16 Visual Transduction by Rod

Transcription

1 16 Visual Transduction by Rod and Cone Photoreceptors MARIE E. BURNS AND TREVOR D. LAMB VISUAL PROCESSING begins with the conversion of light to an electrical signal in retinal photoreceptors. In this review we describe the current state of knowledge of this conversion process. First we summarize the remarkable performance of rod and cone photoreceptors and briefly describe their salient anatomical features. Then we present a brief overview of the transduction process before delving into the details. We divide transduction into activation and inactivation phases because there is a clear distinction between the molecular processes underlying the onset and recovery phases of the light response. Finally, we consider how the balance between activation and inactivation is achieved, and how these processes contribute to photoreceptor adaptation. PERFORMANCE OF RODS VERSUS CONES Under most conditions, our vision is mediated by cones. Cones operate over an enormous range of intensities; indeed, at any level of illumination where we can make out color, our cones are providing useful signals. Cones also operate very rapidly, a factor that is likely to be of great importance to survival. In addition, our photopic (cone) visual system is extremely sensitive to small changes, being able to detect a contrast (Weber fraction, I/I) of ~.5%. Finally, cones give us color vision, which, among other benefits, helps to break camouflage, again providing a survival advantage. In evolutionary terms, rods have arisen more recently than cones, and in the retina the specialized rod circuitry gives the appearance of having been superimposed on preexisting cone circuitry. Despite their more recent origin, rods are numerically the dominant class of photoreceptor in the human eye by far, outnumbering the cones by about 2-fold over most of the retina. Yet, rods do not function over most of the intensity range that we normally experience (they are saturated); instead, they only function at exceedingly low intensities, corresponding to starlight or moonlight conditions. What is their role then, and why are they so numerous? The answer seems to be that rods are specialized for the detection of individual photons of light (Hecht et al., 1942), and the subsequent circuitry of the scotopic visual system is designed to be able to detect just a few photon hits scattered over large areas (Barlow et al., 1971). This enables the visual system to operate in a photon-counting mode (Rose, 1948). The enormous number of rods is employed primarily to catch just about every photon that reaches the retina. Rods do not need to continue operating at intensities much above the level at which their responses to different photons begin overlapping appreciably because the primordial system (of cones) is sufficiently sensitive to operate at those levels. As long ago as the 193s, it was proposed that the physical limit to stimulus detection in the scotopic and photopic visual systems was set by the existence of an equivalent background light within the photoreceptors (Stiles and Crawford, 1932). Perhaps the greatest difference between rods and cones lies in the magnitude of this so-called dark light. In cones, the dark light corresponds to some hundreds of photon hits per cone per second, whereas for rods, the level is around four orders of magnitude lower, corresponding to just one photon hit per rod every 1 to 2 seconds (see Barlow, 1964). One of the great challenges in phototransduction is to understand how rhodopsin molecules in the rod outer segment are able to achieve such a great advantage over their counterparts in the cones. It seems that this major advantage of rods over cones brings with it a great cost (Lamb, 1981): namely, the inability of the scotopic system to recover speedily after exposure to extremely intense bleaching illumination. This, it seems, may be an inevitable consequence of some sacrifice that has been made in order to achieve the incredibly low dark light that is needed to give the organism a survival advantage in conditions of near-darkness. PHOTORECEPTOR STRUCTURE IN RELATION TO FUNCTION The ultrastructure of photoreceptors is remarkably similar across the animal kingdom. Photoreceptors consist of three primary functional domains: outer segment, inner segment, and synaptic ending. For the purposes of this chapter, we will focus our brief discussion on the general ultrastructural properties and functions of the outer and inner segments. By far the most abundant protein in the outer segment is the light-capturing protein, rhodopsin. Historically, the term rhodopsin has been applied to the pigment found in those rods that use vitamin A 1 (such as mammalian rods), whereas the term porphyropsin has been applied to vitamin A 2 -based rod pigments, and iodopsin to cone pigments. Here, however, we 215

2 Outer segment Inner segment will adopt the recent loose convention whereby the term rhodopsin will be generalized to refer to all visual pigments; the more explicit terms for rod or cone pigments will be applied only when necessary. Rhodopsin and many of the other components of the transduction cascade reside in (or at the surface of) the lipid material comprising the hundreds to thousands of membrane pairs that are stacked along the outer segment (Fig. 16.1). In cones, these sac-like structures are simply foldings of the plasma membrane, but in rods, the corresponding structures become completely pinched off from the plasma membrane, forming intracellular organelles called discs. This topological difference between rods and cones represents a major anatomical disparity between the two types of cell, yet we do not at present understand its functional significance. One interesting possibility is that the enormous increase in stability of the rhodopsin molecule is aided in some way by the sealing-off of the discs; thus, perhaps it is advantageous to expose the amino-terminal region, not to the extracellular solution, but instead to the intradiscal medium. For cones, the greatly increased surface-to-volume ratio that is possible with patent sacs may be important, as this speeds the rate of change of intracellular ion concentrations mediated by plasma membrane fluxes; this is likely to be especially