Fly strains

We used adult wild type Drosophila melanogaster (Oregon R or w¹¹¹⁸ strains) for basic immunocytochemistry. For correlation with various neuronal phenotypes we performed immunocytochemistry on a variety of Gal4 lines crossed with UAS-GFP, as specified below. L2 monopolar interneurons were visualized by the 21D-Gal4 driver [53] (from Tomas Raabe, University Würtzburg, Germany). C3 neurons were identified by 5-6-8/CyO;TM2/TM6B-Gal4 (abbreviated 5-6-8-Gal4; from Larry Zipursky, Howard Hughes Medical Institute at UCLA, Los Angeles, CA). Other Gal-4 lines used were: rdl-Gal4 (4.7 kb upstream rdl-gene; from Julie Simpson, Howard Hughes Medical Institute, Janelia Farm, VA), Cha-Gal4 [54] (from Bloomington Stock Center at Indiana University, Bloomington, IN), and OK371-Gal4 ([55], from Hermann Aberle, University of Münster, Germany). These were used to visualize expression of the GABA_A receptor subunit RDL, choline acetyltransferase (Cha), and vesicular glutamate transporter (vGluT) gene products, respectively. To visualize Gal4-expression with GFP, we crossed these lines with flies expressing UAS-mCD8-GFP (from Bloomington Stock Center). Presynaptic sites were visualized by driving a neuronal synaptobrevin-GFP fusion line (w[*];P{w[!mC] # UAS-nsyb.egfp}2; Bloomington stock center) with either the OK371- or 21D-Gal4 lines (see [56], [57]).

Antisera

Several antisera were used to detect neurotransmitters and other signaling components in the lamina. The antisera and their corresponding antigens are listed separately (Table 1). Antisrum specificities have been carried out for all antisera in earlier publications (listed in Table 1). A comprehensive description of antiserum production and specificity tests is given below.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Antisera used for immunocytochemistry.

https://doi.org/10.1371/journal.pone.0002110.t001

Immunocytochemistry

For glutamate immunolabeling, brains were dissected out of the head capsule in modified Zamboni's fixative (4% paraformaldehyde, 1.6% glutaraldehyde, 0.2% saturated picric acid, in 0.1M sodium phosphate buffer, pH 7.4) and left for between 1 h and overnight at 4°C. They were washed in sodium phosphate buffered saline (PBS), and then sectioned at 50–80 µm slices on a Vibratome. The sections were washed in PBS, blocked with normal goat serum (NGS), transferred to 0.5% Triton X-100 in PBS for 30 min prior to primary antibody incubation, and then and incubated overnight at 8°C in one of two glutamate antisera (Table 1), a rabbit polyclonal antibody at a dilution of 1∶10,000, or a monoclonal antibody at a dilution of 1∶5000. After the primary antibody, the tissue was washed several times in PBS, and then incubated with the corresponding secondary antibody (goat anti-rabbit, goat anti-mouse: Jackson ImmunoResearch Labs, West Grove, PA) conjugated to a fluorochrome (Cy3 or FITC).

For all other immunocytochemistry the adult fly heads were dissected in PBS-TX (0.01M phosphate-buffered saline with 0.5% Triton X-100, pH 7.2) before fixation. For DmGluRA, DvGluT and DLG immunolabeling, opened heads were fixed 2 h in freshly prepared ice-cold Zamboni's fixative (4% paraformaldehyde and 0.5% picric acid in 0.1 M phosphate buffer, pH 7.2) or 4% paraformaldehyde in 0.1 M sodium phosphate buffer (PB) at pH 7.4. For GABA_BR2 immunolabeling, fly heads were fixed for 2 h in ice-cold 4% paraformaldehyde in 0.1 M PB (pH 7.2). For GABA and GAD immunolabeling, the heads were fixed in freshly prepared ice-cold Bouin's fixative for 30 min (tissues were also fixed in Zamboni's fixative to obtain better GFP preservation). For ChAT and vAChT immunolabeling, adult fly heads were fixed for 2 h in ice-cold 4% paraformaldehyde (pH 7.4) or in Zamboni's fixative. Additional labeling with anti-GFP was necessary to amplify GFP fluorescence partly quenched after Bouin fixation.

Tissues were thoroughly washed with 0.1 M PB and incubated overnight at 4°C in 20% sucrose in 0.1 M PB. 20 µm thick sections of the head were cut on a Leitz 1720 Cryostat at −23°C and collected on chromalum-gelatin–covered microscope slides. After washing in PBS-TX, tissues were incubated with primary antibodies in 4°C overnight or for 48 h. Brains were washed in PBS-TX at room temperature (about 22°C) and incubated with fluorophore-tagged secondary antibodies (Cy2- or Cy3-tagged IgGs, raised in goat; Jackson ImmunoResearch) diluted 1∶1000, either overnight at 4°C or 2 h at room temperature (around 22°C). After washing in PBS-TX and rinsing in 0.01 M PBS, tissue was mounted under a coverslip in 20% glycerol in 0.01 M PBS.

