A programmable rhodopsin photocurrent in a synthetic membrane

What the work claims

This is a primary experimental result in acellular biophysics, not a modeling study or a review. The authors express TARA76, a member of the proteorhodopsin family of light-driven microbial ion pumps, in a bacterial host, reconstitute single copies of it into lipid nanodiscs, and embed those into a black lipid membrane suspended across a micro-aperture in a two-chamber microfluidic device. Under laser illumination they record light-gated ionic currents with picoampere sensitivity, peaking near 1 picoampere, and characterise how those currents respond to voltage, light power, pH and the ions in solution.

Two claims make the paper more than a competent reconstitution. The first is a strong and, the authors say, previously unreported dependence of the photocurrent on sodium ions, which they attribute to sodium stabilising the protein in its active shape. The second is that the orientation of the protein inside the artificial membrane, and therefore the direction of the current it drives, can be controlled from outside by applying a direct-current voltage during bilayer formation. It is worth stating plainly what this is not: TARA76 here is a purified protein in a synthetic bilayer, with no neuron, no synapse, and no living tissue anywhere in the system. The relevance to organoid intelligence is that of an engineerable transduction primitive, not a demonstration of computation.

How it works

A rhodopsin is a membrane protein wrapped around a retinal molecule, the same light-absorbing cofactor the eye uses. When a photon is absorbed the retinal changes shape, and that isomerisation drives a cycle of internal proton transfers that pushes charge across the membrane. Proteorhodopsins are a large, diverse family of such pumps, mostly moving protons outward. To get one working outside a cell, the authors first place it in a nanodisc, a nanometre-scale patch of lipid held flat by two belt-like scaffold proteins, which by design carries a single rhodopsin per disc. Those discs are then delivered into a black lipid membrane, a single free-standing lipid bilayer spanning the aperture between the two fluid chambers, each fitted with a silver-chloride electrode.

Before adding protein, the team confirm the bilayer electrically: an intact membrane behaves like a capacitor, and a benchmark pore-forming peptide produces the expected stepwise channel openings of about 2 picoamperes. With TARA76 present, switching the laser on produces a fast rise in current, a decay to a steady stationary level, and a return to baseline when the light goes off. The authors are careful to note that the roughly 80 millisecond rise and 85 millisecond decay constants reflect the limited bandwidth of a setup tuned to detect sub-picoampere signals, not the intrinsic speed of the protein. The steady current is proton pumping; the initial spike reflects the early charge movements of retinal isomerisation. The current-voltage relation is essentially linear, with a reversal potential near positive 75 millivolts, which means the polarity of the applied field can reverse the direction of proton flow. The measured photoconductance is on the order of 1 picosiemens in direct current and about 20 picosiemens under alternating drive, where the wider bandwidth reveals more of the signal.

The sodium result comes from swapping the cation in the buffer. At matched conditions the stationary current was about 0.35 picoamperes with sodium present, against 0.06 with potassium and 0.02 with calcium. A structural model built with AlphaFold 3.0 superimposes cleanly on bacteriorhodopsin, the textbook proton pump, and every residue known to carry the proton in that pump has a counterpart in TARA76. The model also predicts sodium binding sites, one of them adjacent to a glutamate that is the counterpart of a proton-pathway residue in bacteriorhodopsin, supporting the reading that sodium holds the protein in a productive conformation. Whether sodium is also carried across, in a coupled sodium-proton flux, is left as an open question the authors say needs direct transport assays, mutagenesis and a real structure. Finally, the orientation control: the alternating-current analysis shows the light-induced conductance is asymmetric between the two halves of the voltage cycle, evidence that the proteins sit with a preferred orientation, and poling the membrane with a direct-current voltage as it forms biases that orientation on purpose, fixing which way the photocurrent runs.

Where a skeptic should push

The single most load-bearing assumption is the one the abstract invites without stating: that a light-gated current in a synthetic bilayer is a meaningful step toward interfacing living neural tissue. Stress that assumption and the distances become obvious. The peak current is about 1 picoampere, spread across an estimated 335,000 active molecules at a density around 65 per square micrometre, with the discs replacing under 1 percent of the membrane. That is a beautifully clean measurement of a very small effect, and nothing in it involves a cellular electrical load, let alone a neuron that would need tens of picoamperes to depolarise.

