The Extraretinal Eyelet of Drosophila: Development, Ultrastructure, and Putative Circadian Function

Bolwig's organ transforms into eyelet during metamorphosis

In larvae, the Bolwig's nerve projects from the larval photoreceptors toward the larval optic neuropil (for review, seeMeinertzhagen and Hanson, 1993) over the face of the forming medulla (Fig. 1A). A difficulty in tracing a possible developmental relationship between it and eyelet is that expression of photoreceptor antigens turns off in Bolwig's organ during early pupal life, making it difficult to examine the possibility that Bolwig's organ might persist thereafter. Indeed, Bolwig's nerve was stained only faintly at 20% of pupal development (P + 20%) (Fig. 1B) and not at all in later stages (Fig. 1C–F). This was true not only for the expression of the GFP under the control of the strong photoreceptor-specific glass multiple reporter construct (gmr-GAL4;UAS-gfp line) (Fig.1C) but also for photoreceptor-specific antibody staining with 24B10 (Zipursky et al., 1984) (Fig.2B). We therefore examined the reporter line BS23 (Kr-BO-lacZ; Schmucker et al., 1992) driven by a promoter of the developmental gap geneKrüppel. This drives lacZ expression that is specific to Bolwig's organ (Schmucker et al., 1992), and the resultant immunoreactivity to β-gal indeed labeled Bolwig's organ and nerve from P + 0% until eclosion (Fig.3). In the white prepupa intense immunofluorescent labeling highlighted bilateral clusters of 12 cells lying near the larval mouthhooks and the axon bundles emerging from these (Fig. 3A). Profiles 5–6 μm in diameter detected in Bolwig's organ were probably photoreceptor nuclei (Fig. 3A,inset). In prepupal control preparations the labeling was seen neither in Bolwig's organ nor in Bolwig's nerve (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Bolwig's nerve and the eyelet tract both have a close spatial relationship to the lateral neurons in the developing optic lobe. Shown are confocal images of the right optic lobe in frontal view. Green channel, Photoreceptors (gmr-driven expression of GFP inA–F; antibody 24B10 in G–I);red channel, lateral neurons (anti-PDH). Some ommatidia of the developing retina are visible in D andE; other images depict the terminals of photoreceptors R1–R6 in the lamina (La) and R7, R8 in the second neuropil, the medulla (Me). In larvae (A) and P + 20% pupae (B), Bolwig's nerve (BN) traverses the medulla and terminates directly in the larval optic neuropil (arrowhead), which transforms into the accessory medulla during metamorphosis. In the larval optic neuropil the terminals overlap putative dendrites from the small and large lateral neurons (s-LN_v and l-LN_v). Additional fibers parallel to Bolwig's nerve (small green arrows) are possibly neurites of tangential neurons. Expression in the latter persists in early pupae (B) when expression in Bolwig's nerve starts to disappear (C). At approximately P + 50% (D) faint labeling first appears in the eyelet tract (Ey-Tr) together with the first labeling in l-LN_v. At P + 80% (E) the labeling pattern mainly resembles that of adult flies (F). The tract from eyelet now shows a ventral extension overlapping arborizations of the l-LN_v(arrowheads). Fibers arising from l-LN_v form a network over the surface of the medulla in close vicinity to the tract from eyelet (arrow in F). Bolwig's tract (G) and eyelet (H) are present in eyeless sine oculis (so¹) mutants contacting the LN_v. so¹ flies with tiny eye remnants possess, in addition, a well established axonal connection of their ommatidial photoreceptors to the medulla (I). Scale bar: (in I)A–I, 20 μm.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Bolwig's organ and the developing projection of eyelet. A, Frontal 8 μm wax section of anterior region of third instar larva showing immunoreactivity to Rhodopsin 6 in Bolwig's organ (BO), situated dorsally to the cephalopharyngeal skeleton (arrowheads). Differential interference contrast microscopy was used.B, C, Horizontal cryostat sections 10 μm thick of pupal optic lobe immunolabeled with antibody 24B10.B, Pupa at approximately P + 45%. In early pupal stages the developing medulla (Me) lies orthogonal to the lamina (La), and the photoreceptor terminals of eyelet are not seen. C, Pupa at approximately P + 70%. At approximately P + 50% the medulla rotates, becoming concentric with the lamina, and the terminals of eyelet immunolabel at a position anterior to the medulla neuropil (arrow) corresponding to the accessory medulla. R, Retina. Cell bodies are not easily seen in this preparation. In the late pupa and adult the axons from the cells of eyelet project through the outer chiasm, and their terminals in the accessory medulla all label with photoreceptor-specific antibodies. Scale bars: A, 10 μm; (in B) B, C, 50 μm.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Krüppel-drivenlacZ expression in Bolwig's organ and nerve persists throughout metamorphosis. Bolwig's organ and nerve throughout pupal metamorphosis are detected by immunoreactivity to β-gal in the transgenic line BS23 (Kr-BO-lacZ) in whichlacZ expression is driven by the promoter forKrüppel. A, Dorsal view of both eye discs and hemispheres in a whole-mount preparation revealing Bolwig's organ (arrowhead) and the trajectory of its nerve (BN) in the white prepupa. B–D, Horizontal Vibratome slices (100 μm) of the right hemisphere of the pupal optic lobe at different stages: B, P + 20%;C, P + 40%; D, P + 80%. Persistent β-gal expression reveals that Bolwig's organ is transformed into eyelet, the axon bundle of which is shown relative to the lamina (La) and medulla (Me) neuropils (inD), retina (R in B,D), and basement membrane (arrows inD). Cell number in Bolwig's organ decreases from 12 (inset, A) to four in the adult (inset, D) after separation from the mouthhooks at P + 1 hr (inset, B) and retraction of the cell bodies that are distributed in a trail of two clusters by P + 5% (inset, C). Scale bars: A, 250 μm; B, C,inset in C, 50 μm; D,insets in A and B, 25 μm; inset in D, 20 μm.

