Piezo Is Essential for Amiloride-Sensitive Stretch-Activated Mechanotransduction in Larval Drosophila Dorsal Bipolar Dendritic Sensory Neurons

Abstract

Stretch-activated afferent neurons, such as those of mammalian muscle spindles, are essential for proprioception and motor co-ordination, but the underlying mechanisms of mechanotransduction are poorly understood. The dorsal bipolar dendritic (dbd) sensory neurons are putative stretch receptors in the Drosophila larval body wall. We have developed an in vivo protocol to obtain receptor potential recordings from intact dbd neurons in response to stretch. Receptor potential changes in dbd neurons in response to stretch showed a complex, dynamic profile with similar characteristics to those previously observed for mammalian muscle spindles. These profiles were reproduced by a general in silico model of stretch-activated neurons. This in silico model predicts an essential role for a mechanosensory cation channel (MSC) in all aspects of receptor potential generation. Using pharmacological and genetic techniques, we identified the mechanosensory channel, DmPiezo, in this functional role in dbd neurons, with TRPA1 playing a subsidiary role. We also show that rat muscle spindles exhibit a ruthenium red-sensitive current, but found no expression evidence to suggest that this corresponds to Piezo activity. In summary, we show that the dbd neuron is a stretch receptor and demonstrate that this neuron is a tractable model for investigating mechanisms of mechanotransduction.

Fig 1. Stretch-activated features of dbd neuron receptor potentials are similar to those in muscle spindles.

(A) Immunostaining of yw; Gal4^109-80, UAS-mCD3-GFP larval thoracic segment PNS (peripheral nervous system) with α-GFP. The dbd neuron is readily identifiable by its distinctive morphology–a cell body (arrow), present in each thoraco-abdominal segment, with two long dendrites (arrowheads) projecting longitudinally along the anterior/posterior axis. These dendrites are thought to be the sensory transduction elements. (B) Stretch-evoked receptor potential recording from dbd neurons exhibit a complex receptor potential profile strongly resembling that in mammalian muscle spindles (sample trace from one such recording shown) (C). The numbers indicate common features with the mammalian potential. The corresponding ramp-and-hold stretch stimulus is indicated by the black line. (C) Distinctive receptor potential profile of stretch-activated mammalian muscle spindle. During dynamic stretching, there is a large initial depolarisation (1–2, defined here as dEm), followed by a partial repolarisation (3). Upon transitioning to static stretch, the receptor repolarises to a stable hold potential (4–6), which is maintained throughout static stretch. Following release, the receptor hyperpolarises (7), and returns to resting membrane potential. Again, the corresponding ramp-and-hold stimulus profile is indicated by the lower line [modified from 7]. Most of these features are present in the dbd neuron receptor potentials (B).

Fig 2. Key dbd receptor potential relationships resemble those of mammalian muscle spindles.

(A) Receptor potential recordings from intact dbd neuron preparations vary in their initial resting membrane potential (E _mrest), reflecting differing basal levels of pelt stretch prior to stimulation (sample traces from typical recordings at each level of pelt stretch shown). The three most distinctive features of the stretch-evoked response profile to a standard mechanical stimulus were analysed for their relationship to this difference in initial E _mrest in a range of such dbd neurons: peak depolarisation (E _p; blue square, corresponding to 2 in Fig 1B), E _hold (yellow inverted triangle, 4–6 in Fig 1B) and E _h (orange diamond, 7 in Fig 1B). (B) The maximum amplitude of E _p and E _hold varied directly in proportion to the pre-stimulus potential, whilst E _h was relatively independent of E _mrest (n = 10). (C) Whilst the absolute voltage of E _p significantly varies in proportion to the pre-stimulus potential (Pearson correlation = 0.76, n = 10), the relative change in depolarisation amplitude from baseline (dEm = E _p-E _mrest, see also Fig 1) shows only a weak relationship to E _mrest in dbd neurons (Pearson correlation = 0.42, n = 10), consistent with previous findings in muscle spindles [7]. (D) E _p also shows a high degree of direct correlation with the amplitude of mechanical displacement (p<0.0001, n = 5).

Fig 3. Improved in silico model of the Drosophila mechanosensory receptor potential.