It is worth separating what is demonstrated from what is asserted. Demonstrated: reproducible, protein-dependent, light-gated picoampere currents; a clear cation preference for sodium; and direction control by a poling field. The dehydration robustness that motivates the work is a structural property reported in earlier work on this protein class, not a photocurrent measurement made here after drying. Asserted or extrapolated: the framing of TARA76 as a candidate for single-photon or quantum light sensing and for neuromorphic components, which rests on analogy to the retina and to the family rather than on any measurement here. The sodium role is well supported by the cation series and the model, but the model is a prediction, not a crystal structure, and the coupled-transport idea is explicitly a hypothesis. The single-protein-per-disc figure is taken from the literature, and the reversal potential is noted to shift with conditions. None of these caveats undo the result; they bound it to what it is, a careful acellular characterisation of a transducer.

A light-to-current write channel for tissue

Optogenetics, the standard way to write light into neural tissue, works by putting a gene for an opsin into each target cell and waiting for the cell to build the protein in its own membrane. That demands viral delivery, expression, and tolerance of whatever mosaic of expression results. This work is the mirror image of that approach: an opsin-class transducer operating entirely outside cells, in a membrane that can be fabricated on a surface, and carrying two properties that in-cell optogenetics does not offer. Its polarity is deterministic, set by the poling field during assembly rather than left to the accidents of protein insertion, and its structure is reported to survive complete drying, which points toward a component that could be manufactured and stored rather than cultured. The non-obvious implication is that the interesting object for organoid intelligence is not a photocurrent record but a fabricable, orientation-programmable, light-addressable current source that could sit as an actuator layer against tissue without genetically modifying the tissue at all.

The opportunity is grounded in three specific mechanisms in the paper. Because current direction follows protein orientation, and orientation is fixed by an external field, a patterned, oriented rhodopsin membrane would have a defined write polarity, which is exactly the property an interface needs to push charge one way on demand. Because the sodium dependence ties activity to the surrounding ionic environment, the physiological saline that bathes an organoid is a natural operating medium and a possible tuning knob. And because the protein's structure is reported to tolerate dehydration, the transducer layer is a candidate for shelf-stable fabrication rather than a fragile live culture, once photocurrent function after a dry-store cycle is actually demonstrated. That is a coherent, mechanism-anchored case for a bio-hybrid write channel that avoids viral opsin delivery.

The threat, and it is the same size as the opportunity, is the gap the currents expose. One picoampere from a synthetic photodiode is far below what would move a neuron, and there is no demonstration anywhere in the paper of coupling to a living membrane, so the entire step from acellular transducer to writing into a spiking network remains unsolved. Present closed-loop tissue systems still rely on an external microelectrode array to read spikes out and deliver stimulation in,² and this protein does neither of those things yet. The dual-use caution is more about language than misuse: calling an acellular photocurrent neuromorphic or quantum borrows an authority the data do not supply, and for a field already fighting hype, the honest label is a promising write-side primitive with an unproven path to tissue.

The bottom line

What is established: TARA76 generates reproducible, light-gated picoampere currents in a cell-free bilayer; those currents are predominantly proton based with a strong and newly reported sodium dependence; and their direction can be fixed by a poling field.¹ The dehydration robustness that motivates the design is a structural observation carried over from earlier work, not a demonstration that the photocurrent survives drying and rehydration. What remains hypothesis or aspiration: any role as a neural write channel, a single-photon detector, or a neuromorphic element. The claim would be confirmed by coupling oriented TARA76 membranes to a real neuron or organoid and showing light-evoked, direction-controlled changes in membrane potential or firing. It would be broken, for organoid purposes, by an inability to raise the current into a physiologically relevant range once a cellular load is attached, or by the loss of orientation control at that interface. As a piece of biophysics it is clean and honest; as a step toward computing on living tissue it is a component looking for its coupling.