Pre-embedding immuno-electron microscopy

We also used both wild-type and white eye mutants for electron microscopy of preparations immunolabeled for glutamate by the pre-embedding method using the polyclonal rabbit anti-glutamate [62]. Tissue incubated as above in this primary antibody was next incubated in a biotinylated goat anti-rabbit antibody (Vector Labs) and then in a solution of peroxidase conjugated Avidin Biotin Complex (ABC complex, Vector Labs). Labeling was detected with 3,3′-diaminobenzidine (DAB) as the substrate. The sections were then osmicated, dehydrated in graded ethanol series, changed into propylene oxide and then flat-embedded between Aclar sheets (Ted Pella Inc, Redding, CA) in Poly/Bed 812 resin. The tissue was sliced at 80 µm using a Vibratome and selected slices subsequently resectioned at 60 nm for electron microscopy. Ultrathin sections were then viewed at 60 kV with a Philips 201 C electron microscope, photographed on 35 mm film at primary magnifications of between 5,000 and 20,000, and the prints then labeled and scanned.

Imaging

For glutamate-immunolabeled specimens, Vibratome slices were mounted in Vectashield (Vector Labs, Burlingame, CA) and viewed with a Zeiss LSM 410 confocal microscope (Zeiss, Jena, Germany). All other specimens were imaged with a Zeiss LSM 510 confocal microscope. Confocal images were obtained at an optical section thickness from 0.1–0.35 µm and were processed with Zeiss LSM software and edited for contrast and brightness in Adobe Photoshop CS3 Extended version 10.0.

Acetylcholine signaling components in the lamina

Acetylcholine is a major excitatory neurotransmitter in Drosophila and other insects [49], [78], [79]. Choline acetyltransferase (ChAT) and the vesicular acetylcholine transporter (vAChT) are essential for cholinergic neurotransmission and antisera to these proteins are phenotypic markers for cholinergic neurons [54], [73]. Several papers have used ChAT antisera or in situ hybridization for Cha transcript to localize presumed cholinergic neurons to the Drosophila visual system [49], [51], [70], [80], but to our knowledge none has yet employed vAChT antiserum to this part of the brain. We examined the Drosophila lamina with antisera to both proteins. ChAT-immunolabeling reveals several types of lamina neuron (Fig. 5A). The cell bodies of large and small monopolar neurons are ChAT-immunolabeled (Fig. 5A,E), and what appear to be the axons and tripartite lamina collaterals of L4 monopolar neurons also react with ChAT antiserum (Fig 5A, see also 5B). The axons of other monopolar neurons were not seen. Using 21D-Gal4 to drive GFP in L2 cells we could also show that anti-ChAT labeled L2 cell bodies, but no immunolabeling was visible in their dendritic processes in the lamina (Fig. 5E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Monopolar neurons in the lamina labeled with different antisera or Gal4 lines.

A ChAT immunoreactive lamina neurons. Asterisk: layer with large monopolar cells (L1–L3); triangle: layer with small monopolar cells (L4 and L5); square: layer with processes of Cha-Tan neurons (see Fig. 8 and 9). Arrow labels level of branching of L4 neurons in the proximal lamina. Scale bar = 10 µm (for all images, except panel B). B GFP expression in lamina driven by the rdl-Gal4 reveals L4 neurons. Triangle: L4 cell body; arrow: characteristic branching of L4 collaterals in the proximal lamina. GFP is also seen in branches of a wide-field tangential neuron, in the distal lamina. Scale bar = 10 µm. C Weak immunolabeling in cell bodies of large monopolar neurons with antiserum to vGluT. Strong immunolabeling in the lamina neuropil is seen in processes of amacrine neurons. D1–3 Distributions of vesicular acetylcholine transporter (vAChT) immunoreactivity and rdl-Gal4 driven GFP expression co-localize to arborizations of the L4 neurons (arrow) in the proximal lamina, but not in their cell bodies (cb) and not in processes of rdl-Tan neurons (asterisk in D1) in the distal lamina. However, vAChT immunoreactivity is seen in enlarged boutons of another tangential neuron in this dorsal layer (large arrow). Scale bar = 5 µm. E1–3 Cell bodies of L2 monopolar neurons are ChAT immunoreactive, revealed by 21D-Gal4 driven GFP (green) in L2 neurons labeled with anti-ChAT (α-Cha; magenta). Co-localization of label is seen in cell bodies, but not clearly in their neurites. Scale bar = 10 µm. F1–3 Anti-vAChT labeling (α-vAChT; magenta) is not co-localized in L2 monopolar cells displayed by GFP driven by 21D-Gal4. Scale bar = 10 µm.

https://doi.org/10.1371/journal.pone.0002110.g005

Additional to the cell bodies and presumed L4 processes, the ChAT antiserum also labeled enlarged boutons at the level of the C2 terminals (Fig. 5A, 6A,C). These structures seem to be associated with tangential neuronal elements having boutons in the distal lamina neuropil. Using Cha-Gal4 to drive GFP we obtained strong fluorescence in tangential neurons with similar boutons (Fig. 6A), but no labeling of any monopolar neurons. It is not clear whether these tangential processes seen with Cha-Gal4 are derived from the Tan tangential neurons (see Fig. 2B) or a novel type of tangential neurons (or even new amacrine neurons, like Am2 of other flies), both with more pronounced varicosities distal to the lamina neuropil than Tan. Arguing against the amacrine neuron possibility, the Cha-Gal4 expressing neurons appear to derive from neurons with axons projecting towards or even connecting to the medulla (Fig. 6A). Thus they are most likely to be a form of wide-field tangential neuron. For simplicity we will therefore refer to these neurons henceforth as Cha-Tan neurons. We found co-localization of Cha-Gal4 driven GFP and anti-ChAT immunolabeling in the tangential processes and enlarged boutons of these cells (Fig. 6C), but ChAT-immunolabeling was detected mainly in the GFP-labeled processes in the distal lamina, not in their cell bodies or axons (data not shown). Our immunocytochemistry thus confirms that the lamina neurons seen in the Cha-Gal4 reporter line actually express ChAT-immunoreactivity. We could exclude that the ChAT-immunolabeling in this distal layer is derived from C2 neurons, although, as we show later, C2 neurons may contact the Cha-expressing tangential neurons on these enlarged boutons. Seen in cross section, it appears that the overlapping Cha-Tan neurons form aggregates of boutons, each aggregate associated with an underlying cartridge (see section on GABA receptors).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Tangential and amacrine neurons in lamina, and markers for GABA, glutamate and acetylcholine signaling.