During pupal development the cluster of cells in Bolwig's organ lost its rosette shape. Starting as soon as P + 1 hr, labeled somata occupied progressively more caudal positions, as if the photoreceptors either were migrating caudally or were being retracted along the original pathway of their axons (Fig.3B,C, inset). By P + 5% the somata were split into two clusters retracting along this former path (Fig. 3C, inset) and were no longer adjacent to the mouthhooks. The number of cells in each cluster was not consistent (data not shown). At P + 10% the development of the retina in the eye disc and the formation of the optic neuropils were both apparent (data not shown), contributing additional axons to the pathway taken by Bolwig's nerve through the optic stalk between eye disc and brain (Steller et al., 1987). Bolwig's organ still contained ∼12 cells but now comprised two quite distinct clusters like those seen at P + 5%. By P + 20% (Fig. 3B) immunofluorescence labeled only three to six cells; however, the position of Bolwig's organ and nerve relative to the developing compound eye now anticipated that of eyelet. Then Bolwig's organ was detected below the expanding retina in an area that subsequently would become the posterior margin of the compound eye. At this stage the terminals of Bolwig's nerve were metamorphic, sometimes extending beyond the anterior medulla and occasionally (n = 2 of >50) so that their neurites met symmetrically at the midline (data not shown).

From P + 20% on Bolwig's organ and nerve maintained their positions while the compound eye and optic neuropils formed around them (Fig.3C,D). There was some defasciculation of Bolwig's nerve as it crossed the chiasma between lamina and medulla. At P + 80% Bolwig's organ and nerve corresponded to eyelet and its tract when viewed in a horizontal plane (Fig. 3D). In newly eclosed BS23 flies Bolwig's organ, now transformed into eyelet, was still immunoreactive to β-gal, but the axons of Bolwig's nerve were no longer as strongly labeled and in some cases were not detected (data not shown).

Bolwig's organ and eyelet both express Rh6 immunoreactivity

Bolwig's organ is reported to express immunoreactivity to three of six Drosophila rhodopsins that have been tested so far, Rh1, Rh3, and Rh4 (Pollock and Benzer, 1988), whereas eyelet expresses Rh6 (Yasuyama and Meinertzhagen, 1999). To examine whether this difference indicates a change in rhodopsin expression, we therefore examined immunoreactivity to Rh6 in Bolwig's organ. Clear expression of immunoreactivity to Rh6 was seen (Fig. 2A) in preparations made identically to those revealing Rh6 expression in eyelet (data not shown). Thus eyelet retains the expression found in Bolwig's organ.

Bolwig's nerve and projection of eyelet both enjoy a close association with lateral neurons

In the larval brain the circadian pacemaker center appears to consist of five small ventral lateral neurons (s-LN_v; Kaneko et al., 1997; Blanchardon et al., 2001). Four of these contain the neuropeptide PDF, a homolog of the crustacean neuropeptide PDH (Park and Hall, 1998). Arborizations of the PDF-expressing s-LN_v visualized with an antibody against crab PDH (Helfrich-Förster, 1997) innervate the larval brain hemispheres and have fine dendrite-like arborizations in the larval optic neuropil. There they overlap the terminals from Bolwig's nerve (Fig. 1A) (Kaneko et al., 1997).