important for calcium ions, because calcium levels control negative feedback to the cascade (see Calcium Feedback Regulation of Phototransduction ). The lipids and proteins that comprise the discs and sacs eventually deteriorate, so that the photoreceptor needs to perform continual maintenance during its life span. Rods periodically shed old discs from the distal tip of the outer segment, and continually generate new foldings at the base of the outer segment, that in due course pinch off to form discs (reviewed in Young, 1976; see also Steinberg et al., 198). Thus, rod outer segments are not of uniform age, but rather the newer components move up as they achieve seniority. A similar process of renewal and shedding occurs in cones; however, because the entire membrane is continuous, protein components of different age are distributed along the length of the outer segment. Connecting the outer segment to the inner segment is the ciliary stalk, with a arrangement of microtubules and abundant actin filaments. All proteins destined for the outer segment (whether soluble or membrane bound) must pass through this narrow constriction. The mechanisms by which the proteins are transported and the discs are formed are not well understood. The outermost portion of the inner segment contains a high density of mitochondria called the ellipsoid, which supplies adenosine triphosphate (ATP) via diffusion to the metabolically demanding outer segment. The high concentration of mitochondria in the ellipsoid also serves an optical purpose; this region has a high refractive index and helps to funnel photons into the outer segment, thereby increasing the probability that they will be absorbed. This phenomenon is CONE Connecting cilium Ellipsoid Nucleus Synaptic terminal ROD FIGURE Anatomical features of rod and cone photoreceptors. The light-sensitive outer segment comprises a very large area of lipid membrane in which the photopigment molecules are densely packed. In cones, the pigment-containing membrane comprises the highly folded plasma membrane, which forms sacs. In rods, the rhodopsin-containing membrane becomes pinched off into discrete discs that are separated from the plasma membrane and from each other. In both cases, new membrane is synthesized at the base of the outer segment, just above the ciliary stalk, where new foldings are continually being formed. The inner segment contains the conventional cellular metabolic machinery. Organelles include the ellipsoid, which is packed with mitochondria, the myoid, endoplasmic reticulum, and the nucleus. At the base of the cell is the synaptic terminal that mediates communication with postsynaptic neurons. The path of the circulating electrical current is shown for the rod. especially marked in cones, and gives rise to the Stiles Crawford directional sensitivity effect. The inner segment also contains the endoplasmic reticulum, Golgi apparatus, and nucleus; thus, it performs a cell s usual housekeeping chores, albeit with higher than average rates of protein synthesis and unusual membrane-trafficking demands. The plasma membrane of the inner segment contains potassium channels, which, in darkness, permit the efflux of potassium ions that balances the influx of cations into the outer segment, completing the loop that gives rise to the dark current. OVERVIEW OF PHOTOTRANSDUCTION Phototransduction in rods and cones is mediated by closely related cascades, which are broadly similar in many respects to the transduction processes utilized in a variety of other G protein coupled Circulating current 216 RETINAL MECHANISMS AND PROCESSES

3 signaling systems. A receptor protein is activated, causing activation of a G protein, which, in turn, activates a third protein, a so-called effector protein. This brings about a change in the cytoplasmic concentration of a second messenger substance, and thereby leads to modulation of ion channel opening in the plasma membrane. Thereafter, each of the active species must be turned off in order to permit recovery. Phototransduction is the essential first step in vision; thus, it is not surprising that defects in the phototransduction machinery are associated with a variety of blinding diseases in humans (reviewed in Dryja, 2). Phototransduction differs from chemoreceptive systems in that the receptor protein does not need to bind a chemical substance before activation can occur; instead, that chemical (a retinoid in the photoreceptor) is already prebound to the receptor. The absorption of light converts the retinoid from a powerful antagonist into a powerful agonist. Another difference from some other systems is that the response to stimulation is upside-down that is, illumination leads to a reduction of a standing current and, hence, to hyperpolarization of the cell. Why vertebrate photoreceptors should have chosen this inverted response polarity is not entirely clear. Activation of phototransduction In this section, we will consider those steps that lead to activation of the light response, and we will postpone until the next section ( Inactivation of the Phototransduction Cascade ) those steps that contribute to inactivation (or recovery) of the light response. First we will outline each of the molecular mechanisms known to underlie activation. Then we will examine the onset phase of the electrical response that is predicted from theoretical analysis of the activation steps, and we will compare that prediction with experimental results. MOLECULAR MECHANISMS OF ACTIVATION The molecular reactions underlying the activation stages of vertebrate phototransduction are shown schematically in Figure 16.2 (modified from Lamb and Pugh, 1992, Fig. 1). Upon absorption of a photon of light, the rhodopsin molecule is transformed into an enzymatically active state (R*) that catalyzes the activation of the G protein transducin (to G*). Transducin, in turn, activates the effector protein phosphodiesterase (PDE) to PDE*. The activated PDE hydrolyzes the diffusible messenger cyclic guanosine monophosphate (cgmp) so that the cytoplasmic concentration of cgmp decreases, leading to closure of cgmp-gated ion channels in the plasma membrane. Closure of these channels has the dual effect of generating the photoreceptor s electrical response (by reducing the circulating current