signal coding in cockroach photoreceptors is tuned to dim

doi:10.1152/jn.00588.2012 108:2641-2652, 2012. First published 29 August 2012;J NeurophysiolVähäsöyrinki and M. WeckströmK. Heimonen, E.-V. Immonen, R. V. Frolov, I. Salmela, M. Juusola, M.dim environmentsSignal coding in cockroach photoreceptors is tuned to

You might find this additional info useful...

76 articles, 24 of which can be accessed free at:This article cites /content/108/10/2641.full.html#ref-list-1

3 other HighWire hosted articlesThis article has been cited by

[PDF] [Full Text] [Abstract]

, August 15, 2014; 112 (4): 913-922.J NeurophysiolRoman Frolov and Matti Weckström

): better performance at higher costGerris lacustrisstrider (Developmental changes in biophysical properties of photoreceptors in the common water

[PDF] [Full Text] [Abstract], November 1, 2014; 217 (21): 3853-3861.J Exp Biol

Aaron E. Schirmer, Frederick R. Prete, Edgar S. Mantes, Andrew F. Urdiales and Wil BogueHierodula patelliferalocomotion of the praying mantis

Circadian rhythms affect electroretinogram, compound eye color, striking behavior and

[PDF] [Full Text] [Abstract], December 1, 2014; 217 (23): 4262-4268.J Exp Biol

Anna Honkanen, Jouni Takalo, Kyösti Heimonen, Mikko Vähäsöyrinki and Matti WeckströmCockroach optomotor responses below single photon level

including high resolution figures, can be found at:Updated information and services /content/108/10/2641.full.html

can be found at:Journal of Neurophysiologyabout Additional material and information http://www.the-aps.org/publications/jn

This information is current as of May 20, 2015.

American Physiological Society. ISSN: 0022-3077, ESSN: 1522-1598. Visit our website at http://www.the-aps.org/.(monthly) by the American Physiological Society, 9650 Rockville Pike, Bethesda MD 20814-3991. Copyright © 2012 by the

publishes original articles on the function of the nervous system. It is published 12 times a yearJournal of Neurophysiology

on May 20, 2015

Dow

nloaded from on M

ay 20, 2015D

ownloaded from

http://jeb.biologists.org/content/217/23/4262.abstract.html

http://jeb.biologists.org/content/217/23/4262.full.html

http://jeb.biologists.org/content/217/23/4262.full.pdf

http://jeb.biologists.org/content/217/21/3853.abstract.html

http://jeb.biologists.org/content/217/21/3853.full.html

http://jeb.biologists.org/content/217/21/3853.full.pdf

http://jn.physiology.org/content/112/4/913.abstract.html

http://jn.physiology.org/content/112/4/913.full.html

http://jn.physiology.org/content/112/4/913.full.pdf

http://www.the-aps.org/publications/jn

Signal coding in cockroach photoreceptors is tuned to dim environments

K. Heimonen,1 E.-V. Immonen,1 R. V. Frolov,1 I. Salmela,1 M. Juusola,2,3 M. Vähäsöyrinki,1

and M. Weckström1

1University of Oulu, Department of Physics, Oulu, Finland; 2University of Sheffield, Department of Biomedical Science,Sheffield, United Kingdom; and 3State Key Laboratory of Cognitive Neuroscience, Beijing Normal University, Beijing, China

Submitted 9 July 2012; accepted in final form 24 August 2012

Heimonen K, Immonen E-V, Frolov RV, Salmela I, Juusola M,Vähäsöyrinki M, Weckström M. Signal coding in cockroach photore-ceptors is tuned to dim environments. J Neurophysiol 108: 2641–2652,2012. First published August 29, 2012; doi:10.1152/jn.00588.2012.—Indim light, scarcity of photons typically leads to poor vision. None-theless, many animals show visually guided behavior with dim envi-ronments. We investigated the signaling properties of photoreceptorsof the dark active cockroach (Periplaneta americana) using intracel-lular and whole-cell patch-clamp recordings to determine whetherthey show selective functional adaptations to dark. Expectedly, dark-adapted photoreceptors generated large and slow responses to singlephotons. However, when light adapted, responses of both phototrans-duction and the nontransductive membrane to white noise (WN)-modulated stimuli remained slow with corner frequencies 20 Hz.This promotes temporal integration of light inputs and maintains highsensitivity of vision. Adaptive changes in dynamics were limited todim conditions. Characteristically, both step and frequency responsesstayed effectively unchanged for intensities 1,000 photons/s/photo-receptor. A signal-to-noise ratio (SNR) of the light responses wastransiently higher at frequencies 5 Hz for 5 s after light onset butdeteriorated to a lower value upon longer stimulation. Naturalisticlight stimuli, as opposed to WN, evoked markedly larger responseswith higher SNRs at low frequencies. This allowed realistic estimatesof information transfer rates, which saturated at 100 bits/s atlow-light intensities. We found, therefore, selective adaptations ben-eficial for vision in dim environments in cockroach photoreceptors:large amplitude of single-photon responses, constant high level oftemporal integration of light inputs, saturation of response propertiesat low intensities, and only transiently efficient encoding of light con-trasts. The results also suggest that the sources of the large functionalvariability among different photoreceptors reside mostly in phototrans-duction processes and not in the properties of the nontransductive mem-brane.

vision; systems analysis; adaptation; temporal resolution; photons

SENSORY SYSTEMS PROVIDE ANIMALS with necessary informationfor survival and reproduction. Like all senses of differentspecies, visual systems are thought to have selectively adaptedfor functioning under their prevailing environmental condi-tions during their evolution and development (Laughlin 1989,1990, 1996; Warrant 1999, 2004; Weckström and Laughlin1995). However, for species living in dim environments, ex-tracting visual information can be a self-contradictory task,because there may be little reliable information to be gatheredwithin behaviorally relevant integration times. With only a fewphotons to absorb, vision becomes unreliable (Warrant 1999,2006). The random (Poisson-distributed) arrival of photonstriggers single-photon responses, so-called quantum bumps, in

the photoreceptors, generating photon shot noise and inevitablylowering the signal-to-noise ratio (SNR). The macroscopicphotoreceptor responses are formed when the bumps are inte-grated by the nontransductive part of the photoreceptor mem-brane (Wong and Knight 1980; Wong et al. 1980). An intrinsicbut single-photon-related source of noise, called transducernoise, varies timing, shapes, and sizes of the bumps, furtherlowering the SNR of the macroscopic response (Laughlin andLillywhite 1982; Lillywhite and Laughlin 1979). Dark noise(spontaneous bumps without light), which may be significantin vertebrate vision (Aho et al. 1988), is of minor importancein invertebrates (Heimonen and Weckström, unpublished ob-servations; Laughlin and Lillywhite 1982; Lillywhite 1977;)and likely to be suppressed specifically by a molecular mech-anism within the phototransduction cascade (Katz and Minke2012).

Vision generally improves with increasing brightness. Lightadaptation (LA) makes phototransduction faster and quantumbumps smaller and their latencies shorter (Howard et al. 1987;Juusola et al. 1994). Concomitantly, transduction gain (re-sponse amplitude/unit of stimulus intensity) decreases, al-though contrast gain (response amplitude/unit of stimulus con-trast) increases. Such signaling dynamics have been describedin detail in several insects (Faivre and Juusola 2008; Juusolaand Hardie 2001; Juusola et al. 1994; Niven and Laughlin2008; Niven et al. 2003). With LA, noise becomes less signif-icant, and the SNR of responses improves. In dim conditions,the only way for the visual system to make sense of the visualenvironment is to increase sensitivity by integrating moresignals over space and time (Laughlin 1990; Warrant 1999,2006, 2008; Warrant and McIntyre 1993). This includes lightcollection from wider receptive fields, pooling of signals fromseveral photoreceptors, and longer integration times of thephotoreceptors. At the level of photoreceptors, there are someapparent general strategies to cope with the problem of photonscarcity: high gain of phototransduction and temporal andspatial summation.

Good vision in dim light is achieved by large-sized cameraeyes or in insects, generally by superposition compound eyeswith specialized features for efficient gathering of light (Laughlin1990; Warrant 1999, 2004). In contrast, apposition compoundeyes, such as those in the cockroach (Butler 1971, 1973b), maybe thought to be inadequate for night vision. However, cock-roaches are active mainly under dim conditions (Guthrie andTindall 1968; Lipton and Sutherland 1970; Roberts 1965).They tend to avoid light when given a choice (Guthrie andTindall 1968; Halloy et al. 2007; Kelly and Mote 1990),directing their escape responses toward dark places (Riemay1984; Ye et al. 2003). In addition, there are other nocturnal

Address for reprint requests and other correspondence: M. Weckström,Univ. of Oulu, Dept. of Physics, Biophysics, FIN-90014, Oulu, Finland(e-mail: [email protected]).

J Neurophysiol 108: 2641–2652, 2012.First published August 29, 2012; doi:10.1152/jn.00588.2012.

26410022-3077/12 Copyright © 2012 the American Physiological Societywww.jn.org

on May 20, 2015

Dow

nloaded from

species, which have apposition eyes [see review by Warrant(2008)]. Contrary to compound eyes in general, cockroach eyeshave a highly irregular structure (Butler 1973a, 1973b; Ernstand Füller 1987; Ribi 1977). Functionally, they show variableresponsiveness of the photoreceptors (Heimonen et al. 2006)and spike coding in the photoreceptor axons (Weckström et al.1993).

