A flexible photoacoustic retinal prosthesis

Citation: Leong A, Li Y, Ruikes TR, et al. A flexible photoacoustic retinal prosthesis. Nat Commun 17, 815 (2026). https://doi.org/10.1038/s41467-025-67518-6

• Received: 3 February 2025 | Accepted: 3 December 2025 | Published: 23 December 2025 | VoR: 21 January 2026
• Open access: CC BY-NC-ND 4.0
• Funding: NIH and Axorus SAS (COI: Moulet & Losasco are major Axorus stakeholders; Yang & Cheng are minor stakeholders)

Authors & Affiliations

Equal contributions: Audrey Leong, Yueming Li, T.R. Ruikes, Serge Picaud, Hélène Moulet.

• Ji-Xin Cheng (jxcheng@bu.edu) — BU College of Engineering, Distinguished Professor
• Chen Yang (cheyang@bu.edu) — BU, Professor of Chemistry and ECE
• Yueming Li — BU (partial Axorus funding)
• Hélène Moulet (helene.moulet@axorus.com) — Axorus SAS; major stakeholder
• Serge Picaud (serge.picaud@inserm.fr) — Institut de la Vision, Sorbonne Université, INSERM
• Jean-Damien Losasco (J-DL) — major Axorus stakeholder

Abstract

Flexible PDMS/candle-soot/PDMS film converts 1030-nm NIR pulsed laser into localised 51-µm acoustic field. Stimulates mechanosensitive retinal cells in wild-type and degenerated (P23H) rat retinae ex vivo. Subretinal implantation: pulsed laser → neural modulation along visual pathway to superior colliculus (fUSI). Biosafety: temperature increase <0.52 °C at film surface; within FDA ophthalmic ultrasound guidelines.

Key Technical Details

Film composition

• PDMS/CS/PDMS (thick, 115 µm): candle soot (CS) absorber sandwiched between two PDMS layers. Young’s modulus 2.12 ± 0.10 MPa (minimises immune response).
• PDMS-CNT (thin, 40 µm): uniformly mixed; approaches PRIMA’s 30 µm clinical thickness.
• Fabrication: flame-synthesised CS on glass, PDMS spin-coated (5:1 ratio, 500 rpm, 110 °C, 15 min); oxygen plasma treatment for hydrophilicity.
• Absorbs >99% of 1030-nm laser energy (IR does not pass through to activate photoreceptors).

Acoustic characterisation

• 4.2-ns pulses at 7 µJ/pulse → peak-to-peak pressure 146.2 kPa at 0.9 mm; conversion efficiency ~26 kPa/µJ.
• Central acoustic frequency: 42.2 MHz; −6 dB bandwidth: 29.6–59.9 MHz.
• Lateral resolution (50-µm fibre illumination): 51 µm FWHM at Z=34 µm from film surface.
• Pressure linear with laser energy (R²=0.9945); precise modulation by energy adjustment.
• Temperature increase: 0.52 ± 0.09 °C (max at film surface); cumulative baseline drift 0.21 °C over 40 s — far below thermal neural modulation threshold.

Ex vivo Results

Wild-type Long-Evans (LE) rats (n=4 rats)

• 74% of sites showed RGC modulation; 100/129 RGCs (78%) within 300 µm responded.
• Predominantly excited responses (92% of responders).
• Mean firing rate: 66 ± 3.7 Hz; mean latency: 51 ± 34.2 ms.

P23H degenerated rats (RP model, n=4 rats)

• 39% RGC modulation (significantly lower than LE, P<0.001 Mann-Whitney).
• 89/229 RGCs (39%) responded; predominantly excited (93%).
• Mean firing rate: 29 ± 2.88 Hz; mean latency: 89 ± 65 ms; only 36% with latency <45 ms.
• Higher stimulation thresholds in degenerated retina (consistent with ultrasound literature).