and hyperpolarizing the membrane voltage), and of causing a reduction in cytoplasmic calcium concentration. Each of these molecular steps will now be described. ACTIVATION OF RHODOPSIN BY A PHOTON Absorption of a photon by the bent 11-cis retinal isomer within rhodopsin has a high probability (~.67) of triggering the cis to trans isomerization of the retinoid. Apparently, the straighter alltrans isomer no longer fits neatly into the hydrophobic pocket so the rhodopsin molecule is strained internally, 4Na + R* G R*-G G* PDE** GC K + Exch. Ca 2+ hn b a g GDP GTP b a g GDP a GTP a g GMP a b g a cg GCAP GTP cg cg cg cg Ca 2+ Na + Open Closed cg FIGURE Schematic representation of molecular reactions underlying activation. Following absorption of a photon (hn), the activated rhodopsin (R*) repeatedly contacts molecules of the heterotrimeric G protein (G), catalyzing the exchange of GDP for GTP, producing the active form G* (= Ga -GTP). Two G*s bind to the two inhibitory g subunits of the phosphodiesterase (PDE), thereby activating the corresponding a and b catalytic subunits, forming PDE**, which then catalyzes the hydrolysis of cgmp (cg). The consequent reduction in the cytoplasmic concentration of cgmp leads to closure of the cyclic nucleotide gated channels and blockage of the inward flux of Na + and Ca 2+, thereby reducing the circulating electrical current. The Na + /Ca 2+,K + exchanger continues to pump Ca 2+ out, so that the cytoplasmic Ca 2+ concentration declines during illumination. BURNS AND LAMB: VISUAL TRANSDUCTION 217

4 undergoing a series of molecular rearrangements that lead to its activation (as R*). The crystal structure of the ground state of rhodopsin was determined by Palczewski et al. (2); what is now awaited is a determination of the crystal structure of R* so that we can better understand the molecular nature of photoactivation. Although the precise structure of R* is not yet known, many studies have contributed clues to the kinds of changes that must occur. Historically, the molecular rearrangements underlying photoactivation have been studied spectrophotometrically because each intermediate has a different absorption spectrum, or color. Several transitions occur extremely rapidly, leading within ~1 ms to the formation of an intermediate named metarhodopsin I (or M I ). The next transition, to the form named metarhodopsin II (or M II ), is crucial to the photoresponse, as M II is the form of the protein that is enzymatically active. The transition from M I to M II takes 1 to 1 ms (depending on temperature) and is thought to involve a substantial structural rearrangement of the protein (recently reviewed by Okada et al., 21), thereby exposing the site for R* interaction with the G protein. The conversion to M II involves both a proton translocation and a proton uptake. The H + that had, until this stage, been protonating the Schiff base (linking the retinoid to Lys296) jumps across to the Glu113 residue that had been forming the negatively charged counterion that stabilized this configuration; in addition, another proton binds to Glu134 (Arnis and Hofmann, 1993). CATALYTIC ACTIVATION OF THE G PROTEIN TRANSDUCIN As indicated in Figure 16.2, the molecular basis for activation of transducin is fundamentally the same as for any other G- protein cascade. Upon binding of the quiescent G protein to R*, the Ga subunit releases its bound molecule of guanosine diphosphate (GDP) and takes up a molecule of guanosine triphosphate (GTP) from the cytoplasm. The GTP-bound form of Ga represents the active entity, denoted as G*, which carries forward the activation signal. The activated rhodopsin, R*, separates from the activated transducin and, unaltered by the interaction, is free to interact with additional molecules of the G protein, thereby catalyzing the activation of many G* during its active lifetime. We denote the gain of this step by the parameter n G,representing the rate of G* activation by a single R*. Over the last few decades, estimates of the magnitude of n G have varied widely. At room temperature, a number of biochemical experiments have reported values of 1G*s -1 or lower, whereas light-scattering approaches have yielded values exceeding 1G*s -1 (e.g., Vuong et al., 1984), and theoretical analysis has suggested that the encounter rate between R* and G is around n enc 5s -1 (Lamb and Pugh, 1992). It is only recently that the different approaches have been reconciled, and it now seems clear that a rate of n G 15G*s -1 at room temperature is consistent with biochemical, light-scattering, and electrophysiological measurements (Leskov et al., 2). We shall return to these rates after we have developed the theoretical framework in the section entitled Predicted Form of the Onset Phase of the Light Response. R* can only activate the G protein when the two molecules come into physical contact. This occurs via lateral diffusion of the molecules (within the membrane for the integral protein rhodopsin, and at the membrane s cytoplasmic surface for transducin). On this basis, rapid lateral diffusion at the membrane has long been thought to be crucially important to activation. By studying rhodopsin hemizygote mice, Calvert et al. (21) have provided evidence that the overall amplification of transduction, and hence the rate of G* activation, depends strongly on membrane fluidity. Their results are consistent with the idea that numerous encounters between R* and G are required (on average) before the two molecules are in the precise configuration required for binding. Hence, the activation rate (n G ) is indeed proportional to, although far lower than, the rate (n enc ) of molecular encounters. Thus, lateral diffusion is crucial to activation. Furthermore, as we discuss later, the results of Calvert et al. (21) show that lateral diffusion is important not only to activation, but also to inactivation. COUPLING OF G* TO THE PDE Although the activated Ga-GTP (= G*) is weakly water-soluble, it probably only dissociates from the disc membrane on a relatively slow time scale, or at very high intensities. The usual means by which it encounters molecules of the effector protein PDE is again via lateral diffusion at the surface of the disc membrane. The PDE is dimeric in structure, comprising a pair of closely