In the present work, we explore the response dynamics ofcockroach photoreceptors to investigate whether they are se-lectively adapted for operating in dim light and whether cock-roaches differ in this respect from other species studied withsimilar techniques. The results, obtained by intracellular re-cordings of intact photoreceptors and patch-clamping of iso-lated ommatidia and with the use of both white noise (WN)-modulated and naturalistic stimulation, reveal exclusive phys-iological adaptations of cockroach photoreceptors to dark andemphasize the value of naturalistic stimulation as a tool ofchoice for studying early visual encoding.

METHODS

Animals and preparation. Recordings were performed from green-sensitive photoreceptors of adult male cockroaches (Periplanetaamericana). Some females were also tested to ensure that their voltageresponses to light stimuli did not differ significantly from those ofmales. Cockroaches were maintained for all of the experiments andprepared for the in vivo intracellular experiments as described inHeimonen et al. (2006). After preparing the animal and positioningthe recording and indifferent electrodes, the preparation was darkadapted for 30 min. All intracellular recordings were done in theretina, where the somata of the photoreceptors are located. Under astereomicroscope, the retina of the cockroach is seen below the corneaas a black, highly pigmented area, and the location of recordingelectrode in the preparation could be identified on these grounds. Thislocation is also characterized by graded-depolarizing responses, whichlack the superimposed spikes typical for cockroach photoreceptoraxon recordings (Heimonen et al. 2006; Weckström et al. 1993). Noattempts were made to distinguish between the responses of the short-and long-type photoreceptors, which terminate in the first and secondneuropile, respectively (Ribi 1977). For the in vitro whole-cell patch-clamp experiments on dissociated cockroach ommatidia, including thesomata of the eight photoreceptors, the preparation was done similarlyas described for Drosophila melanogaster (Hardie 1991). Shortly afterdissection in Ringer’s solution, both retinae were cut into pieces smallenough to barely fit into a glass capillary, fire polished to an openingdiameter of 500 m. Subsequently, the retina pieces were trans-ferred into a drop of Ringer’s solution supplemented with 0.2 mg/mlcollagenase type 2 (Worthington Biochemical, Lakewood, NJ) and 0.2mg/ml pankreatin (Sigma-Aldrich, St. Louis, MO) for 10 min incu-bation before the actual dissociation. The dissociation of ommatidiawas done by triturating the enzymatically treated pieces with a seriesof three to four fire-polished glass capillaries (from 500 to 250 mopening diameter). The dissociated ommatidia were then transferredinto the recording chamber and allowed to settle in darkness for a fewminutes.

Setup for in vivo experiments. Intracellular voltage responses ofphotoreceptors to light or current stimulation were recorded at roomtemperature (20°C) with borosilicate glass capillary microelectrodes(resistances 50–150 M), filled with a 2-M KCl solution, whose pHwas adjusted to 6.84 with a K-phosphate buffer. The electrodecapacitance was compensated and the signals amplified with anintracellular amplifier (SEC-05L; NPI, Germany). The light stimuliwere produced with a green, high-intensity light-emitting diode (LED;Roithner LaserTechnik, Austria; peak wavelength, 525 nm). Thecomputer-generated stimulus waveforms for the LED were fed to acustom-made voltage-to-current driver, which controlled the current

and thus the intensity modulation of the LED output. The driver wasequipped with a voltage output for monitoring the LED current.Additional background intensity control (7 log units) was achieved byplacing neutral density filters (Kodak Wratten, Los Angeles, CA)between the stimulating light and the eye. A Cardan arm systemallowed the LED to be moved along the surface of an imaginarysphere (radius 50 mm) with the eye in the center. In this configu-ration, the LED spatially subtended 2° at the photoreceptor level.The current stimulus waveforms were also computer generated andinjected into the cells through the amplifier and the recording elec-trode (in current-clamp mode, using 8 kHz switching frequency).The stimulus waveforms were produced, the measurements con-trolled, and the signals sampled and analyzed with a Matlab (Math-Works, Natick, MA)-based, custom-written software (M. Juusola),running on a computer equipped with a data acquisition card (NationalInstruments, Austin, TX; PCI-MIO-16E-4). The monitored stimuliand the amplified responses were low-pass filtered at 250, 500, or700 Hz (Kemo VBF/23) and sampled at 0.5, 1.0, or 1.2 kHz.

Setup for in vitro experiments. The whole-cell patch-clamp record-ings were performed at room temperature (20°C) using borosilicatemicroelectrodes pulled to have a resistance of 3–8 M. The elec-trodes were filled with a solution containing (in mM) 140 KCl, 10 NTES, 2 MgCl2, 4 Mg ATP, 0.4 Na GTP, and 1 NAD (pH adjusted to7.15 with potassium hydroxide). The cockroach Ringer (extracellularsolution) contained (in mM) 120 NaCl, 5 KCl, 4 MgCl2, 1.5 CaCl2, 10TES, 25 proline, and 5 alanine (pH adjusted to 7.15 with sodiumhydroxide). The liquid junction potential between the solutions (4mV) was considered negligible for these recordings. Signals wereamplified with Axopatch one-dimensional amplifier (Molecular De-vices, Sunnyvale, CA), filtered with an eight-pole Bessel filter, andrecorded, sampled, and analyzed with pCLAMP 9 software (Molec-ular Devices). The light stimulation was provided by a green LED(525 nm), driven by a similar system as described above. Therecordings were sampled at 2 kHz and low-pass filtered at 1 kHz.

Stimulus protocols. In the in vivo experiments, all light stimuliwere given along the optical axis of each photoreceptor. This axis waslocalized before actual recordings by finding the maximal response toa test light flash when moving the light source across the receptivefield of each cell. In the in vitro experiments, the light stimuli weregiven to isolated ommatidia in a bath solution through the microscopeoptics. For each photoreceptor, the intensity of the stimulating LEDwas first calibrated by counting the number of single-photon re-sponses (quantum bumps) during constant dim illumination (back-ground light). This gave a cell-specific relative measure of intensity inso-called effective photons/s (Heimonen et al. 2006; Juusola et al.1994). Thereafter, the light stimulus waveforms were chosen accord-ing to the experiment in question, the cell’s functional category(Heimonen et al. 2006), and the need for additional results. The usedwaveforms were pulses with different intensities, contrasts, and du-rations or pseudorandomly modulated contrast sequences (filteredGaussian WN), both at different intensities of background light (amore complete account of these stimuli is reported in Juusola 1993,1994; Kouvalainen et al. 1994). Alternatively, the waveforms could bea time series of naturalistic light-intensity variation [naturalistic in-tensity series (NIS)] selected from van Hateren’s (1997) natural imagedatabase. When LA responses were recorded, the originally dark-adapted (DA) photoreceptors were exposed to each background lightlevel for at least 90 s before introducing contrast stimuli (steps, WN,or NIS). This aimed to ensure that the sensitivity of the photoreceptorshad reached a steady-state and that most adaptation processes werecompleted, including possible palisade formation and pigment migra-tion (Butler 1971, 1973b; Ferrell and Reitcheck 1993; Snyder andHorridge 1972). Long periods of light stimulation, especially withbright intensities, were always followed by DA for several minutes.The length of DA was considered to be adequate when the responseto a test flash had attained its original amplitude and overall shape.The light contrast (c) was defined as the change in light intensity (I)

2642 SIGNALING IN COCKROACH PHOTORECEPTORS

J Neurophysiol • doi:10.1152/jn.00588.2012 • www.jn.org

on May 20, 2015

Dow

nloaded from

divided by the mean light background (I). In the case of pseudoran-dom contrast modulation, the SD of the stimulus intensity was takenas I. The SD of WN contrast modulation was 0.32. The experimentswere performed first at the lowest level of background light beforeproceeding to the brighter ones. Additionally, to investigate theproperties of the photoinsensitive cell membrane, current pulses orpseudorandom WN current sequences were injected through therecording electrode into the photoreceptors in the discontinuous cur-rent-clamp (switched-clamp) mode of the amplifier. The SD of thepseudorandom current was 0.16 nA.

Quality of recordings. In vivo intracellular recordings had to fulfillthe following quality requirements to be included in this study: 65mV resting potential 55 mV; input resistance 60 M (testedwith a hyperpolarizing 0.3-nA current pulse at the dark restingpotential), and they have clearly visible quantum bumps with ampli-tudes of several millivolts (Fig. 1) at low-light intensities. Typically,the resting potential was 60 mV and the input resistance 70–80M, occasionally even 100 M. Recordings were also rejected ifafter a period of light stimulation, the sensitivities and time courses oftest flash responses failed to return to their initial values withinminutes of DA. In vitro whole-cell recordings had to fulfill thefollowing criteria to be selected into this study: resting potential 50 mV (recorded under current clamp in I 0 mode of theamplifier), series resistance 30 M before compensations, andinput resistance 500 M (tested under voltage clamp below restwith a voltage pulse from 70 to 80 mV). The resting potential was,on average, 59 6 mV (SD; n 20), the series resistance 5M after compensation, and the input resistance always 1 G.

