Surprising gene expression patterns within and between PDF-containing circadian neurons in Drosophila

Abstract

To compare circadian gene expression within highly discrete neuronal populations, we separately purified and characterized two adjacent but distinct groups of Drosophila adult circadian neurons: the 8 small and 10 large PDF-expressing ventral lateral neurons (s-LNvs and l-LNvs, respectively). The s-LNvs are the principal circadian pacemaker cells, whereas recent evidence indicates that the l-LNvs are involved in sleep and light-mediated arousal. Although half of the l-LNv-enriched mRNA population, including core clock mRNAs, is shared between the l-LNvs and s-LNvs, the other half is l-LNv- and s-LNv-specific. The distribution of four specific mRNAs is consistent with prior characterization of the four encoded proteins, and therefore indicates successful purification of the two neuronal types. Moreover, an octopamine receptor mRNA is selectively enriched in l-LNvs, and only these neurons respond to in vitro application of octopamine. Dissection and purification of l-LNvs from flies collected at different times indicate that these neurons contain cycling clock mRNAs with higher circadian amplitudes as well as at least a 10-fold higher fraction of oscillating mRNAs than all previous analyses of head RNA. Many of these cycling l-LNv mRNAs are well expressed but do not cycle or cycle much less well elsewhere in heads. The results suggest that RNA cycling is much more prominent in circadian neurons than elsewhere in heads and may be particularly important for the functioning of these neurons.

Fig. 1.

Gene expression can be directly assessed in the purified clock neurons. (A) l-LNvs and s-LNvs can be identified by size differences of their somas. (B) Experimental overview. l-LNvs and s-LNvs were purified separately. To identify circadian genes, we used a panneuronal driver, elav-GAL4. Enrichment analysis was performed relative to elav/EGFP neurons. (C) Pdf mRNA is highly enriched in both l-LNvs and s-LNvs relative to generic neurons (elav-GAL4 driver in combination with UAS-EGFP). All comparisons are based on chip data that were assayed in triplicate (elav/EGFP, large/EGFP cells) or duplicate (small/EGFP cells). (D) qPCR validation of Pdf mRNA enrichment in the l-LNvs and s-LNvs. Gene expression was assayed in triplicate and normalized to calmodulin (11). Enrichment is shown as fold-change relative to ELAV cells. Error bars represent SEM. (E) Clock genes are highly enriched in LNvs. Enrichment analysis was performed relative to elav-GAL4/EGFP neurons. Clock gene mRNAs rank high in both l-LNvs and s-LNvs and essentially show similar enrichment in these two cell populations (1 indicates top enriched, 2 indicates next most enriched, etc.). Essentially identical results were obtained when the chips were divided in half and analyzed separately. An arrow points to Pdp1, an outlier among clock genes.

Fig. 2.

Number of l-LNv– and s-LNv–specific mRNAs identified from purified clock neurons. (A) Examples of candidate genes showing high enrichment in both l-LNvs and s-LNvs (DH31R) and enrichment in s-LNvs but not in large cells (PDFR, sNPF) as well as two octopamine receptor mRNAs (Oa2 and Oamb) specifically enriched in l-LNvs but not in small cells; ranking in l-LNvs and s-LNvs is relative to ELAV cells, whereas ranking in s-LNvs and L-LNvs directly compares the expression levels in small vs. large PDF cells [in parentheses, a reverse ranking (i.e., l-LNvs/s-LNvs)] (SI Materials and Methods). (B) l-LNvs but not s-LNvs respond potently to octopamine. A star indicates statistically significant differences (t test, P < 0.05).

Fig. 3.

Most clock gene mRNAs cycle with much higher amplitude in pacemaker cells than in heads. per and vri mRNA (A), Pdp1 and cwo (B), Clk (C), and tim probes (D) are shown across four time points in clock neurons (ZT0, ZT6, ZT12, and ZT18) and in heads (ZT3, ZT9, ZT15, and ZT21). Cell and head data represent independent samples [GeneSpring (Agilent Technologies); one-way ANOVA at P < 0.05, 5% false discovery rate, and fold change = at least 1.8; SI Text].

Fig. 4.

Assay of previously identified cycling head genes in clock neurons. Ugt35b (A) and CG17386 (B) are examples of genes cycling robustly in heads but not in the clock neurons. (C) CG5798 is the only exception because it shows robust cycling in clock neurons. Cell and head data are from independent samples.

Fig. 5.

Many more cycling genes identified in clock neurons than in heads. (A) Microarray data collected from large PDF neurons at the indicated time points are shown. GeneSpring (Agilent Technologies) analysis was applied to identify cycling transcripts: one-way ANOVA at P < 0.05, 5% false discovery rate (FDR), and fold change = at least 1.8. To analyze circadian gene expression, we looked at four time points and compared gene expression levels between three pairs: ZT0 vs. ZT12, ZT6 vs. ZT18, and ZT0 vs. ZT18. Each comparison yields a set of genes (e.g., ANOVA showed that there were 1,321 differentially expressed genes between ZT0 vs. ZT12). To identify genes overlapping between different time point comparisons, we cross-referenced the datasets. This analysis showed, for example, that there were 225 mRNAs that were identified as differentially expressed between the above time points. Also, 564 genes were only significant between ZT0 vs. ZT12. (B) Data analysis on head mRNA at two time points revealed many fewer cycling genes than the same comparison on clock neuron mRNA (two time points examined: ZT0 and ZT12). After normalization, microarray data were analyzed to determine the number of genes differentially expressed in heads and PDF neurons (GeneSpring; one-way ANOVA at P < 0.05, 5% FDR, and FC = at least 1.8). The bars represent the number of differentially expressed genes between ZT0 and ZT12 in heads and cells. There are about 13 times more differentially expressed mRNAs in PDF neurons than in heads: CG4726 (C), Pka-R1 (D), Pka-R2 (E), and Ir (F). (C–F) Examples of genes cycling only in the clock neurons.