Photomechanical Responses in Drosophila Photoreceptors

See allHide authors and affiliations

Science  12 Oct 2012:
Vol. 338, Issue 6104, pp. 260-263
DOI: 10.1126/science.1222376

Feeling the Light

There has been a link missing in our understanding of the biochemical signaling mechanism initiated when photons are absorbed by rhodopsin in the photoreceptor cells of Drosophila eyes. Photoisomerization of rhodopsin activates a heterotrimeric guanine nucleotide—binding protein which leads to activation of phospholipase C. But how phospholipase C activates the transient receptor potential (TRP) or the transient receptor potential–like (TRPL) ion channels that produce the rest of the cellular response has been unclear. Hardie and Franze (p. 260; see the Perspective by Liman) propose that there is a physical mechanism at work. Atomic force microscopy revealed light-induced contractions of photoreceptor cells. Consistent with a physical coupling mechanism, manipulations of the membrane altered responses of the photoreceptor cells to light. Thus, physical changes in the membrane appear to couple phospholipase C activity to the opening of mechanosensitive TRP and TRPL channels.


Phototransduction in Drosophila microvillar photoreceptor cells is mediated by a G protein–activated phospholipase C (PLC). PLC hydrolyzes the minor membrane lipid phosphatidylinositol 4,5-bisphosphate (PIP2), leading by an unknown mechanism to activation of the prototypical transient receptor potential (TRP) and TRP-like (TRPL) channels. We found that light exposure evoked rapid PLC-mediated contractions of the photoreceptor cells and modulated the activity of mechanosensitive channels introduced into photoreceptor cells. Furthermore, photoreceptor light responses were facilitated by membrane stretch and were inhibited by amphipaths, which alter lipid bilayer properties. These results indicate that, by cleaving PIP2, PLC generates rapid physical changes in the lipid bilayer that lead to contractions of the microvilli, and suggest that the resultant mechanical forces contribute to gating the light-sensitive channels.

In most invertebrate photoreceptor cells, the visual pigment (rhodopsin) and other components of the phototransduction cascade are localized within tightly packed microvilli (tubular membranous protrusions), together forming a light-guiding rod-like stack (rhabdomere; Fig. 1). After photoisomerization, rhodopsin activates a heterotrimeric guanine nucleotide–binding protein (Gq protein), releasing its guanosine triphosphate–bound α subunit, which in turn activates phospholipase C (PLC; Fig. 1C). How PLC activity leads to gating of the light-sensitive transient receptor potential channels (TRP and TRPL) in the microvilli is unresolved (13). PLC hydrolyzes the minor membrane phospholipid phosphatidylinositol 4,5-bisphosphate (PIP2), yielding soluble inositol 1,4,5-trisphosphate (InsP3), diacylglycerol (DAG, which remains in the inner leaflet of the microvillar lipid bilayer), and a proton. The light-sensitive channels in Drosophila photoreceptors can be activated by a combination of PIP2 depletion and protons (4), but it remains unclear how PIP2 depletion might contribute to channel gating. It has been speculated that changes in membrane properties play a role (4, 5), and members of the TRP ion channel family have been repeatedly, although controversially, implicated as mechanosensitive channels (6, 7). This led us to ask whether cleavage of the bulky, charged inositol head group of PIP2 (Fig. 1C) from the inner leaflet might alter the physical properties of the lipid bilayer in the microvilli, resulting in mechanical forces that contribute to channel gating.

Fig. 1

AFM measurements of photomechanical responses. (A) An AFM cantilever contacts distal tips of ommatidia in an excised Drosophila retina. (B) An ommatidium, containing photoreceptors (orange) and pigment cells (red). Elements of the phototransduction cascade are contained within microvillar rhabdomeres (two shown in longitudinal section, seven in cross section), which are rodlike stacks ~80 μm in length containing ~30,000 microvilli. Right: Electron micrograph cross section of a rhabdomere (scale bar, 1 μm), showing tubular microvilli, each ~50 nm in diameter, with lumen in diffusional continuity with the cell body. (C) Phototransduction cascade. Rhodopsin (R) is photoisomerized to metarhodopsin (M*), which catalyzes release of the Gq protein α subunit to activate PLC. PLC hydrolyzes PIP2 (red), leaving DAG (green) in the membrane. Ca2+ influx via TRP channels inhibits PLC. (D) Lower traces: AFM measurements of contractions (cantilever z-position) in a wild-type retina in response to 5-ms flashes, with intensity increased from ~200 to 8000 effectively absorbed photons per photoreceptor. Blue traces: Whole-cell current-clamped voltage responses to the same stimuli recorded from a dissociated photoreceptor cell. (E) Contractions evoked by 5-ms flashes covering the full intensity range (~200 to 106 photons) in a wild-type retina. (F) Same on faster time base. (G) Response versus intensity (R/I) functions of contractions (nm) from wild-type retinae (mean ± SEM, N = 13), and peak voltage (mV) recorded from dissociated photoreceptors (blue; means ± SEM, N = 6). (H) Responses to flashes (~5 × 104 photons) in trpl mutant before and after (red) channel block by 50 μM La3+ and 10 μM ruthenium red (RR), which prevents inhibition of PLC by Ca2+ influx. Blue trace: Lack of response in norpAP24 (PLC mutant) despite using flashes of higher intensity (~2 × 105 photons; N = 3). (I) R/I function of contractions from trpl retinae before and after (red) channel block by La3+ and RR (means ± SEM, N = 5). For intensity calibration, see fig. S2.