In the pupa the medulla rotated from an anterior position (Fig.2B) to a position between the lobula complex, lying beneath it, and the lamina (Fig. 2C) (Hofbauer and Campos-Ortega, 1990). It reached its final position by approximately P + 50% at a time when faint GFP expression reappeared in a tract describing the same path toward the s-LN_v as the former Bolwig's nerve (Fig. 1D). This tract was traced during subsequent pupal development and corresponded to the projection of eyelet. At the same time, somata of the large LN_v (l-LN_v) and their neurites first became immunoreactive to anti-PDH (Fig.1D). As revealed in Figure 1D, these extended neurites into the accessory medulla, which originates from the larval optic neuropil (Ehnbohm, 1948; Hagberg, 1986;Helfrich-Förster, 1997). Whereas the larval optic neuropil was oval in shape, however, the accessory medulla formed a ventral elongation containing fibers arising from the l-LN_v (Fig.1E,F) (cf.Helfrich-Förster, 1997). The eyelet tract followed this ventral elongation to form a ventral hook (Fig. 1F). At P + 80% the immunolabeling of LN_v cells and the eyelet tract were both very intense, closely resembling the adult pattern. In all brains that were investigated (n = 12) the terminals of the eyelet tract completely overlapped the LN_v fibers in the accessory medulla. Furthermore, in ∼50% of the brains the eyelet tract lay in close vicinity to fibers from the l-LN_v cells arborizing on the surface of the medulla (Fig. 1F). The spatial relationship between the eyelet tract and LN_vfibers strongly suggests their functional connection and suggests that eyelet serves to transfer photic information toward the circadian pacemaker cells. A similar functional connection also may exist between photoreceptor cells R7 and R8 of the compound eyes, the axonal terminals of which are in close vicinity to the varicose network of the l-LN_v cells on the surface of the medulla (Fig.1F).

Ultrastructural differentiation of the eyelet cells commences in midpupal stages

In adults eyelet contains between three and six neurons (Robinow and White, 1991), but usually four, all of which have an ultrastructure typical for photoreceptor cells (Yasuyama and Meinertzhagen, 1999). The four cells first were identified at approximately P + 60% when rhabdomeric structures were first visible (Fig.4A). At earlier stages we were not able to recognize the cells morphologically. At this stage of differentiation the beginning of rhabdomere formation is indicated by microvillar structures protruding into an intercellular space formed by the eyelet cells. This shows that the elaborate rhabdomeres as well as the pigmentation typical of the cells of the adult eyelet (Fig.4B) are formed in the second half of pupal development, comparable in time course to that for the retinal photoreceptor cells of the compound eye. The time course of this differentiation indicates that the larval photoreceptor cells dedifferentiate in early metamorphosis and then redifferentiate to attain their imaginal structure.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Ultrastructural differentiation of eyelet. Shown are the photoreceptors of eyelet at different stages of differentiation in pupal (A) and adult (B) flies. A, Pupa with amber eyes corresponding to approximately P + 60% (Bainbridges and Bownes, 1981). The first microvilli of developing rhabdomeres have become distinct, but no pigment is yet visible. B, Fully differentiated eyelet photoreceptors with rhabdomeres and black pigment granules.Arrows show developing microvillar structures inA and fully developed rhabdomeres in B. Scale bar, 1 μm.

Eyelet is present in eyeless sine oculis¹ flies but is absent in the double-mutant sine oculis¹;glass^60j

Compatible with the formation of eyelet from Bolwig's organ, eyeless flies (so¹ ) have a normal Bolwig's organ (Cheyette et al., 1994) and have been shown previously to retain eyelet (Hofbauer and Buchner, 1989). As in wild-type flies, Bolwig's nerve and the eyelet tract both are associated closely with the LN_v cells (Fig.1G,H). By comparison, mutantglass (gl^60j ) flies lack Bolwig's organ (Kunes et al., 1993) and eyelet (Helfrich-Förster et al., 2001). To reveal whether eyelet plays an active role in circadian entrainment, we therefore could compare the light sensitivity of the circadian system between flies that retained and flies that lacked eyelet but that were otherwise similar. To do this, we generated double-mutantso¹;gl^60j flies that were identical in the shielding properties of their head with so¹ flies (see Materials and Methods). Whole-mount preparations and cryostat sections of mutantso¹ andso¹;gl^60j flies then were checked for the presence of eyelet by using the photoreceptor-specific antibody 24B10 that reliably marks this structure and, in the case of so¹ , additionally checked with anti-TAU in flies expressing TAU under control of the Rh6 promotor. Eyelet and its tract were clear in mostso¹ flies (31 of 34 brain hemispheres labeled with 24B10 and 24 of 27 hemispheres with anti-TAU). Inso¹;gl^60j mutants eyelet was absent. In only one brain hemisphere of 60 was a 24B10-immunoreactive structure found that resembled eyelet in size, but this was neither at its normal location nor did its axons project to the LN_v; we therefore concluded that this structure did not represent eyelet.