homologous hydrolytic subunits (denoted a and b; see Fig. 16.2). Each of the catalytic subunits binds a small g subunit which, under resting conditions, inhibits the PDE, preventing hydrolysis of cgmp. When G* encounters PDEg, the two bind, thereby relieving the g subunit s inhibitory effect and permitting the hydrolysis of cgmp to proceed. Although there have been claims that G*-PDEg dissociates from the remainder of the PDE molecule, it is now generally thought more likely that the g subunits remain closely associated with the ab hydrolytic units, ensuring that inhibition can rapidly be restored once G* is inactivated. One detail that has not yet been clarified is whether or not there is cooperativity in the activation of PDE by G*, that is, whether the singly activated form G*-PDE (which we denote PDE*) has exactly half the activity of the doubly activated form G*-PDE-G* (denoted as PDE**), or whether the singly activated form might, for example, have much lower activity. However, this detail may be relatively unimportant, as the high density of G*s in the neighborhood of an R* is likely to favor the formation of the doubly activated form, PDE**. How efficient is the coupling of G* to PDE? Or, in other words, at some arbitrary time during activation, what pro- 218 RETINAL MECHANISMS AND PROCESSES

5 portion of the G*s that have been produced will have been able to bind to PDEs? Put another way, what is the ratio of PDE* activation rate to G* activation rate (c GE =n G /n E; see Fig. 16.3)? Stochastic simulations have shown that, even if G* were activated at the very high rate of several thousand per second set by encounters with R*, then a PDE subunit density of 25 mm -2 (i.e., just 1 PDE subunit per 1Rh) would be sufficient to achieve a coupling efficiency of 67% (c GE.67). At the much lower rate (n G ) that is now known to apply to G* activation, the coupling efficiency c GE should be very close to unity. This is just what has been reported by Leskov et al. (2), who found the mean rates of activation of G* and PDE* (n G and n E ) both to be around 15s -1 in frog rods at room temperature. Hence it appears that the coupling from activated transducin to activation of the PDE has an efficiency approaching unity (in terms of PDE* subunit activation, equivalent to.5 in terms of PDE** holomer activation). For our subsequent calculation of the overall gain of phototransduction, we will need a value for the rate of PDE* activation per R*; as noted earlier, this appears to be around n E 15 PDE* s -1 in amphibian rods. HYDROLYSIS OF CCMP BY THE PDE The activated PDE** is a highly effective enzyme, and recent measurements indicate that it hydrolyzes cgmp at a rate close to the limit set by aqueous diffusion (Leskov et al., 2). Its maximal catalytic rate has, for some years, been known to be around k cat 4cGMPs -1, but what has only recently become clear is that its Michaelis constant is far lower than previously thought, at K m 1mM cgmp (Leskov et al., 2). Because the cytoplasmic concentration of cgmp is always lower than this value, the rate constant of cgmp hydrolysis is directly proportional to the ratio k cat /K m, a value that is far higher than previously realized. This reaction, the hydrolysis by activated PDE of cgmp in the cytoplasm, represents the second amplifying step in phototransduction. In the theoretical analysis presented in the section on Predicted Form of the Onset Phase of the Light Response, the gain of the PDE step will be shown to be directly proportional to the parameter k cat /K m. As the gain achieved by rapid cgmp hydrolysis is cascaded with the first amplifying step the R*-catalyzed formation of G* (and hence of PDE*) the overall amplification during the activation phase of phototransduction can be extremely high. CHANNEL OPENING/CLOSING BY CGMP In darkness, there is a resting concentration of cgmp of several micromolar, and this has the action of directly gating open cyclic nucleotide gated ion channels in the plasma membrane. For reviews of the channel structure and gating, see Flynn et al. (21). The decline in cgmp concentration leads to closure of the channels, and this has been shown to occur with a sub-millisecond delay (Karpen et al., 1988) as a combined result of very rapid radial diffusion in the aqueous medium and extremely fast gating of channels by cgmp. The concentration dependence of channel opening is cooperative, with a Hill coefficient of n chan 3 and a dissociation constant around K chan 2mM cgmp (although this value depends on calcium concentration; see the section entitled Calcium Regulation of cgmp-gated Ion Channels: Calmodulin ). At the low cytoplasmic concentrations of cgmp that pertain both in darkness and in the light, the number of open channels can accurately be described as proportional to the cgmp concentration raised to the power n chan, that is, as the cube of the cgmp concentration. This parameter n chan can also be viewed as a gain factor (see the section that follows). PREDICTED FORM OF THE ONSET PHASE OF THE LIGHT Response From analysis of the molecular scheme set out in Figure 16.2, and with knowledge of much of the information just presented, a mathematical formulation for the onset phase of the light response was developed by Lamb and Pugh (1992). In order to keep the analysis tractable, all inactivation reactions were ignored, restricting the window of validity to relatively short times after illumination. Several other simplifying assumptions were also made, but were shown not to compromise the validity of the solution under most circumstances. The predictions of the simplified analysis are presented graphically in Figure In response to a single photoisomerization, a single R* is created (after a short delay corresponding mainly to the M I to M II transition); because inactivation reactions are being ignored, it simply remains present (Fig. 16.3A). The presence of R* catalyzes the activation of G* at a constant rate, so that the quantity G*(t) ramps as a function of time, t, with a slope of n G 15G*s -1. Because of the very efficient coupling to PDE, the quantity of activated PDE* subunits, E*(t), also ramps with time, with a similar slope of n E 15E*s -1 (Fig. 16.3B). Figure 16.3C illustrates the predicted time course for the decline of cgmp concentration, cgmp(t), and for the decline of the fractional inward current, F(t). As the rod photocurrent is relatively insensitive to changes in membrane voltage (Baylor and Nunn, 1986), F(t) also represents the fractional circulating electrical current. Both cgmp(t) and F(t) decline with an accelerating time course owing to the continually increasing quantity of PDE*. If E*(t) indeed ramps with time, then it can be shown (Lamb and Pugh, 1992; Pugh and Lamb, 1993) that the electrical response should, to a good approximation, follow a delayed Gaussian time course: F(t ) = exp {- 1 /2 F A (t-t eff ) 2 } (1) where F is the number of photoisomerizations (the flash intensity), A is a fixed parameter termed the amplification constant of transduction, and t eff is an effective delay time BURNS AND LAMB: VISUAL TRANSDUCTION 219

