness and is not limited to deafness incurred during development.

Results

In this study, we investigated visual plasticity by recording VEPs before and after adult-onset deafness (Fig. 1 see details in materials and methods). After deafening, animals were recorded each week for the first month, biweekly for the 2 months afterwards, and then monthly from the third month onwards for 12–13 months after deafening. Some recordings were not conducted in the first month to ease recovery after deafening in certain animals. Data was collected using two active electrodes: one positioned near the occipital brain region, where VEPs are typically generated in hearing subjects, and the other near the temporal brain region, corresponding to the location of the reorganized auditory cortex in deaf subjects. Data from both electrodes were analyzed and reported in parallel.

Fig. 1.

Experimental methodology. (A) Timeline of subjects from birth to post adult-onset deafness (B) Top-down illustration of a cat undergoing recording relative to the stimulus presenting monitor 17.2 cm away with a 124-degree horizontal viewing angle and to a 8-cm-diameter loudspeaker 10 cm away. Coloured dots demonstrate the position of electrodes on the subject. Red, occipital. Blue, temporal. Green, reference. Grey, ground. (C) Stimulus used during recording with leftward moving dots for 200 ms per trial with a ~ 1000 ms inter-trial interval and 20 trials per motion-onset speed. (D) Representative example of VEPs generated across all motion-onset speeds for one timepoint of recording. (E) Representative timeline of 32 deg/s stimulus generated VEPs and signal power for subject 1 before and after deafening.

Comparisons of VEP signal power and P1 amplitude

To confirm the validity of motion-related VEP signal power and P1 amplitude as a previously demonstrated biomarker of visual plasticity in cats⁴⁰, we derived median values from VEPs evoked by stimuli of 10 varying speeds for signal power and P1 amplitude. Then, we compared the VEPs from the “initial” and the “last” sets of recording days following deafness to the set of VEPs measured “before” deafening (Table 1). To make groups of equal populations for comparison, in both the “initial” and the “last” sets of data, we included the same number of recording days as in the “before” set, making them a “comparison among equipopulated groups”. One subject had 3 recording days included in both the “initial” and the “last” sets (i.e., the two sets overlapped for this subject) due to a slower recovery leading to fewer recordings during the first post-deafening month (Table 1). The analysis was performed on both original signal powers and P1 amplitudes as well as those normalized within each subject (See Figs. 2 and 3).

Table 1.

Age of deafening and timeline of recordings for each animal.

Subject	Age at deafening	Recording timepoints of VEPs in the “Before” set prior to deafening (days)	Recording timepoints of VEPs after deafening (days)
S1	5.5y	− 98, − 84, − 70, − 56, − 40	8, 16, 21, 30, 43, 58, 73, 108, 135, 173, 196, 231, 261, 289, 315
S2	3.8y	− 250, − 236, − 220, − 209, − 195, − 59, − 46	44, 57, 71, 92, 120, 151, 177, 199, 229, 331, 361
S3	6.7y	− 211, − 195, − 183, − 169, − 147, − 33, − 19	18, 28, 41, 55, 70, 106, 134, 160, 190, 225, 255, 296, 357, 387
S4	4.8y	− 258, − 244, − 229, − 214, − 202, − 52, − 39	9, 18, 23, 29, 44, 58, 72, 100, 128, 157, 183, 205, 235, 276, 338, 368

Included in “initial” set, Included in “last” set, Included in both sets

Fig. 2.

Comparison among equipopulated groups for signal power. (A) Occipital and (B) temporal electrode unnormalized data violin plot comparisons of median VEP signal power generated from the 10 different stimuli speeds with equipopulated groups between the before, initial, and last categories. (C) Occipital and (D) temporal electrode violin plot comparisons of normalized median VEP signal power generated from the 10 different stimuli speeds with equipopulated groups. Data was grouped for comparison among equipopulated groups for based on categories demonstrated in Table 1. Mann–Whitney U tests were used for comparisons between each group.

Fig. 3.

Comparison among equipopulated groups for P1 amplitude. (A) Occipital and (B) temporal electrode unnormalized data violin plot comparisons of median VEP P1 amplitude generated from the 10 different stimuli speeds with equipopulated groups between the before, initial, and last categories. (C) Occipital and (D) temporal electrode violin plot comparisons of normalized median VEP P1 amplitude generated from the 10 different stimuli speeds equipopulated groups. Data was grouped for comparison among equipopulated groups based on categories demonstrated in Table 1. Mann–Whitney U tests were used for comparisons between each group.

We quantified the noise-corrected RMS as a measure of signal power in the 400-ms window post stimulus-onset for VEPs averaged for each recording day. In a comparison among equipopulated groups for the signal power before normalization, the occipital electrode showed a significant (p < .05) RMS increase in the “initial” set of measurements after deafening as compared to those “before” deafening, and a highly significant (p < .001) increase in the “last” set of measurements after deafening as compared to those “before” deafening (Fig. 2A). The temporal electrode showed no significant changes in RMS between all three comparison groups (Fig. 2B). After normalization, the occipital electrode showed a significant (p < .01) increase in RMS from “before” deafening to the “initial” set of measurements after deafening, a highly significant (p < .001) increase from “before” deafening to the “last” set of measurements after deafening, and a significant (p < .05) increase from the “initial” set of measurements to the “last” set of measurements after deafening (Fig. 2C). For the temporal electrode, the amplification of VEP signal power was only significant (p < .05) during the “last” set of measurements compared to the “before” set (Fig. 2D).

