To profile gene expression throughout the nervous system, we generated a stable, chromosomally integrated transgenic line expressing an epitope-tagged poly-A binding protein (FLAG::PAB-1) throughout the nervous system. Pan-neuronal expression was confirmed by immunostaining with a FLAG-specific antibody (Figure 1). We selected the second larval stage (L2) to test the application of the mRNA-tagging method. At this stage, the nervous system is largely in place and should, therefore, express a broad array of transcripts that define the development and function of most neurons. Sub-microgram quantities of mRNA isolated by the mRNA-tagging method were amplified and labeled for application to an Affymetrix chip representing approximately 90% of predicted C. elegans genes. Neuron-enriched transcripts in these samples were detected by comparison to a reference profile of all larval cells (see Materials and methods). We reasoned that this approach should detect a significant fraction of known neuronal transcripts and thus provide an initial test of the specificity of this strategy.

Comparisons of independently derived datasets for both the experimental (larval pan-neural) and reference samples showed that individual replicates for each condition are highly reproducible (Figure 2a,b). For example, an average coefficient of determination (R²) of approximately 0.96 was calculated from pairwise combinations of each individual reference dataset (Figure 2d). The pan-neural datasets were similarly reproducible (R² of approximately 0.96; Figure 2e). The overall concurrence of these data is graphically illustrated in the scatter plots shown in Figure 2a,b.

Scatter plots comparing larval pan-neural versus reference data revealed a substantial number of transcripts with significant differences in hybridization intensities (Figure 2c). Statistical analysis detected 1,562 transcripts with elevated expression (≥ 1.5-fold, ≤ 1% false discovery rate (FDR)) in the larval pan-neural sample (Additional data file 1). Strikingly, we found that 92% of the 443 genes with known expression patterns included in the larval pan-neural enriched dataset (409/443) are listed in WormBase 15 as neuronally expressed (Figure 3a; Additional data file 1). By contrast, only 57% of all genes (1,612/2,837) with defined expression patterns in WormBase are annotated as expressed in neurons (see Materials and methods; Figure 3a; Additional data files 2 and 3). Moreover, genes with key roles in neuronal function are highly represented in this list. For example, 55 transcripts encoding ion channels, receptors or membrane proteins with known expression in the C. elegans nervous system are enriched (Figure 3b; Additional data file 7). The enrichment of transcripts known to be expressed in neurons demonstrates that the larval pan-neural profile is largely derived from neural tissue. This conclusion is also substantiated by the finding that mRNAs highly expressed in other cell types are preferentially excluded from this dataset (Figure 2c). For example, microarray profiling experiments identified a total of 1,926 transcripts enriched in either larval germline, muscle or intestinal cells (GMI; Additional data file 5) 13. This set of genes is significantly under-represented (97/1,562) in the larval pan-neural dataset (representation factor 0.6, p < 2.033e^-9; a representation factor <1 indicates under-representation; see Materials and methods). Of the 97 genes that intersect our larval pan-neural profile and the GMI set, 35 have a previously characterized spatial expression pattern. Of these, 89% (31/35) are also expressed in neurons. A comparison of the top 50 most significantly enriched transcripts in a MAPCeL profile of embryonic body wall muscle cells (RMF, DMM, unpublished data) detected only four transcripts that also show elevated expression in the larval pan-neural profile (Figure 4a; Additional data file 6). Independent results have confirmed that at least one of these, the acetylcholine receptor subunit acr-16, is expressed in both muscle and neurons 1617. The apparent low frequency of false positives empirically defined by these comparisons is consistent with the estimated FDR of ≤ 1% for this dataset. The stringent exclusion of non-neuronal transcripts has been achieved, however, while retaining sensitivity to transcripts that may be expressed in limited numbers of neurons (Figure 5). For example, our methodology identifies genes that are expressed in only two neurons; daf-7 (transforming growth factor (TGF)-beta-like peptide expressed in ASIL and ASIR) 18 and gcy-8 (guanylate cyclase expressed in AFDL and AFDR) 19 (Figure 5).

