Report

Choice Behavior of Drosophila Facing Contradictory Visual Cues

See allHide authors and affiliations

Science  16 Nov 2001:
Vol. 294, Issue 5546, pp. 1543-1547
DOI: 10.1126/science.1058237

Abstract

We studied the underlying neural mechanism of a simple choice behavior between competing alternatives in Drosophila. In a flight simulator, individual flies were conditioned to choose one of two flight paths in response to color and shape cues; after the training, they were tested with contradictory cues. Wild-type flies made a discrete choice that switched from one alternative to the other as the relative salience of color and shape cues gradually changed, but this ability was greatly diminished in mutant (mbm1 ) flies with miniature mushroom bodies or with hydroxyurea ablation of mushroom bodies. Thus,Drosophila genetics may be useful for elucidating the neural basis of choice behavior.

Most animals can make a rapid and rational choice among alternative behaviors by assessing the advantages and disadvantages on the basis of previous experience, but the underlying neural mechanism is largely unknown (1). Studies of simple organisms have made important contributions to our understanding of the cellular and molecular basis of learning and memory, as well as other cognitive functions (2–5). Here, we used the visual learning paradigm developed by Wolf and Heisenberg (6) to examine choice behavior inDrosophila. Individual flies were presented with two conflicting cues, which they had been previously conditioned to follow, and their choice behavior was examined.

An individual fly in a flight simulator was trained to associate a particular visual pattern with a punishment (heat). In a typical protocol, the fly was first examined during a test period (three 2-min blocks) for its directional preference for various patterns in the flight arena. This was followed by two training sessions (two 2-min blocks each, spaced by one 2-min test block) during which the heat was switched on whenever a particular pattern entered the frontal 90° sector of the fly's visual field. The posttraining test sessions consisted of four 2-min blocks without heat application. In experiments in which wild-type Berlin (WTB) flies (5) were trained to follow a color cue (7), the flies showed no color preference between a green T and a blue T on a dark background during the pretraining test (preference index PI1–3 ∼ 0) (6) (Fig. 1A). After the training to associate the heat punishment with the blue T, they exhibited persistent preference for the green T (positive PI9–12). The flight path angular histograms in Fig. 1B (average of 26 flies) depict, at 0.5-min intervals, the relative amount of time spent by the flies in different directions between –90° and +90° relative to the location of the green T. However, when the same experiment was carried out with colored Ts on a white background, the flies failed to learn a preferred color T (Fig. 1C). White background illumination may result in a loss of color sensitivity of fly photoreceptors (8); thus, a dark background was used in all experiments. In the absence of color cues, the WTB flies could also learn to choose correctly between a white upright T and an inverted T (Fig. 1D). Thus, WTB flies can use either color or shape cue alone in visual learning.

Figure 1

Visual learning of Drosophila using single visual cues. (A) WTB flies were trained to prefer a green upright T over a blue upright T in the flight simulator. Histograms of preference index (PI) showed mean positive values during training (PI4–5 and PI7–8, SEM, n = 26 flies) and after training (PI6 and PI9–12). Similar results were found when the flies were trained to prefer a blue T over a green T. (B) Flight path angular histograms of the relative time spent by flies in different flight directions between –90° and +90° relative to the location of the green upright T, estimated over a 0.5-min interval for each histogram and averaged for 26 flies [same set as in (A)]. (C) Failure of visual learning for a colored cue on a white background (n = 20). PI showed high positive value during training (PI4–5 and PI7–8), but memory retention near zero after training (PI6 and PI9–12). (D) Visual learning with the shape cue. The flies were trained to prefer a white upright T over a white inverted T on a dark background. Histograms of PI showed positive values during and after training (n = 25). (E) The flies use the color rather than the intensity of the colored light for visual learning. The flies were trained to prefer color cues of high or low intensity; after training, they were tested with color cues with intensity values reversed. Left panel: Spectra of blue and green light and the relative values used for the high- and low-level intensities. Right panel: Histograms summarizing the data for all experiments using color/intensity reversal during the posttraining test. The positive values of PI10–12 (n = 21) indicate flight behavior in favor of the color cue.