We then quantified the amplitude of the positive deflection that is the largest in the waveform which we refer to as the P1 amplitude to better understand the timing and morphology of the VEPs in addition to their overall activity captured by signal power. Waveforms sometimes had other small components before or after P1, but these were not analyzed further due to inconsistencies both between and within subjects. In a comparison among equipopulated groups for the P1 amplitudes before normalization, the occipital electrode showed a significant (p < .01) increase in amplitude only from “before” deafening to the “last” set of measurements after deafening (Fig. 3A). The temporal electrode showed no significant changes in P1 amplitude between all three comparison groups (Fig. 3B). After normalization, the occipital electrode showed a significant (p < .05) increase in P1 amplitude from “before” deafening to the “initial” set of measurements after deafening, a significant (p < .01) increase from “before” deafening to the “last” set of measurements after deafening, and a significant (p < .01) increase from the “initial” set of measurements to the “last” set of measurements after deafening (Fig. 3C). The temporal electrode showed a significant (p < .05) increase in P1 amplitude from the set of measurements “before” deafening to the “last” set of measurements after deafening and a significant increase (p < .05) between the “initial” set of measurements to the “last” set of measurements after deafening (Fig. 3D).

In summary, our statistical analysis on the equipopulated group data, original and normalized, demonstrated that adult-onset deafness led to an increase in the signal power and P1 amplitude captured by both the occipital and temporal electrodes. This increase was more prominently reflected from the occipital electrode, and signal power appeared to increase further by the last set of measurements after deafening.

Effect of adult-onset deafness on signal power across recording timepoints

To refine and characterize the time course of the emergence of the visual plasticity from adult-onset deafness, normalized median signal power values were derived in the same way as previously mentioned and these values were plotted across the recording days for both recording sites (Fig. 4A, B). These data points were then binned within the hearing baseline as well as using the boundaries of every 100 days post deafening (Fig. 4C, D), and Mann–Whitney U tests were subsequently used to compare the bins (Table 2). Compared to the baseline, the VEPs measured during the first 100 days after deafness were significantly (p < .01) greater (Fig. 4C, D; Table 2). This amplification of VEPs sustained through the post-deafening measurements, as observed from the second (p < .001), the third (p < .001), and the fourth (p < .001) 100-day periods (Fig. 4C, D; Table 2). This amplification of VEPs plateaus by the fourth 100-day period. This increase was not as apparent from the temporal electrode, where a comparison of RMS between the hearing baseline and the first 100-day period post-deafening was not significant (Fig. 4C, D; Table 2). Statistically significant (p < .05) increases were not found in the measurements made until the second 100-day period after deafness (Fig. 4C, D; Table 2). This amplification then sustained throughout the remaining post-deafening measurements observed at the third (p < .05) and the fourth (p < .01) 100-day periods (Fig. 4C, D; Table 2). A Tau-U test was employed on the RMS before and after the deafening intervention for all subjects, resulting in significance (p < .001; Tau = 0.6502) for the occipital electrode and marginal-significance (p = .078; Tau = 0.2501) for the temporal electrode.

Fig. 4.

Timeline of signal power and bins for comparison. (A) The median normalized VEP signal power generated from the 10 different stimuli speeds used in each recording was plotted in a timeline for both (A) the occipital electrode and (B) the temporal electrode. These signal powers were binned for the hearing baseline recordings and every 100-day period post deafening for comparison in both (C) the occipital electrode and (D) the temporal electrode. Mann–Whitney U tests were used for comparisons between bins and significance levels were recorded in Table 2. On 100-day period bins, *, **, and *** represent unadjusted significance levels of p < .05, p < .01, and p < .001 compared to the baseline hearing bin respectively.

Table 2.

Multiple comparisons between timeline bins for signal power.

Electrode	Comparison pair	Unadjusted significance level	Bonferroni-adjusted significance level
Occipital	Hearing vs. 0–100 (days)	p < .01	p < .005
	Hearing vs. 100–200 (days)	p < .001	p < .0001
	Hearing vs. 200–300 (days)	p < .001	p < .0001
	Hearing vs. 300–400 (days)	p < .001	p < .0001
	0–100 (days) vs. 100–200 (days)	n.s	n.s
	0–100 (days) vs. 200–300 (days)	p < .05	n.s
	0–100 (days) vs. 300–400 (days)	p < .01	p < .005
	100–200 (days) vs. 200–300 (days)	n.s	n.s
	100–200 (days) vs. 300–400 (days)	p < .05	n.s
	200–300 (days) vs. 300–400 (days)	n.s	n.s
Temporal	Hearing vs. 0–100 (days)	n.s	n.s
	Hearing vs. 100–200 (days)	p < .05	n.s
	Hearing vs. 200–300 (days)	p < .05	n.s
	Hearing vs. 300–400 (days)	p < .01	p < .005
	0–100 (days) vs. 100–200 (days)	n.s	n.s
	0–100 (days) vs. 200–300 (days)	n.s	n.s
	0–100 (days) vs. 300–400 (days)	n.s	n.s
	100–200 (days) vs. 200–300 (days)	n.s	n.s
	100–200 (days) vs. 300–400 (days)	n.s	n.s
	200–300 (days) vs. 300–400 (days)	n.s	n.s

In summary, our analysis found that adult-onset deafness led to an increase in the signal power of VEPs as shown by RMS recorded by both the occipital and temporal electrodes. This increase was more prominently reflected from the occipital electrode, and signal power appeared to increase further throughout the time course.