Remarkably, Drosophila photoreceptors responded to light flashes with small (<1 μm) but rapid contractions that were directly visible in dissociated cells via bright-field microscopy (see movie S1). To obtain improved temporal and spatial resolution, we recorded these photomechanical responses with an atomic force microscope (AFM) (8), positioning the AFM cantilever on the distal tips of photoreceptors in a whole excised retina glued to a coverslip (Fig. 1A). Contact force (~100 pN) was maintained constant, so that changes in sample height resulted in immediate, matching changes in the cantilever’s z-position. Contractions were elicited indefinitely by repeated brief flashes of modest intensity, with kinetics similar to those of electrical responses recorded from dissociated photoreceptors (Fig. 1D and fig. S1). The latencies of contractions induced by the brightest stimuli (4.9 ± 0.9 ms, mean ± SEM, N = 11; Fig. 1F) were significantly shorter than the latencies of voltage responses to the same stimuli (6.6 ± 0.6 ms, N = 6; P = 0.002, unpaired two-tailed t test). Only a few hundred effectively absorbed photons per photoreceptor were required to elicit detectable contractions, which saturated with flashes containing ~106 photons, corresponding to ~30 effectively absorbed photons per microvillus (Fig. 1, E and G). This intensity dependence overlapped with that of the electrical response (Fig. 1G and supplementary text). Like the electrical responses, the contractions were eliminated in mutants lacking PLC (norpAP24), which shows that they too required PLC activity (Fig. 1H).

Because PLC activity is normally terminated by Ca2+ influx through the light-sensitive channels, net PIP2 hydrolysis is enhanced when the light-sensitive current is blocked (9, 10). We therefore measured contractions in trpl mutants expressing only TRP light-sensitive channels before and after blocking TRP channel activity with La3+ and ruthenium red (RR). Indeed, after blocking the light-sensitive channels the contractions were enhanced, more sensitive to light, and saturated at lower intensities, corresponding to only ~1 to 5 effectively absorbed photons per microvillus (Fig. 1, H and I). In the absence of Ca2+ influx, such intensities deplete virtually all microvillar PIP2, resulting in temporary loss of sensitivity to light (10). After such saturating flashes, the photomechanical response was also temporarily refractory, recovering sensitivity with a time course (t1/2 ~ 40 s) similar to that of PIP2 resynthesis (10). By contrast, without channel blockers, sensitivity recovered within ~10 s (fig. S3).

Blockade of all light-sensitive current in these experiments also shows that the contractions cannot result from any downstream effects of Ca2+ influx or osmotic changes caused by ion fluxes associated with the light response. Therefore, these results indicate that the contractions result from hydrolysis of PIP2. Although we do not exclude other downstream effects of PLC, the speed of the contractions supports a simple and direct mechanism. Cleavage of the bulky head groups from PIP2 molecules, which represent 1 to 2% of lipids in the plasma membrane, leaves DAG in the membrane, which occupies a substantially smaller area than PIP2. This should increase membrane tension, leading to shrinkage of the microvillar diameter, as reported for the action of PLC on the diameter of artificial liposomes (11). Integrated over the stack of ~30,000 microvilli, such a mechanism seems capable of accounting for the observed macroscopic contractions, which represent at most ~0.5% of the rhabdomere length (see supplementary text). Within each microvillus, we suggest that the alteration to the mechanical properties of the lipid bilayer may contribute to channel gating.