A GFP expression driven by Cha-Gal4 reveals putative choline acetyltransferase containing lamina neurons. GFP expression in cell bodies near the medulla (arrow), and lamina morphology together suggests that Cha-Gal4 reports Tan neurons (La wf1) in the lamina. Scale bar = 20 µm. A1 Magnification of the putative Tan varicosities in the distal lamina. Scale bar = 10 µm. B GFP expression in Rdl-Gal4 reveals tangential neurons in lamina with cell bodies localized above medulla (arrow). Scale bar 10 µm. B1 Enlarged view of the Rdl-Gal4-expressing lamina neurons. Scale as in A1. C(1–3) Distribution of ChAT immunoreactivity (C1, α-Cha; magenta), in relation to Cha-Gal4 driven GFP (C2, green) in lamina cross section. Co-localization (C3) is seen in the distal rosette-like structures (see magnifications in insets). Scale bar = 10 µm. D (1–3) Cross section of lamina showing co-localization of GFP in Cha-Gal4 and anti-vAChT (D1, α-vAChT: magenta) in distal boutons of Cha-Tan neurons. Scale bar = 10 µm. E (1–3) GABA_BR2 immunoreactive neurons (E1, α-GBR2; magenta) in relation to Cha-Gal4 driven GFP in tangential neurons (E2, green). Close contacts and some co-localization (E3) between labels are seen (insets show magnified views), suggesting localization of GABA_BR2 on these tangential neurons. Scale bar = 5 µm.

https://doi.org/10.1371/journal.pone.0002110.g006

The antiserum to the vAChT confirmed most of the ChAT-immunolabeling in the lamina. We could detect vAChT-immunolabeling in basal processes likely to represent collaterals of L4 neurons (Fig. 5D) and in the dilated boutons of the Cha-Tan neurons (Fig. 6D). Double-labeling with rdl-Gal4 and vAChT antiserum showed the close match between the two in the morphology of the L4-like profiles (Fig. 5D). This double-labeling also clearly showed the distinction between the rdl-Gal4 (described below) and Cha-Gal4 expressing tangential processes in the distal lamina. Whereas the Cha-Gal4 tangential profiles co-localize another acetylcholine marker, vAChT (Fig. 6D), the rdl-Gal4 processes did not (see Fig. 5D). Furthermore, the vAChT antiserum did not label neuronal cell bodies in the lamina cortex (Fig. 5D), and as a result we could not match the rdl-Gal4 signal with ChAT-immunolabeling in monopolar cell bodies, even those of L4 cells.

Glutamate signaling components in the lamina

Immunocytochemistry has previously suggested the presence of glutamate in the large lamina monopolar neurons, L1 and L2, in the flies Drosophila, Musca, Calliphora and Phaenicia sericata [46], [47] as well as in type 1 amacrine neurons of the latter fly [47]. Here, we examined the Drosophila lamina for evidence of glutamate neurotransmission by applying antisera to two essential molecules, the neurotransmitter glutamate and the Drosophila vesicular glutamate transporter (vGluT). The presence of glutamate is a requirement for its candidacy as a neurotransmitter, but given the widespread availability of glutamate as an intermediary metabolite, this evidence alone is unacceptably weak. On the other hand, vGluT is required to load synaptic vesicles with glutamate and is a highly specific marker for sites of glutamate neurotransmission so that, for example, vGluT antisera label motoneuron varicosities [55], [61] that are known to utilize glutamate as a neurotransmitter [81].

Drosophila vGluT immunolabeling.