Effect of adult-onset deafness on P1 amplitude and latency across recording timepoints

To further explore the time course of visual plasticity after adult-onset deafness, we marked the normalized median P1 amplitude values which were derived in the same way as previously described and these values were plotted across the recording days for both recording sites (Fig. 5A, B). These data points were then binned within the hearing baseline as well as using the boundaries of every 100 days post deafening (Fig. 5C, D), and Mann–Whitney U tests were subsequently used to compare the bins (Table 3). Compared to the baseline, the VEPs measured during the first 100 days after deafness were significantly (p < .01) greater (Fig. 5C, D; Table 3). This amplification of VEPs sustained through the post-deafening measurements following a linear trend, as observed from the second (p < .001), the third (p < .001), and the fourth (p < .001) 100-day periods (Fig. 5C, D; Table 3). This increase was not as apparent from the temporal electrode, where a comparison of P1 amplitude between the hearing baseline and the first 100-day period post-deafening bin was not significant (Fig. 5C, D; Table 3). Statistically Significant (p < .05) increases were not found in the measurements made until the second 100-day period after deafness (Fig. 5C, D; Table 3). This amplification then sustained throughout the remaining post-deafening measurements observed at the third (p < .01) and the fourth (p < .01) 100-day periods (Fig. 5C, D; Table 3). A Tau-U test was employed on the P1 amplitudes before and after the deafening intervention for all subjects, resulting in significance (p < .001; Tau = 0.5539) for the occipital electrode and significance (p < .05; Tau = 0.2878) for the temporal electrode.

Fig. 5.

Timeline of P1 amplitude and bins for comparison. (A) The median normalized VEP P1 amplitude generated from the 10 different stimuli speeds used in each recording was plotted in a timeline for both (A) the occipital electrode and (B) the temporal electrode. These P1 amplitudes were binned for the hearing baseline recordings and every 100-day period post deafening for comparison in both (C) the occipital electrode and (D) the temporal electrode. Mann–Whitney U tests were used for comparisons between bins and significance levels were recorded in Table 3. On 100-day period bins, *, **, and *** represent unadjusted significance levels of p < .05, p < .01, and p < .001 compared to the baseline hearing bin respectively.

Table 3.

Multiple comparisons between timeline bins for P1 amplitude.

Electrode	Comparison pair	Unadjusted significance level	Bonferroni-adjusted significance level
Occipital	Hearing vs. 0–100 (days)	p < .01	n.s
	Hearing vs. 100–200 (days)	p < .001	p < .0001
	Hearing vs. 200–300 (days)	p < .001	p < .0001
	Hearing vs. 300–400 (days)	p < .001	p < .0001
	0–100 (days) vs. 100–200 (days)	n.s	n.s
	0–100 (days) vs. 200–300 (days)	p < .05	n.s
	0–100 (days) vs. 300–400 (days)	p < .05	n.s
	100–200 (days) vs. 200–300 (days)	p < .05	n.s
	100–200 (days) vs. 300–400 (days)	p < .05	n.s
	200–300 (days) vs. 300–400 (days)	n.s	n.s
Temporal	Hearing vs. 0–100 (days)	n.s	n.s
	Hearing vs. 100–200 (days)	p < .05	n.s
	Hearing vs. 200–300 (days)	p < .01	n.s
	Hearing vs. 300–400 (days)	p < .01	p < .005
	0–100 (days) vs. 100–200 (days)	n.s	n.s
	0–100 (days) vs. 200–300 (days)	n.s	n.s
	0–100 (days) vs. 300–400 (days)	p < .05	n.s
	100–200 (days) vs. 200–300 (days)	n.s	n.s
	100–200 (days) vs. 300–400 (days)	n.s	n.s
	200–300 (days) vs. 300–400 (days)	n.s	n.s

The time course of normalized median P1 latencies was subsequently analyzed and plotted in a similar manner (Fig. 6A, B). These data points were again binned between every 100 days post deafening as well as during the hearing baseline, (Fig. 6C, D) and a Mann–Whitney U test was subsequently used to compare the bins (Table 4). In contrast to P1 amplitudes, compared to their hearing baselines, P1 latencies did not significantly differ by both the first and the second 100-day periods post-deafening, while a significant decrease in P1 latency (p < .05) was determined by the third 100-day period and sustained through the fourth (p < .01) in the occipital electrode location (Fig. 6C, D; Table 4). P1 latencies for the temporal electrode were significantly (p < .001) lower than the hearing baseline for all four 100-day periods post deafening (Fig. 6C, D; Table 4). A Tau-U test was employed on the P1 latencies before and after the deafening intervention for all subjects, resulting in significance (p < .01; Tau = − 0.4157) for the occipital electrode and significance (p < .001; Tau = − 0.7839) for the temporal electrode.

Fig. 6.

Timeline of P1 latency and bins for comparison. (A) The median normalized VEP P1 latency generated from the 10 different stimuli speeds used in each recording was plotted in a timeline for both (A) the occipital electrode and (B) the temporal electrode. These P1 latencies were binned for the hearing baseline recordings and every 100-day period post deafening for comparison in both (C) the occipital electrode and (D) the temporal electrode. Mann–Whitney U tests were used for comparisons between bins and significance levels were recorded in Table 4. On 100-day period bins, *, **, and *** represent unadjusted significance levels of p < .05, p < .01, and p < .001 compared to the baseline hearing bin respectively.

Table 4.

Multiple comparisons between timeline bins for P1 latency.