In vivo intracellular recordings. The frequency response (presentedas the gain and the phase functions) and the coherence function

between the WN-modulated contrast stimulus and the induced pho-toreceptor voltage responses (French 1980a, b), as well as the SNR ofthe responses in frequency domain [SNR(f)], were all recorded andcomputed as reported earlier (Juusola et al. 1994; Kouvalainen et al.1994). Typically, 16-s-long data sequences sampled at 1 kHz wereused. The frequency domain functions, including membrane imped-ance (frequency response between injected current and recordedmembrane voltage), were calculated as averages of 10–20 adjacentresponses to a pseudorandom stimulus sequence. The linear-impulseresponses (first-order Wiener kernels) were obtained from the fre-quency responses via the inverse fast Fourier transform. Details ofhow to measure cell impedances in vivo, using WN-modulated currentinjections, are reported earlier (Juusola and Weckström 1993; Weck-ström et al. 1992). The coherence functions between NIS and photo-receptor voltage responses and the SNR(f) of these responses werealso recorded and analyzed as described earlier (Heimonen et al. 2006;Juusola et al. 1994; Kouvalainen et al. 1994). Here, 10 10-s-long datasequences were sampled at 1.2 kHz and averaged. Finally, the sig-naling performance of cockroach photoreceptors was quantified (inbits/s) from the voltage responses to a 100–200 repeated NIS, usingthe triple extrapolation method (Juusola and de Polavieja 2003). Thismethod measures the rate of information transfer as the differencesbetween the response entropy and the noise entropy rates. Here, theduration of the recorded sequences was 2 s, the sampling frequency500 Hz, and both stimuli and responses were low-pass filtered at250 Hz.

In vitro whole-cell recordings. Responses to bright, 10-s-long lightpulses were recorded both in I 0 and voltage-clamp mode (at aholding potential of 70 mV) of the amplifier in the same cells tojustify direct comparison of their waveforms. SNR of responses

"slowly/non-adapting""adapting""hyper-adapting"

200 ms

20 mV

A

2 s

20 mV

C

200 ms

2 mVB

Fig. 1. Dark-adapted (DA) cockroach photoreceptors produce variable voltage responses to light steps in vivo. Cells with different types of responses can becategorized according to their adaptation dynamics: “hyper-adapting” (left column), “adapting” (middle column), and “slowly/non-adapting” (right column).A: typical saturated voltage responses to 300 ms light pulses in each category. B: examples of single-photon responses (quantum bumps) during constantillumination. Even during bright illumination, hyperadapting responses often show bump-like events (left), while repolarizing close to the dark resting potential(see box and arrow). Adapting and slowly adapting cells (middle and right) generate clearly distinguishable single-photon responses (here, 3 and 17 bumps/s)only during dim illumination. C: saturated responses to bright, 10-s-long light pulses (timing shown between the response rows) exemplify the variability inadaptation time course and level of depolarization among different types of cells. The dotted lines below the responses indicate the dark resting potential (ca.60 mV in each cell).

2643SIGNALING IN COCKROACH PHOTORECEPTORS


on May 20, 2015

Dow

nloaded from

(recorded in I 0 mode) to a repeated (10–15) 10-s-long, WN-modulated contrast sequence (c 0.32) was estimated in two ways.The initial peak transient (the first second of the response) was left outof the analysis, and the rest of the responses (1–10 s) was detrendedby using a sixth-order polynomial fit. The SNR(f) was first analyzed asreported previously (Kouvalainen et al. 1994). Then, in SNR timedomain [SNR(t)], was estimated in the following way: 1) the de-trended responses were divided into 10 equal length-time segments(10; 0.9-s bins); 2) the corresponding response segments were aver-aged to give an estimated signal; and 3) the variance of this signal wasdivided by that of noise (the difference between the signal and anoriginal detrended response) to obtain SNR(t) of each segment. In theSNR recordings, the mean light intensity was always bright enough toproduce saturated responses in DA cells.

RESULTS

The most remarkable and unusual feature in cockroachphotoreceptor function is the large variability in their voltageresponses (Heimonen et al. 2006). This functional diversity,which seems not to follow any clear retinotopical organization,can be used to categorize the photoreceptors according to theiradaptation profile, or rate of light-response decay, during lightstimulation (Fig. 1). The three photoreceptor categories: “hy-peradapting”, “adapting”, and “nonadapting” typify two ex-tremes and an intermediate behavior of an actually continuousdistribution of response variability (Heimonen et al. 2006).

In the present study, we examined how this functionaldiversity affects visual encoding. We first present results fromin vivo (intracellular) and in vitro (whole-cell patch-clamp)recordings in DA photoreceptors. The light-induced current(LIC) responses (Fig. 2) were examined to find out whether thesource of the variability of the light-voltage responses is in theLIC. Then, we examined how LA photoreceptors can codecontrast steps into voltage responses (Fig. 3) at various lightlevels and if this shows functional specialization to dark. Thedynamics of contrast coding was investigated further by re-

cording and analyzing responses to WN-modulated contrastsequences at different adapting backgrounds (Fig. 4). Theelectrical properties of the light-insensitive photoreceptormembrane were quantified to examine their contribution tocoding (Fig. 5). Finally, to determine the photoreceptor per-formance at various light levels, we measured the SNR, boththe SNR(f) and the SNR(t), of voltage responses after the lightonset (Fig. 6) and computed the rate of information transfer forNIS (Fig. 7). These results quantify how the response proper-ties of cockroach photoreceptors are selectively adapted forencoding visual signals in scotopic environments.

Light responses of DA photoreceptors. The DA voltageresponses were recorded first to show the variability as de-scribed before (Heimonen et al. 2006) and secondly, to furtherstudy the LA as it is reflected in response shapes with long lightpulses of up to 10 s. When dark adapted, the impulse responsesof different types of photoreceptors were often similar. How-ever, the cells could be distinguished by their 100-foldsensitivity differences, as determined by their individual volt-age vs. log of intensity (V-logI) curves (Heimonen et al. 2006;Weckström et al. 1993). With longer light pulses, differencesin adaptation emerged (Figs. 1, A and C).

Whereas the early transient responses of different cells (Fig.1C) had similar shapes, they repolarized differently duringconstant light. Cells that were categorized earlier as nonadapt-ing, using 300 ms stimuli (Heimonen et al. 2006), actuallyresponded with slow repolarization toward dark resting poten-tial when longer, 10-s light pulses were used (Fig. 1). Hence,they are better termed “slowly adapting”. Fig. 1 also showsresponses of the hyperadapting cells. Surprisingly, these pho-toreceptors produced bump-like events with more or less un-attenuated amplitudes in bright light when the initial stepresponse had repolarized close to the dark resting potential(Fig. 1B). Thus the “hyperadaptation” is more like reducedexcitation efficiency than adaptation of bumps. Although here

20 mV

0.5 nA

10 s

Current clamp

Voltage clampFig. 2. In vitro whole-cell patch-clamp recordings of DA photoreceptors show similar variability both in voltage responses (top row) and in light-induced current(LICs; bottom row; at 70 mV) when stimulated with bright (saturating), 10-s-long light pulses (timing shown with bars between the response rows). Theresponses in each column are produced by the same (repeated) stimulus and recorded from the same cell. The time courses of the voltage responses are closelymirroring the shape and waveform of the corresponding LICs. Scales are the same for all of the responses. The transient early responses of the LICs are clippedbecause of their large amplitude.



on May 20, 2015

Dow

nloaded from

found to be a misnomer, the hyperadapting shall be used in thefollowing to be consistent with the existing nomenclature(Heimonen et al. 2006) when referring to those kinds ofphotoreceptor responses.

In very dim conditions, the light responses of all cockroachphotoreceptors consisted only of single-photon-induced dis-crete events—quantum bumps (Fig. 1B). At stimulus intensi-ties of 10 photons/s, the bumps were large (up to 5–10 mV).In contrast to the 100-fold variation in sensitivity of gradedlight responses (Heimonen et al. 2006; Weckström et al. 1993),the differences in the absolute sensitivity were 10 times less.With the same dim background, the average bump frequencyof different cells varied between two and 20 bumps/s (9 4bumps/s; mean SD; n 18).

The findings thus showed that cockroach photoreceptors cangenerate large voltage responses to single photons and that themacroscopic responses tend to adapt toward the restingpotential.

LICs and corresponding voltage responses. To search forthe sources of functional variability, we recorded LICs fromisolated photoreceptors in voltage-clamp mode (Fig. 2; notethat the large, unreliable, initial 6-nA transients are trun-cated). When comparing voltage and current step responses forsaturating responses, we found that their waveform dynamicswere nearly equivalent in each of the cells. For example, thevoltage response of a hyperadapting cell closely followed the timecourse of its LIC. These results suggest that 1) the variability involtage responses reflects the variability in the phototransductionprocess (LIC) and that 2) the role of the nontransductive mem-brane properties in shaping the response waveforms is limitedcompared with the LIC.

Contrast coding with step responses in time domain. Wenext examined how the different cells encoded light contrasts

at different adapting backgrounds. Photoreceptors were firstlight adapted to different light levels and then stimulated withcontrast pulses, i.e., light increments or decrements (Fig. 3A).