6 Numbers of activated proteins Fractional cgmp & current A B C F (t ) cgmp (t ) Time (msec) that lumps together a number of very short delays in transduction. Equation 1 is important because it provides what is essentially a single-parameter description of the onset phase of the light response; thus, apart from the short delay term (t eff ), there is only one parameter involved in describing the onset of the electrical response to any arbitrary flash intensity (F). All of the physical and biochemical parameters inherent in the scheme of Figure 16.2 have collapsed down into this single parameter, A. Furthermore, this overall amplification constant is specified by the theory as the product of the gains of the underlying stages: A =ne bsub nchan (2) where n E is the rate of PDE* activation in response to a single photoisomerization, b sub is a single PDE* subunit s rate constant for cgmp hydrolysis, and n chan is the Hill coefficient of channel opening by cgmp. Finally, b sub is given by 1 cat m b sub = 2 k K (3) N V B AV cyto cg R *(t ) G *(t ) E *(t ) 1 F (t )=exp{- F A (t - t eff ) 2 2 } FIGURE Predicted time course of activation when inactivation reactions are ignored. A, Following photoisomerization, a single R* is activated. B, The quantities G*(t) and E*(t) of activated G protein and PDE increase linearly with time after R* activation; in both cases, the slope is estimated to be around 15s -1 per R* (see section entitled Estimated Rate of Activation of the PDE ). C, Activation of E* causes the cgmp concentration to decline. Consequently, cgmp-gated channels close, and the fractional circulating current F(t) declines according to equation 1 (in text). Responses in C are shown for a flash delivering F =25 photoisomerizations, with A =.1s -2. where k cat /K m is the ratio of enzyme parameters for the PDE (described earlier), N Av is Avagadro s number, V cyto is the cytoplasmic volume of the outer segment, and B cg is the cytoplasmic buffering power for cgmp. The factor of 1 /2 provides the conversion from a k cat value for the holoenzyme PDE** to a b sub value for a PDE* subunit. In Figure 16.3C, the traces are drawn for a flash intensity of F=25 photoisomerizations, in conjunction with an amplification constant of A =.1s -2, typical of an amphibian rod. The theory predicts that, for any other flash intensity, the expected curve would have the same Gaussian shape as drawn, simply with a different time scaling. How does this prediction compare with experiment? COMPARISON OF THEORY WITH EXPERIMENT In Figure 16.4, we compare the predictions of equation 1 with suction pipette responses from an amphibian rod (A) (data from Torre et al., 1986) and from a human rod (B) (data from Kraft et al., 1993). In each panel, a single value for the amplification constant A has been chosen, and curves are calculated for the set of measured flash intensities. Clearly, the onset phase of the response is reasonably well described over a wide range of flash intensities. The values of A were.65s -2 for the salamander rod and 2s -2 for the human rod. The smaller value of A for the amphibian rod arises primarily because of the much greater cytoplasmic volume of its outer segment ( V cyto ), which, in the theory, appears in the denominator of equation 3. For the salamander rod (Fig. 16.4A), the theory traces provide a good fit that extends to later times than one might reasonably expect for a model that ignores inactivation. The reason for the extended period of good fit in this cell is that it had been loaded with the calcium buffer BAPTA for the express purpose of slowing down the onset of inactivation reactions. In the same cell, prior to BAPTA incorporation, the experimental traces began deviating from theory by about 3ms after the flashes (see Fig. 7 of Lamb and Pugh, 1992). As far as we are aware, equation 1 has always been capable of providing a good description of the onset phase of rod responses at suitably early times. ESTIMATED RATE OF ACTIVATION OF THE PDE In an earlier section, entitled Catalytic Activation of the G Protein Transducin, we referred to disparities between different methods for estimating the rates n G and n E at which G* and PDE* are activated in response to a single R*. Recently, these discrepancies appear to have been reconciled by Leskov et al. (2). Their biochemical measurements of n G and n E were broadly consistent with earlier measurements using comparable methods; both were found to be around 15s -1. However, in addition, they made new measurements of the PDE hydrolytic parameters and found the value of the Michaelis constant K m to be far lower than previously 22 RETINAL MECHANISMS AND PROCESSES