Electrode	Comparison pair	Unadjusted significance level	Bonferroni-adjusted significance level
Occipital	Hearing vs. 0–100 (days)	n.s	n.s
	Hearing vs. 100–200 (days)	n.s	n.s
	Hearing vs. 200–300 (days)	p < .05	n.s
	Hearing vs. 300–400 (days)	p < .01	n.s
	0–100 (days) vs. 100–200 (days)	n.s	n.s
	0–100 (days) vs. 200–300 (days)	n.s	n.s
	0–100 (days) vs. 300–400 (days)	p < .05	n.s
	100–200 (days) vs. 200–300 (days)	n.s	n.s
	100–200 (days) vs. 300–400 (days)	n.s	n.s
	200–300 (days) vs. 300–400 (days)	n.s	n.s
Temporal	Hearing vs. 0–100 (days)	p < .001	p < .0001
	Hearing vs. 100–200 (days)	p < .001	p < .0001
	Hearing vs. 200–300 (days)	p < .001	p < .0001
	Hearing vs. 300–400 (days)	p < .001	p < .0001
	0–100 (days) vs. 100–200 (days)	n.s	n.s
	0–100 (days) vs. 200–300 (days)	n.s	n.s
	0–100 (days) vs. 300–400 (days)	p < .05	n.s
	100–200 (days) vs. 200–300 (days)	n.s	n.s
	100–200 (days) vs. 300–400 (days)	p < .05	n.s
	200–300 (days) vs. 300–400 (days)	n.s	n.s

To summarize, our analysis found that adult-onset deafness led to an increase in amplitude and decrease in latency of the P1 component captured by both the occipital and temporal electrodes. The increase in P1 amplitude was more prominently reflected from the occipital electrode, while the shortening of P1 latency was more distinctly reflected from the temporal electrode. In both recording sites, the effects of adult-onset deafness appeared more pronounced later into the time course.

The statistical significance of the increase in RMS and P1 amplitude survived Bonferroni correction, while the decrease in P1 latency did not. Given the limited sample size, the statistical power for the effect size we observed in RMS and P1 amplitude is highlighted. Meanwhile, our observation in P1 latency will benefit confirmatory replication in future studies.

To ensure that the observed post-deafening changes were not influenced by pre-existing trends during the hearing period, we assessed the effect of deafening on VEPs using Tau-U analysis⁵³, which takes into consideration the baseline trends in VEP trajectories. All animals except Subject 1 demonstrated stable hearing baselines with no significant trends across recording days. Subject 1 exhibited a marginally significant increasing trend in signal power at the occipital recording site (p = 0.050, Tau = 0.8000). In accordance with Tau-U conventions, a baseline correction was applied before the calculation of aggregate group-level effect sizes⁵³.

As a control, we examined pre-deafening hearing recordings extended with earlier data from⁴⁰. Signal power, P1 amplitude, and P1 latency were plotted over a timeline spanning up to 306 days (Figs. S1–S3) with no significant changes observed across time. (Figs. S4–S6).

Permutation test across motion-onset stimuli speeds

Finally, our analysis used the median signal power and P1 amplitude resulting from the waveforms generated by the 10 stimulus dot speeds as a representation of each point in the timeline. To study the effect of adult-onset deafness on VEPs at each of the speeds used, we performed permutation tests⁵⁴ on the signal power and P1 amplitude of each waveform (Table 5). Seven timepoints before deafening and seven timepoints after deafening underwent 10,000 randomized permutations for each moving stimuli speed. The probability that the true median differences are greater or smaller than those obtained from permutation is determined by the number of occurrences out of 10,000 permutations, which serves as the significance level p. (Table 1). The results of our permutation test showed that the RMS and P1 amplitude of the VEPs before deafening were not equal to those after deafening for stimulus motion above 5.8 deg/s in both recording sites with exception to 16 and 23 deg/s for the temporal electrode (Table 5). As such, the permutation test showed that, for both recording sites, a medium-to-high stimulus speed is needed for the effect of adult-deafening on VEP signal power or P1 amplitude to be observed. Perhaps there is a lack of statistical sensitivity when VEPs were evoked by low speed moving stimuli.

Table 5.

Permutation test for signal power and P1 amplitude.

Stimulus motion-onset speed (deg/s)	Bonferroni-adjusted significance level
	Signal power		P1 amplitude
	Occipital	Temporal	Occipital	Temporal
64	p < .005	p < .005	p < .005	p < .005
32	p < .005	p < .005	p < .005	p < .005
23	p < .005	n.s	p < .005	n.s
16	p < .005	n.s	p < .005	n.s
11	p < .005	p < .005	p < .005	p < .005
8.0	p < .005	n.s	p < .005	n.s
5.8	n.s	n.s	n.s	n.s
4.0	n.s	n.s	p < .005	n.s
2.8	n.s	n.s	p < .005	n.s
2.0	n.s	n.s	n.s	n.s

Overall, the results of the permutation tests were consistent with our other analyses, again demonstrating that adult-onset deafness has a significant effect on both the RMS and P1 amplitude of VEPs.

We also provide the VEP waveforms for all subjects across each of the stimulus motion-onset speeds in both recording sites (Fig. S7).

Discussion

In the present study, we investigated the effects of hearing loss on motion-onset VEPs in cats deafened in adulthood. Specifically, we analyzed the changes in RMS, P1 amplitude, and P1 latency on a longitudinal timeline after deafening. Demonstrated by amplified VEP amplitude/signal power and shortened peak latency after deafness compared to their hearing baselines, our results support our hypothesis that compensatory plasticity would still occur in cats even after deafness in adulthood rather than only in youth. Furthermore, our results suggest plasticity following adult-onset deafness associated with the occipital and temporal brain regions, which are known to be associated with visual and auditory processing, respectively. The findings from our macro-level recordings using VEPs were consistent with prior extracellular recordings from adult-deafened ferrets²¹ in suggesting that sensory loss, even after the maturation of the sensory system, can still result in significant compensatory plasticity. Additionally, our investigation reinforces this concept, as changes in VEPs were observed at a substantially later stage, with an average age of approximately 5.2 years in cats which was well beyond the 0.42 years of age at which the ferrets were deafened²¹.