The strong enrichment of known neuronal genes in the larval pan-neural dataset indicates that other previously uncharacterized transcripts in this list are also likely to be expressed in the nervous system. To test this prediction, we evaluated GFP reporter genes for representative transcripts in this profile. As shown in Table 1 and Additional data file 17, all but one of the transgenic lines (24 of 25) derived from these promoter GFP fusions show expression in neurons (Figure 6). Of the GFP reporters tested, 56% (14/25) are exclusively detected in neurons (Additional data file 17). For example, the stomatin gene sto-4 is highly expressed in ventral cord motor neurons, touch neurons and in head and tail ganglia (Table 1; Figure 6d,h). Our GFPreporter analysis demonstrates that the remaining 11 genes tested are expressed in other tissues in addition to neurons. For instance, the GFP reporter for C04E12.7 (phospholipid scramblase), which is expressed widely throughout the nervous system, is also expressed in muscle cells (Table 1; Figure 6c). Thus, these results indicate that the genes identified in the larval pan-neural profile largely fall into two classes; those that are exclusively expressed in neurons, and those that are expressed in multiple tissues, including neurons. Our finding of neuronal GFP expression for transcripts exhibiting a wide range of enrichment (1.5- to 8.3-fold) predicts that most of the genes in this list that have not been directly tested are also likely to be expressed in neurons. Together, these results demonstrate that our pan-neural mRNA-tagging approach enriches for bona fide neuronally expressed transcripts and effectively excludes transcripts expressed exclusively in other tissues.

Protein-encoding genes in the enriched larval pan-neural profile were organized into groups on the basis of KOGs and other descriptions that identify functional or structural categories (Table 2; Additional data file 4) 20. Over half (880/1,562) are homologous to proteins in at least one other widely diverged eukaryotic species (that is, KOGs and TWOGs), 49 of which are classified as uncharacterized conserved proteins. Homologs for an additional 225 pan-neural enriched proteins are limited to other nematode species (that is, LSEs).

Transcripts encoding proteins with fundamental roles in neuronal activity or signaling are highly represented in this dataset (for a comprehensive list see Additional data file 4). For example, in addition to the 34 synaptic vesicle (SV) associated transcripts from Figure 3b (Additional data file 7), transcripts for 19 proteins with potential roles in synaptic vesicle function are identified (Figure 7). These include six members of the synaptotagmin family of calcium-dependent phospholipid binding proteins (snt-1, snt-4, snt-5, snt-6, DH11.4, T10B10.5), only one of which, snt-1, has been previously shown to function in neurons 21. Expression of the additional synaptotagmin genes in the nervous system may account for the residual synaptic vesicle function of snt-1 mutants 21. Three members of the copine family (B0495.10, tag-64, T28F3.1), a related group of calciumbinding proteins with potential roles in synaptic vesicle fusion (listed as part of endocytosis machinery in Figure 7), are also enriched 22.

In addition to genes with general functions in synaptic vesicle signaling, the larval pan-neural profile includes transcripts encoding proteins with roles specific to particular neurotransmitters. For example, the plasma membrane and vesicular transporters for choline and acetylcholine (cho-1 and unc-17), GABA (snf-11 and unc-46, unc-47), dopamine (dat-1 and cat-1), and glutamate (glt-3 and eat-4) are included (Figure 7) 2324252627. The corresponding families of neurotransmitter-specific ligand-gated ion channels are highly represented, including 22 members of the ionotropic nicotinic acetylcholine (ACh) receptor family (Additional data file 4). Other classes of ion channels with key neural functions are also abundant, such as potassium channels (24), voltage-gated calcium channels (10) and DEG/ENaC sodium channels (10) (Table 2).

The wide range of neurotransmitter-specific genes in the larval pan-neural dataset reflects the diverse array of neuron types in C. elegans (Figure 5). This point is underscored by the detection of a large number of transcription factors with established roles in neuronal specification (Table 3). These include UNC-86, the POU homeodomain protein that regulates the differentiation of a broad cross-section of neuron classes 282930, as well as transcription factors that define specific neuronal subtypes, such as the canonical LIM homeodomain MEC-3 (mechanosensory neurons) 313233 and the UNC-4 homeodomain (A-class ventral cord motor neurons, see below) 34353637. Transcription factors with undefined roles in the nervous system are also identified. Of particular note are 15 members of the nuclear hormone receptor (NHR) family, only one of which, fax-1, has been previously shown to regulate neuronal differentiation 38.