To confirm that immunoreactivity to glutamate signified a capacity for glutamatergic transmission in the monopolar cells, we also applied four different antisera to the Drosophila vGluT and obtained identical labeling with each (Fig. 7). Strong vGluT immunolabeling was detected in profiles similar to α-processes of the amacrine neurons or possibly like β-processes of T1 neurons (Fig. 7; Figure S4). Weak vGluT immunolabeling of cell bodies was seen in the chiasma between the lamina and medulla, in a position corresponding to those of amacrine cells (Fig. 2A), but it was not possible to connect these to lamina processes (Figure S4A). The vGluT immunosignal in the lamina was mostly distinct from that seen with the OK371-Gal4 [55], representing vGluT promoter expression (Fig. 7A,B). The OK371-Gal4 drove GFP in smaller or larger populations of large monopolar neurons (Fig. 7A–E). Thus, instead of a complete co-localization of OK371 driven GFP and anti-vGluT expression, we saw neurons expressing vGluT lying adjacent to the GFP-labeled large monopolar neurons (Fig. 6A3,B3). In cross-section, six profiles in each cartridge expressed vGluT immunolabeling and surrounded the GFP labeled monopolar neurons (Fig. 7B,C2,D2,E2). To confirm the failure of vGluT immunolabeling to localize to processes of monopolar neurons, we investigated the relation between this label and OK371-driven GFP in terminals of monopolar neurons in the medulla (Fig. 7C3, D3 and E3). No clear co-localization was detectable. However, some vGluT-immunolabeling can be seen in cell bodies of large monopolar neurons (Fig. 5C) and we could not rule out low levels of vGluT immunolabeling in dendrites of monopolar cells that also express OK371-Gal4 (see Fig. 7A3,B3). Thus there is a lack of correspondence between data from the antisera and data from the Gal4 driver. OK371 expression indicates that at least the large monopolar cells express vGluT (vglut-promoter), just as they also contain glutamate, whereas at best the vGluT antisera only weakly label the corresponding cell bodies and tips of dendrites. On the other hand the vGluT immunolabeling seen probably in amacrine cell processes is not matched by a similar pattern of GFP-labeling for the vGluT promoter. Part of this discrepancy may reflect the different intraneuronal distribution of the markers: vGluT antibodies label predominantly presynaptic sites while GFP (cd8-GFP) expression is distributed in the plasma membrane throughout the neuron. There may also be very small amounts of highly localized vGluT protein in L1 and L2 compared with the surrounding amacrine cell processes. To investigate this possibility we used a neuronal synaptobrevin-GFP fusion (nsyb-gfp) to target GFP primarily to presynaptic sites (Figure S5). Using the 21D-Gal4 to drive nsyb-GFP resulted in fluorescence localized predominantly or exclusively to the medulla terminals of the L2 neurons (Figure S5A), but still no co-localization was seen with vGluT immunolabeling (Figure S5 B–D). Moreover with OK371-driven nsyb-GFP there was no co-localization to vGluT immunolabeling (Figure S5 E1–3). Finally, we cannot exclude that the OK371-Gal4 expression in neurons is incomplete because it lacks promoter/enhancer elements in the construct.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. Distribution of vesicular glutamate transporter (vGluT) immunolabeling.

Different antisera were combined with OK371-Gal4 driven GFP expression, which reports vGluT-expressing neurons. A1–3 Frontal sections of the lamina (vGluT antiserum) revealing lack of co-localization between vGluT immunolabeling (magenta) and GFP expression (A3 merged) in L2 monopolar cells (A1, green). Scale bar = 5 µm. B1–3 Cross-section of the same lamina region with enlargements in insets. B1 GFP in L2 neurons, B2 vGluT immunolabeling, B3 merged. Note that the vGluT immunolabels six structures surrounding the margin of the cartridge and at the extensions of L2 dendrites. C–E Similar structures labeled with three other antisera to vGluT. All vGluT antisera display the same immunolabeling in the lamina and medulla. C1–3 Affinity purified antiserum to vGluT (magenta) applied to OK371-Gal4 driven GFP (green). C1 Frontal section of lamina. Scale bar = 10 µm. C2 Cross-section of lamina (same magnification). C3 Frontal section of medulla showing that vGluT immunolabeling is not in GFP-labeled terminals of L2 cells in stratum M2 (arrow). Scale bar = 10 µm. D1–3 Similar structures labeled with antiserum to N-terminus of vGluT. Same scales as C. E1–3 Similar images using antiserum to the C-terminus of vGluT. Scales as in C.

https://doi.org/10.1371/journal.pone.0002110.g007

To reveal more clearly the relationship between vGluT-immunolabeled amacrine cell processes and the terminals of photoreceptors R1–R6, we used antibodies to Discs large (DLG) as a marker. The DLG protein is a membrane associated guanylate kinase (MAGUK) family protein located at the pre and postsynaptic area of functional glutamatergic synapses, at least in the Drosophila neuromuscular junction [84]. The vGluT immunolabeled structures are likely to be amacrine α-processes that seem to make contacts with DLG immunolabeled photoreceptors (Figure S4B1).

GABA signaling components in the lamina

GABA and GAD immunolabeling.

GABA is a major inhibitory neurotransmitter in Drosophila and other insects and distributed in large numbers of neurons [68], [86], [87]. Proven markers for GABAergic neurons are antisera to GABA, vesicular GABA transporter (vGAT) and the biosynthetic enzyme GAD. Here we employed GABA, vGAT and GAD (GAD-1) antisera to label lamina neurons. To identify C3 neurons we employed the 5-6-8-Gal4 line (Fig. 8A). Previous studies have shown that the centrifugal neurons C2 and C3 in different fly species display GABA immunoreactivity [42], [43], [44], [45]. Our study confirmed GABA and GAD immunoreactivity in C2 and C3 neurons in Drosophila (Fig. 8B–D, J). In a recent report from our laboratory [68] vGAT immunolabeling was also detected in C2 and C3 neurons. This suggests that the C2 and C3 neurons indeed both contain and utilize GABA as a neurotransmitter in the lamina: C2 probably releasing the transmitter from presynaptic sites that localized to enlarged boutons in a distal layer and C3 from similar sites at varicosities along their length in the lamina [9].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 8. Centrifugal neurons in the lamina and markers for GABA signaling.