VEPs are gradually amplified after adult-onset deafness

Our findings reveal a significant increase in the signal power of VEP waveforms following adult-onset deafness. The noise-corrected RMS, which is a measure of VEP signal power, was significantly enhanced post-deafening with more pronounced changes in the occipital electrode. This enhancement occurred after auditory inputs were deprived for 100 days and continued to increase throughout deafness as far as data was collected. Meanwhile, the signal power for the temporal electrode was also enhanced, but in a less pronounced way. It was not statistically significant until auditory inputs were deprived for 200 days rather than 100 days. Overall, the abrupt hearing loss gradually amplified the signal power of VEPs acquired from both the occipital and temporal electrodes with moderate and small effect sizes, respectively, as revealed by Tau-U tests⁵⁵. These findings were unlike those seen in perinatally deafened cats, where the pattern of different effect sizes between the two recording sites was not observed⁴⁰. Therefore, our present results suggest that compensatory intramodal changes may occur more rapidly in the occipital lobe compared to cross-modal changes in temporal lobe of cat cortex. The auditory cortex may require more time to reorganize itself to become responsive to visual stimuli as compared to occipital regions, which are inherently specialized for visual processing, and could explain the slower, albeit eventual increase in signal power in the temporal electrode.

Following our analysis on RMS, we examined the P1 component of the VEPs, which also exhibited significant changes post-deafening. Like RMS, P1 amplitude was significantly increased after adult-onset deafness, with the occipital electrode demonstrating more pronounced effects likely for the same reasons. Specifically, this increase in P1 amplitude occurred after auditory inputs were deprived for 100 days and continued to increase throughout deafness as far as data was collected. The P1 amplitude for the temporal electrode was also increased, but again in a less pronounced way. It was not statistically significant until auditory inputs were deprived for 200 days. Overall, the abrupt hearing loss gradually amplified the P1 amplitude of VEPs acquired from both the occipital and temporal electrodes with small effect sizes, respectively, as revealed by Tau-U tests⁵⁵. Together with the analysis on RMS, these results further support the notion that cross-modal enhancements in the temporal lobe may occur later than intramodal enhancements in the occipital lobe.

Our results provided further evidence in support of recent studies of compensatory plasticity in adult-onset hearing impairment, which have demonstrated enhanced neural activity in auditory brain areas^11,56–59, visual brain areas^60,61, and audiovisual brain areas^58,62. However, our study is the first to suggest a topographical difference in the time course of more than 1 year, during which these enhancements develop.

Topographical discrepancy in VEPs

Being the first to suggest a possible topographical difference in the time course of plasticity in terms of waveform magnitude, we continued to discuss changes in latency with this consideration. Conversely to VEP signal power and P1 amplitude, P1 latency showed a significant decrease following deafness, indicating that the neural response to motion-onset stimuli became faster over time. This decrease was not statistically significant from the occipital electrode until auditory inputs were deprived for 300 days, whereas the temporal electrode exhibited a statistically significant reduction in P1 latency much earlier after 100 days of hearing loss. Overall, the abrupt hearing loss gradually shortened the P1 latency of VEPs acquired from both the occipital and temporal electrodes with small and moderate Tau-U effect sizes respectively⁵⁵. This decrease in peak latency is consistent with VEP studies in congenitally deaf humans⁶³ and perinatally deafened cats⁵², where shorter latencies of peak components were similarly observed.

Our results again demonstrate a topographical discrepancy in the time course of plasticity for the development of this shortened latency This suggests that while both cortical regions undergo compensatory changes, the temporal lobe may adapt more rapidly in terms of processing speed. Synaptic transmission, due to its cascading cellular and molecular processes⁶⁴, is considered a primary source of neural variability, and the reorganized auditory cortex may subtly contribute to early visual processing involving a reduction in delay during synaptic transmission. Therefore, we propose that a possible explanation for this decrease in latency is a result of the faster synaptic transmission in the reorganized auditory cortex to visual stimuli, allowing for more efficient neural processing as the timeline progresses.

The amplification of the P1 component observed after deafening aligns with human studies reporting enhanced visual responses in deaf participants. P1-equivalent components of human VEPs have been shown to be larger in amplitude among deaf participants⁶³, and the amplitude of the P1 component has been proposed as a predictor of increased visual reactivity in deaf individuals⁶⁵. In contrast, studies of deaf participants who received cochlear implants have reported no robust group differences in P1 amplitude or latency compared to hearing controls, suggesting that the amplification of visual activations diminishes once auditory input is restored^66–68. This parallel supports our interpretation that the P1 enhancements in amplitude and latency observed in the present study reflects compensatory visual plasticity associated with hearing loss.

Comparisons among equipopulated groups and speed sensitivity

Given that the intervals between recording days varied throughout the study, with more recordings concentrated earlier in the timeline, earlier time bins had greater numbers of recording days included than later. As such, our equipopulated group analysis on the P1 amplitude and signal power of VEPs confirms our interpretation of the results from the 100-day period time course bins. Our results from this approach were consistent with the time course, reinforcing our interpretation that adult-onset deafness led to an increase in both signal power and P1 amplitude. Notably, the occipital electrode continued to exhibit more prominent increases compared to the temporal. Furthermore, our comparison among equipopulated groups revealed that the increase in signal power and P1 amplitude was more pronounced later in the time course compared to the beginning. This suggests that the both intramodal and cross-modal adaptations following deafness were progressive throughout the investigation, with long-term accumulation as the timeline extended beyond the initial months of sensory loss.