(A) Our in silico model accurately reproduces the most striking features (1–6) of the dynamic and static stretch responses seen in dbd neurons up to the post-release phase (inset). Modelling this latter aspect is under development. (B) In the in silico model, the values of dEm (dEm = E _p-E _mrest) accurately reflects the mean observed values of dEm in dbd neurons (Fig 2C). As might be expected for natural biological tissues the experimental data show more variability (n = 10, p = 0.5).

Fig 4. Modelling incremental inhibition of the mechanosensory sodium current proportionally inhibits all components of the receptor potential.

(A) Full activation of the MSC reproduces the previously modelled behaviour (red line, as for the full response in Fig 2B). The activation term is reduced by 2μm increments, resulting in a proportional inhibition of depolarisation (sequential nested traces). The after-depolarisations (shoulder) and hold potentials are correspondingly reduced, as well. Mathematically, the model will not accept an activation value of 0, but as MSC _Act → 0, dEm → 0mV. (B) This modelled inhibition closely corresponds to stimulus-depolarisation relationship observed in our in vivo system. In both the in silico (light squares) and in vivo (dark diamonds–data from Fig 2D) systems, dEm appears to vary sigmoidally with stimulus amplitude.

Fig 5. Receptor depolarisation in dbd neurons is inhibited by blocking a sodium-dependent MSC.

(A) Replacing Na⁺ in the extracellular medium with NMDG resulted in a significant reduction in dEm in stretch-evoked responses by 79.8% (±2.7%, p<0.0001, n = 3). (B) Stretch-evoked depolarisation of the receptor ending in dbd neurons was inhibited by amiloride. The change in membrane potential (dEm) is normalised to the pre-drug control. Depolarisation was reduced in a dose-dependent manner (n = 5, p<0.001). (C) Ruthenium red application also reduces stretch-evoked depolarisation dose dependently (n = 7, p<0.0001). (D-G) Representative traces for control (D), NMDG (E), amiloride (F) and ruthenium red (G).

Fig 6. Loss of TRPA1 and dmPiezo function have small and large effects respectively on response to stretch.

(A) Stretch-evoked depolarisation was recorded in response to maximal stretch stimuli in dbd neurons of wild-type (w ¹¹¹⁸), TrpA1 ¹ mutants or RNAi knockdown larvae. Data were compared to the previously-measured effect on w ¹¹¹⁸ treated with 30μM ruthenium red (Fig 5). Loss of TRPA1 inhibited stretch-evoked depolarisation <20% (n = 3, p<0.0001), whereas the inhibition by ruthenium red is at least 3x more profound (n = 7, p<0.0001). RNAi knock-down of TRPA1 produced a similar partial blockade of stretch-evoked depolarisation (dEm = 68% control, n = 3, p<0.0001). However, when DmPiezo expression is reduced via RNAi knock-down, the receptor potential is almost completely abolished (E _p<5mV) compared to corresponding controls (GD993 –dEm = 1.3% controls, n = 3, KK101815 –dEm = 14.9% controls, n = 4, p<0.0001). (B-E) Representative traces for control (B), TrpA1 ¹ mutant (C), dmPiezo knockdown line GD993 (D) and dmPiezo knockdown line KK101815 (E).

Fig 7. Ruthenium red inhibits stretch-evoked firing of muscle spindle afferents, but not via Piezo.

(A) Ruthenium red substantially decreases stretch-evoked afferent firing at 100μM (p<0.001, n = 6), but not at 50μM (n = 7). (B–D) Cryosection of medial portions of deep masseter muscle containing muscle spindles. Section immunolabelled for synaptophysin (B) and Piezo2 (C) showing prominent synaptophysin labelling in spindle sensory terminals (arrows), but nothing above background was seen for Piezo2. (D) Merge image of the above. (E,F) Western blots for the expression of Piezo1 (E) or Piezo2 (F). A very similar banding pattern was seen with the other Piezo2 antibody (AbCam FAM38B). In all three cases, no expression could be detected in muscle spindle homogenates, with two being illustrated here. (G,H) Rat muscle spindle annulospiral primary afferent endings (arrows), identified by synaptophysin-positive labelling of their synaptic-like vesicles (green), show high TRPC1 immunoreactivity (red).