In response to a prolonged (90 s) constant light exposureand adaptation, surprisingly, only 39% of the cells (nine of 23)sustained their steady-state depolarization markedly (5–15mV) above the dark resting potential (Figs. 3, A and B). Inthese photoreceptors, originally categorized either adapting ornon/slowly adapting, the contrast gain (the slope and range oftheir V-logI curves) increased marginally with increasing lightintensity. In other words, the same contrasts evoked largerresponses at brighter backgrounds. In particular, we found thatthe higher the maintained steady-state depolarization, thelarger the contrast-evoked responses. However, this increase incontrast gain took place only at low background intensities(Fig. 3B). When the steady-state depolarization saturated, thecontrast-evoked responses saturated as well. This usually oc-curred at an intensity range of 1,000–3,000 photons/s (eight ofnine cells) and in only one case at 10,000 photons/s. Theresponses to light increments and decrements were onlyroughly symmetrical, with the smallest (0.2–0.4) contrastsreminiscent of those in fly photoreceptors (Juusola 1993). Forthe larger contrasts (0.6–1.0), the asymmetric nonlinearitiesbecame more obvious, whereupon the hyperpolarizing re-sponses to light decrements were typically larger and slowerthan the depolarizing ones to light increments. For low-contrastresponses at low-intensity backgrounds, the SNR was verylow, and individual responses were noisy and difficult todistinguish from each other, even after being averaged 20times. In the slowly adapting cells (sustaining 5–15 mV steady-state depolarization), the capability for contrast coding resem-bled that of adapting cells (Figs. 3, A and B), but theirresponses to positive contrasts were occasionally smaller, con-

10 100 1000 10000 100000

-60

-56

-52

-48

-44

-40"adapting"

320 000 ph/s

4 mV

5 mV

11 mV

14 mV

12 mV

100 ms10 mV

13 mV

BA

C100 ph/s

320 ph/s

1000 ph/s

3200 ph/s

32 000 ph/s

Mem

bran

e vo

ltage

(mV

)

Light intensity (ph/s)

0 mV

15 mV

3000 ph/s100 ms

10 mV2800 ph/s

"hyper-adapting"

"slowly-adapting"

Fig. 3. Light-adapted (LA) photoreceptors of different categories produce different voltage responses to positive and negative contrast pulses in vivo (allrecordings are 10–20 times averaged). A: in a typical adapting cell, responses to contrast stimuli [protocol at the bottom; 1 light contrast 1] change bothin amplitude and in shape as a function of the background intensity [given on the right as photons/s (ph/s)]. Changes are especially obvious among thelow-intensity backgrounds. In addition, the average “steady-state” depolarization (given on the left; from the dark resting potential of 60 mV) increases withthe intensity of the background. B: voltage vs. log of intensity (V-logI) curves for the data in A depicts the changes in the maximum amplitude of the contrastresponse (solid lines connect the data points at each background) and in the background-induced steady-state potential (dotted lines). Note that both thesteady-state depolarization and the contrast responses (whole V-logI curves) saturate already 1,000 photons/s. C: examples of similar contrast-step responsesin different cell categories. Stimulus protocol was the same, and the background intensities and steady-state potentials are presented, as in A. Note that in some(not all) “slowly adapting” cells, the positive responses were very small, and in hyperadapting cells, all of the responses are nearly nonexistent.



on May 20, 2015

Dow

nloaded from

taining randomly overlapping quantum bumps (Fig. 3C). Nat-urally, this noisy nature reduced the amplitude of averagedresponses.

Unexpectedly, most cells (14/23; 61%), including manycategorized originally as adapting or nonadapting, in fact,behaved quite similarly to the hyperadapting cells duringprolonged (90 s) illumination (Fig. 3C). Like the hyperadapt-ing photoreceptors during the first second of their response toconstant light (Fig. 1), all of these cells repolarized during theprolonged illumination within 0–5 mV of the dark restingpotential. Consequently, these photoreceptors completely failedto encode negative contrasts (Fig. 3C) and did this irrespective ofthe background intensity. Additionally, only some of themcould respond to positive contrasts (c 1) with random andsparse bump-like responses (similar to Fig. 1B), which werelost in averaging. As a result, the majority of cockroach

1 10 100

0.0

0.2

0.4

0.6

0.8

1.0

0.1

1

10

0 50 100 150 200

0.0

0.1

0.2

D

Frequency (Hz)

Coh

eren

ce

1 10 100

0.0

0.2

0.4

0.6

0.8

1.0AC

oher

ence

Frequency (Hz)

B

Pha

se (

deg.

)

Gai

n(m

V/u

nit c

ontr

ast)

Frequency (Hz)

-1200

-900

-600

-300

0

1 10 100

C

Time (ms)

K*(

mV

/uni

t con

tras

t)

Fig. 4. Linear filtering properties of LA cockroach photoreceptors investigatedwith white noise (WN)-modulated light stimulation also show differencesamong different cell categories. A: coherence functions show increase withbrightening background (100, 320, 1,000, 10,000, 32,000, and 100,000 pho-tons/s) in this adapting cell, as it did in all cells that maintained theirsteady-state depolarization during the whole illumination period at differentlevels of light adaptation. Here, the order of the traces follows brightening (red:the lowest intensity). Note that cells behaving like this were only a minority ofadapting and slowly adapting photoreceptors. B: linear frequency-responsefunction (gain and phase) for the example cell in A remains nearly constantwith brightening (red: the lowest intensity). C. linear-impulse responses[1st-order Wiener kernels] were calculated for the data in B and are almostidentical. The poor coherence at the lowest background (red in A) makes theestimates in B and C (red) unreliable. D: like this example, coherence functionsof hyperadapting cells always had very low values, irrespective of the adaptingbackground intensity (14, 44, 140, 440, 1,400, 4,400, and 14,000 photons/s).Note that the trace order does not follow the order of increasing intensity; forinstance, the highest coherence is for 1,400 photons/s and the second highest(gray) for 14,000 photons/s. Note also that a majority of cells, regardless oftheir initial behavior, also responded like this—i.e., hyperadapted close to darkresting potential during prolonged illumination—and after this, always hadvery low coherence values.

1 10 100

10

100

1 10 100

0.0

0.2

0.4

0.6

0.8

1.0

A

100 ms

10 mV

0.4 nA

B

Frequency (Hz)

Impe

danc

e (M

Ω)

CC

oher

ence

Frequency (Hz)

Fig. 5. Functional properties of the light-insensitive membrane in vivo aresimilar in all cockroach photoreceptors. A: a typical current-clamp recording(resting potential 62 mV) shows slow charging of nearly passive membranewhen hyperpolarized with current pulses (protocol below) and outward recti-fication when depolarized. B: examples of impedance functions of a typicalphotoreceptor illustrate this same behavior. In relation to the dark restingpotential (61 mV), the average membrane potentials from top to bottom are25 mV, 0 mV, 8 mV (red), and 10 mV. Hyperpolarization was inducedby current injection and depolarization by constant illumination (8 mV; red)and current injection (10 mV; black). C: coherence functions for the data inB show high linearity and signal-to-noise ratio (SNR) in all voltages (25, 0,and 10 mV; all black). Depolarization by light ( 8 mV; red) decreases thecoherence at lower frequencies by adding photon shot noise.



on May 20, 2015

Dow

nloaded from

photoreceptors (here, 61% of tested cells) could not encodecontrast steps below unity (1 c 1) when LA for a longenough time. Because these cells could encode light changesafter DA, generating large impulse and step responses, weconclude that prolonged light exposure selectively abolishestheir encoding abilities.

Contrast-coding experiments showed that the coding ofstep-like contrasts does not increase in cockroach photoreceptorsbeyond background intensities of 1,000–10,000 photons/s.

Coding of WN-modulated contrast in frequency domain. Toquantify the response speed and reliability in terms of lineardynamics of contrast coding, we computed the coherence andfrequency response functions between the WN-modulated lightstimulus and the voltage response at different light back-

grounds (Fig. 4). Because WN stimulation contains fast lightchanges around the mean, it has a tendency to linearize theoutput of insect photoreceptors (Faivre and Juusola 2008;Juusola and Hardie 2001; Juusola et al. 1994; Niven et al.2004; Spekreijse and Oostings 1970). Here, the stimulus waschanging fast enough to keep most responses within the linearrange while still being sufficiently variable to induce occa-sional larger responses (10–15 mV). Similar to the results inthe contrast-step experiments, we found that the contrast gainimproved only marginally, if at all, with brightening intensity.Furthermore, the speed of responses was also practically in-variant; the 3-dB cutoff frequency and the phase of the re-sponses remained unchanged across the tested light backgrounds.

Like in contrast-step experiments above, a minority of thecells studied (six of 20; 30%) typically maintained 10–15 mVsteady-state depolarization during prolonged constant lightexposure (1,000 photons/s). Compared with the majority ofcells, these photoreceptors had also markedly higher values ofcoherence (0.5) at low stimulus frequencies (0.5–30 Hz)—inthe best case, 0.8–0.9 at the brightest background (Fig. 4A).Thus apart from the dimmest photon shot noise-dominatedbackgrounds (1,000 photons/s), their frequency-response es-timates (Fig. 4B) were sufficiently reliable for comparativeanalysis. Their contrast gain (mV/unit contrast) generally in-creased only slightly, if at all, at low frequencies as a functionof light intensity (Fig. 4B), and the variation of the 3-dB cornerfrequency of the cell (19–24 Hz) was small and failed to risesystematically with a brightening background. This suggestedthat the frequency response withstood LA largely unchanged.Because of lack of a systematic rise with brightness, the smallvariance in corner frequency is likely to be a result of thevariance inherent in the computed estimate, due to the limitedamount of data (Bendat and Piersol 1971), and hence, calcu-lating the average corner frequency as a function of intensitywould be misleading. The conclusion—that there were nosystematic differences in temporal properties of the voltageresponses among different levels of LA—was supported fur-ther by the astonishingly constant-phase functions (Fig. 4B),which are normally very sensitive to adaptive changes (Juusolaet al. 1994). Finally, this constancy was quantified in timedomain by calculating the linear-impulse responses from thefrequency-response functions (Fig. 4C). When comparing dif-ferent cells at the brightest backgrounds, where the coherencehad the highest values, and the frequency responses were mostreliable, the corner frequency varied between 15 and 25 Hz.