Our prior analysis studied the medians of the signal power and P1 amplitude over varying stimuli speeds. To also assess the effects at a specific stimulus speed, we conducted permutation tests at all the different stimuli speeds. The results of our permutation tests demonstrated that significant increases in signal power and P1 amplitude were evident for any stimulus speed above 5.8 deg/s in both recording sites. The lack of significance in slower stimuli speeds were likely due to insufficient signal-to-noise ratio, which compromised statistical sensitivity. The only exceptions were found for 16 and 23 deg/s in VEPs at the temporal electrode, which did not show a significant increase likely due to individual variability in the subjects. Overall, our VEP results are consistent with previous extracellular study in ferrets with adult-onset deafness²¹ showing that compensatory plasticity is possible in a deprived auditory cortex even after the maturation of the sensory system.

Potential modulators on VEPs and methodological considerations

To focus on the changes along the time course of plasticity after adult-onset deafness, we pursued an exploratory single-arm experimental design^69,70. We considered aging as a possible factor which could potentially contribute to the significant increases in VEP signal power and P1 amplitude as well as the significant decreases in P1 latency we observed in our results. Despite not being included in the current study, previous VEP studies in humans have unanimously demonstrated that aging resulted in significantly diminished waveform amplitudes and increased the latency of VEP components, which starkly contrasts our observations following deafness^71–75. Given the opposite directions for VEP change during deafness and during aging, we did not attribute our findings to aging. Instead, we interpreted the VEP changes following adult-onset deafness to visual plasticity induced by hearing loss.

Our use of two subdermal recording electrodes, positioned over the occipital and temporal cortices, enabled us to explore potential topographical distinctions in plasticity. Compared to single-electrode approaches⁵², this dual-site setup better approximates activity in the underlying visual and auditory regions. However, due to the relatively low impedance of subdermal needle electrodes in comparison to extracellular recording electrodes, it is still difficult to pinpoint the exact neural sources of these changes. Unlike our prior work in perinatally deafened animals⁴⁰, where both recording sites showed similar VEP amplification without any latency shifts, adult-onset deafness produced earlier increases in signal power and P1 amplitude at the occipital electrode compared to the temporal, and earlier latency shortening at the temporal site compared to the occipital. Apart from a shortage of statistical power in the latency that may warrant confirmatory replication in further studies. These findings suggest the possibility that intramodal and cross-modal plasticity may develop differently after adult-onset deafness compared to perinatal deafness. A quantitative comparison between perinatally-deafened and adult-deafened cats could also be the major aim of a future study. While the spatial resolution of this method is limited compared to human EEG systems^76,77, it offers a viable bridge between less invasive monitoring and future studies using extracellular recording methods in animal models.

In this study, we opted to use dexmedetomidine for sedation during EEG recordings. Recording high-quality EEG signals is challenging in head-free moving animals, so sedation was crucial to ensure stable recordings. Unlike anesthetics such as propofol and isoflurane, which are known to have a substantial impact on VEPs³⁹, dexmedetomidine has been shown to have minimal effects on VEPs. While anesthetics like propofol and isoflurane can decrease VEP amplitudes in a dose-dependent manner^78,79, previous studies in humans undergoing spine surgery⁸⁰ and in dogs⁸¹ have shown that dexmedetomidine does not interfere with VEPs. Based on these findings, we concluded that the use of dexmedetomidine did not affect our VEP recordings^40,51,52 and our conclusion regarding visual plasticity after adult-onset deafness in cats.

Time course of functional compensatory plasticity: from electrophysiology to behaviour

The research of adult-onset deafness began with extracellular single-unit recordings in ferrets, offering a high-resolution, microscale view of neuronal activity changes following hearing loss²¹. Somatosensory conversion in 84% of neurons from 225 recording sites was identified in the auditory cortex of 7 ferrets, one of which was recorded only 16 days post-deafening and the other six at 76 ± 9 days post deafening²¹. This was in stark contrast to their control group made of 4 ferrets with intact hearing, where 96% of 100 recording sites were responsive only to auditory stimuli and not to tactile/visual stimuli²¹. Meanwhile, EEG recordings in the present study which are derived from gross neural responses exhibited a more delayed timeline in capturing compensatory plasticity (Fig. 7). In adult-deafened cats, plasticity was only detected around 100 days post-deafening despite continuous weekly and biweekly recordings during this period, indicating that broader macro-level neural reorganization takes longer to manifest at a detectable magnitude. Unfortunately, there is no literature documenting a time course of behavioral adaptation after adult-onset deafness. However, both the micro and macro-level data from these electrophysiological studies can serve as a bridge for estimating the time required for changes in neural responses to be translated into behavioral enhancements. Given this delay in VEP/EEG-detected plasticity, it is likely that behavioral enhancements in these animals will also progress slowly, as functional reorganization at the neural level precedes the integration of these changes into perceptual and motor behaviors (Fig. 7).

Fig. 7.

Framework incorporating compensatory plasticity observed in various study approaches. Schematic of the extent of cross-modal plasticity observed in a time course after adult-onset deafness using extracellular single-unit recording²¹, EEG/VEP recording (present study), and a predicted time course of behavioral enhancement after adult-onset deafness.