A GFP expression in columnar C3 neuron terminals in the lamina (La) and medulla (Me) driven by the 5-6-8-Gal4. The C3 cell bodies (cb) proximal to the medulla are also visible. Ch: optic chiasma. Scale bar = 10 µm. B Distribution of glutamic acid decarboxylase-1 (GAD) immunoreactivity in C2 and C3 neurons (overview of lamina and part of medulla). Scale as in A. C Higher magnification of C2 and C3 terminals in lamina revealed by anti-GAD antiserum; note C3 varicosities along the axons. Scale bar = 10 µm. D Both C3 and C2 (labeled by arrow) terminals can be visualised in the lamina by anti-GABA antiserum. Same magnification as in C. E GFP in C3 neurons (in green) driven by 5-6-8-Gal4 with anti-GABA_BR2 immunolabeling (magenta) in horizontal section of lamina. The C3 axons traverse the optic chiasma. Note that much of the receptor immunolabeling is in neurons other than C3, but that some appears co-localized (white). Scale bar = 10 µm. F–H Details of double-labeling with GFP expression in C3 neurons and antiserum to GABA_BR2. Some GABA_B receptor expression is in C3 neurons (arrows in F, G and H). At other sites the receptor is expressed on neuronal structures that appear to be closely adjacent to C3 neurons (long arrowheads in G and H) or in C2 terminals indicated by short arrowhead in F. Scale bar for F–H (F1): 5 = µm. I C3 neurons (5-6-8-Gal4 driven GFP, green) do not co-localize ChAT immunoreactivity (magenta) but appear to be located close to ChAT immunolabeled profiles. Scale bar = 5 µm. J1–3 GFP-labeled C3 cell bodies (J1, green) express GABA-immunoreactivity (J3, magenta) as seen in merged image (J2). Scale as in F.

https://doi.org/10.1371/journal.pone.0002110.g008

We analyzed the relations between the GFP-labeled C3 neurons (5-6-8-Gal4) and ChAT-immunolabeling and found no co-localization of markers (Fig. 8I). However, the C3 neurons were seen close to the ChAT-immunolabeled monopolar axons (which are most likely L4 neurons) and terminated close to the enlarged boutons of Cha-expressing tangential neurons, Cha-Tan.

GABA receptors.

The localization of the metabotropic GABA_B receptor 2 (GABA_BR2) has previously been demonstrated in the brain of Drosophila by means of a specific antiserum [65], [68]. Here we show the distribution of GABA_BR2 immunoreactivity (GBR-IR) in relation to the different lamina neurons visualized by GFP driven by specific Gal4-lines (Fig. 8E–H). The major expression of GBR-IR was seen on the distal varicosities of C2 neurons (Fig. 8F–H) and in boutons of C3 neurons (Fig. 8E–H), as well as on enlarged boutons of Cha-Gal4-expressing tangential neurons, Cha-Tan (Fig. 6E). GBR-IR expression in the lamina is thus likely to be localized presynaptically in C2 and C3 boutons and postsynaptically on the tangential neuron boutons. This would explain why the distribution of GBR-IR signal in this region appears in coherent aggregates larger than the C2 terminals and larger than the Cha-Gal4-expressing boutons (Fig. 6E). To investigate the relationship between C2 neurons and Cha-expressing neurons further, we also double-labeled tissues with anti-GABA_BR2 antibodies and anti-ChAT (Fig. 9D). Again we saw that the immunolabeled Cha-Tan neuron boutons co-expressed GBR-IR material.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 9. Comparison between rdl-Tan and Cha-Tan neurons in distal part of lamina.

Arrows indicate differences in structures between the two types of tangential neurons. rdl-Tan have thin varicose processes hanging down into the lamina, whereas the Cha-Tan have enlarged boutons in the same layer. A (1–3) rdl-Tan neurons (A2, green), visualised by GFP expression driven by rdl-Gal4, contact GABA_BR-immunolabeled cells at arrows, but do not coexpress the receptor (A1, magenta). B (1–3) Cha-Tan cells visualised by GFP expression driven by Cha-Gal4 (B2, green) co-express GABA_BR immunoreactivity (B1, magenta) in their boutons (arrow). C (1–3) rdl-Tan neurons in cross-section are organized in widely branched network with arborizations in each cartridge (C2, green). Close contacts with GABA_BR-immunopositive cells (C1, magenta) are visible. D (1–3) Cha-Tan neurons also are organized in a network but they have more distinct aggregates of boutons in each cartridge and these boutons co-express GABA_BR-immunolabeling (schown in Fig. 8E) and co-localize anti-ChAT (D1, magenta, D3 – merged). Scale bar for images A to D = 10 µm. E (1–2) Higher magnification of the rdl-Tan processes distally in the lamina in cross section (E1) and contacts with GABA_BR-immunolabeled neurons (E2). Scale bar = 10 µm.

https://doi.org/10.1371/journal.pone.0002110.g009

The GBR-IR material associated with the C3 neurons appear to be predominantly co-localized within the membranes of the C3 boutons throughout the depth of the lamina (Fig. 8F, G). This we interpret to represent presynaptic GABA_BR2 in GABAergic C3s. However, some immunoreactivity seemed to lie in structures very close to these boutons, not co-localized to them (Fig. 8G, H). Thus, there is a possibility that a neuron postsynaptic to C3 also expresses GABA_BR2 at sites located close to the contacts with C3 neurons. Given the restricted distribution of the immunoreactivity however we could not identify this neuron type. From EM analysis, possible candidates postsynaptic to C3 would be L1–L3 and amacrine cell processes [9], [13].