Pharmacological dissection of mechanosensitivity

• L-AP4 (blocks photoreceptor → ON bipolar synapse): suppressed short-latency (<45 ms) responses → photoreceptors generate most short-latency RGC responses to ultrasound.
• ACET added to L-AP4: further suppressed → inner retinal cells upstream of RGCs also contribute.
• CPP+CNQX in P23H (blocks all glutamate transmission): nearly abolished PA responses → mechanosensitive cells are upstream of RGCs; glutamate neurotransmission required.

Spatial resolution (ex vivo)

• 73% of LE and 70% of P23H RGCs modulated within 100 µm of laser spot.
• Firing rate negatively correlated with distance from laser spot.
• Different cell populations activated at different sites (spatial selectivity demonstrated).

In vivo Results

Biocompatibility (subretinal implantation, LE and P23H rats)

• OCT follow-up: 7, 15, 30, 90 dpi. Correct positioning, no retinal tearing, no major inflammation.
• Retinal thinning above implant in LE (photoreceptors detach from RPE — same as PRIMA and all subretinal prostheses); stable in P23H up to 120 dpi.
• Immunohistochemistry: activated microglia at implant site (expected); Müller glia activation in LE only. No inflammation in surrounding area except implant insertion track (surgical artefact).
• Thermal safety: <1 °C; MI <0.03 (PDMS/CS/PDMS) and <0.1 (PDMS-CNT); ISPTA <0.06 and 0.9 mW/cm² respectively → within FDA ophthalmic ultrasound guidelines.

Superior colliculus activation (fUSI, LE rats)

• 1030-nm laser (8 × 125-ms bursts over 2 s, repeated every 15 s, 15 total stimulations per recording).
• PDMS/CS/PDMS: activated 25 ± 4% of full-field-white-light-activated cSC area (n=4 recordings).
• PDMS-CNT: activated 38 ± 10% (n=5 recordings).
• Comparable rCBV amplitude to 400-µm visible (595-nm) laser spot on healthy retina.
• Control (direct 1030-nm on healthy retina, no film): no significant cSC activation — confirms prosthetic effect is film-mediated, not direct IR activation.

Discussion Highlights

• PA pressure thresholds (0.05 MPa) are two orders of magnitude lower than external transducer ultrasound stimulation (Lu et al., Nat Commun 2024) — subretinal proximity explains efficiency.
• Film is spatially continuous and photon-multiplexable → can scale to whole macula coverage (25 mm²) with pixel density up to 2500 px/mm² — not limited to rigid chip footprint.
• Challenges: mechanosensitivity structures unknown (likely Piezo1/2 channels); long-term stimulation effects need study; need sub-20 µm pixel pitch for 20/20-equivalent resolution; surgery artefact (insertion track) needs refinement; RPE/choroid effects of bidirectional acoustic waves not yet studied.

Comparison with PRIMA (photovoltaic)

Parameter	PRIMA (photovoltaic)	PA film (this study)
Mechanism	NIR light → photovoltaic → electrical stimulation of bipolar cells	NIR laser → photoacoustic → ultrasound → mechanosensitive cells
Resolution (pixel)	100 µm (378 pixels, 2×2 mm)	51 µm lateral US field (from 50 µm fibre)
Achieved acuity (human)	20/460–20/565 in AMD patients	Pre-clinical; fUSI colliculus activation demonstrated in rats
Film thickness	~30 µm	115 µm (thick) or 40 µm (thin CNT version)
Stage	CE-marked, commercial EU launch July 2026	Pre-clinical; first-in-human safety study planned
Company	Science Corp. (US)	Axorus (France, BU spin-out)

Commercial Translation

• BU patents licensed to Axorus SAS (Paris, France) — developing a contact lens form factor for AMD.
• Chen Yang plans BU start-up for US commercialisation.
• Next step: first-in-human safety study; then clinical trial in blind patients.
• Timeline estimate (Yang): ~decades to clinic (prior comparable technology took >20 years).