In transitioning to clinical settings, micro-level approaches like extracellular recordings are not feasible for use with humans, as they require invasive electrode implantation. Instead, non-invasive methods are necessary to track compensatory plasticity following adult-onset deafness. Behavioral approaches, while valuable, present significant challenges in quantification and implementation, as patients who have recently lost their hearing are unlikely to engage in extensive visual training tasks. In contrast, VEPs provide a more clinically applicable tool for monitoring neural reorganization after deafness, being an objective, non-invasive measure of visual cortical activity. Furthermore, VEPs are recorded using EEG similarly as in ABRs, and since ABRs are the standard clinical method for assessing hearing loss^82–84, incorporating both would be straightforward. This dual approach would enable a comprehensive evaluation of auditory deficits and potential compensatory changes with minimal disruption to patients, making it a practical and efficient method for tracking neuroplasticity in response to adult-onset deafness.

Conclusion

In conclusion, our study of motion-onset VEPs in cats with adult-onset deafness suggests compensatory plasticity in the visual and auditory systems, resulting in significantly increased VEP signal power, increased P1 amplitude, and shortened latency. Overall, our study makes an essential contribution to characterizing the time course for neuroplasticity following adult-onset deafness that is currently understudied. This is also a necessary first step in forming better rehabilitation strategies and interventions for those affected by late-onset hearing loss.

Materials and methods

Subjects

Four cats (sourced from Liberty Research, now Marshall BioResources), all of which were hearing at birth and subsequently deafened as adults, were included. Animals used in this study were female and of similar age at deafening (x̄ = 5.2y). All procedures conducted followed the National Research Council’s Guide for the Care and Use of Laboratory Animals (8th edition; 2011) and the Canadian Council on Animal Care’s Guide to the Care and Use of Experimental Animals (1993). Furthermore, all procedures and experiments were approved by the Animal Care Committee of the Faculty of Medicine and Health Sciences at McGill University.

Ototoxic deafening of adult cats

The animals used in this study came from a selection based on availability from our colony without any bias on their sensory functions. The selected animals ranged from 3.8 to 6.7 years of age at the time of deafening (age at deafening: S1, 5.5y; S2, 3.8y; S3, 6.7y; S4, 4.8y) (Fig. 1A). This variation falls within the range of ages of early adulthood in cats. As studies in human participants with hearing loss inevitably investigate subjects with varying ages, we did not consider using an exact age for deafening in the animal model to be more practical or comparable to human disorders. Most importantly, as this age range is well beyond the period of sensory and cortical maturation in cats while before most geriatric factors manifest, no developmental differences in VEPs due to their ages were expected across subjects. The subjects were ototoxically deafened through the systemic co-administration of subcutaneous kanamycin (300 mg/kg) along with intravenous furosemide (100 mg/kg, Valent Pharmaceuticals). This combination of aminoglycoside antibiotic and diuretic has been demonstrated to result in profound bilateral deafness through damaging cochlear hair cells⁸⁵. During this procedure, animals were sedated using an intramuscular injection of dexmedetomidine (0.04 mg/kg, Dexdomitor, Zoetis) and anesthesia was maintained with 0.5 ~ 1.0% isoflurane (AErrane, Baxter) using an isoflurane vaporizer (Isotec 4, Smiths Medical) mixed with oxygen. The animal was given a clear eye lubricant (Optixcare, Aventix) to each eye and placed on a water-circulated heating pad (TP-400, Gaymar) to help maintain their body temperature during the procedure.

Auditory brainstem responses (ABRs) were obtained using an active electrode positioned near the midpoint of the subject’s interaural line, while the reference electrode was placed beneath the left ear, contralateral to an 8-cm diameter loudspeaker (Fostex). Click stimuli at 80 dB sound pressure level (SPL) were programmed on a stimulus/recording system (Synapse, TDT) and played from this loudspeaker placed 10-cm away perpendicular to the ear (Fig. 1B). Every 3–5 min, the click stimuli generated using a processor (RZ2, TDT) were played to obtain ABRs to monitor the progress of deafening throughout the procedure. Impedance in both electrodes were maintained below 1 kOhm during the procedure. The signal from the electrodes was amplified and digitized using a pre-amplifier (TDT, Medusa4Z) and routed to the same processer (TDT, RZ2) to provide real-time feedback on the progress of deafening. Prior to deafening, the normal hearing of all subjects was tested with clicks played from 80 dB SPL to 10 dB SPL to obtain an ABR threshold. Furosemide was infused intravenously in 1–2 boluses (5 mg/kg per bolus) until the monitored ABRs began to decrease, after which furosemide was administered at 1 mg/kg/min for approximately 60–90 min until an ABR threshold higher than 80 dB SPL was permanently reached. The isoflurane was then stopped, while oxygen continued to be administered until the animal was recovered from anesthesia. Deafness was also re-confirmed in all cats at 1 month after deafening by observing another absence of ABR at 80 dB SPL.