The majority of studied cells (14/20; 70%) was again, like incontrast-step experiments, found to behave like the hyper-adapting photoreceptors. Irrespective of the background bright-ness, they repolarized within 0–5 mV of the dark restingpotential during long-lasting constant stimulation. Their coher-ence values were invariably very low (0.4) in all tested lightbackgrounds, covering 3–5 log units and starting at about 10photons/s (Fig. 4D). This implies that their responses either hada low SNR or were highly nonlinear (Bendat and Piersol 1971).It is reasonable to assume here that their low coherence mainlyresulted from the low SNR, because when similarly LA, mostphotoreceptors (61%) failed to generate proper graded re-sponses even to contrast steps (Fig. 3C). If anything, thesephotoreceptors produced only quantum bumps to positive con-trasts. Because of the low coherence, the frequency-response

1 10

0.1

1

0 1 2 3 4 5 6 7 8 9 10 11

0.0

0.1

0.2

0.3

0.4

0.5

1 s

10 mV

1 mV

B

A

Frequency (Hz)

SN

R

C

Time interval (s)

SN

R

Fig. 6. SNR of cockroach photoreceptors decreases markedly during prolongedlight stimulation in vitro. A: a typical example voltage response (middle trace)to a bright WN-modulated stimulus [bottom trace; light background (Imean)5,000 photons/s] after dark adaptation resembles the waveform of a responseto a light step (cf. Figs. 1 and 2). The transient response (1st s) is omitted fromthe SNR analysis. The subsequent overall waveform (110 s) is fitted, and theresult is subtracted from the original response to leave 9 s of the WNmodulation-induced response with 0 mean (top trace). B: mean SNR of theresponses in frequency domain [SNR(f)] with SD bars (n 5) is shown for 3time periods: 1–5.5 s (black); 5.5–10 s (gray), and 1–10 s (red; SD bars omittedhere for clarity). Only photoreceptors with higher SNR (1) at low frequencieswere chosen for this analysis. C: the mean SNR time domain (presented withSD bars; n 13) during different time intervals (10 0.9-s bins) of the WNstimulus shows a clear and continuous drop after the first 5 s.



on May 20, 2015

Dow

nloaded from

estimates were highly unreliable (Bendat and Piersol 1971) andwere not evaluated further.

Experiments with WN stimulation again showed that thecoding properties of photoreceptors do not change beyond1,000 photons/s and that with WN stimulation, the codingreliability at all light levels is relatively poor.

Membrane filtering. To find out how the properties of thenontransductive membrane are related to signal coding, weinjected currents into the photoreceptors through the recordingelectrode using the switched-clamp technique. The input resis-tances of DA cells, when hyperpolarized (with a 0.3-nAminus 0.5-nA pulse) below the resting potential, were rela-tively large (60–180 M) and increased with the level ofhyperpolarization (Fig. 5A). The membrane charged very slowly;the time constants were between 20 ms and 50 ms in responseto the negative current pulses. Conversely, depolarizing pulsesrevealed a strong rectification, which is probably caused by theactivation of voltage-gated K-channels, as in other studiedinsect photoreceptors (Hardie 1991; Laughlin and Weckström1993; Niven et al. 2003; Vähäsöyrinki et al. 2006; Weckströmet al. 1991; Weckström and Laughlin 1995).

The membrane impedance or the frequency response be-tween the current input and voltage output (Fig. 5B) wasdetermined with WN-modulated current injections (Juusolaand Weckström 1993; Weckström et al. 1992). The impedancebehaved as expected from the current pulse experiments (Fig.5A). When the membrane was hyperpolarized 20–30 mVbelow the dark resting potential, the impedance function (max-

imum 70–120 M) followed the shape of a single-pole resis-tor–capacitor (RC) filter, with a corner frequency of 9 1 Hz(mean SD; n 19; Fig. 5B). At the dark resting potential,the maximum impedance was 40–60 M and the cornerfrequency 20 2 Hz (n 22). When the membrane wasdepolarized, either by current injection or by constant illumi-nation, the impedance at low frequencies dropped to 10 M,and the average corner frequency rose to 48 7 Hz (n 11).Again, these changes probably resulted from gradual activationof voltage-gated K-channels. The corner frequency of the LAmembrane (Fig. 5B) was about two times larger than that of thephototransduction output, i.e., light-induced voltage response(Fig. 4B). This suggested that the membrane does not limit thespeed of the voltage responses in LA. No systematic associa-tion between the adapting properties (category) of the photo-receptor and the impedance of its membrane was found.Instead, all of the photoreceptor membranes behaved similarlywithin the above-given variation in corner frequency. Thecoherence functions indicated high linearity in all of the im-pedance recordings (Fig. 5C). When membrane was depolar-ized by constant light stimulation, the small drop in coherencevalues at low frequencies was most likely caused by photonshot noise.

With WN-modulated current and computation of the imped-ance functions, we could show that the DA membrane is veryslow but does not limit the light-induced voltage responsesunder LA conditions.

0.1

1

10

100

0.1 1 1010-8

10-6

10-4

10-2

1 100.0

0.2

0.4

0.6

0.8

1.0

100

101

102

103

104

105

106

0

20

40

60

80

100

120C

SN

R

LA

DA

A

LA

DA2 s

10 mV 15,000 photons/s

NIS Rel

ativ

e po

wer

Frequency (Hz)

B

Coh

eren

ce

LA

Frequency (Hz)

DA

Info

rmat

ion

rate

(bits

/s)

Average light intensity (photons/s)

WN limit

Linear limit

Fig. 7. Stimulation with naturalistic light intensity series (NIS) evokes noticeably higher SNR and information transfer rate than WN stimulation. A: voltageresponses from DA (black) and LA (red) photoreceptors to NIS (Imean 5,000 photons/s) carried much larger voltage variations than corresponding WN-inducedresponses (cf. Fig. 6A). The power spectrum of the stimulus (on the right) followed roughly the 1/f shape of intensity variation found in natural scenes.B: coherence functions and SNR(f)s (averages for 10 recordings as in A; color as in A) had markedly higher values, especially at low frequencies, thancorresponding functions for WN stimulation (cf. Figs. 4, A and B, and 6, B and C). C: information transfer rate (mean with SD bars) as a function of averagelight intensity in all 7 different photoreceptors is clearly higher than ever in WN experiments. Cell-specific intensity values were obtained with bump calibrationfor each photoreceptor separately. Gray symbols and lines represent 3 rapidly and strongly adapting cells, of which 1 was hyperadapting (the lower-most trace).Black symbols and lines represent 4 more slowly adapting cells, similar to the example cell in A. “WN limit” for the information transfer rate [13 bits/s; Shannon(1948) information capacity] is calculated for the highest mean SNR(f) in Fig. 6B (1–5.5 s; black) and the “Linear limit” (45 bits/s) for the higher SNR(f) in B(DA; black in lower panel).



on May 20, 2015

Dow

nloaded from

SNR. Our previously published results (Heimonen et al.2006) indicated that after a prolonged LA, the SNR(f) would beconsistently very low (0.1 in the majority of cells andmaximally, 0.3 in some cells at low frequencies). However,less was known about how SNR actually changes over timeimmediately following light stimulation. To address this ques-tion, DA cells were stimulated repeatedly by a 10-s-longWN-modulated contrast sequence. As seen in the responsesequence (Fig. 6A), its modulation amplitude (i.e., the signal)was largest in the beginning and then diminished gradually.Further analysis in frequency domain, using only cells with thehighest SNRs (Fig. 6B; n 5), showed that the SNR(f) peakedfor a few seconds after the stimulus onset, thereby implyingthat apart from the initial transient, the best performance ofcockroach photoreceptors is in the early phases of LA. Duringthe first 5 s of light stimulation, the SNR(f) had maximal values(1) up to 5 Hz, after which, it fell below 0.1 at 10–11 Hz.Additional analysis in the time domain (Fig. 6C; n 13)quantified that the SNR(t) peaked on average for the first 5 safter stimulus onset and thereafter, rapidly decreased to the lowlevels reported previously (Heimonen et al. 2006) as the cellsadapted to the mean light intensity.

These experiments showed the rapid decline of the SNR inphotoreceptors when stimulated with WN stimulation, consis-tent with the results of experiments with contrast pulses.

Responses to naturalistic stimuli. Finally, to estimate thesignaling performance of cockroach photoreceptors for morerealistic stimuli, we recorded in vivo voltage responses of 12photoreceptors to NIS. Fig. 7A shows representative responsesof a slowly adapting photoreceptor, both after prolonged(90 s) dark and LA. After DA, the NIS onset evoked aprominent depolarization, which decreased slowly over time,while being strongly modulated by the larger contrasts in thestimulus (NIS). After LA, the same cell produced a smallerresponse to the same NIS but still clearly followed the stimuluswaveform and encoded the larger contrasts in it. In general, wefound that faster-adapting cells reached this LA-encoding statemore rapidly and hyperadapting cells immediately after thetransient response or within the first second of stimulation.Thus all photoreceptors produced prominent voltage responsesto NIS. Even the LA-hyperadapting cells generated up to 10mV responses to the largest contrasts (1) in it.