Action spectra in mutant flies

To study the function of eyelet in detail, we determined action spectra for entraining the activity rhythm of adultso¹ andso¹;gl^60j flies. An action spectrum plots the magnitude of a photobiological response as a function of the wavelength of light eliciting the response. Ideally, the action spectrum matches the absorption spectrum of the photoactive pigment. To test the hypothesis that eyelet provides a photic input to entrain the fly's activity rhythm, we quantified entrainment from the percentage of entrained flies at different irradiances and wavelengths (see Materials and Methods). To ensure that flies entrained to the LD cycles and simply did not free-run with a period of 24 hr, we subjected them to a 6 hr phase advance of the LD cycle (Fig. 5); only flies that shifted their activity rhythm within 7 d were judged to be entrained. First we determined dose–response curves between irradiance and the percentages of resynchronized flies for each wavelength (Fig.6A). The slopes of the semilog curves for different wavelengths were essentially equal. Action spectra were derived from the intersection of the curves with the 60% synchronization value (Fig. 6B). Forso¹ flies the action spectrum showed two peaks, at 420 and ∼480 nm. The action spectrum forso¹;gl^60j flies was similar to the so¹ spectrum at wavelengths shorter than 450 nm, but the peak at 480 nm was virtually absent. Wavelengths longer than 550 nm failed to synchronize both mutants. This differed from wild-type flies (Fig. 6B; derived from data in Blaschke et al., 1996) that, like the mutants, showed two peaks in their action spectrum but that were ∼100 times more sensitive and showed considerable sensitivity in the red part of the spectrum. In summary, these results suggest that the compound eyes contribute to entrainment over a broad spectral range extending into the red, whereas cryptochrome is responsible for the sensitivity peak at ∼420 nm and eyelet is responsible for that at 480 nm. If true, then mutants that retain only eyelet but lack compound eyes and cryptochrome should maintain the sole peak at 480 nm. To test this hypothesis, we generatedso¹;cry^b double mutants and tested their entrainability at different wavelengths. To judge the contribution of the compound eyes, we also testedso¹;cry^b flies that possessed tiny remnants of the compound eyes containing between 10 and 50 ommatidia. Histologically, such flies showed well established axonal connections of their few ommatidial photoreceptors to a small underlying lamina and to the medulla, which was enlarged at the site of photoreceptor innervation (Fig. 1I).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Entrainment of flies by spectral light. Shown are representative actograms from wild-type flies (WT), so¹ flies, andso¹;gl^60j flies subjected to 12 hr light/dark cycles (LD) of different wavelengths, the phase of which was advanced twice by 6 hr. One-half of the flies subsequently were recorded under constant darkness (DD). Light was blue-green (top left panel, 460–540 nm), red (top right panel, >600 nm), blue (middle and bottom left panels, 420 nm), or green (middle and bottom right panels, 486 nm). Irradiances are indicated in themargin in photons · cm⁻² · sec⁻¹. Wild-type flies re-entrained quickly to the 6 hr phase advances, showing bimodal activity patterns with morning and evening peaks. Activity was restricted mainly to the light phase of the LD cycle. Mutants extended activity into the dark phase and needed several days to re-entrain to phase advances. Some flies failed to re-entrain to green light (so¹;gl^60j and so¹ after second phase advance). The morning peak was visible in so¹ flies only during the blue LD cycle after the first phase shift and faintly during the green LD cycle before the first phase shift. Theso¹ fly in green light showed antidromic phase shifting.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Dose–response curves for re-entrainment and corresponding action spectra. The dose–response curves (A) for so¹,so¹;gl^60j , andso¹;cry^b flies show the percentage of re-entrained flies for all wavelengths as a function of the irradiance. Forso¹;cry^b mutants both the flies with eye remnants (●) and completely eyeless flies (○) were tested. From left toright, the wavelengths for eyeless flies were 472, 486, 517, and 456 nm. Action spectra (B) were derived from these dose–response curves by determining the light intensities that were necessary to entrain 60% of the flies at each wavelength. The action spectrum for wild-type flies is derived from data byBlaschke et al. (1996).