7 Fractional current, F (t ) Fractional current, F (t ) A Salamander rod Torre et al. (1986) B Human rod Kraft et al. (1993) Time (ms) FIGURE Comparison of experiment with predicted onset, for families of flash responses from a salamander rod (A) and a human rod (B), recorded using the suction pipette method. In both panels, the black traces show experiments, whereas the gray traces plot equation 1 (in text) using the calculated number of photoisomerizations, F, and a fixed value of the amplification constant A and effective delay time t eff. A, Recordings from a salamander rod, following incorporation of the calcium buffer BAPTA, for flashes delivering from 9 to 2 isomerizations. (Data from Fig. 2B of Torre et al., 1986.) Dark current, -29 pa; amplification constant, A =.65s -2 ; delay time, t eff = 2ms. B, Recordings from a human rod for flashes delivering from 34 to 38 isomerizations. (Data from Fig. 4 of Kraft et al., 1993.) Dark current, -13.5pA; amplification constant, A = 2s -2 ; delay time, t eff = 2.2ms. In both panels, the origin of time has been set to the middle of the flash stimulus, after allowance for the delay introduced by electronic filtering. reported. Substitution of this value into equations 2 and 3 yielded estimates for n E from electrophysiological measurements that were consistent with the biochemical values. Thus, with k cat = 44s -1, K m = 1mM, and V cyto =.85pl for a toad rod outer segment, and assuming B cg = 1 (no cgmp buffering), the parameter b sub in equation 3 was calculated to be b sub = s -1. Then, substituting an amplification constant of A.1s -2 for a toad rod, and a Hill coefficient of n chan = 2 in equation 2, the rate of PDE* subunit activation was found to be n E 12E*s -1, which is in reasonable agreement with the biochemical estimate. Finally, Leskov et al. (2) suggested that the estimates from light scattering could also be brought into line if it were assumed that the signal measured by Vuong et al. (1984) originated from the binding of G* to PDE, rather than from the release of G*. Such a proposal for the origin of the lightscattering signal had, in fact, already been made by Kamps et al. (1985). If their explanation is correct, then the estimates of protein activation rates from biochemistry, light scattering, and electrophysiology all fall neatly into line with each other. Inactivation of the phototransduction cascade Rapid time resolution in the visual system is crucial for survival. However, the overall visual system cannot respond appreciably faster than its photoreceptors, and it is therefore important that the photoreceptor response be as brief as possible, consistent with the achievement of a sufficiently high amplification. Speed of response is most important at increased light intensities at which amplification can be sacrificed. At the very lowest intensities, when the visual system operates in a photon-counting mode and photon hits are few and far between, the speed of response is less critical. What, then, determines how long the photoreceptor s response to a brief flash of light lingers? Is it determined passively, by the inherent rates of thermal degradation of activated cascade components? Recent results have shown the answer to this question to be a resounding no, with shutoff being an active process in both rods and cones. In this section, we will focus on what is currently known about rhodopsin inactivation in rods, mentioning relevant cone data when it is available. MOLECULAR MECHANISMS OF RHODOPSIN INACTIVATION Inactivation of the photopigment molecule is arguably the most important aspect of recovery owing to the high amplification achieved by R* (see previous section). As long as rhodopsin maintains some level of catalytic activity, an amplified response persists, resembling the presence of steady light and contributing to desensitization and/or saturation of the cell. At exceedingly low scotopic intensities, at which the visual system operates in its photon-counting mode, it can be argued that reproducibility of both the amplitude and the time course of the single-photon response is important, in order that the system should be better able to count the photons. However, this argument does not apply at higher scotopic intensities, or in the photopic range, where the shape of responses to individual photon hits is irrelevant; at these intensities, all that matters is the shape of the average response. The mechanisms that limit the duration of BURNS AND LAMB: VISUAL TRANSDUCTION 221

8 rhodopsin s active lifetime, that contribute to its reproducibility under dark-adapted conditions, and that shorten the lifetime even further under light adaptation, are the subject of intense study. Fast and slow mechanisms contribute to the shut-off of activated rhodopsin. First, R* is phosphorylated; second, it is capped by the protein arrestin (e.g., Kühn and Wilden, 1987). These two steps rapidly lead to an enormous reduction in rhodopsin s catalytic activity, as described in the next subsections. Subsequent to these fast inactivation steps, the photoisomerized all-trans retinal is hydrolyzed from the opsin protein and, in due course, is replaced by 11-cis retinal, leading eventually to the total shut-off of catalytic activity. These slower steps, which contribute to dark adaptation and pigment regeneration, are beyond the scope of this review. PHOSPHORYLATION AS ARAPID FIRST STEP IN SHUT-OFF The first step in shut-off phosphorylation of rhodopsin s COOH-terminal residues is performed by rhodopsin kinase (RK) (Kühn and Wilden, 1987), although other protein kinases, such as protein kinase C (e.g., Udovichenko et al., 1997), may play a role under some conditions. Phosphorylation of rhodopsin reduces its catalytic activity substantially (e.g., Wilden et al., 1986) and is essential for recovery of the photoresponse (Nakatani and Yau, 1988b; Sagoo and Lagnado, 1997; see also the section on Comparison of Effects of Mutations of Inactivation Proteins ). There are multiple potential sites for phosphorylation by RK in rhodopsin s COOH-terminal domain, six such sites in mice and seven in humans. However, mass spectroscopy of COOH-terminal peptides of rhodopsin following light exposure has suggested that rhodopsin is predominantly monophosphorylated after a flash (Ohguro et al., 1995), raising the interesting possibility that the large number of sites merely serve to increase the rate at which phosphorylation occurs. However, experiments on individual transgenic mouse rods have suggested that multiple phosphorylation is essential for rapid recovery of the singlephoton response (Mendez et al., 2; see also the section on Comparison of Effects of Mutations of Inactivation Proteins ). Furthermore, mutant mouse rhodopsins containing fewer than three of the six potential phosphorylation sites were found to display first-order inactivation kinetics, suggesting that, under these conditions, a single, stochastic process is responsible for turning off unphosphorylated rhodopsin, and that it is rate-limiting for recovery. Similar behavior is also observed when phosphorylation