The log-normal distribution of sensitivities also suggests that more synapses will be matched to the luminance values most prevalent in the image falling on the retina. The tuning curve of a sensory neuron is a key determinant of the information that it can transmit about a stimulus. Several theoretical studies have suggested that sharper tuning curves within individual neurons can improve the overall efficiency of population

codes, in part because the finest discrimination occurs over the range of stimulus strengths that most rapidly alter the neurons response (Brunel and Nadal, 1998, Pouget et al., 1999, Seriès et al., 2004 and Butts and Venetoclax molecular weight Goldman, 2006). Tuning curves similar to Hill functions or Gaussians can only provide this advantage at the cost of signaling over a narrower range of stimulus strengths, but we found a subset of bipolar cell synapses in which the dynamic range of signaling was increased by an unexpected mechanism: switching the polarity of the exocytic response as a function of luminance. Examples of sypHy signals from such terminals are shown in Figure 6A (ON) and Figure 6B (OFF): the response to a dim light was of the opposite polarity to the larger response to a brighter light. We

examined the tuning curves of linear and nonlinear synapses more closely by normalizing the relation measured in individual terminals to I1/2 and then averaging within the linear and nonlinear classes (Euler and Masland, 2000). The response of nonlinear ON synapses did not saturate check details as light Resminostat intensity increased but passed through a minimum (transition from phase one to two) and then a maximum (transition from phase two to three) before reaching a steady state (Figure 6C). The response of nonlinear OFF synapses was roughly an inversion of this triphasic shape (Figure 6D). A good empirical description of triphasic tuning curves could be obtained by considering them as the sum of two components, which we termed “intrinsic” (black traces in Figures 6E and 6F), and “antagonistic” (blue traces). The expression fitted to these curves is equation(Equation 3) Vexo=A+Int(I′hI′h+1)+Antagσ2π∫0I′exp[−(ln(I′)2σ)2]dI′where

I′ is the intensity normalized to I1/2, A is an offset, Int is a scaling factor for the “intrinsic” component described by a Hill function, Antag is the scaling factor for the “antagonistic” component, described by the cumulative density function of a log-normal distribution, and 2σ is the width of that distribution in log units. The value of σ varied between 3.0 and 4.5 log units and was therefore similar to the distribution of sensitivities across the population of terminals shown in Figure 5C. The growth of the antagonistic component in parallel with the number of bipolar cells activated suggests that this signal may originate from neighboring bipolar cells that are progressively recruited as the light intensity increases.