As expected, the sensitivity peak at 420 nm was lacking inso¹;cry^b flies, whereas that at ∼480 nm remained (Fig. 6B). This shows unequivocally that the 420 nm peak in the action spectrum derives from cryptochrome and indicates that the 480 nm peak stems from eyelet. However, only very few eyelessso¹;cry^b flies at all were able to follow the 6 hr phase advance of the LD cycle. At 456 nm one of 20 flies (5%), at 471 nm three of 31 flies (9.7%), at 486 nm three of 24 flies (12.5%), and at 517 nm one of 16 (6.3%) flies re-entrained. At the other wavelengths we did not find any re-entrainment in eyelessso¹;cry^b flies, not even at the highest irradiance possible with our light source (10¹⁷photons · cm⁻² · sec⁻¹). Although the few cases of entrained flies occurred at the putative sensitivity maximum of eyelet, eyelet alone thus appears not to be capable of reliably entraining the activity rhythm. Rather, the additional presence of either cryptochrome or the compound eyes may be required. As soon as only a few ommatidia were present inso¹;cry^b flies, the 480 nm sensitivity peak became more prominent and the sensitivity of the circadian system extended from the blue into the red, but the sensitivity peak at 420 nm was absent (Fig.6B). This shows clearly that the compound eyes contribute to the flies' sensitivity throughout the spectrum and are especially responsible for that at longer wavelengths. Probably because of the small number of ommatidia present, however, the overall sensitivity ofso¹;cry^b flies with eye remnants was considerably less than that of wild-type flies (Fig. 6B).

The pattern of activity in eyeless mutants

Wild-type flies showed a bimodal activity pattern, with a morning peak at approximately lights-on, an evening peak at approximately lights-off, and little activity during the dark phase even at the higher and medium irradiance levels of the wavelengths that were tested (e.g., 10¹¹–10¹³photons · cm⁻² · sec⁻¹of blue-green light) (Fig. 7). However, under spectral light of low intensities (e.g., at 10¹⁰photons · cm⁻² · sec⁻¹blue-green light) at the limits of entrainment, wild-type flies frequently lacked the morning peak (I. Blaschke, unpublished observations). Eyeless flies (so¹ andso¹;gl^60j mutants) already lacked the morning peak at rather high irradiances (Fig. 5). At the highest irradiances that were applied here (e.g., 10¹⁴photons · cm⁻² · sec⁻¹of blue-green light), only 50% of the so¹ mutants and 55% of theso¹;gl^60j mutants still had a normal bimodal activity pattern with morning and evening peaks, whereas at intermediate irradiances (e.g., 2 × 10¹³photons · cm⁻² · sec⁻¹blue-green light) only 25% of the so¹ and 24% of theso¹;gl^60j mutants were bimodal; at the lowest irradiances that still led to entrainment (3 × 10¹²photons · cm⁻² · sec⁻¹blue green light) all flies were unimodal, lacking the morning peak completely. This behavior was independent of the wavelength in both mutants (χ²_(9,182/119)< 8.95; p > 0.442), and no significant differences in the presence of the morning peak were found between the two mutants (χ²_(1,301) = 0.54;p = 0.461).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Phase plot of evening activity peaks and average activity profiles of entrained flies at different wavelengths. Theblack and whitebars along the top of each plot indicate the light/dark regimen. The maxima of the evening peaks occurred before lights-off in wild-type flies (▴) and inso¹;cry^b mutants with eye remnants (○) but after lights-off in bothso¹;gl^60j (●) and so¹ (■) mutants. Average activity profiles at wavelengths of 420, 486, and 600 nm reveal that the activity of wild-type flies andso¹;cry^b mutants with eye remnants is restricted mainly to the light phase of the LD cycle, whereas it extends into the dark phase in the eyeless mutants so¹ andso¹;gl^60j (black portion of the curves). Inso¹ mutants the phase of the evening peak was dependent on the wavelength and occurred later at longer wavelengths (see phase plot and arrows in the activity profiles of so¹ flies). Except for a single fly, the eyeless so¹ andso¹;gl^60j were not entrainable at 600 nm, so average activity profiles could not be calculated for these flies at that wavelength. ZT, Zeitgeber time (ZT 0 = lights-on; ZT 12 = lights-off).