The coherence functions and the SNR(f)s of responses to NIS(Fig. 7B) were determined from nine photoreceptors both inDA and LA. In all of these cells, both functions were similar tothe shown examples, although the hyperadapting photorecep-tors had somewhat smaller values. Even though the coherencehad higher values in LA, the SNR(f) was higher after DA. Thiswas also seen during the first 5 s of the responses (Fig. 7A),where the modulation caused by the same stimulus sequencewas notably larger after DA than in the LA state. This corre-sponds well with the WN results in Fig. 6. However, there werealso prominent differences. The SNR(f) of responses to WNstimulation was always low—one to two at largest in lowfrequencies (5 Hz; Fig. 6B)—whereas the SNR(f) of re-sponses to NIS, presented here on a logarithmic scale (Fig. 7B),had 100-fold higher low-frequency content (5 Hz) and inmany cases, contained more signal than noise (SNR 1) up to10 Hz.

The rate of information transfer was estimated at five toseven different average light intensities in seven different

cockroach photoreceptors (Fig. 7C) using the triple extrapola-tion method (Juusola and de Polavieja 2003). In brighterstimulation (1,000 photons/s), hyperadapting cells had mark-edly lower information transfer rates than the more slowlyadapting cells. One of the photoreceptors reached a relativelyhigh rate of over 100 bits/s, already at an astonishingly lowlevel of light intensity (200 photons/s). The signaling per-formance of all recorded cells saturated in intensities between100 and 10,000 photons/s.

To finalize the results, we compared the photoreceptorperformance estimates obtained under different stimulationconditions and with different analyzing methods. In addition toinformation transfer rates during NIS stimulation, Fig. 7C showsestimates of Shannon (1948) information [C log2(1 SNR) df], achievable with the WN used [“WN limit”, calcu-lated for the highest mean SNR(f) in Fig. 6B, and “linear limit”for the experiments with naturalistic stimuli, calculated for thehigher SNR(f) in Fig. 7B]. Therefore, WN limit approximatesthe achievable information capacity during WN stimulation,whereas the linear limit does the same when linear analysis isapplied to the results of the NIS experiments.

These sets of experiments showed that whatever stimulationmethod is used, the contrast coding in cockroach photorecep-tors seems to saturate near intensities of 1000 photons/s.

DISCUSSION

With the use of various experimental and analytical ap-proaches, we quantified functional properties of cockroachphotoreceptors and showed how their signaling is selectivelyadapted for encoding light changes in dim naturalistic condi-tions. In the following, we consider the physiological adapta-tions of cockroach photoreceptors in specializing for the sco-topic conditions.

Phototransduction gain. The gain of cockroach phototrans-duction is high, leading to production of large (5–10 mV)quantum bumps (Fig. 1). Large bumps may be sufficient todrive the photoreceptor synapses, transmitting signals todeeper layers of the nervous system. In contrast, the quantumbumps are often small (1 mV) in the few studied insectphotoreceptors, including most diurnal flies (Dubs et al. 1981;Hardie 1979; Laughlin 1981; Wu and Pak 1975). Large quan-tum bumps have been found in rods of some vertebrate species(Baylor et al. 1979); in photoreceptors of nocturnal or crepus-cular invertebrates, such as horseshoe crab Limulus (Dodgeet al. 1968; Fuortes and Yeandle 1964), bee Megalopta(Frederiksen et al. 2008; Warrant 2008), and crane-fly Tipula(Laughlin 1996); and in locusts (Lillywhite and Laughlin 1979)that are active also in daylight. Hence, a large gain of photo-transduction in dim conditions is obviously a necessary prop-erty for all animals active in such environments.

At bright light intensities, high phototransduction gain hasthe unwanted property of creating more voltage noise andthereby decreasing the SNR. In cockroach photoreceptors, thecoherence and the SNR were poor at all light levels with WNstimulation (Figs. 4 and 6) (Heimonen et al. 2006), especiallyafter long-lasting LA. The time course, during which thephotoreceptor gain and SNR are decreased markedly whenilluminated, is 10 s (Fig. 6). The contrast gain both in stepand frequency responses was essentially unchanged aboveintensities of 1,000 photons/s (Figs. 3B and 4B), indicating that



on May 20, 2015

Dow

nloaded from

phototransduction is saturated in some manner. A comparisonof well-studied diurnal or crepuscular insects, such as Calli-phora (Juusola et al. 1994) and Drosophila (Juusola andHardie 2001; Niven et al. 2003), with a nocturnal bee Mega-lopta (Frederiksen et al. 2008) reveals a striking contrast. In allof these species, the bumps become smaller and faster, the gaindiminishes, and the signaling bandwidth increases, while theSNR rises with increasing light up to daylight intensities. In thecockroach, the absence of these processes above 1,000 pho-tons/s means that the photoreceptors adapt to bright light levelsineffectively. For the cockroach, the high gain at or near thequantum bump level seems to be paramount.

Phototransduction speed and tuned filtering. In dim condi-tions, visual signals are integrated over time, and accordingly,the photoreceptor response dynamics are slow (van Hateren1992; Warrant 1999). This is also the case with the cockroach.The speed of voltage responses (Fig. 4C) and the chargingproperties of the DA photoreceptor membrane (Fig. 5B) aretuned similarly, both having 20 Hz corner frequencies. Thistranslates to an integration time constant of 8 ms ( 1/2f,where f 3 dB corner frequency) if a single-pole RC filter isassumed. Thus the temporal dynamics of phototransductionand the nontranductive membrane are matched approximately.

The slowness of light responses, as such, is not exceptional,since photoreceptors of other insect species are about as slowwhen dark adapted (Faivre and Juusola 2008; Frederiksen et al.2008; Howard et al. 1984; Juusola and Hardie 2001; Juusolaet al. 1994; Laughlin and Weckström 1993; Niven et al. 2003).Surprising in the cockroach is that the speed of the voltageresponses in bright light equals that in dim conditions. Theusual finding of responses accelerating with increasing bright-ness was not observed. The response speed might increaseslightly in some photoreceptors from a single-photon level upto 1,000 photons/s but not beyond that. The constancy inresponse speed is best seen in the phase functions (Fig. 4B),which are normally the most sensitive indicators of adaptivechanges. Concomitantly, also, the steady-state depolarizationin constant light saturates at the same light level (Fig. 3B).Hence, cockroach photoreceptors are able to transduce themaximum of only 1,000 photons/s, with only a minority ofcells slightly exceeding that number. On the basis of theanatomy of cockroach retina, the number of microvilli iscertainly much larger than 1,000 (Butler 1973b; Ferrell andReitcheck 1993). The reason for the saturation of informationrate at dim-light level has, therefore, to be sought from thebump production dynamics. Again, the well-studied photo-transduction processes of the blowfly Calliphora (e.g., Howardet al. 1987; Juusola et al. 1994) and fruitfly Drosophila (e.g.,Hardie and Raghu 2001; Juusola and Hardie 2001; Niven et al.2003) contrast with this, showing clear changes in transductionspeed and gain, at least up to intensities of 106 photons/s.Accordingly, adaptive regulation of the gain and speed ofcockroach phototransduction is limited to very low intensities.This conclusion is also supported by a recent study on flyphotoreceptors, which provided strong evidence that bumps arelikely to be generated in individual microvilli (transduction orsampling units), after which, these remain refractory for avariable period—unable to produce another bump even whenhit by new photons (Song et al. 2012). According to thoseresults, the ratio between activated and available microvilli andthe timing and size of the bumps that they produce may be

crucial in determining the photoreceptor’s information transferrate and the intensity where it saturates. Accordingly, if mi-crovilli in a cockroach photoreceptor generate large bumps andhave long refractory periods, then even a relatively dim-lightbackground would gradually render most of the microvillirefractory, saturating the photoreceptor output. This couldexplain why the signaling performance of cockroach photore-ceptors to WN stimulation is so poor and improves onlymarginally with brightening.

However, when using NIS stimulation (van Hateren 1997),containing much higher contrasts and nearly 1/f frequencycontent, the photoreceptor performance becomes much better(Fig. 7, A and B), reaching a SNR of 100 and coherence of1 at low frequencies. Moreover, in contrast to the mostlyindecisive results of WN experiments, the photoreceptor per-formance improves with increasing brightness of NIS stimula-tion. Information transfer rates can reach 100 bits/s with NISstimulation, whereas they are only 20 bits/s during WNstimulation. For a purely linear measure, such as the oneobtained from the SNR(f) by using the Shannon (1948) formulafor information capacity, the estimated information transferrate is limited below 45 bits/s. The high rates (100 bits/s)were obtained only with NIS stimulation and a novel informa-tion rate estimator (Juusola and de Polavieja 2003; Takalo et al.2011), which does not assume linearity of the system orGaussian distribution of the signals. Hence, the signalingmechanisms in cockroach photoreceptors appear to be highlyadapted to naturalistic light changes, which contain long,darker periods that are likely to foster recovery of microvillifrom the refractory state. Since the novel information rateestimates were clearly larger than the linear estimates, thisadaptation is most likely achieved by using nonlinearities whentuning the photoreceptor signaling to code the changes of lightintensity in natural environments. One such tuning mechanismis obviously formed by adaptation dynamics, and clearly, onefuture challenge is to elucidate optimal stimulus statistics anddynamics for the cockroach photoreceptors.