To test whether phototransduction generates sufficient mechanical forces to gate mechanosensitive channels (MSCs), we made whole-cell patch-clamp recordings from dissociated photoreceptors lacking all light-sensitive channels (trpl;trp double mutants, or trpl mutants exposed to La3+ and RR). We then perfused the photoreceptors with gramicidin, a monovalent (Ca2+-impermeable) cation channel and one of the best-characterized MSCs, which is known to be regulated by changes in bilayer physical properties (12, 13). Incorporation of gramicidin channels into the membrane generated a constitutive inward current that stabilized after a few minutes (Fig. 2A). Despite having replaced the native light-sensitive channels with MSCs, the photoreceptors still responded to light, with a rapid increase in the gramicidin-mediated current (Fig. 2, B and C). Like the photomechanical responses recorded after blocking Ca2+ influx through the light-sensitive channels (Fig. 1H), these gramicidin-mediated responses inactivated slowly (Fig. 2C), were temporarily refractory to further stimulation and had a similar intensity dependence (Fig. 2D).

Fig. 2

Light-induced modulation of gramicidin channels. (A) Whole-cell recordings from trpl;trp mutant photoreceptor cell lacking all native light-sensitive channels. Perfusion with gramicidin induced a constitutive inward current. (B) Flashes of increasing intensity (5 × 103 to 4 × 105 effective photons), each 1 s (denoted by bar at upper left), up-regulated the current. (C) Averaged responses (±SEM) to 100-ms flashes containing 1.3 × 104 (middle trace, N = 4) and 3.7 × 104 (lower trace, N = 10) effective photons (data pooled from trpl;trp and trpl mutants recorded in the presence of La3+ and RR). The same flashes delivered before gramicidin application (controls) induced residual, noise-free transient currents of uncertain origin. (D) R/I function (after subtracting control responses measured before gramicidin perfusion) expressed as a fraction of the steady-state gramicidin current I/ISS (means ± SEM, N = 4). Dotted curve: R/I function of contractions measured by AFM in trpl mutant in the presence of La3+ and RR, replotted from Fig. 1I.

To test whether the light-sensitive channels were mechanically sensitive, we manipulated membrane tension osmotically. Channels were not directly activated by perfusing cells with hyper- or hypo-osmotic solutions; however, we reasoned that it would be impossible to mimic the exact physical effects of PIP2 hydrolysis, which would include a specific combination of changes in membrane tension, curvature, thickness, lateral pressure profile, charge, and pH. We therefore tested whether osmotic manipulation could enhance or suppress light-sensitive channel activity. In wild-type photoreceptors, light-induced currents were rapidly and reversibly facilitated by ~50% after perfusion with hypo-osmotic solutions (200 mOsm), which, like PIP2 depletion, would be expected to alleviate crowding between phospholipids, increase tension, and reduce membrane thickness. Conversely, responses in hyperosmotic (400 mOsm) solutions were about half those in control solutions (Fig. 3). Analysis of single-photon responses (quantum bumps) indicated that modulation resulted from changes in both quantum efficiency (fraction of rhodopsin photoisomerizations generating a quantum bump) and bump amplitude (fig. S4 and supplementary text). Recordings from trp and trpl mutants showed that both TRP and TRPL channels were modulated, although facilitation of currents mediated by TRPL channels (in trp mutants) was more pronounced (Fig. 3). Modulation of the light response by osmotic manipulation was at least as pronounced in Ca2+-free bath (Fig. 3D), indicating that facilitation by membrane stretch did not result from leakage of Ca2+ into the cell from the extracellular space.

Fig. 3

Modulation of light-sensitive channels by osmotic pressure. (A to D) Whole-cell voltage-clamped responses to 1-ms flashes (~50 effective photons) in control bath (300 mOsm) were reversibly increased by perfusion with 200 mOsm solution and suppressed by 400 mOsm in wild-type (A), trpl (B), and trp (C) mutants as well as in wild-type photoreceptors recorded in Ca2+-free solutions (D). (E) Response amplitudes (I/I300) after hyperosmotic (400 mOsm) and hypo-osmotic (200 mOsm) challenges normalized to control responses in 300 mOsm bath. Data are means ± SEM; N = 4 to 8 cells. All conditions plotted were significantly increased (200 mOsm) or decreased (400 mOsm) relative to control responses from the same cells [P < 0.005; analysis of variance (ANOVA) followed by posttest for trend]. (F) Spontaneous TRP channel activity (from trpl mutant) after several minutes of recording, using pipettes lacking nucleotide additives. Perfusion with 400 or 200 mOsm (bar) reversibly suppressed and facilitated this “rundown current” (RDC). (G) Channel noise resolved on a faster time base, plus trace recorded in dark before onset of RDC. (H) Left: Amplitude of steady-state RDC normalized to value at 300 mOsm. Right: Effective single-channel conductance (γ) estimated by variance/mean ratio (means ± SEM, N = 7). Although macroscopic RDC was substantially modulated (P < 0.001; ANOVA, posttest for trend), single-channel conductance was not significantly affected by osmotic manipulation (P > 0.2).