The Drosophila eye is differentially sensitive to green and blue light (9). We made use of this fact to investigate whether the fly indeed uses color rather than light intensity in its visual learning of colored patterns. The WTB flies were trained to follow a high-intensity blue T (Bh) instead of a low-intensity green T (Gl). After the training, the flies were tested for their preference between a high-intensity green T (Gh) and a low-intensity blue T (Bl). The light intensity was set quantitatively equal between blue and green lights at high and low intensities, respectively, and the intensity of Gh was set above that of Bl, and that of Bh above Gl, at all wavelengths (Fig. 1E). Posttraining tests showed that the flies chose to follow the color cue in favor of the low-intensity blue T. Similar preference for color rather than intensity cue was found when the flies were trained to prefer Gh over Bl or to prefer Gl over Bh (Fig. 1E). Thus, the flies were indeed using color vision in visual learning in the present experimental protocol.

Can flies use color and shape cues simultaneously in visual learning? If so, can memory be selectively retrieved independently by a single cue, and is there an overshadowing effect of one cue over the other during training? To address these questions, we trained flies to choose between a green upright T and a blue inverted T, with color intensity (CI) set at a maximal value of 1.0 (7). After training with the double cues, the flies were tested with a single cue (Fig. 2A). For memory retrieval with the color cue, flies were tested for preference between a green upright T and a blue upright T. For retrieval with the shape cue, flies were tested for preference between a white upright T and a white inverted T. Significant retrieval of memory was found in both cases (Fig. 2, B and C), with PI values close to those observed after single-cue training (Fig. 1, A and D). Furthermore, the shape cue was retrieved equally well over the entire range of CI from 1.0 to 0 (Fig. 2D), suggesting the absence of any overshadowing of the shape cue by the color cue for both WTB and mbm1 mutant flies (described below). In addition, over a large range of CI values (1.0 to 0.4), the color cue could also be retrieved (Fig. 2E), which suggests that there was no overshadowing of the color cue by the shape cue (10). Thus, flies can simultaneously use two types of cues for visual learning, and each cue can be used for memory retrieval independently.

Figure 2

Visual learning using double cues and the test of the overshadowing effect. (A) WTB flies were trained with the color and shape cues at a high color intensity (CI = 1.00) and were tested with a single cue during the posttraining session at the same CI. (B) After training to prefer a green upright T over a blue inverted T, flies were tested for their behavior based only on the color cue (to choose between a green upright T and a blue upright T). The PI10–12 values for WTB flies were positive (n = 26). (C) After training to prefer a green upright T over a blue inverted T, flies were tested for their behavior based only on the shape cue (to choose between a white upright T and a white inverted T). The PI10–12 values for WTB flies were positive (n = 21). (D) The experiment described in (C) was repeated for WTB (n = 131) and mbm1 (n = 93) flies over the entire range of CI values (1.0 to 0). The number of flies for each CI value tested ranged from 10 to 25. No difference was found between WTB and mbm1 flies (P = 0.66, Student's paired t test). (E) The experiment described in (B) was repeated for WTB (n = 96) and mbm1 (n = 80) flies over the range of CI from 1.0 to 0. No difference was found between WTB andmbm1 flies (P = 0.89, Student's paired t test). (F) Control experiments that examined the color sensitivity and the role of light intensity in the choice behavior of WTB and mbm1 flies. The flies were trained with both color and shape cues at CI = 0.80 (as in Fig. 3A) and were presented with only the color cue (upright green T versus upright blue T) at CI = 1.0 to 0 during the posttraining test (T = 18 to 24 min). The dependence of PI10–12 on CI was identical for WTB (n = 54) and mbm1 (n = 55) flies (n = 8 to 10 for each CI value).

The use of multiple cues in visual learning may result in a dilemma for the fly when it encounters conflicting information. We presented a “color/shape” dilemma to the fly after the training by reversing the matching rules for the color and shape of T patterns. The fly was first trained to follow the green upright T and avoid the blue inverted T. During the posttraining test, the green upright T was changed to a green inverted T, whereas the blue inverted T became a blue upright T. The fly may choose the green inverted T on the basis of color, or may choose the blue upright T on the basis of shape. In making a choice, the fly must evaluate the relative weight of the color and shape cues. We found that the outcome of the fly's choice depended critically on the color intensity of the patterns during the test. When color intensity was high (CI = 1.0), the WTB flies made their choice according to the color cue, whereas at a lower color intensity (CI = 0.7) the WTB flies chose to follow the shape cue, with a preference for the blue upright T. At CI = 0.8, however, the flies showed equal preference for color and shape cues, which suggests that these cues exerted equal weight in choice behavior at this particular CI.