Whereas the maximum for the evening peak was ∼2 hr before lights-off in wild-type flies and inso¹;cry^b flies with eye remnants, it occurred after lights-off in the eyeless flies (Fig. 7). Such an activity phase lag is expected for animals having a free-running period of >24 hr (Moore-Ede et al., 1982). Indeed, both groups of eyeless flies had periods of at least 24.6 hr (24.61 ± 0.07 hr forso¹;gl^60j and 24.72 ± 0.07 hr for so¹ ) under constant darkness (DD) conditions, whereas the average period for wild-type andso¹;cry^b flies was close to 24.0 hr. Interestingly, the phase relationship of the evening peak to the LD cycle was dependent on the wavelength inso¹ mutants (ANOVA,F_(9,142) = 4.74; p < 0.001): the longer the wavelength, the larger the phase lag (Fig. 7). We did not see such a dependency inso¹;gl^60j mutants nor inso¹;cry^b flies with compound eye remnants nor in wild-type flies (ANOVA,F_(9,63) = 0.72; p = 0.675).

Flies with or without eyelet differ in the pattern of their resynchronization

Of the two ways to resynchronize to a 6 hr phase advance of the LD cycle, the most likely is simply for the clock to follow the shift in the LD cycle and advance its phase by 6 hr (orthodromic phase shifting). However, it is also possible to delay the phase by 18 hr (antidromic phase shifting). Although mostso¹;gl^60j flies phase advanced their clock, as expected, manyso¹ flies reacted unexpectedly and phase-delayed their clock (Fig. 5, Table1). Interestingly, the antidromic phase shifting of so¹ depended on the wavelength, occurring predominantly at longer wavelengths close to the putative sensitivity maximum of eyelet as inferred from our results (480 nm). In contrast, the few cases of antidromic phase shifting inso¹;gl^60j flies occurred at ∼420 nm (Table 1). Wild-type flies tested in a parallel experiment for their capability to entrain in the red (n = 19) or blue-green (n = 14) all advanced their clock after a 6 hr phase advance (Fig. 5). Similarly, all so¹;cry^b flies with eye remnants phase advanced their clock. The aberrant behavior of so¹ flies corresponds to the phase lag of the evening peak observed inso¹ mutants at longer wavelengths and could result from either a very long period or an altered phase–response curve in so¹ flies. Because so¹ andso¹;gl^60j flies had the same long period, however, they should react similarly, yet this was obviously not the case. We therefore propose that eyelet mainly mediates phase delays but that this property becomes evident only in the absence of the compound eyes.

The origin of eyelet from Bolwig's organ

To examine these questions, we first needed to determine whether eyelet is a remnant of the larval Bolwig's organ or whether it develops anew. Our observations on the Krüppelreporter line clearly support the first hypothesis. Although the photoreceptors of Bolwig's organ differ from those of eyelet (Yasuyama and Meinertzhagen, 1999) in number (Steller et al., 1987), position, and ultrastructure (Melzer and Paulus, 1989; Paulus, 1989), our observations reveal that eyelet photoreceptors stem from the 12 larval photoreceptors, the surviving four members of which retract to occupy the position of eyelet in the adult. This explains why the eyelet tract coincides so precisely with the previous pathway of Bolwig's nerve. During their retraction in the young pupa the photoreceptors of Bolwig's organ lose downstream photoreceptor-specific expression such as immunoreactivity to 24B10, which they regain later in pupal life with the differentiation of the photoreceptors of eyelet. Coexpression of immunoreactivity to two fast neurotransmitter systems, histamine and acetylcholine, already has been reported in eyelet (Yasuyama and Meinertzhagen, 1999). This rather perplexing finding thus could be reconciled as persistent immunoreactivity to ChAT, retained from Bolwig's organ (Yasuyama et al., 1995) overlapping the expression of histamine immunoreactivity newly arisen in eyelet. Expression of Rh6 either is retained from Bolwig's organ or is reestablished in the photoreceptors of eyelet.

The finding that only approximately four of the original 12 cells from Bolwig's organ survive pupation may reflect their embryonic origin from approximately four atonal and rhomboid(rho)-positive founder cells (Daniel et al., 1999; Suzuki and Saigo, 2000). Via rho activity these four also expressSpitz, which then signals neighboring cells to activateSpitz transcription and in turn become Bolwig's organ cells. Spitz expression then maintains the fate of all 12 cells (Daniel et al., 1999). Possibly in the young pupaSpitz is silenced in Bolwig's organ, with the loss of the laterally specified cells and a reversion in the structure of Bolwig's organ to the original four founder cells.