To test whether modulation might have been mediated by effects on upstream components of the cascade such as PLC (14), we measured the activity of spontaneously active TRP channels in recordings made with pipettes lacking adenosine triphosphate. Under these conditions, PIP2 becomes depleted (thereby removing PLC’s substrate), sensitivity to light is lost, and the TRP channels enter a constitutively active (“rundown”) state uncoupled from the phototransduction cascade (15, 16). Nonetheless, the channels were still similarly modulated by osmotic manipulation, whereas single-channel conductance, estimated by noise analysis, was unaffected (Fig. 3, F to H). These results indicate that osmotic pressure directly modulated the open probability of both TRP and TRPL channels.

MSCs such as gramicidin are sensitive to amphiphilic compounds, which insert into the lipid bilayer. Because they are attracted to anionic phospholipids, cationic amphipaths insert preferentially into the inner leaflet, where they increase crowding, promote negative (concave) curvature, and decrease membrane stiffness (17, 18). We found that four structurally unrelated cationic amphipaths were all effective, reversible inhibitors of the light-induced current. Neither light-induced PLC activity (measured using a genetically targeted PIP2-sensitive biosensor to monitor PIP2 hydrolysis) nor single-channel conductance were substantially affected (fig. S5). The 50% inhibitory concentrations (IC50 values) were much higher than those of their traditional drug targets and ranged over approximately three orders of magnitude. However, after correcting for pKa and partitioning, the effective concentration of the compounds in the membrane was similar (~5 mM) in each case (Fig. 4). Thus, their mode of action is likely related to their physicochemical properties rather than conventional drug-receptor interactions. Because cationic amphipaths are also lipophilic weak bases, and because we propose that protons are also critical for activating the light-sensitive channels (4), an alternative but not mutually exclusive possible mechanism of action is as lipophilic pH buffers of the membrane environment. We also note that polyunsaturated fatty acids (PUFAs) such as arachidonic and linolenic acid—which are effective activators of both TRP and TRPL (5, 19)—are not only anionic amphipaths (predicted to have opposite effects to cationic amphipaths) but also, as weak acids, natural protonophores; such a dual action could account for their agonist effect.

Fig. 4

Cationic amphipaths inhibit the light response. (A) Response to 1-ms flashes in trp mutants in the presence of procaine (PROC), imipramine (IMP), trifluoperazine (TFP), and chlorpromazine (CPZ). Control responses before (turquoise) and after washout (blue) are superimposed. (B) Dose response functions (means ± SEM: PROC, N = 4; IMP, N = 3; TFP, N = 3; CPZ, N = 6) fitted by inverse Hill curves (slope constrained at n = 2), based on raw values (IC50: PROC, 3.1 mM; IMP, 27 μM; TFP, 4.9 μM; CPZ, 4.5 μM). (C) Same as in (B) after correction for pKa and octanol partition coefficients, reflecting predicted concentration in the lipid membrane (IC50: PROC, 2.7 mM; IMP, 7 mM; TFP, 3.4 mM; CPZ, 2.3 mM).

The mechanism of activation of the light-sensitive channels in invertebrate microvillar photoreceptors has long remained an enigma (2, 3, 20). Neither InsP3 nor DAG—the two obvious products of PIP2 hydrolysis—are reliable agonists for the light-sensitive channels. Although PUFAs are effective agonists and might be generated from DAG, a DAG lipase with the appropriate specificity has not been found in the photoreceptors (21). By contrast, two neglected consequences of PLC activity—the depletion of its substrate (PIP2) together with protons released by PIP2 hydrolysis—were recently shown to potently activate the light-sensitive channels in a combinatorial manner (4). Our results support the hypothesis that the effect of PIP2 depletion is mediated mechanically by changes to the physical properties of the lipid bilayer, thereby introducing the concept of mechanical force as an intermediate or “second messenger” in metabotropic signal transduction.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S5

Movie S1

References (2231)

References and Notes

  1. Acknowledgments: We thank D. G. Stavenga, S. B. Laughlin, and M. Postma for comments on the manuscript; O. Andersen for advice on the use of gramicidin; and J. Grosche (Effigos AG) for artwork for Fig. 1A. Supported by UK Biotechnology and Biological Sciences Research Council grant BB/G0068651 (R.C.H.) and a UK Medical Research Council Career Development Award (K.F.). Author contributions: project initiation, R.C.H.; whole-cell electrophysiology experiments, R.C.H.; AFM measurements, K.F., R.C.H.; paper written by R.C.H. with contribution from K.F.
View Abstract

Navigate This Article