To further examine the choice behavior, we trained WTB flies to follow a green upright T and to avoid a blue inverted T at CI = 0.8, thus allowing learning of color and shape cues with “equal” weight. The flies were then tested at different levels of color intensity (CI = 1.00 to 0.00) with color and shape cues reversed.Figure 3A depicts the mean PI10–12 as a function of CI, obtained from a total of 165 WTB flies. When the intensity of the color cue was strong during the posttraining test, the flies made their choice in favor of the color. However, they chose to follow the shape cue when the color intensity dropped below CI = 0.80. These WTB flies exhibited a sharp and complete transition in flight behavior within a narrow range of CI values from 0.84 to 0.76. Beyond the transition zone, the mean PI10–12 exhibited a relatively constant positive or negative value, suggesting stable behavior in accordance with color or shape cue, respectively. The existence of a discrete transition point in the flight behavior suggests that the fly can make firm and stable choices on the basis of small differences in the relative salience of competitive cues.

Figure 3

Choice behaviors of WTB,mbm1 , and HU-treated flies facing the “color/shape dilemma.” (A) The flies were trained at CI = 0.8 to prefer a green upright T over a blue inverted T, but were presented with conflicting cues (green inverted T and blue upright T) during the posttraining session over the entire range of CI values (1.00 to 0.00). Data points represent mean PI10–12values (SEM; n = 10 to 38 for each point). (B) Angular histograms during posttraining sessions for WTB and mbm1 flies. The flies were presented with original nonconflicting cues during the first posttraining test session (T = 16 to 18 min), followed by conflicting cues in subsequent test sessions (T = 18 to 24 min). Data belong to the same set as shown in (A).

To gain insight into the underlying neural mechanism of choice behavior in Drosophila, we have examined the role of mushroom bodies (MBs), structures proposed to endow the insect with a degree of “free will” or “intelligent control” over instinctive actions (11). MBs are involved in multimodal sensory processing (12) and context generalization (5), and so may be involved in making choices. We found thatmbm1 mutant flies, which have miniature MBs, exhibited a choice behavior distinctly different from that of WTB flies (Fig. 3A). There was no sharp transition in flight behavior ofmbm1 flies as CI was gradually reduced (Fig. 3A), indicating indecisive choice-making over a wide range of color intensity (CI = 0.92 to 0.52) when conflicting cues were presented. Similar results were observed for flies in which the MB neuroblasts were ablated by applying the cytostatic drug hydroxyurea (HU) during the early first larval instar (13) (Fig. 3A). The absence of MBs in HU-treated flies [P-GAL4 line OK107 crossed to UAS-GFP (14)] was confirmed by fluorescence microscopy. The flight path angular histograms ofmbm1 flies during the posttraining test (T = 18 to 24 min) with conflicting cues showed equally strong preference for the color and shape cues (Fig. 3B), which suggests that the flies failed to make a firm choice between two alternatives. A posttraining test (T = 16 to 18 min) with the original nonconflicting cues showed thatmbm1 flies exhibited normal learning and memory (Fig. 3B).

The difference in choice behavior was not due to any difference in color sensitivity between WTB and mbm1 flies. When the flies were trained with double cues but tested after training with only the color cue, the resulting PI10–12 showed a gradual reduction for tests with CI = 1.0 to 0, with no difference between WTB and mbm1 flies (Fig. 2F). Furthermore, over the critical range of CI = 1.0 to 0.6, both WTB and mbm1 flies showed normal flight behavior based on the color cue, indicating that the difference in color sensitivity cannot account for the difference in choice behavior.

In summary, we have shown that (i) flies can make discrete and firm choices among behavioral alternatives on the basis of small differences in the saliency of visual cues; (ii) the behavioral choice is not based on overshadowing or selective visual attention, but is due to a process that can assess the relative saliency of conflicting visual cues; and (iii) MBs are the likely site for such choice behavior. Hence, our results support the notion that the core function of MBs in insects is to mediate intelligent behavior (11).

  • * To whom correspondence should be addressed. E-mail: akguo{at}ion.ac.cn

REFERENCES AND NOTES

View Abstract

Navigate This Article