The spatiotemporal association between photoreceptors of eyelet and circadian pacemaker neurons

As described previously (Kaneko et al., 1997), we found that terminals of Bolwig's nerve overlap the putative dendritic field of the s-LN_v. This relationship is not simply anatomical because, together with the blue-light photoreceptor cryptochrome, the larval photoreceptors also mediate entrainment of the molecular oscillations of the pacemaker (Kaneko et al., 2000). The circadian function of Bolwig's organ in the larva apparently is resumed in the adult fly, by eyelet, which likewise mediates entrainment, apparently together with cryptochrome and, now, the compound eyes. The anatomical data presented here are consistent with such a role. The photoreceptors of eyelet differentiate concurrently with the l-LN_v, and the neurites of both cell types maintain a close spatial relationship to each other during pupal morphogenesis. They overlap partially on the surface of the medulla and completely in the accessory medulla. Although the eyelet tract arises from Bolwig's nerve, it differs from the latter in one important detail, namely the ventral elongation of the terminal of the eyelet nerve, which follows the fibers of the l-LN_v in the accessory medulla. In the latter, the axon terminals of eyelet also overlap with putative dendrites of the s-LN_v that are the most important in controlling adult behavioral rhythmicity (for review, see Kaneko, 1998; Edery, 2000) (see also Emery et al., 2000b). These anatomical credentials qualify eyelet to transfer photic information to the fly's circadian clock.

Action spectra imply a role for the compound eyes, for cryptochrome, and for eyelet in circadian entrainment

Wild-type flies have major sensitivity peaks at 420 and 500 nm. The peak at 420 nm obviously is caused by the blue-light photopigment cryptochrome and that at 500 nm by the compound eyes and eyelet combined (see below). In addition to their action spectrum, wild-type flies are simply more sensitive than so¹ or so¹;gl^60j flies by ∼10² times between 360 and 550 nm and by 10³ to 10⁴ times at wavelengths >550 nm. Thus the compound eyes provide a strong input to entrainment of the circadian clock, especially at longer wavelengths. Accordingly,so¹;cry^b flies with tiny eye remnants are entrainable by wavelengths >550 nm, whereas their eyeless siblings are not. The compound eye photopigment conferring sensitivity to long wavelengths could be Rh6, which expresses in a subgroup of the R8 photoreceptors of the ommatidia and has an absorption maximum (λ_max) at 530 nm (Salcedo et al., 1999). This sensitivity also could derive from Rhodopsin 1 (Rh1; λ_max = 486 nm), which is expressed in photoreceptors R1–R6 of the ommatidia. Rh1 has a metarhodopsin that is sensitive in the red (λ_max = 566 nm). At illumination the rhodopsins are photoconverted into metarhodopsins that then photoreisomerize back into rhodopsins (Hamdorf, 1979). The physiologically derived spectral sensitivity of R1–R6 should reflect the mixed absorption of both rhodopsin forms and therefore extend into the red, as in the action spectra for wild-type flies andso¹;cry^b mutants with eye remnants. Anatomically, the compound eyes may provide direct light input to the l-LN_v via photoreceptors R7 and R8, the terminals of which are in close vicinity to the fiber network of the l-LN_v on the surface of the medulla. This is true for wild-type flies as well forso¹ mutants with eye remnants.

In addition to the compound eyes, the action spectra clearly confirm cryptochrome as a circadian photopigment that is still active in flies lacking both compound eyes and eyelet (Emery et al., 1998, 2000a,b;Stanewsky et al., 1998; Helfrich-Förster et al., 2001). The sensitivity peak at 420 nm inso¹;gl^60j flies that retain cryptochrome as the sole circadian photoreceptor coincides nicely with the absorption peak of cryptochrome (Selby and Sancar, 1999). Furthermore, flies lacking cryptochrome (so¹;cry^b mutants) lack the sensitivity peak at 420 nm.