Variability of cockroach phototransduction. To explain thelarge variability of light responses, a specific form of popula-tion coding has been suggested to take place when cockroachphotoreceptors are pooled for joint signal transmission (Hei-monen et al. 2006). Most of the variability seems to originatein the phototransduction process, as was shown by the strikingsimilarity of the dynamics of the voltage and current responsesto light (Fig. 2). In addition, the photoreceptor sensitivityvaried much less at the level of quantum bumps than at thelevel of full-fledged macroscopic responses. This suggests thatcell-to-cell variation in the expression level of the light-gatedchannels is probably not the reason. The variability in proper-ties of the nontransductive membrane of cockroach photore-ceptors (Fig. 5) was also negligible. These findings suggest thatmuch of the variability in macroscopic responses is introducedduring so-called “bump summation”, where graded receptorpotentials are summed up from single-photon signals and in theadaptation processes therein, rather than in the single-photonresponses themselves. This summation process is well ana-lyzed only in the case of Limulus photoreceptors (Wong andKnight 1980; Wong et al. 1980, 1982).

In the bump summation process, the generated bumps, eachpresumably arising from a single microvillus (Hardie andRaghu 2001; Howard et al. 1987; Song et al. 2012), and the



on May 20, 2015

Dow

nloaded from

calcium signals, arising from the calcium influx during photo-transduction (Hardie 2001; Hardie and Raghu 2001; Minke andParnas 2006), are integrated by the cell soma and its mem-brane. For changes in the speed of this summation to occur, amechanism is needed to change communication of microvilliwith each other. Different calcium-dependent processes areprobable candidates for this. They are also likely to be themajor source of variability in the cockroach photoreceptors.Adaptation processes that could produce the variability in-clude: translocation of the light-gated channels (Bähner et al.2002; Meyer et al. 2006), concomitant changes in pH causedby protons released from their buffers by an increased intra-cellular calcium concentration (Huang et al. 2010), other ca-clium ion (Ca2)-dependent enzymatic processes in photo-transduction (Hardie 2001), and regulation of nontransductivemembrane conductances (Krause et al. 2008).

Light responses of photoreceptors in cockroach and trpmutants. When observing the rapid repolarization of the cock-roach photoreceptor light responses during continuing stimu-lation, one cannot avoid the comparison of this behavior withthat of transient receptor potential (trp) mutants of D. melano-gaster (Cosens and Manning 1969; Minke et al. 1975) and thenss-mutant (no steady-state) of the blowfly Lucilia (Barash et al.1988; Howard et al. 1984; Suss-Toby et al. 1991). Both of theselack TRP channels, and consequently, the LIC is caused onlyby TRP-like (TRPL) channel activation. In these mutants, thevoltage response to light drops rapidly toward the dark restingpotential when stimulated with moderate-to-high light intensi-ties. This phenomenon has been explained with rapid depletion ofphosphatidylinositol 4,5-bisphosphate (PIP2), a crucial membranecomponent in phototransduction that is the substrate of PLC,which is activated, via G-protein, by light (Hardie and Postma2008). Normally, rising intracellular Ca2 inhibits PLC (Har-die et al. 2001). In both mutants, the rapid decline of the lightresponses is due to the lack of TRP channels and the fact thatlight-activated current is flowing through TRPL channels only,leading to reduced influx of Ca2 and therefore, unattenuatedPIP2 depletion (Hardie et al. 2001). This line of reasoningsuggests that in the cockroach, the light-gated channels couldbe more like trpl channels in D. melanogaster, but this remainsto be investigated.

Conclusions. We have shown that in cockroach photorecep-tors, 1) the temporal properties of both phototransduction andnontransductive membrane are slow, 2) the adaptive changesare mostly limited to intensities 1,000 photons/s, 3) the highvariability of the voltage responses originates in the variabilityof the LIC, and 4) voltage responses are nonlinearly tuned tonaturalistic stimulation and perform poorly with step or WNstimulation.

GRANTS

Support for the study was provided by grants from the Academy of Finlandand Sigrid Juselius Foundation (to M. Weckström and M. Vähäsöyrinki) andthe Biotechnology and Biological Sciences Research Council (BBF0120711and BBD0019001 to M. Juusola).

DISCLOSURES

The authors declare no potential conflicts of interest, financial or otherwise.

AUTHOR CONTRIBUTIONS

Author contributions: K.H. and M.W. conception and design of research;K.H., E-V.I., R.F., and I.S. performed experiments; K.H., E-V.I., R.F., I.S.,M.J., and M.W. analyzed data; K.H., E-V.I., R.F., I.S., M.J., M.V., and M.W.interpreted results of experiments; K.H. prepared figures; K.H. and M.W.drafted manuscript; K.H., E-V.I., R.F., I.S., M.J., M.V., and M.W. edited andrevised manuscript; K.H., E-V.I., R.F., I.S., M.J., M.V., and M.W. approvedfinal version of manuscript.

REFERENCES

Aho AC, Donner K, Hyden C, Larsen LO, Reuter T. Low retinal noise inanimals with low body temperature allows high visual sensitivity. Nature334: 348–350, 1988.

Bähner M, Frechter S, Da Silva N, Minke B, Paulsen R, Huber A.Light-regulated subcellular translocation of Drosophila TRPL channelsinduces long-term adaptation and modifies the light-induced current. Neuron34: 83–93, 2002.

Barash S, Suss E, Stavenga DG, Rubinstein CT, Minke B. Light reduces theexcitation-efficiency in the nss mutant of the sheep blowfly Lucilia. J GenPhysiol 92: 307–330, 1988.

Baylor DA, Lamb TD, Yau KW. Responses of retinal rods to single photons.J Physiol 288: 613–634, 1979.

Bendat JS, Piersol AG. Random Data: Analysis and Measurement Proce-dures. New York: John Wiley & Sons, 1971.

Butler R. The anatomy of the compound eye of Periplaneta americana L. 1.General features. J Comp Physiol 83: 223–238, 1973a.

Butler R. The anatomy of the compound eye of Periplaneta americana L. 2.Fine structure. J Comp Physiol 83: 239–262, 1973b.

Butler R. The identification and mapping of spectral cell types in the retina ofPeriplaneta americana. Z Vergl Physiol 72: 67–80, 1971.

Cosens DJ, Manning A. Abnormal electroretinogram from a Drosophilamutant. Nature 224: 285–298, 1969.

Dodge FA, Knight BW, Toyoda J. Voltage noise in Limulus visual cells.Science 160: 88–90, 1968.

Dubs A, Laughlin SB, Srinivasan MV. Single photon signals in fly photo-receptors and first order interneurones at behavioral threshold. J Physiol317: 317–334, 1981.

Ernst A, Füller H. Zur Feinstruktur der Lamina Ganglionaris von Periplanetaamericana (L.). Zool Jahrb Anat 115: 393–416, 1987.

Faivre O, Juusola M. Visual coding in locust photoreceptors. PLoS One 3:e2173, 2008.

Ferrell BR, Reitcheck BG. Circadian changes in cockroach ommatidialstructure. J Comp Physiol A 173: 549–555, 1993.

Frederiksen R, Wcislo WT, Warrant EJ. Visual reliability and informationrate in the retina of a nocturnal bee. Curr Biol 18: 349–353, 2008.

French AS. Coherence improvement in white noise analysis by the use of arepeated random sequence generator. IEEE Trans Biomed Eng 27: 51–53,1980a.

French AS. The linear dynamic properties of phototransduction in the flycompound eye. J Physiol 308: 385–401, 1980b.

Fuortes MG, Yeandle S. Probability of occurrence of discrete potential wavesin the eye of Limulus. J Gen Physiol 47: 443–463, 1964.

Guthrie DM, Tindall AR. The Biology of the Cockroach. London: EdwardArnold, 1968.

Halloy J, Sempo G, Caprari G, Rivault C, Asadpour M, Tache F, Said I,Durier V, Canonge S, Ame JM, Detrain C, Correll N, Martinoli A,Mondada F, Siegwart R, Deneubourg JL. Social integration of robots intogroups of cockroaches to control self-organized choices. Science 318:1155–1158, 2007.

Hardie R, Postma M. Phototransduction in microvillar photoreceptors ofDrosophila and other invertebrates. In: The Senses—A ComprehensiveReference, Vision. Oxford, UK: Academic, 2008, vol. 1, p. 77–130.

Hardie RC. Electrophysiological analysis of fly retina. I: Comparative prop-erties of R1–6 and R7 and 8. J Comp Physiol A 129: 19–33, 1979.

Hardie RC. Phototransduction in Drosophila melanogaster. J Exp Biol 204:3403–3409, 2001.

Hardie RC. Voltage-sensitive potassium channels in Drosophila photorecep-tors. J Neurosci 11: 3079–3095, 1991.

Hardie RC, Raghu P. Visual transduction in Drosophila. Nature 413: 186–193, 2001.

Hardie RC, Raghu P, Moore S, Juusola M, Baines RA, Sweeney ST.Calcium influx via TRP channels is required to maintain PIP2 levels inDrosophila photoreceptors. Neuron 30: 149–159, 2001.



on May 20, 2015

Dow

nloaded from

Heimonen K, Salmela I, Kontiokari P, Weckström M. Large functionalvariability in cockroach photoreceptors: optimization to low light levels. JNeurosci 26: 13454–13462, 2006.