of R* is completely prevented (see section cited above). Stochastic shut-off of this kind is quite abnormal, as rhodopsin inactivation in WT rods is fairly reproducible (see the section entitled Rhodopsin Shut-off: Reproducibility of the Single- Photon Response ). This suggests that reproducible rhodopsin inactivation requires multiple phosphorylation of R* (Mendez et al., 2, Figs. 5 and 6). ARRESTIN BINDING IS REQUIRED FOR COMPLETION OF THE RAPID PHASE OF R* INACTIVATION Following phosphorylation, the protein arrestin binds with high affinity to phosphorhodopsin (Wilden et al., 1986). In mouse rods that do not express arrestin (Arr-/-; Xu et al., 1997), the dim flash response rises to a peak amplitude comparable to that of normal rods, and begins to recover roughly halfway back to baseline levels (see Figs and 16.6). However, it then enters a period of extremely slow recovery, declining with a final time constant that is on the order of 4 seconds. The time course of the final decline of the response in the Arr-/- rods is comparable to that for the disappearance of metarhodopsin II (Cone and Cobbs, 1969). The failure of normal recovery in Arr-/- rods suggests that the binding of arrestin is essential for the rapid quench of rhodopsin s activity, and that in the absence of arrestin, phosphorylated R* must await hydrolysis of the retinoid. The fact that the failure of normal recovery occurs fairly late in the response suggests that significant rhodopsin activity normally persists throughout the duration of the singlephoton response. Slowing the rate of guanosine triphosphate (GTP) hydrolysis by transducin (see the section entitled Inactivation of the G Protein and the Effector (Transducin and PDE) ) also selectively slows the final recovery of the dim flash response (Sagoo and Lagnado, 1997; Chen et al., 2). This raises the interesting and yet unanswered question of whether deactivation of rhodopsin or of transducin/pde is rate-limiting for the recovery of the singlephoton response. Inactivation of cone pigments is also likely to involve phosphorylation and arrestin binding. Because recovery of the photoresponse occurs roughly 1 times faster in cones than in rods, one might expect these biochemical shut-off steps likewise to occur more quickly in cones. Consistent with this possibility was the discovery of a cone-specific isoform of opsin kinase (GRK7), which was first cloned from ground squirrel (Weiss et al., 1998), and has since been identified in other mammals as well. Whether or not this cone-specific isoform is specialized for more rapid inactivation of R* is unclear, though, partially because of the apparent differences in expression of RK and cone opsin kinases in different species. For example, cones of dogs and pigs exclusively express cone opsin kinase (Weiss et al., 21), whereas mouse and rat cones express only RK (Lyubarsky et al., 2; Weiss et al., 21), and primate cones appear to express both (Chen et al., 21; Weiss et al., 21). A second cone-specific isoform of the regulatory machinery for transduction is cone arrestin (Craft et al., 1994). It 222 RETINAL MECHANISMS AND PROCESSES

9 exhibits no detectable binding to light-activated or phosphorylated rod rhodopsin (Maeda et al., 2), suggesting that it has fundamentally different biochemical properties from rod arrestin. Its role in recovery of the cone photoresponse has yet to be determined. RHODOPSIN SHUT-OFF: REPRODUCIBILITY OF THE SINGLE- PHOTON RESPONSE One electrophysiological approach that has been used to estimate the number of stages involved in R* shut-off is an examination of the degree of reproducibility, or variability, in the shape of a rod s response to different single-photon hits. Rieke and Baylor (1998) and Whitlock and Lamb (1999) conducted experiments of this kind on toad rods and, despite recording apparently similar events, came to quite different conclusions. Rieke and Baylor reported that the single-photon events were highly reproducible in shape, concluding that 1 to 2 stages in R* s shutoff needed to be invoked in order to explain such a high degree of reproducibility. In the same species, Whitlock and Lamb reported that the apparent variability in shut-off timing was twice as large as that found by Rieke and Baylor. They concluded that six or seven stages in R* shut-off could explain the observed degree of kinetic variability. They went on to speculate that, if a powerful mechanism of calcium feedback were to modulate the R* lifetime within the single-photon response, then the variability might even be explained by a single stage of shut-off. However, that possibility has now been excluded by the GCAPs (guanylate cyclase activating proteins)-/- experiments of Burns et al. (22), which have ruled out calcium feedback onto R* lifetime within the single-photon response. Accordingly, the electrophysiology of singlephoton response is consistent with an apparent number of stages involved in R* shut-off of around six. Although this number corresponds neatly with the number of sites available for phosphorylation, the agreement may be no more than fortuitous. INACTIVATION OF THE G PROTEIN AND THE EFFECTOR (TRANS- DUCIN AND PDE) Despite the importance of R* inactivation, that alone is not sufficient for rapid recovery of the electrical response to a flash. It is also important that G* and PDE* (the activated transducin and cgmp PDE) be turned off rapidly. As in most heterotrimeric G protein cascades, the rate-limiting step for turning off both the G protein and its effector is the hydrolysis of GTP, that is, the rate at which the GTP s terminal phosphate is cleaved by the Ga subunit. For many years, it has been noted that the basal GTPase activity of purified transducin is exceedingly low, with a turnover time of minutes rather than seconds. Thus, it has long been anticipated that mechanisms for speeding the rate of GTP hydrolysis must exist in order to explain the rapid physiological shut-off. In the last few years, an entirely new family of proteins, called RGS proteins (regulators of G protein signaling; reviewed in Ross and Wilkie, 2), has been discovered. One member of this family, RGS9-1 (He et al., 1998), is expressed in rods and even more abundantly in cones (Cowan et al., 1998). In photoreceptors, RGS9-1 is obligatorily