The fact that the ultrastructural reorganization of eyelet is not complete until eclosion already suggests that eyelet assumes a function only in adult flies and may serve as a third photoreceptor of the circadian system. Our determinations of the action spectra of circadian entrainment support this view. As in wild-type flies,so¹ flies have two sensitivity peaks in their action spectrum, at 420 and 480 nm. In contrast, the action spectrum ofso¹;gl^60j flies virtually lacks the 480 nm peak, suggesting that this peak stems from eyelet. Like the earlier Bolwig's organ (see Results), eyelet is immunoreactive to Rh6 (Yasuyama and Meinertzhagen, 1999), which has λ_max at 530 nm and a metarhodopsin λ_max at 468 nm (Salcedo et al., 1999). Furthermore, some eyelet photoreceptor cells may express Rh5 (Malpel et al., 2002) with λ_max at 442 nm and a metarhodopsin λ_max at 500 nm. If all eyelet photoreceptor inputs carry equal weight, the physiologically derived spectral sensitivity should reflect the mixed absorption of all four rhodopsin forms and lie at ∼480 nm, exactly as observed in the action spectrum of so¹ flies. Consistent with this, flies with only eyelet show entrainment solely at ∼480 nm. Conviction on this point is frustrated only because so few flies actually entrain at this wavelength, indicating that, acting alone, eyelet is a rather weak photoreceptor of the circadian system. We cannot exclude, however, that this weak influence is attributable to less extensively developed eyelet terminals in the mutantso¹ compared with the wild type (compare Fig. 1H,I).

Our action spectra clearly speak for an active role of eyelet in circadian entrainment, but we have to consider possible side effects of the glass mutation on central brain neurons, which also might cause spectral sensitivity differences betweenso¹ andso¹;gl^60j flies. This is because, besides photoreceptor cells, glassalso expresses in a few central brain neurons (Moses and Rubin, 1991;Ellis et al., 1993; Vosshall and Young, 1995) that thegl^60j mutation therefore may eliminate. Indeed, we found recently that a subgroup of the period gene expressing dorsal neurons (DN1) is missing ingl^60j mutants (Helfrich-Förster et al., 2001). Like the H–B eyelets, these dorsal neurons send axonal projections toward the LN_v (Kaneko and Hall, 2000) through which they theoretically could entrain the latter. Furthermore, flies bearing a mutation in the disconnected(disco) gene, which mainly lack the LN_v but not the DN1 cells, are able to entrain to LD cycles (Zerr et al., 1990; Hardin et al., 1992;Helfrich-Förster, 1998). The DN1 are present inso¹ mutants (Kaneko, 2000; B. Wisotzky and A. Hofbauer, unpublished observations) and therefore could account for differences in the action spectra of so¹ and so;gl^60j flies. However, it is not known whether the DN1 cells express rhodopsins at all and, if so, whether the λ_max of such DN1 rhodopsins fits the action spectrum as do Rh5 and Rh6 in eyelet.

Eyelet appears to play a complex role in the circadian system of flies

Although we cannot exclude a role for the DN1 cells as circadian photoreceptors, the anatomical and action spectrum data presented here support a role for eyelet in the entrainment pathway of D. melanogaster. Several observations indicate that this putative role is complex, however.

Obviously, eyelet functions only in concert with cryptochrome (inso¹ flies) or with the compound eyes (inso¹;cry^b flies with eye remnants). Depending on which additional photoreceptor is present, eyelet seems to have quite different effects on the activity rhythm.

In the presence of cryptochrome but without the compound eyes eyelet appears to mediate phase delays. The evening peak ofso¹ flies is phase delayed, andso¹ flies frequently show antidromic phase shifting. Interestingly, both phase delaying effects occur only at wavelengths close to the putative λ_max of eyelet, implicating eyelet as the source of these behaviors. Nevertheless, in the presence of tiny eye remnants but in the absence of cryptochrome no phase lag of the evening peak is observed, and the flies show no antidromic phase shifting.

In summary, our results indicate that eyelet is a pupal transformation of Bolwig's organ, that both receptor systems have an association with the PDH-immunoreactive lateral neurons, and that both mediate circadian entrainment exhibiting an action spectrum, which for eyelet in the adult fly is based in part on Rh6 expression possibly also inherited from the photoreceptors of Bolwig's organ in the larva. Apparently, cryptochrome and rhodopsins of eyelet and the compound eyes collude in a complex manner to entrain the adult fruit fly's circadian activity to the 24 hr day. This dual nature of circadian photoreception encompassing opsins and cryptochromes as photopigments resembles the situation in mammals in which melanopsin and cryptochromes of the inner retina apparently team up with rhodopsin in the rods to entrain the circadian system of mice (Freedman et al., 1999; Selby et al., 2000;Lucas et al., 2001; Thompson et al., 2001; Berson et al., 2002; Hattar et al., 2002).