Howard J. Calcium enables photoreceptor pigment migration in a mutant fly.J Exp Biol 113: 471–475, 1984.

Howard J, Blakeslee B, Laughlin SB. The intracellular pupil mechanism andphotoreceptor signal: noise ratios in the fly Lucilia cuprina. Proc R Soc LondB Biol Sci 231: 415–435, 1987.

Howard J, Dubs A, Payne R. The dynamics of phototransduction in insects:a comparative study. J Comp Physiol A 154: 707–718, 1984.

Huang J, Liu CH, Hughes SA, Postma M, Schwiening CJ, Hardie RC.Activation of TRP channels by protons and phosphoinositide depletion inDrosophila photoreceptors. Curr Biol 20: 189–197, 2010.

Juusola M. Linear and nonlinear contrast coding in light-adapted blowflyphotoreceptors. J Comp Physiol A 172: 511–521, 1993.

Juusola M, de Polavieja GG. The rate of information transfer of naturalisticstimulation by graded potentials. J Gen Physiol 122: 191–206, 2003.

Juusola M, Hardie RC. Light adaptation in Drosophila photoreceptors: I.Response dynamics and signaling efficiency at 25 degrees C. J Gen Physiol117: 3–25, 2001.

Juusola M, Kouvalainen E, Järvilehto M, Weckström M. Contrast gain,signal-to-noise ratio, and linearity in light-adapted blowfly photoreceptors. JGen Physiol 104: 593–621, 1994.

Juusola M, Weckström M. Band-pass filtering by voltage-dependent mem-brane in an insect photoreceptor. Neurosci Lett 154: 84–88, 1993.

Katz B, Minke B. Phospholipase C-mediated suppression of dark noiseenables single-photon detection in Drosophila photoreceptors. J Neurosci32: 2722–2733, 2012.

Kelly K, Mote MI. Avoidance of monochromatic light by the cockroachPeriplaneta americana. J Insect Physiol 36: 287–291, 1990.

Kouvalainen E, Weckström M, Juusola M. A method for determiningphotoreceptor signal-to-noise ratio in the time and frequency domains witha pseudorandom stimulus. Vis Neurosci 11: 1221–1225, 1994.

Krause Y, Krause S, Huang J, Liu CH, Hardie RC, Weckström M.Light-dependent modulation of Shab channels via phosphoinositide deple-tion in Drosophila photoreceptors. Neuron 59: 596–607, 2008.

Laughlin SB. Invertebrate vision at low luminances. In: Night Vision, editedby Hess RF, Sharpe LT, Nordby K. Cambridge, UK: Cambridge UniversityPress, 1990, p. 223–250.

Laughlin SB. Matched filtering by a photoreceptor membrane. Vision Res 36:1529–1541, 1996.

Laughlin SB. Neural principles in the peripheral visual systems of inverte-brates. In: Handbook of Sensory Physiology, edited by Autrum H. Berlin,Heidelberg, New York: Springer-Verlag, 1981, vol. VII/6B, p. 133–280.

Laughlin SB. The role of sensory adaptation in the retina. J Exp Biol 146:39–62, 1989.

Laughlin SB, Lillywhite PG. Intrinsic noise in locust photoreceptors. JPhysiol 332: 25–45, 1982.

Laughlin SB, Weckström M. Fast and slow photoreceptors—a comparativestudy of the functional diversity of coding and conductances in the Diptera.J Comp Physiol A 172: 593–609, 1993.

Lillywhite PG. Single photon signals and transduction in an insect eye. JComp Physiol A 122: 189–200, 1977.

Lillywhite PG, Laughlin SB. Transducer noise in a photoreceptor. Nature277: 569–572, 1979.

Lipton GR, Sutherland DJ. Activity rhythms in the American cockroach,Periplaneta americana. J Insect Physiol 16: 1555–1566, 1970.

Meyer NE, Joel-Almagor T, Frechter S, Minke B, Huber A. Subcellulartranslocation of the eGFP-tagged TRPL channel in Drosophila photorecep-tors requires activation of the phototransduction cascade. J Cell Sci 119:2592–2603, 2006.

Minke B, Parnas M. Insights on TRP channels from in vivo studies inDrosophila. Annu Rev Physiol 68: 649–684, 2006.

Minke B, Wu CF, Pak WL. Induction of photoreceptor voltage noise in thedark in Drosophila mutant. Nature 258: 84–87, 1975.

Niven JE, Laughlin SB. Energy limitation as a selective pressure on theevolution of sensory systems. J Exp Biol 211: 1792–1804, 2008.

Niven JE, Vähäsöyrinki M, Juusola M, French AS. Interactions betweenlight-induced currents, voltage-gated currents, and input signal properties inDrosophila photoreceptors. J Neurophysiol 91: 2696–2706, 2004.

Niven JE, Vähäsöyrinki M, Kauranen M, Hardie RC, Juusola M, Weck-ström M. The contribution of Shaker K channels to the informationcapacity of Drosophila photoreceptors. Nature 421: 630–634, 2003.

Ribi WA. Fine structure of the first optic ganglion (lamina) of the cockroach,Periplaneta americana. Tissue Cell 9: 57–72, 1977.

Riemay H. Verhaltensphysiologische Untersuchungen zur Einwirkung desLichtes auf den Fluchtlauf der Schabe Periplaneta americana (L.). Zool JbPhysiol 88: 491–497, 1984.

Roberts SK. Photoreception and entrainment of cockroach activity rhythms.Science 148: 958–959, 1965.

Shannon CE. A mathematical theory of communication. Bell Syst Tech J 27:379–423, 623–656, 1948.

Snyder AW, Horridge GA. The optical function of changes in the mediumsurrounding the cockroach rhabdom. J Comp Physiol A 81: 1–8, 1972.

Song Z, Postma M, Billings SA, Coca D, Hardie RC, Juusola M. Stochastic,adaptive sampling of information by microvilli in fly photoreceptors. CurrBiol 22: 1371–1380.

Spekreijse H, Oostings H. Linearizing: a method for analysing and synthe-sizing non-linear systems. Kybernetik 7: 2–30, 1970.

Suss-Toby E, Selinger Z, Minke B. Lanthanum reduces the excitationefficiency in fly photoreceptors. J Gen Physiol 98: 848–868, 1991.

Takalo J, Ignatova I, Weckström M, Vähäsöyrinki M. A novel estimatorfor the rate of information transfer by continuous signals. PLoS One 6:e18792, 2011.

Vähäsöyrinki M, Niven JE, Hardie RC, Weckström M, Juusola M.Robustness of neural coding in Drosophila photoreceptors in the absence ofslow delayed rectifier K channels. J Neurosci 26: 2652–2660, 2006.

van Hateren JH. A theory of maximizing sensory information. Biol Cybern68: 23–29, 1992.

van Hateren JH. Processing of natural time series of intensities by the visualsystem of the blowfly. Vision Res 37: 3407–3416, 1997.

Warrant EJ. Invertebrate vision in dim light. In: Invertebrate Vision, editedby Warrant EJ, Nilsson D-E. Cambridge, UK: Gambridge University Press,2006, p. 83–126.

Warrant EJ. Seeing better at night: life style, eye design and the optimumstrategy of spatial and temporal summation. Vision Res 39: 1611–1630,1999.

Warrant EJ. Seeing in the dark: vision and visual behaviour in nocturnal beesand wasps. J Exp Biol 211: 1737–1746, 2008.

Warrant EJ. Vision in the dimmest habitats on earth. J Comp Physiol A 190:765–789, 2004.

Warrant EJ, McIntyre PD. Arthropod eye design and the physical limits tospatial resolving power. Prog Neurobiol 40: 413–461, 1993.

Weckström M, Hardie RC, Laughlin SB. Voltage-activated potassiumchannels in blowfly photoreceptors and their role in light adaptation. JPhysiol 440: 635–657, 1991.

Weckström M, Järvilehto M, Heimonen K. Spike-like potentials in theaxons of nonspiking photoreceptors. J Neurophysiol 69: 293–296, 1993.

Weckström M, Kouvalainen E, Juusola M. Measurement of cell impedancein frequency domain using discontinuous current clamp and white-noise-modulated current injection. Pflügers Arch 421: 469–472, 1992.

Weckström M, Laughlin SB. Visual ecology and voltage-gated ion channelsin insect photoreceptors. Trends Neurosci 18: 17–21, 1995.

Wong F, Knight BW. Adapting-bump model for eccentric cells of Limulus. JGen Physiol 76: 539–557, 1980.

Wong F, Knight BW, Dodge FA. Adapting bump model for ventral photo-receptors of Limulus. J Gen Physiol 79: 1089–1113, 1982.

Wong F, Knight BW, Dodge FA. Dispersion of latencies in photoreceptors ofLimulus and the adapting-bump model. J Gen Physiol 76: 517–537, 1980.

Wu CF, Pak WL. Quantal basis of photoreceptor spectral sensitivity ofDrosophila melanogaster. J Gen Physiol 66: 149–168, 1975.

Ye S, Leung V, Khan A, Baba Y, Comer CM. The antennal system andcockroach evasive behavior. I. Roles for visual and mechanosensory cues inthe response. J Comp Physiol A 189: 89–96, 2003.



on May 20, 2015

Dow

nloaded from

signal coding in cockroach photoreceptors is tuned to dim

Documents