associated with an orphan G protein b subunit, Gb5L (Makino et al., 1999). The resulting RGS9/Gb5L complex is essential for achieving normal rapid shut-off of the light response, both in rods (Chen et al., 2; see Fig. 16.5) and in cones (Lyubarsky et al., 21). Under physiological conditions, it appears that RGS9/ Gb5L only affects GTPase activity when G* is bound to the PDE g subunit, that is, while the PDE is in its active form, PDE* (see the earlier section on Coupling of G* to the PDE ). It has been known for some time that the acceleration of GTPase activity is aided by PDEg (Arshavsky and Bownds, 1992; Antonny et al., 1993), but the mechanism by which both PDEg and RGS9/Gb5L might work together has been unclear. Recently, it has been found that PDEg increases the affinity of RGS9/Gb5L for Ga (Skiba et al., 2), thereby speeding the rate of GTP hydrolysis and the rate of recovery of the flash response in vivo. Thus, the inactivation of G*/PDE* utilizes an elegant method of selfregulation, whereby transducin is prevented from turning off before it has bound to and activated PDE. The crystal structure of the complex between Ga-GTP, PDEg, and RGS9 has recently been presented by Slep et al. (21). Details of the molecular mechanisms involved in the regulation of transducin s GTPase activity are reviewed by Cowan et al. (21) and Arshavsky et al. (22). COMPARISON OF EFFECTS OF MUTATIONS OF INACTIVATION PROTEINS Collaborations between researchers in molecular biology, biochemistry, and physiology are revealing the many levels of regulation that actively turn off the phototransduction machinery, thus speeding recovery of the response to an increment in illumination. Although a great deal has been learned about mechanisms of response shut-off from conventional approaches over several decades, there has been a recent explosion in the application of transgenic approaches to photoreceptors, some of which we will illustrate here. One striking observation to have emerged from gene manipulations in mouse photoreceptors is that rods and cones appear to lack sophisticated compensatory mechanisms to oppose alterations in gene expression of many of the molecules of transduction. Thus, there have been few problems associated with significant changes in expression levels of proteins other than the one under examination. Instead, the most serious and consistent difficulty in creating these transgenic and knockout mice has been the degeneration of photoreceptors, especially in cells with prolonged responses that make them sensitive to steady backgrounds BURNS AND LAMB: VISUAL TRANSDUCTION 223

10 (Makino et al., 1998). In addition, overexpression or underexpression of rhodopsin can cause degeneration (Li et al., 1996; Lem et al., 1999). These issues can usually be resolved either by careful titration of transgene expression levels (e.g., Mendez et al., 2) or by rearing the animals in darkness (e.g., Xu et al., 1997). As examples of the mutations that have been studied in mouse photoreceptors, we present in Figures 16.5 and 16.6 a composite set of responses from cells of different transgenic and knockout mice. In addition to wild-type responses, we illustrate five mutants: three involving disruption of rhodopsin inactivation and two involving the slowing of subsequent stages of recovery. Both figures plot responses to repeated presentations of a very dim flash. In Figure 16.5, the raw responses are shown on a slow time scale, whereas in Figure 16.6, the averaged responses are compared on a faster time scale. In the following descriptions, many of the features we discuss can be seen in these two figures. R* inactivation Genetic perturbations of rhodopsin inactivation have yielded a wide range of phenotypes that reveal much about the inherent time course of rhodopsin inactivation in intact cells. Phosphorylation of rhodopsin can be prevented by several transgenic techniques: by deleting the entire COOH-terminal domain of rhodopsin (S334ter; Chen et al., 1995), by abolishing the expression of RK (Chen et al., 1999), or by mutating all of the carboxyterminal serine and threonine residues to alanines (CSM [completely substituted mutant]; Mendez et al., 2). Under each of these conditions, rhodopsin s activity, measured electrophysiologically, persists for several seconds, with the single-photon response reaching a plateau level of around double the normal amplitude, and then decaying abruptly back to baseline after a long stochastic interval of around 3 to 5 seconds (see Fig. 16.5). Because this interval is considerably shorter than the time course of hydrolysis of FIGURE Dim flash responses recorded from mouse rods of different genotypes. Flashes were delivered at the times indicated by the flash monitor below each recording. Mouse rods expressing mutant rhodopsin with three of the six potential phosphorylation sites (Serine Triple Mutant, STM; Mendez et al., 2) show responses that are qualitatively similar to wild-type (WT) responses, whereas responses from rods lacking RK (RK-/-; Chen et al., 1999) are prolonged steps of variable duration. Mouse rods lacking arrestin (Arr-/-; recording by AF Almuete in the Burns lab) initially recover, but require much longer times for final inactivation (Xu et al., 1997). Note the sixfold longer time base for the Arr-/- recording. Deficiencies in the rate of GTP hydrolysis (RGS9-/-; Chen et al., 1999) and calcium feedback to guanylate cyclase (GCAPs-/-; Burns et al., 22) likewise yielded prolonged responses. Note the increased cellular dark noise and very large elementary response amplitude in the GCAPs-/- recording. all-trans retinal, it can be surmised that mammalian rods possess a backup mechanism that is capable of turning off unphosphorylated R*. Ironically, these large and long-lasting responses are occasionally observed in normal mammalian rods (e.g., Baylor et al., 1984). This suggests that, in normal rods, as many as 1% of all rhodopsin molecules are not capable of being phosphorylated. The similarity of responses from RK-/-, S334ter, and CSM rods suggests that the sole defect in each case is a lack of rhodopsin phosphorylation, and that the binding of RK has little effect on R* s catalytic activity. Furthermore, pa Time (s) 2 WT STM RK -/- RGS9 -/ Arr -/- GCAPs -/ RETINAL MECHANISMS AND PROCESSES