Ionotropic GABA_A type receptors are the major receptors mediating fast inhibitory transmission in the insect brain (see [87]). We tried to analyze the distribution of this type of receptor in the Drosophila visual system, but an antiserum to one of the Drosophila GABA_A receptor subunits, RDL, failed to produce distinct immunolabeling in the lamina, even though the medulla displayed layers of strong RDL-immunoreactivity (Figure S4E). An earlier report indicated diffuse immunolabeling possibly associated in part with large monopolar neuron dendrites [68]. Given this uncertainty, we resorted to using an rdl-promotor-Gal4 to drive GFP expression in neurons and reveal lamina expression of possible GABA_A receptors.

In the fly visual system, rdl-Gal4 expression (amplified by anti-GFP immunolabeling) was seen in two types of neurons in the lamina (Fig. 5B, 6B, 9A). It appears that one type is the L4 monopolar neurons. This is based on the small cell bodies just above the lamina neuropil and the short collateral branches in the proximal layer (Fig. 5B,D). The second type is a wide-field tangential neuron morphologically similar to the one designated La wf2 by Fischbach and Dittrich [12] (Figs. 5B, 6B, and 9B). We choose to refer to the latter neurons as rdl-expressing tangential (rdl-Tan) neurons. Since the RDL-antiserum does not yield strong immunolabeling we cannot confirm the Gal4 expression pattern in the lamina as representing RDL protein expression. However, in other parts of the Drosophila brain this Gal4 line seems to produce GFP expression that matches RDL-immunolabeling quite well (Enell and Nässel, unpublished). In double-labeled specimens, we detected no apparent co-localization of GABA_BR2 immunolabeling and rdl-Gal4 expression in the distal portion of the lamina. Rather we noted a close association between the boutons of the rdl-Tan neurons and GBR-IR signal in C2 varicosities and/or large boutons of Cha-Tan neurons (Fig. 9A,C). If the rdl-Gal4 does in fact represent RDL distribution, there must be a differential distribution of GABA_A and GABA_B receptors in lamina neurons visualized in our study.

Another important finding was obtained by labeling with GABA signaling markers. The two Gal4 expression patterns in tangential processes seen in the distal lamina with Cha-Gal4 and rdl-Gal4 lines could be clearly distinguished. Using antiserum to GABA_BR2 on rdl- and Cha-Gal4 flies it is clear that the rdl-Tan neurons are distinct from the tangential processes of the Cha-Tan neurons (Fig. 9A–E). Whereas the Cha-Gal4 expressing boutons co-express GABA_BR2, the rdl-Gal4 ones do not. The boutons of these two types of tangential neurons are, however, in close proximity suggesting that they could both receive input from the same GABAergic C2 neurons. Confirmation of this possibility must await EM examination.

Acetylcholine signaling components

The best evidence for neurons qualified to use acetylcholine for signaling was obtained for the wide-field Cha-Tan neurons. Both Cha-Gal4 expression and the ChAT- and vAChT- antisera identify these neurons. The Cha-Tan neurons give rise to enlarged boutons, most probably associated with distal C2 neuron terminals. Thus, we believe they were mistaken for C2 neurons in earlier reports on ChAT-immunolabeling in flies [89]. Especially with Gal4-driven GFP expression it is clear that these large boutons are parts of the Cha-Tan neurons, and may thus be partly regions receiving input from centrifugal neurons. The Cha-Tan neurons also produce varicose processes that run between the boutons and that have short branches hanging down into the lamina synaptic neuropil. It therefore seems that a portion of the cholinergic neurotransmission from Cha-Tan neurons is confined to a shallow layer in the distal lamina. The wide spread of these processes was the reason that the synaptic connections of wide-field tangential neurons were not investigated by Meinertzhagen and O'Neil [9], so the synaptic targets or inputs of these neurons are still unknown in Drosophila. Another layer of cholinergic neurotransmission may occur in the proximal portion of the lamina, by means of collaterals of the L4 monopolar neurons.

Published reports on the immunocytochemical localization of acetylcholine receptors in the CNS of Drosophila have also shed some light on the lamina circuitry. Two nicotinic receptor subunits and the muscarinic receptor have previously been detected in the lamina [90], [91], [92]. Whereas the muscarinic receptor [92] and the ALS subunit were only seen weakly and diffusely distributed in this neuropil, the ARD subunit was revealed distally in the lamina in bouton-like clusters [90]. Thus, the ARD distribution closely matches that of the boutons of the cha-Tan neurons, but it is not clear what neuron type(s) expresses the receptor.

L4 monopolar neurons have three collateral processes in the basal portion of the lamina [12]. These interconnect the L4 neurons in adjacent cartridges, as well as the L2 cell and photoreceptor terminals within the neighboring and native cartridges, and appear to be the major outputs from the L4s within the lamina [9]. At intermediate pattern contrasts, L2 in Drosophila recruits L4 as the substrate for detection of front-to-back motion [93]. We find that the collateral branches of L4s strongly express ChAT and vAChT immunoreactivities, suggesting that a cholinergic pathway may be responsible for this recruitment in the lamina. In two earlier reports on ChAT-immunoreactivity in Drosophila [49], [89] the L4 collaterals are visible in the figures, but did not receive specific comment. Interestingly the rdl-Gal4 drives GFP in what appears to be L4 neurons (and a set of tangential neurons, rdl-Tan). The failure of our antiserum to provide matching immunocytochemical evidence for RDL expression in these neurons, means that it is not clear whether the neurons express this GABA_A receptor subunit, or – more likely – whether they may merely do so at levels too low to detect immunocytochemically. Overall, it is tempting to speculate that acetylcholine is used for lateral connections over larger or smaller areas of the lamina mosaic (respectively: Cha-Tan distally and L4 proximally).