Recording of VEPs

Each animal in this study received 5–7 baseline recordings when they had intact hearing, and repeated recordings for approximately 1 year after deafening. EEG recordings for VEPs were conducted in a dark soundproof chamber (IAC Acoustics) to decrease both environmental auditory and visual noise. Subjects were sedated at the start of each procedure using an intramuscular injection of dexmedetomidine (0.04 mg/kg). Animals were placed in a sphinx-like position perpendicular to a 30-inch 2560-by-1600 LED screen with 124-degree horizontal viewing angle (Dell, U3014) placed 17.2 cm away from both eyes. The stimuli were presented binocularly. Phenylephrine (Mydfrin, Alcon) was applied to each eye to dilate the pupil and to retract the nictitating membrane, and a clear eye lubricant was applied (Optixcare, Aventix). Subdermal needle electrodes (25G) were then inserted for EEG recording, where the active electrodes were placed over the occipital and temporal lobes, the reference electrode inferior to the left ear, and the ground electrode placed on the dorsum (Fig. 1B). For optimal recording quality, the impedance of all electrodes was maintained below 1kOhm during the recording and frequently assessed. The signals from the electrodes were digitized and amplified using a pre-amplifier (TDT, Medusa4Z). Subsequently, the signals were routed to a digital signal processor (TDT, RZ2) and saved to the computer’s hard drive for storage and offline analysis. Minimal artifacts were present in the EEG from movement due to the stability of the animal under anesthesia. The recordings were terminated 45 min post sedation or at any point with a sign of the animal moving. Electrodes were removed, and the animals were recovered from anesthesia through an intramuscular injection of antisedan (0.04 mg/kg, Atipamezole, Zoetis).

Visual stimulus

The visual stimuli from Zhu and colleagues (2025) were employed for the present study. These stimuli consisted of 200 ms of coherently leftward-moving dots (Fig. 1C) at 10 different speeds (2.0-, 2.8-, 4.0-, 5.8-, 8.0- 11.3-, 16-, 22.8-, 32-, and 64-deg/s) (Fig. 1D). Each trial was followed by a 0.9–1.1 s intertrial interval when the dots were stationary but still present on the screen. Every recording block consisted of 20 trials from each of the 10 speeds played in randomized order, and a total of 7–8 blocks were presented during each day of recording for all recording days before and after deafening (Fig. 1E). Each dot presented was soft-edged, 0.30° visual angle in size, and 258 cd/m² in luminance. There were 400 dots with randomly generated locations in a black background on the full screen. When moving towards the edge of the screen, they faded out and reappeared on the right side. VEPs were produced from the average of 140–160 trials for each stimulus speed.

The stimuli were programmed in MATLAB with the Psychophysics Toolbox^86–88. All images were rendered by a graphics card (AMD, Radeon HD 6800 Series) and processed using a video processing unit (Cambridge Research System, Bits#) for the generation of a timestamp before each set of stimulus motion.

Data analysis

Digital filtering was applied offline to the EEG signal, where a band-pass filter with a range of 1 to 30 Hz was used to isolate the desired frequency components. Furthermore, the signal was digitally notched at a frequency of 60 Hz to eliminate unwanted interference. Epochs were extracted from 200 ms prior to stimulus onset and 600 ms post stimulus onset, grouped by the motion-onset speed, and subsequently averaged together to create averaged traces i.e. VEP waveforms. To correct for slow fluctuations in EEG signal, all averaged traces were shifted by the mean of their 200 ms pre-stimulus baselines. Post baseline correction, the amplitudes and latencies of the dominant positive peak i.e. “P1”, were automatically detected using a customized script in MATLAB for statistical analysis.

To study the signal power of the waveform, we computed root-mean-square (RMS) values using the built-in MATLAB function rms(). Each averaged waveform was separated to produce a 400 ms RMS post-stimulus response window (RMSresponse) and a 200 ms RMS pre-stimulus zero-mean baseline window (RMSbaseline). The corrected RMS value (RMScorrected) was derived using the following equation:

Statistics

Given the sample size of 4 subjects employed in this study, non-parametric statistics were used. Motion-onset of stimuli of 10 different speeds were used to generate 10 different VEPs of varying power respectively. The median of these 10 VEPs were used for statistical comparison, as it focuses on the more stable middle values of the dataset and is not affected by extreme high or low power VEPs generated by the slowest or fastest stimulus speeds. In the time course analysis, the datapoints for signal power, P1 amplitude, and P1 latency were binned within all hearing baseline measurements before deafening as well as using the bin edges of every 100-day period post deafening. This forms the “H” baseline bin, the 0–100-day period bin, 100–200-day period bin, 200–300-day period bin, and 300–400-day period bin, respectively. The Mann–Whitney U test (ranksum() in MATLAB) was used to compare VEPs at different bins.

Additionally, a Tau-U test was employed as a quantitative approach to measure the nonoverlap between the hearing and deaf phases as well as the deaf phase trend using an online calculator⁵³. The data from each subject is first analyzed individually, then an aggregate value is derived and weighted by the length of the data series. The interpretation of the Tau-U effect size followed the guideline⁵⁵, where values ranging from 0 to 0.62 signify a small effect, values between 0.63 and 0.92 represent a moderate effect, and values from 0.93 to 1 indicate a strong effect. The Tau-U test is used in studying single-case experimental data where intensively repeated observations are taken for individual subjects during several phases^55,89.

Permutation tests⁵⁴ were also performed on the signal powers and P1 amplitudes for each of the 10 moving-dot speeds. Seven recording timepoints were taken from before and after deafening and were compared in their medians across subjects for each speed. 10,000 randomized permutations of these fourteen post-deafening timepoints were performed for each stimulus speed condition. The chance of real median differences being greater or smaller than those derived from permutation is a number out of 10,000 permutations, and this acts as the significance level p. Multiple comparisons were corrected using the Bonferroni correction method⁹⁰, where the Bonferroni-adjusted significance criterion for each test would be .005 (.05/10, single tailed).

Author contributions

The research question and experimental design were conceived by SZ, XB, and SGL. Data collection was conducted by SZ and XB. SZ and XB analyzed the data. SZ and XB drafted the figures. SZ drafted the manuscript. XB, and SGL edited the manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.