As reported previously (see [49]), and confirmed in our study, the cell bodies of the large monopolar neurons L1 and L2 also express ChAT-immunoreactivity, and Cha-transcript [51] although we could not detect vAChT immunoreactivity in these neurons. Thus, it is not clear whether the large monopolar neurons utilize acetylcholine as a neurotransmitter, even though they may have a capacity to synthesize it, or whether the vesicular transporter is expressed at too low levels to detect.

Glutamate signaling components

We validated previously published data, including our own [46] on glutamate signaling components by using different antisera to the Drosophila vesicular glutamate transporter (vGluT), as well as analyzing vGluT-Gal4 expression. Glutamate signaling seems to be performed at two main candidate sites in the lamina, large monopolar cells and amacrine neurons.

We obtained clear evidence for glutamate-like immunoreactivity in the large monopolar cells L1–L3 in the lamina and medulla of two fly species (Musca and Calliphora), whereas only two of these neurons, L1 and L2, were detected in Drosophila. There was, however, some variation in the latter species, L2 alone being invariably labeled. It most plausible to attribute this variation to different levels of cytoplasmic glutamate that could have existed under different functional states prior to preparation for immunolabeling. Compatible with these neurons having the capacity to store vesicular glutamate, the OK371-Gal4 line, specific for vglut expression, also drives GFP in the large monopolar neurons, but we did not detect clear vGluT immunolabeling in monopolar neurons. However, low levels of vGluT immunolabeling were seen in cell bodies of the large monopolar neurons and immunolabeling in dendrites of these neurons may be masked by the stronger immunolabeling seen in amacrine processes. Another more likely possibility is that the amount of vesicle-bound glutamate (and vGluT) is simply too low to detect. We thus do not have conclusive evidence that L1 and L2 have the capacity to store vesicular glutamate, and consequently that they are glutamatergic in Drosophila. These neurons have most of their synaptic output in the medulla and either no (L1) or a limited number (L2) of output synapses in the lamina [9], [13]. Insofar as L1 sometimes clearly expresses a glutamate phenotype but lacks output synapses in the lamina, we would predict the absence of glutamate containing vesicles and corresponding vGluT in the lamina, at least in L1. On the other hand, as revealed by the 21D-Gal4 line, we did not detect vGluT immunolabeling in the L2 medulla terminals either, again possibly because there was insufficient protein to yield a clear immuno signal. Thus it still cannot be entirely excluded that the glutamate immunoreactivity seen previously [46], [47] may represent non-vesicular amino acid stored as a metabolic intermediate. On the other hand, the possibility that these monopolar neurons, the major output neurons of the fly's lamina, might use two fast neurotransmitters, glutamate and acetylcholine, may not be unprecedented [94]. However, our evidence provides no support for the possibility that the cells might release these at different sites, or even in different neuropils (L1 in the medulla, and L2 also in the lamina).

While this paper was in revision an elegant study appeared on the distribution of a vesicular glutamate transporter in Drosophila [95]. The authors of that report used a different vglut promoter Gal4, but one of the vGluT antisera [61] also used in our investigation. Although the paper did not report on vGluT distribution in the lamina, the authors report expression in the medulla, where, as in our study, they found no conclusive evidence for vglut or vGlut expression in the terminals of L1 to L3.

In addition to the monopolar neurons, strong vGluT immunolabeling was seen in structures resembling the α-processes of amacrine neurons, and this could be correlated with immunoreactivity to glutamate seen by electron microscopical analysis. Sinakevitch and Strausfeld [47] also detected such immunoreactivity in the fly Phaenicia sericata, thus providing some measure of support for a glutamatergic phenotype in the lamina amacrine cells.

Overall, there are some incongruencies in the data for glutamate signaling: the processes of monopolar neurons express immunoreactivity for glutamate but not the vesicular transporter, while the amacrine cells express immunolabeling for the transporter but not the expected Gal4 expression. To resolve some of these issues, we had hoped to see DmGluRA expression in lamina circuits, but very weak labeling was seen and this could not readily be assigned to any specific neuron type. Since we detected very strong DmGluRA immunolabeling in neurons of the medulla, we presume the expression level is just very low in the lamina neurons. It was therefore surprising that the antiserum to the NMDA1 receptor subunit used in a previous study [47] labeled neither neither in the lamina nor elsewhere in the visual system. The antiserum was raised against a sequence of a mammalian NMDA1 receptor protein with limited similarities to that in Drosophila and thus not likely to display much cross-reactivity in Drosophila. However, we could show rather strong immunolabeling of neurons in the mushroom body lobes, suggesting that again the lack of signal could be a matter of low levels of expression in the visual system of Drosophila.

Adopting cautious criteria, we can summarize the positive findings on glutamate signaling components in the lamina as follows. We find evidence that the α-processes of lamina amacrine neurons express vGluT, and glutamate. These neurons, which we might therefore predict to be glutamatergic, have many outputs onto β-profiles of T1 neurons, and onto R1–R6 and L1–L3 neurons [9]. Compatible with this suggestion, Sinakevitch and Strausfeld [47] reported NMDA1 receptor-like immunoreactivity on T1 neurons in P. sericata. Glutamate may thus be used as a transmitter in amacrine neurons for wide-field interconnections (see also [29], [30]). We also entertain the possibility that lamina monopolar neuron L2 may use glutamate for signaling within the lamina at some of its many minority classes of synapses, but that neither L1 nor L2 shows clear evidence of doing so at their chief output terminals in the medulla.

GABA signaling

Our immunocytochemical data show that C2 and C3 neurons (identified by Gal4-driven GFP) express both GABA and GAD. Neither of these neurons was detected using a GAD1-Gal4 line [96] tested here, and no other lamina neuron clearly expressed GAD1 or GABA immunoreactivity. An exclusive GABA phenotype among centrifugal neurons is confirms earlier reports on Drosophila and other flies [42], [43], [44], [45]. Previously we have also shown that the C2 and C3 neurons express the Drosophila vesicular GABA transporter [68], further suggesting that these neurons signal by means of GABA.

We localized GABA_BR immunoreactivity in relation to various identified neurons. For this we used an antiserum to the GABA_BR2, a G-protein coupled receptor known to dimerize with the GABA_BR1, to form a functional receptor complex [97], [98]. Thus our observations are likely to reveal functional GABA_B receptor sites (see [68]). At least three neuron types seem to express GABA_BRs: C2, C3, and the tangential neuron Cha-Tan. Possibly there is an additional neuron type not identified that may be postsynaptic to the C3 neurons that express GABA_BRs, since we also see immunoreactivity adjacent to C3's boutons. The likely contacts between GABAergic C2 neurons and large boutons of Cha-Tan neurons are quite distinct and express high levels of GABA_B receptor immunoreactivity. The presence of GABA_BR on C2 terminals in the distal lamina indicates the presence of presynaptic GABA receptors at a GABA output site of these neurons. Similarly GABA_BR immunoreactivity is associated with the varicosities of the GABAergic C3 neurons. These varicosities can be assumed to be GABA release sites, and are known to provide input to L1–L3 and amacrine cell processes and to receive no inputs themselves [9], [13]. Thus the GABA_B receptor may be presynaptic in both the C2 and C3 neurons. Both pre- and postsynaptic locations of GABA_BRs have in fact been identified in mammals (see [99]). There, presynaptic GABA_BR activation inhibits transmitter release by inhibiting voltage-gated Ca²⁺ channels via the β/γ subunit of the G-protein, or by inhibiting adenylate cyclase via G_i/o proteins [99]. In this way, GABA release from C2 or C3 may be negatively regulated.

The distribution of GABA_A type receptors in the lamina is still not clear, because the antiserum to the Drosophila GABA_A receptor subunit RDL failed to produce distinct lamina immunolabeling. An earlier study suggested that at least part of the RDL-immunolabeling may be localized to L2 monopolar cells [68]. Here we utilized an rdl-Gal4 line to drive GFP, and although we were unable to demonstrate that lamina neurons revealed by rdl-Gal4 actually produce RDL, a good match between the markers has been seen in many parts of the larval CNS (Enell and Nässel, unpublished). The rdl-Gal4 labels L4 monopolar neurons and rdl-Tan neurons. At least rdl-Tan neurons may be targets of GABAergic C2 neurons as seen in our study, whereas the L4 neurons are not known to be postsynaptic to either C2 or C3 neurons [9], so that receptor expression on these monopolar cells is unexplained and may be targeted to the medulla terminals.

In summary, GABA seems to be primarily (or exclusively) used by centrifugal neurons from the medulla with outputs in the lamina, one of which (C2) may signal to wide-field tangential neurons of the lamina.

Conclusions

This study has increased the number of lamina neurons for which a putative neurotransmitter has now been identified and has also localized GABA_B receptors to identified neurons (see Table 2). There are, of course, still many neuron types for which transmitters remain unknown. Perhaps the greatest mystery of all remains whether the large monopolar neurons utilize glutamate or acetylcholine as neurotransmitters, or whether they may possibly release both. They appear qualified to use either, but it is neither clear which they actually use, nor whether release is the same at sites in the lamina and medulla, or in different strata of these neuropils. It is also evident that glutamate receptors and RDL subunits of GABA_A receptors are expressed at levels too low to be reliably detected in the lamina. Our study now prompts the complete morphological characterization of the possibly novel types of tangential neurons Cha-Tan and rdl-Tan. These are perhaps variants of the La wf1 and 2 neurons already described. It is also urgent to determine the neurotransmitter of the L1 and L2 neurons and to localize ionotropic receptors for acetylcholine, GABA and glutamate in the lamina circuits.

Since, in contrast to the lamina of locusts, cockroaches or other non-dipteran insects [100], [101], [102], it appears that lamina interneurons in flies express neither monoamines such as histamine, dopamine, octopamine or serotonin (see [35], [103]) nor identified neuropeptides, further work will be required to screen for small-molecule neurotransmitters in those neurons not yet assigned a signal molecule. Co-expression of yet unidentified neuromodulators clearly remains an additional possibility, revealed for example by dense-core vesicles in C2 ([9]: their Fig. 36A), alongside the clear vesicles which we may now presume to contain GABA. Thus, the complete neurotransmitter repertoire of even the tiny constituency of neurons in the fly's lamina cartridge still awaits final identification.